Hospital Medicine

Boughton et al.¹ reported a high incidence of hypoglycemia resulting from glucose-with-insulin (GwI) infusion used to treat acute hyperkalemia. This has been reported by other investigators—particularly in subjects without preexisting diabetes² and resonates with the experiences of clinicians practicing in Internal Medicine or Diabetes.

The authors suggested that patients at risk of hypoglycemia be identified and offered a regimen containing less insulin. However, for subjects without preexisting diagnosis and not at high risk of diabetes, we question the physiological logic and the safety basis for administering insulin.

Infusion of glucose only (GO) to subjects with intact pancreatic function and insulin sensitivity stimulates endogenous insulin secretion in a dose-dependent manner, resulting in a reduction in extracellular fluid potassium with no risk of hypoglycemia.^3,4

It is unclear why GwI historically entered mainstream practice rather than GO, but the rationale may have been based on the potential risks of paradoxical hyperglycemia-mediated hyperkalemia (HMK) being induced by GO. In practice, HMK was only observed in subjects with diabetes.⁵

As there is an ongoing need to reduce the impact of iatrogenic hypoglycemia, revisiting of the prematurely abandoned GO regimen in hyperkalemia management is warranted. Such approach may offer a safe and physiological alternative to GwI in nondiabetic patients with hyperkalemia.

We advocate that GO be prospectively evaluated against GwI for the treatment of hyperkalemia in subjects without diabetes, against the endpoints being noninferiority in respect of efficacy and maintenance of euglycemia in respect of safety.

Disclosures

Nothing to declare.

Boughton et al.¹ reported a high incidence of hypoglycemia resulting from glucose-with-insulin (GwI) infusion used to treat acute hyperkalemia. This has been reported by other investigators—particularly in subjects without preexisting diabetes² and resonates with the experiences of clinicians practicing in Internal Medicine or Diabetes.

The authors suggested that patients at risk of hypoglycemia be identified and offered a regimen containing less insulin. However, for subjects without preexisting diagnosis and not at high risk of diabetes, we question the physiological logic and the safety basis for administering insulin.

Infusion of glucose only (GO) to subjects with intact pancreatic function and insulin sensitivity stimulates endogenous insulin secretion in a dose-dependent manner, resulting in a reduction in extracellular fluid potassium with no risk of hypoglycemia.^3,4

It is unclear why GwI historically entered mainstream practice rather than GO, but the rationale may have been based on the potential risks of paradoxical hyperglycemia-mediated hyperkalemia (HMK) being induced by GO. In practice, HMK was only observed in subjects with diabetes.⁵

As there is an ongoing need to reduce the impact of iatrogenic hypoglycemia, revisiting of the prematurely abandoned GO regimen in hyperkalemia management is warranted. Such approach may offer a safe and physiological alternative to GwI in nondiabetic patients with hyperkalemia.

We advocate that GO be prospectively evaluated against GwI for the treatment of hyperkalemia in subjects without diabetes, against the endpoints being noninferiority in respect of efficacy and maintenance of euglycemia in respect of safety.

Disclosures

Nothing to declare.

Boughton et al.¹ reported a high incidence of hypoglycemia resulting from glucose-with-insulin (GwI) infusion used to treat acute hyperkalemia. This has been reported by other investigators—particularly in subjects without preexisting diabetes² and resonates with the experiences of clinicians practicing in Internal Medicine or Diabetes.

The authors suggested that patients at risk of hypoglycemia be identified and offered a regimen containing less insulin. However, for subjects without preexisting diagnosis and not at high risk of diabetes, we question the physiological logic and the safety basis for administering insulin.

Infusion of glucose only (GO) to subjects with intact pancreatic function and insulin sensitivity stimulates endogenous insulin secretion in a dose-dependent manner, resulting in a reduction in extracellular fluid potassium with no risk of hypoglycemia.^3,4

It is unclear why GwI historically entered mainstream practice rather than GO, but the rationale may have been based on the potential risks of paradoxical hyperglycemia-mediated hyperkalemia (HMK) being induced by GO. In practice, HMK was only observed in subjects with diabetes.⁵

As there is an ongoing need to reduce the impact of iatrogenic hypoglycemia, revisiting of the prematurely abandoned GO regimen in hyperkalemia management is warranted. Such approach may offer a safe and physiological alternative to GwI in nondiabetic patients with hyperkalemia.

We advocate that GO be prospectively evaluated against GwI for the treatment of hyperkalemia in subjects without diabetes, against the endpoints being noninferiority in respect of efficacy and maintenance of euglycemia in respect of safety.

Disclosures

Nothing to declare.

With advancement in imaging techniques, incidental pulmonary nodules (IPNs) are routinely found on imaging studies. Depending on the size, an IPN has diagnostic uncertainty. Is it a benign finding? Will it progress to cancer? These questions have the potential to create anxiety for our patients. Between 2012 and 2014, 19,739 patients were discharged from hospitals in the United States with a diagnosis of a solitary pulmonary nodule.¹ Roughly 7,500 were discharged after an inpatient stay; the remainder from the emergency room. Aggregate costs for these visits totaled $49 million. The exact number of nodules receiving follow-up is unknown.

The Fleischner guidelines, updated in 2017, outline management for IPNs.² Depending on nodule size and patient risk factors, repeat imaging is either not indicated or one to two follow-up scans could be recommended. In this issue of the Journal of Hospital Medicine^®, two reports assess provider awareness of the Fleischner guidelines and examine the proportion of patients receiving follow-up.

Umscheid et al. surveyed hospitalists to understand their approach IPN management. Of 174 respondents, 42% were unfamiliar with the Fleischner guidelines.3 The authors proposed methods for improving provider awareness, including better communication between hospitalists and primary care providers, better documentation, and in the case of their institution, the development of an IPN consult team. The IPN consult team is composed of a nurse practitioner and pulmonologist. They inform primary care providers of patient findings and need for follow-up. If no follow-up is made, the team will see the patients in an IPN ambulatory clinic to ensure follow-up imaging is obtained.

Kwan et al. found that fewer than 50% of patients with high-risk new pulmonary nodules received follow-up.4 Although a single-site study, the study is consistent with prior work on tests pending at discharge, which essentially show that there are poor follow-up rates.^5,6 Follow-up was more likely when the IPN was mentioned in the discharge summary. This conclusion builds on previous work showing that IPNs are more likely to be included in a discharge summary if the nodule is noted in the report heading, the radiologist recommends further imaging, and the patient is discharged from a medicine service as opposed to a surgical service.7 IPN follow-up is less likely if results are mentioned in the findings section alone.⁵

IPN follow-up is a piece of a larger issue of how best to ensure appropriate follow-up of any tests pending after discharge. A systematic review of discharge interventions found improvement in follow-up when discharge summaries are combined with e-mail alerts.⁶ A study of the effects of integrated electronic health records (EHR) web modules with discharge specific instructions showed an increase in follow-up from 18% to 27%.8 Studies also consider provider-to-patient communication. One intervention uses the patient portal to remind patients to pick up their medications,9 finding a decrease in nonadherence from 65.5% to 22.2%. Engaging patients by way of patient portals and reminders are an effective way to hold both the physician and the patient accountable for follow-up. Mobile technologies studied in the emergency department show patient preferences toward texting to receive medication and appointment reminders.10 Given wide-spread adoption of mobile technologies,11 notification systems could leverage applications or texting modalities to keep patients informed of discharge appointments and follow-up imaging studies. Similar interventions could be designed for IPNs using the Fleischner guidelines, generating alerts when patients have not received follow-up imaging.

The number of IPNs identified in the hospital will likely remain in the tens of thousands. From the hospitalist perspective, the findings presented in this month’s Journal of Hospital Medicine suggest that patients be educated about their findings and recommended follow-up, that follow-up be arranged before discharge, and that findings are clearly documented for patients and primary care providers to review. More study into how to implement these enhancements is needed to guide how we focus educational, systems, and technological interventions. Further study is also needed to help understand the complexities of communication channels between hospitalists and primary care physicians. As hospitalist workflow is more integrated with the EHR and mobile technology, future interventions can facilitate follow-up, keeping all providers and, most importantly, the patient aware of the next steps in care.

Acknowledgments

Author support is provided by the South Texas Veterans Health Care System. The views expressed are those of the authors and do not reflect the position or policy of the Department of Veterans Affairs.

Disclosures

The authors report no financial conflicts of interest.

With advancement in imaging techniques, incidental pulmonary nodules (IPNs) are routinely found on imaging studies. Depending on the size, an IPN has diagnostic uncertainty. Is it a benign finding? Will it progress to cancer? These questions have the potential to create anxiety for our patients. Between 2012 and 2014, 19,739 patients were discharged from hospitals in the United States with a diagnosis of a solitary pulmonary nodule.¹ Roughly 7,500 were discharged after an inpatient stay; the remainder from the emergency room. Aggregate costs for these visits totaled $49 million. The exact number of nodules receiving follow-up is unknown.

The Fleischner guidelines, updated in 2017, outline management for IPNs.² Depending on nodule size and patient risk factors, repeat imaging is either not indicated or one to two follow-up scans could be recommended. In this issue of the Journal of Hospital Medicine^®, two reports assess provider awareness of the Fleischner guidelines and examine the proportion of patients receiving follow-up.

Umscheid et al. surveyed hospitalists to understand their approach IPN management. Of 174 respondents, 42% were unfamiliar with the Fleischner guidelines.3 The authors proposed methods for improving provider awareness, including better communication between hospitalists and primary care providers, better documentation, and in the case of their institution, the development of an IPN consult team. The IPN consult team is composed of a nurse practitioner and pulmonologist. They inform primary care providers of patient findings and need for follow-up. If no follow-up is made, the team will see the patients in an IPN ambulatory clinic to ensure follow-up imaging is obtained.

Kwan et al. found that fewer than 50% of patients with high-risk new pulmonary nodules received follow-up.4 Although a single-site study, the study is consistent with prior work on tests pending at discharge, which essentially show that there are poor follow-up rates.^5,6 Follow-up was more likely when the IPN was mentioned in the discharge summary. This conclusion builds on previous work showing that IPNs are more likely to be included in a discharge summary if the nodule is noted in the report heading, the radiologist recommends further imaging, and the patient is discharged from a medicine service as opposed to a surgical service.7 IPN follow-up is less likely if results are mentioned in the findings section alone.⁵

IPN follow-up is a piece of a larger issue of how best to ensure appropriate follow-up of any tests pending after discharge. A systematic review of discharge interventions found improvement in follow-up when discharge summaries are combined with e-mail alerts.⁶ A study of the effects of integrated electronic health records (EHR) web modules with discharge specific instructions showed an increase in follow-up from 18% to 27%.8 Studies also consider provider-to-patient communication. One intervention uses the patient portal to remind patients to pick up their medications,9 finding a decrease in nonadherence from 65.5% to 22.2%. Engaging patients by way of patient portals and reminders are an effective way to hold both the physician and the patient accountable for follow-up. Mobile technologies studied in the emergency department show patient preferences toward texting to receive medication and appointment reminders.10 Given wide-spread adoption of mobile technologies,11 notification systems could leverage applications or texting modalities to keep patients informed of discharge appointments and follow-up imaging studies. Similar interventions could be designed for IPNs using the Fleischner guidelines, generating alerts when patients have not received follow-up imaging.

The number of IPNs identified in the hospital will likely remain in the tens of thousands. From the hospitalist perspective, the findings presented in this month’s Journal of Hospital Medicine suggest that patients be educated about their findings and recommended follow-up, that follow-up be arranged before discharge, and that findings are clearly documented for patients and primary care providers to review. More study into how to implement these enhancements is needed to guide how we focus educational, systems, and technological interventions. Further study is also needed to help understand the complexities of communication channels between hospitalists and primary care physicians. As hospitalist workflow is more integrated with the EHR and mobile technology, future interventions can facilitate follow-up, keeping all providers and, most importantly, the patient aware of the next steps in care.

Acknowledgments

Author support is provided by the South Texas Veterans Health Care System. The views expressed are those of the authors and do not reflect the position or policy of the Department of Veterans Affairs.

Disclosures

The authors report no financial conflicts of interest.

With advancement in imaging techniques, incidental pulmonary nodules (IPNs) are routinely found on imaging studies. Depending on the size, an IPN has diagnostic uncertainty. Is it a benign finding? Will it progress to cancer? These questions have the potential to create anxiety for our patients. Between 2012 and 2014, 19,739 patients were discharged from hospitals in the United States with a diagnosis of a solitary pulmonary nodule.¹ Roughly 7,500 were discharged after an inpatient stay; the remainder from the emergency room. Aggregate costs for these visits totaled $49 million. The exact number of nodules receiving follow-up is unknown.

The Fleischner guidelines, updated in 2017, outline management for IPNs.² Depending on nodule size and patient risk factors, repeat imaging is either not indicated or one to two follow-up scans could be recommended. In this issue of the Journal of Hospital Medicine^®, two reports assess provider awareness of the Fleischner guidelines and examine the proportion of patients receiving follow-up.

Umscheid et al. surveyed hospitalists to understand their approach IPN management. Of 174 respondents, 42% were unfamiliar with the Fleischner guidelines.3 The authors proposed methods for improving provider awareness, including better communication between hospitalists and primary care providers, better documentation, and in the case of their institution, the development of an IPN consult team. The IPN consult team is composed of a nurse practitioner and pulmonologist. They inform primary care providers of patient findings and need for follow-up. If no follow-up is made, the team will see the patients in an IPN ambulatory clinic to ensure follow-up imaging is obtained.

Kwan et al. found that fewer than 50% of patients with high-risk new pulmonary nodules received follow-up.4 Although a single-site study, the study is consistent with prior work on tests pending at discharge, which essentially show that there are poor follow-up rates.^5,6 Follow-up was more likely when the IPN was mentioned in the discharge summary. This conclusion builds on previous work showing that IPNs are more likely to be included in a discharge summary if the nodule is noted in the report heading, the radiologist recommends further imaging, and the patient is discharged from a medicine service as opposed to a surgical service.7 IPN follow-up is less likely if results are mentioned in the findings section alone.⁵

IPN follow-up is a piece of a larger issue of how best to ensure appropriate follow-up of any tests pending after discharge. A systematic review of discharge interventions found improvement in follow-up when discharge summaries are combined with e-mail alerts.⁶ A study of the effects of integrated electronic health records (EHR) web modules with discharge specific instructions showed an increase in follow-up from 18% to 27%.8 Studies also consider provider-to-patient communication. One intervention uses the patient portal to remind patients to pick up their medications,9 finding a decrease in nonadherence from 65.5% to 22.2%. Engaging patients by way of patient portals and reminders are an effective way to hold both the physician and the patient accountable for follow-up. Mobile technologies studied in the emergency department show patient preferences toward texting to receive medication and appointment reminders.10 Given wide-spread adoption of mobile technologies,11 notification systems could leverage applications or texting modalities to keep patients informed of discharge appointments and follow-up imaging studies. Similar interventions could be designed for IPNs using the Fleischner guidelines, generating alerts when patients have not received follow-up imaging.

The number of IPNs identified in the hospital will likely remain in the tens of thousands. From the hospitalist perspective, the findings presented in this month’s Journal of Hospital Medicine suggest that patients be educated about their findings and recommended follow-up, that follow-up be arranged before discharge, and that findings are clearly documented for patients and primary care providers to review. More study into how to implement these enhancements is needed to guide how we focus educational, systems, and technological interventions. Further study is also needed to help understand the complexities of communication channels between hospitalists and primary care physicians. As hospitalist workflow is more integrated with the EHR and mobile technology, future interventions can facilitate follow-up, keeping all providers and, most importantly, the patient aware of the next steps in care.

Acknowledgments

Author support is provided by the South Texas Veterans Health Care System. The views expressed are those of the authors and do not reflect the position or policy of the Department of Veterans Affairs.

Disclosures

The authors report no financial conflicts of interest.

Liver transplantation (LT) is “one of the most resource-intense procedures despite significant improvements in procedures and protocols,” say researchers from Seoul National University Hospital in South Korea. But little is known about the “practical aspects of life after liver transplantation,” such as unplanned visits to the emergency department (ED) or readmission for complications. So the researchers conducted a study to find out what health care resources are used after discharge.

Of 430 patients, half visited the ED at least once, and 57% were readmitted at least once. The rate of ED visits rose from 15% at 30 days after discharge to 44% at 1 year. Readmission rates more than tripled, from 16% at 30 days to 49% at 1 year.

Contrary to other research, living donor liver transplantation was not a risk factor of readmission. Emergency LT was a risk factor for ED visits and readmission within 30 days of discharge. And although LT using the left liver lobe and pre-existing hepatitis C are known risk factors for long-term graft failure, at the researchers’ hospital hepatitis B is the most common indication for living donor LT. Most of their patients undergo LT using the right liver lobe.

Some of the identified risk factors were unexpected, the researchers say. One was donor age of < 60 years. Warm ischemic time of 15 minutes or longer was another. The researchers note that prolonged warm ischemic time increases hepatic ischemia and reperfusion injury and is related to postoperative complications, which can be a cause of frequent readmission.

Length of stay (LOS)  > 2 weeks also was a risk factor for readmission. In their institution, the average LOS for patients with a warm ischemic time of < 15 minutes was 15.6 days, shorter than the overall average LOS. Shorter LOS, the researchers add, may reflect fewer immediate postoperative complications.

Although they identified no specific complication as a risk factor for readmission, the researchers found specific conditions that accounted for a relatively high proportion of readmissions and repeated readmission, including abnormal liver function test  (32% of readmissions) and fever (17% of readmissions and 39% of repeated readmissions). The researchers suggest those are conditions to monitor and manage.

Notably, patients who did not require readmission or ED visits in the first 20 months almost never required unplanned health care resources thereafter.

Liver transplantation (LT) is “one of the most resource-intense procedures despite significant improvements in procedures and protocols,” say researchers from Seoul National University Hospital in South Korea. But little is known about the “practical aspects of life after liver transplantation,” such as unplanned visits to the emergency department (ED) or readmission for complications. So the researchers conducted a study to find out what health care resources are used after discharge.

Of 430 patients, half visited the ED at least once, and 57% were readmitted at least once. The rate of ED visits rose from 15% at 30 days after discharge to 44% at 1 year. Readmission rates more than tripled, from 16% at 30 days to 49% at 1 year.

Contrary to other research, living donor liver transplantation was not a risk factor of readmission. Emergency LT was a risk factor for ED visits and readmission within 30 days of discharge. And although LT using the left liver lobe and pre-existing hepatitis C are known risk factors for long-term graft failure, at the researchers’ hospital hepatitis B is the most common indication for living donor LT. Most of their patients undergo LT using the right liver lobe.

Some of the identified risk factors were unexpected, the researchers say. One was donor age of < 60 years. Warm ischemic time of 15 minutes or longer was another. The researchers note that prolonged warm ischemic time increases hepatic ischemia and reperfusion injury and is related to postoperative complications, which can be a cause of frequent readmission.

Length of stay (LOS)  > 2 weeks also was a risk factor for readmission. In their institution, the average LOS for patients with a warm ischemic time of < 15 minutes was 15.6 days, shorter than the overall average LOS. Shorter LOS, the researchers add, may reflect fewer immediate postoperative complications.

Although they identified no specific complication as a risk factor for readmission, the researchers found specific conditions that accounted for a relatively high proportion of readmissions and repeated readmission, including abnormal liver function test  (32% of readmissions) and fever (17% of readmissions and 39% of repeated readmissions). The researchers suggest those are conditions to monitor and manage.

Notably, patients who did not require readmission or ED visits in the first 20 months almost never required unplanned health care resources thereafter.

Liver transplantation (LT) is “one of the most resource-intense procedures despite significant improvements in procedures and protocols,” say researchers from Seoul National University Hospital in South Korea. But little is known about the “practical aspects of life after liver transplantation,” such as unplanned visits to the emergency department (ED) or readmission for complications. So the researchers conducted a study to find out what health care resources are used after discharge.

Of 430 patients, half visited the ED at least once, and 57% were readmitted at least once. The rate of ED visits rose from 15% at 30 days after discharge to 44% at 1 year. Readmission rates more than tripled, from 16% at 30 days to 49% at 1 year.

Contrary to other research, living donor liver transplantation was not a risk factor of readmission. Emergency LT was a risk factor for ED visits and readmission within 30 days of discharge. And although LT using the left liver lobe and pre-existing hepatitis C are known risk factors for long-term graft failure, at the researchers’ hospital hepatitis B is the most common indication for living donor LT. Most of their patients undergo LT using the right liver lobe.

Some of the identified risk factors were unexpected, the researchers say. One was donor age of < 60 years. Warm ischemic time of 15 minutes or longer was another. The researchers note that prolonged warm ischemic time increases hepatic ischemia and reperfusion injury and is related to postoperative complications, which can be a cause of frequent readmission.

Length of stay (LOS)  > 2 weeks also was a risk factor for readmission. In their institution, the average LOS for patients with a warm ischemic time of < 15 minutes was 15.6 days, shorter than the overall average LOS. Shorter LOS, the researchers add, may reflect fewer immediate postoperative complications.

Although they identified no specific complication as a risk factor for readmission, the researchers found specific conditions that accounted for a relatively high proportion of readmissions and repeated readmission, including abnormal liver function test  (32% of readmissions) and fever (17% of readmissions and 39% of repeated readmissions). The researchers suggest those are conditions to monitor and manage.

Notably, patients who did not require readmission or ED visits in the first 20 months almost never required unplanned health care resources thereafter.

Prior healthcare utilization predicts future utilization;¹ thus, providers should know when a child has had a recent healthcare visit. Healthcare providers typically obtain this information from parents and caregivers, who may not always provide accurate information.^2-4

The Hospital to Home Outcomes study (H2O) was a randomized controlled trial conducted to assess the effects of a one-time home nurse visit following discharge on unplanned healthcare reutilization.⁵ We assessed reutilization through two sources: parent report via a postdischarge telephone call and administrative data. In this analysis, we sought to understand differences in reutilization rates by source by comparing parent report with administrative data.

The H2O trial included children (<18 years) hospitalized on the hospital medicine (HM) or neuroscience (Neurology/Neurosurgery) services at Cincinnati Children’s Hospital Medical Center (CCHMC) from February 2015 to April 2016; they had an English-speaking parent and were discharged to home without skilled nursing care.⁶ For this analysis, we restricted the sample to children randomized to the control arm (discharge without a home visit), which reflects typical clinical care.

We used administrative data to capture 14-day reutilization (unplanned hospital readmissions, emergency department [ED] visits, or urgent care visits). CCHMC is the only pediatric admitting facility in the region and includes two pediatric EDs and five urgent care centers. We supplemented hospital data with a dataset (The Health Collaborative⁷) that included utilization at other regional facilities. Parent report was assessed via a research coordinator phone call 14-23 days after discharge. Parents were asked: “I’m going to [ask] about your child’s health since [discharge date]. Has s/he been hospitalized overnight? Has s/he been taken to the Emergency Room/Emergency Department (didn’t stay overnight)? Has s/he been taken to an urgent care?” We report 14-day reutilization rates by source (parent and/or administrative) and visit type.

We considered administrative data the gold standard for documentation of reutilization events for two reasons. First, all healthcare encounters generate billing and are therefore documented with verifiable coding. Second, we had access to data from our center and other regional healthcare facilities. Any parent-reported utilization to a facility not documented in either dataset was considered an unverifiable event (eg, outside our catchment region). Agreement between administrative and parent report of 14-day reutilization was summarized as positive agreement (reutilization documented in both administrative and parent report), negative agreement (no reutilization reported in either administrative or parent report), and overall agreement (combination of positive and negative agreement). We classified discrepancies as reutilization events in administrative data without parent report of reutilization or vice versa. We performed medical record review of discrepancies in our institutional data.

We summarized agreement by using the Cohen’s kappa statistic by reuse type (hospital readmission, ED, and urgent care visit) and overall (any reutilization event). Strength of agreement based on the kappa statistics was classified as poor (<0.20), fair (0.21-0.40), moderate (0.41-0.60), good (0.61-0.80), and very good (0.81-1.00).⁸ We used McNemar’s test to evaluate marginal homogeneity.

Of 749 children randomized to the standard of care arm, 723 parents completed the 14-day follow-up call and were included in this analysis. The median child age was two years (interquartile range: 0.4, 6.9), the median length of stay (LOS) was two days (1, 3), and the majority were white (62%). Payer mix varied, with 44% privately insured and 54% publicly insured. Most patients (83%) were admitted to the HM service, and the most common diagnoses groups for index admission were respiratory (35%), neurologic (14%), and gastrointestinal (9%) diseases.

Administrative data showed 63 children with any reutilization event; parents reported 63 with any reutilization event; 48 children had events reported by both sources. The overall agreement was high, ranging from 95.9% to 98.5% (Table 1) depending on visit type. The positive agreement (ie, parent and administrative data indicated reutilization) ranged from 47.6% to 76.2%. Negative agreement (ie, parent and administrative data agreed no reutilization) was very high, 97.7% to 99.2%. Parents reported three ED visits and four urgent care visits that were unverifiable due to lack of access to administrative data (sites of care reported were not included in our datasets).

By comparing parent report of reutilization after hospital discharge through postdischarge phone calls with administrative data, we demonstrated high overall agreement between sources (95.9%); this finding is similar to prior research investigating the relationship between an established medical home and reutilization.⁹ However, this agreement is largely due to both sources reporting no reutilization. When revisits did occur, the agreement was notably lower, especially with regard to urgent care visits.

Discrepancies between sources have several possible explanations. First, parents may be confused by the framing of reutilization questions, perhaps lacking clarity around which visit we were referencing. Second, parents may experience limitations in health literacy^10,11 with a lack of familiarity with healthcare language, such as the ability to delineate location types (for example, a parent may identify an urgent care visit as an ED visit, given their close proximity at our facility). Finally, our prior work identified that the “fog” of hospitalization,¹² which is often a stressful and disruptive time for families, may linger after admission and could lead to difficulty in recalling detailed events.

Our findings have implications for effective care in a complex healthcare system where parent report may be the most practical method to obtain historical information, both within clinical care and in the context of research or quality measures, such as postdischarge utilization. Given that one of the greatest risk factors for readmission is prior utilization,¹ the knowledge that a patient experienced a reutilization after a prior discharge might prompt the inpatient provider to better prepare families for subsequent transition to home.

To apply our findings practically, it is important to realize that a parent report may be sufficient when reporting that no revisit occurred, if there is also no record of a visit in accessible administrative data (such as an electronic health record). However, further questions or investigation should be considered when parents report a visit did occur or when administrative data indicate a visit occurred that the parent does not recall. Providers and researchers alike should remember to use health literacy universal precautions with all families, employing plain language without medical jargon.¹³ As linked electronic health record use becomes more prevalent, administrative data may be accessible in real-time, allowing for verification of family interview information. Administrative data beyond a single hospital system should be considered to effectively capture reutilization for research or quality efforts.

Our study has several limitations. Similar to most studies using reutilization outcomes, our data may miss a few unverifiable reuse events. By supplementing with additional regional data,⁷ we likely captured most events. Second, we did not include patients with limited English proficiency, although it is unclear how this might have biased our results. Third, while relatively few families did not complete the calls, it is possible that more discrepancies would have been noted in nonresponders. Fourth, research coordinators administering the calls followed a script to determine reutilization information; in clinical practice, a practitioner might not ask questions as clearly, which could negatively impact recall or might add clarifying follow-up questions to enhance recall. Finally, the analysis occurred in the setting of a randomized controlled trial that included children with relatively noncomplex health conditions with short LOS;⁶ thus, the results may not apply to other populations.

In conclusion, parent report and administrative data of reutilization following hospital discharge were usually in agreement when no reutilization occurred; however, discrepancies were noted more often when reutilizations occurred and may have care implications.

Collaborators

On behalf of the H2O Trial study group including: Joanne Bachus, BSN, RN; Andrew F. Beck, MD, MPH; Monica L. Borell, BSN, RN; Lenisa V. Chang, MA, PhD; Patricia Crawford, RN; Jennifer M. Gold, MSN, RN; Judy A. Heilman BSN, RN; Jane C. Khoury, PhD; Pierce Kuhnell, MS; Karen Lawley, BSN, RN; Allison Loechtenfeldt, BS; Colleen Mangeot, MS; Lynn O’Donnell, BSN, RN; Rita H. Pickler, PhD, RN; Hadley S. Sauers-Ford, MPH; Anita N. Shah, DO, MPH; Susan N. Sherman, DPA; Lauren G. Solan, MD, MEd; Karen P. Sullivan, BSN, RN; Susan Wade-Murphy, MSN, RN

Disclosures

Hospital to Home Outcomes team reports grants from the Patient Centered Outcomes Research Institute during the conduct of the study. Dr. White reports personal fees from the Institute for Health Care Improvement, outside the submitted work.

Funding

This work was supported by the Patient Centered Outcomes Research Institute (IHS-1306-0081 to Dr. S. Shah). All statements in this report, including findings and conclusions, are solely those of the authors and do not necessarily represent the views of the Patient-Centered Outcomes Research Institute, its Board of Governors, or the Methodology Committee. Dr Auger’s research is funded by the Agency for Healthcare Research and Quality (1K08HS024735).

Prior healthcare utilization predicts future utilization;¹ thus, providers should know when a child has had a recent healthcare visit. Healthcare providers typically obtain this information from parents and caregivers, who may not always provide accurate information.^2-4

The Hospital to Home Outcomes study (H2O) was a randomized controlled trial conducted to assess the effects of a one-time home nurse visit following discharge on unplanned healthcare reutilization.⁵ We assessed reutilization through two sources: parent report via a postdischarge telephone call and administrative data. In this analysis, we sought to understand differences in reutilization rates by source by comparing parent report with administrative data.

The H2O trial included children (<18 years) hospitalized on the hospital medicine (HM) or neuroscience (Neurology/Neurosurgery) services at Cincinnati Children’s Hospital Medical Center (CCHMC) from February 2015 to April 2016; they had an English-speaking parent and were discharged to home without skilled nursing care.⁶ For this analysis, we restricted the sample to children randomized to the control arm (discharge without a home visit), which reflects typical clinical care.

We used administrative data to capture 14-day reutilization (unplanned hospital readmissions, emergency department [ED] visits, or urgent care visits). CCHMC is the only pediatric admitting facility in the region and includes two pediatric EDs and five urgent care centers. We supplemented hospital data with a dataset (The Health Collaborative⁷) that included utilization at other regional facilities. Parent report was assessed via a research coordinator phone call 14-23 days after discharge. Parents were asked: “I’m going to [ask] about your child’s health since [discharge date]. Has s/he been hospitalized overnight? Has s/he been taken to the Emergency Room/Emergency Department (didn’t stay overnight)? Has s/he been taken to an urgent care?” We report 14-day reutilization rates by source (parent and/or administrative) and visit type.

We considered administrative data the gold standard for documentation of reutilization events for two reasons. First, all healthcare encounters generate billing and are therefore documented with verifiable coding. Second, we had access to data from our center and other regional healthcare facilities. Any parent-reported utilization to a facility not documented in either dataset was considered an unverifiable event (eg, outside our catchment region). Agreement between administrative and parent report of 14-day reutilization was summarized as positive agreement (reutilization documented in both administrative and parent report), negative agreement (no reutilization reported in either administrative or parent report), and overall agreement (combination of positive and negative agreement). We classified discrepancies as reutilization events in administrative data without parent report of reutilization or vice versa. We performed medical record review of discrepancies in our institutional data.

We summarized agreement by using the Cohen’s kappa statistic by reuse type (hospital readmission, ED, and urgent care visit) and overall (any reutilization event). Strength of agreement based on the kappa statistics was classified as poor (<0.20), fair (0.21-0.40), moderate (0.41-0.60), good (0.61-0.80), and very good (0.81-1.00).⁸ We used McNemar’s test to evaluate marginal homogeneity.

Of 749 children randomized to the standard of care arm, 723 parents completed the 14-day follow-up call and were included in this analysis. The median child age was two years (interquartile range: 0.4, 6.9), the median length of stay (LOS) was two days (1, 3), and the majority were white (62%). Payer mix varied, with 44% privately insured and 54% publicly insured. Most patients (83%) were admitted to the HM service, and the most common diagnoses groups for index admission were respiratory (35%), neurologic (14%), and gastrointestinal (9%) diseases.

Administrative data showed 63 children with any reutilization event; parents reported 63 with any reutilization event; 48 children had events reported by both sources. The overall agreement was high, ranging from 95.9% to 98.5% (Table 1) depending on visit type. The positive agreement (ie, parent and administrative data indicated reutilization) ranged from 47.6% to 76.2%. Negative agreement (ie, parent and administrative data agreed no reutilization) was very high, 97.7% to 99.2%. Parents reported three ED visits and four urgent care visits that were unverifiable due to lack of access to administrative data (sites of care reported were not included in our datasets).

By comparing parent report of reutilization after hospital discharge through postdischarge phone calls with administrative data, we demonstrated high overall agreement between sources (95.9%); this finding is similar to prior research investigating the relationship between an established medical home and reutilization.⁹ However, this agreement is largely due to both sources reporting no reutilization. When revisits did occur, the agreement was notably lower, especially with regard to urgent care visits.

Discrepancies between sources have several possible explanations. First, parents may be confused by the framing of reutilization questions, perhaps lacking clarity around which visit we were referencing. Second, parents may experience limitations in health literacy^10,11 with a lack of familiarity with healthcare language, such as the ability to delineate location types (for example, a parent may identify an urgent care visit as an ED visit, given their close proximity at our facility). Finally, our prior work identified that the “fog” of hospitalization,¹² which is often a stressful and disruptive time for families, may linger after admission and could lead to difficulty in recalling detailed events.

Our findings have implications for effective care in a complex healthcare system where parent report may be the most practical method to obtain historical information, both within clinical care and in the context of research or quality measures, such as postdischarge utilization. Given that one of the greatest risk factors for readmission is prior utilization,¹ the knowledge that a patient experienced a reutilization after a prior discharge might prompt the inpatient provider to better prepare families for subsequent transition to home.

To apply our findings practically, it is important to realize that a parent report may be sufficient when reporting that no revisit occurred, if there is also no record of a visit in accessible administrative data (such as an electronic health record). However, further questions or investigation should be considered when parents report a visit did occur or when administrative data indicate a visit occurred that the parent does not recall. Providers and researchers alike should remember to use health literacy universal precautions with all families, employing plain language without medical jargon.¹³ As linked electronic health record use becomes more prevalent, administrative data may be accessible in real-time, allowing for verification of family interview information. Administrative data beyond a single hospital system should be considered to effectively capture reutilization for research or quality efforts.

Our study has several limitations. Similar to most studies using reutilization outcomes, our data may miss a few unverifiable reuse events. By supplementing with additional regional data,⁷ we likely captured most events. Second, we did not include patients with limited English proficiency, although it is unclear how this might have biased our results. Third, while relatively few families did not complete the calls, it is possible that more discrepancies would have been noted in nonresponders. Fourth, research coordinators administering the calls followed a script to determine reutilization information; in clinical practice, a practitioner might not ask questions as clearly, which could negatively impact recall or might add clarifying follow-up questions to enhance recall. Finally, the analysis occurred in the setting of a randomized controlled trial that included children with relatively noncomplex health conditions with short LOS;⁶ thus, the results may not apply to other populations.

In conclusion, parent report and administrative data of reutilization following hospital discharge were usually in agreement when no reutilization occurred; however, discrepancies were noted more often when reutilizations occurred and may have care implications.

Collaborators

On behalf of the H2O Trial study group including: Joanne Bachus, BSN, RN; Andrew F. Beck, MD, MPH; Monica L. Borell, BSN, RN; Lenisa V. Chang, MA, PhD; Patricia Crawford, RN; Jennifer M. Gold, MSN, RN; Judy A. Heilman BSN, RN; Jane C. Khoury, PhD; Pierce Kuhnell, MS; Karen Lawley, BSN, RN; Allison Loechtenfeldt, BS; Colleen Mangeot, MS; Lynn O’Donnell, BSN, RN; Rita H. Pickler, PhD, RN; Hadley S. Sauers-Ford, MPH; Anita N. Shah, DO, MPH; Susan N. Sherman, DPA; Lauren G. Solan, MD, MEd; Karen P. Sullivan, BSN, RN; Susan Wade-Murphy, MSN, RN

Disclosures

Hospital to Home Outcomes team reports grants from the Patient Centered Outcomes Research Institute during the conduct of the study. Dr. White reports personal fees from the Institute for Health Care Improvement, outside the submitted work.

Funding

This work was supported by the Patient Centered Outcomes Research Institute (IHS-1306-0081 to Dr. S. Shah). All statements in this report, including findings and conclusions, are solely those of the authors and do not necessarily represent the views of the Patient-Centered Outcomes Research Institute, its Board of Governors, or the Methodology Committee. Dr Auger’s research is funded by the Agency for Healthcare Research and Quality (1K08HS024735).

Prior healthcare utilization predicts future utilization;¹ thus, providers should know when a child has had a recent healthcare visit. Healthcare providers typically obtain this information from parents and caregivers, who may not always provide accurate information.^2-4

The Hospital to Home Outcomes study (H2O) was a randomized controlled trial conducted to assess the effects of a one-time home nurse visit following discharge on unplanned healthcare reutilization.⁵ We assessed reutilization through two sources: parent report via a postdischarge telephone call and administrative data. In this analysis, we sought to understand differences in reutilization rates by source by comparing parent report with administrative data.

The H2O trial included children (<18 years) hospitalized on the hospital medicine (HM) or neuroscience (Neurology/Neurosurgery) services at Cincinnati Children’s Hospital Medical Center (CCHMC) from February 2015 to April 2016; they had an English-speaking parent and were discharged to home without skilled nursing care.⁶ For this analysis, we restricted the sample to children randomized to the control arm (discharge without a home visit), which reflects typical clinical care.

We used administrative data to capture 14-day reutilization (unplanned hospital readmissions, emergency department [ED] visits, or urgent care visits). CCHMC is the only pediatric admitting facility in the region and includes two pediatric EDs and five urgent care centers. We supplemented hospital data with a dataset (The Health Collaborative⁷) that included utilization at other regional facilities. Parent report was assessed via a research coordinator phone call 14-23 days after discharge. Parents were asked: “I’m going to [ask] about your child’s health since [discharge date]. Has s/he been hospitalized overnight? Has s/he been taken to the Emergency Room/Emergency Department (didn’t stay overnight)? Has s/he been taken to an urgent care?” We report 14-day reutilization rates by source (parent and/or administrative) and visit type.

We considered administrative data the gold standard for documentation of reutilization events for two reasons. First, all healthcare encounters generate billing and are therefore documented with verifiable coding. Second, we had access to data from our center and other regional healthcare facilities. Any parent-reported utilization to a facility not documented in either dataset was considered an unverifiable event (eg, outside our catchment region). Agreement between administrative and parent report of 14-day reutilization was summarized as positive agreement (reutilization documented in both administrative and parent report), negative agreement (no reutilization reported in either administrative or parent report), and overall agreement (combination of positive and negative agreement). We classified discrepancies as reutilization events in administrative data without parent report of reutilization or vice versa. We performed medical record review of discrepancies in our institutional data.

We summarized agreement by using the Cohen’s kappa statistic by reuse type (hospital readmission, ED, and urgent care visit) and overall (any reutilization event). Strength of agreement based on the kappa statistics was classified as poor (<0.20), fair (0.21-0.40), moderate (0.41-0.60), good (0.61-0.80), and very good (0.81-1.00).⁸ We used McNemar’s test to evaluate marginal homogeneity.

Of 749 children randomized to the standard of care arm, 723 parents completed the 14-day follow-up call and were included in this analysis. The median child age was two years (interquartile range: 0.4, 6.9), the median length of stay (LOS) was two days (1, 3), and the majority were white (62%). Payer mix varied, with 44% privately insured and 54% publicly insured. Most patients (83%) were admitted to the HM service, and the most common diagnoses groups for index admission were respiratory (35%), neurologic (14%), and gastrointestinal (9%) diseases.

Administrative data showed 63 children with any reutilization event; parents reported 63 with any reutilization event; 48 children had events reported by both sources. The overall agreement was high, ranging from 95.9% to 98.5% (Table 1) depending on visit type. The positive agreement (ie, parent and administrative data indicated reutilization) ranged from 47.6% to 76.2%. Negative agreement (ie, parent and administrative data agreed no reutilization) was very high, 97.7% to 99.2%. Parents reported three ED visits and four urgent care visits that were unverifiable due to lack of access to administrative data (sites of care reported were not included in our datasets).

By comparing parent report of reutilization after hospital discharge through postdischarge phone calls with administrative data, we demonstrated high overall agreement between sources (95.9%); this finding is similar to prior research investigating the relationship between an established medical home and reutilization.⁹ However, this agreement is largely due to both sources reporting no reutilization. When revisits did occur, the agreement was notably lower, especially with regard to urgent care visits.

Discrepancies between sources have several possible explanations. First, parents may be confused by the framing of reutilization questions, perhaps lacking clarity around which visit we were referencing. Second, parents may experience limitations in health literacy^10,11 with a lack of familiarity with healthcare language, such as the ability to delineate location types (for example, a parent may identify an urgent care visit as an ED visit, given their close proximity at our facility). Finally, our prior work identified that the “fog” of hospitalization,¹² which is often a stressful and disruptive time for families, may linger after admission and could lead to difficulty in recalling detailed events.

Our findings have implications for effective care in a complex healthcare system where parent report may be the most practical method to obtain historical information, both within clinical care and in the context of research or quality measures, such as postdischarge utilization. Given that one of the greatest risk factors for readmission is prior utilization,¹ the knowledge that a patient experienced a reutilization after a prior discharge might prompt the inpatient provider to better prepare families for subsequent transition to home.

To apply our findings practically, it is important to realize that a parent report may be sufficient when reporting that no revisit occurred, if there is also no record of a visit in accessible administrative data (such as an electronic health record). However, further questions or investigation should be considered when parents report a visit did occur or when administrative data indicate a visit occurred that the parent does not recall. Providers and researchers alike should remember to use health literacy universal precautions with all families, employing plain language without medical jargon.¹³ As linked electronic health record use becomes more prevalent, administrative data may be accessible in real-time, allowing for verification of family interview information. Administrative data beyond a single hospital system should be considered to effectively capture reutilization for research or quality efforts.

Our study has several limitations. Similar to most studies using reutilization outcomes, our data may miss a few unverifiable reuse events. By supplementing with additional regional data,⁷ we likely captured most events. Second, we did not include patients with limited English proficiency, although it is unclear how this might have biased our results. Third, while relatively few families did not complete the calls, it is possible that more discrepancies would have been noted in nonresponders. Fourth, research coordinators administering the calls followed a script to determine reutilization information; in clinical practice, a practitioner might not ask questions as clearly, which could negatively impact recall or might add clarifying follow-up questions to enhance recall. Finally, the analysis occurred in the setting of a randomized controlled trial that included children with relatively noncomplex health conditions with short LOS;⁶ thus, the results may not apply to other populations.

In conclusion, parent report and administrative data of reutilization following hospital discharge were usually in agreement when no reutilization occurred; however, discrepancies were noted more often when reutilizations occurred and may have care implications.

Collaborators

On behalf of the H2O Trial study group including: Joanne Bachus, BSN, RN; Andrew F. Beck, MD, MPH; Monica L. Borell, BSN, RN; Lenisa V. Chang, MA, PhD; Patricia Crawford, RN; Jennifer M. Gold, MSN, RN; Judy A. Heilman BSN, RN; Jane C. Khoury, PhD; Pierce Kuhnell, MS; Karen Lawley, BSN, RN; Allison Loechtenfeldt, BS; Colleen Mangeot, MS; Lynn O’Donnell, BSN, RN; Rita H. Pickler, PhD, RN; Hadley S. Sauers-Ford, MPH; Anita N. Shah, DO, MPH; Susan N. Sherman, DPA; Lauren G. Solan, MD, MEd; Karen P. Sullivan, BSN, RN; Susan Wade-Murphy, MSN, RN

Disclosures

Hospital to Home Outcomes team reports grants from the Patient Centered Outcomes Research Institute during the conduct of the study. Dr. White reports personal fees from the Institute for Health Care Improvement, outside the submitted work.

Funding

This work was supported by the Patient Centered Outcomes Research Institute (IHS-1306-0081 to Dr. S. Shah). All statements in this report, including findings and conclusions, are solely those of the authors and do not necessarily represent the views of the Patient-Centered Outcomes Research Institute, its Board of Governors, or the Methodology Committee. Dr Auger’s research is funded by the Agency for Healthcare Research and Quality (1K08HS024735).

Delirium presents as an acute change in mentation characterized by reduced attention, clouding of awareness, and typically an altered level of arousal. It can be caused by a host of medical conditions, medications, or other psychoactive substances and is therefore encountered primarily in acute and postacute medical settings.¹ More than a quarter of all hospitalized patients develop delirium,² with rates up to 80% in the critically ill.³ Similarly, delirium occurs in more than one-third of patients who transition to postacute care.⁴ These high prevalence rates are alarming, especially because delirium is a risk factor for mortality, prolonged hospitalization, institutionalization, and overall higher cost of care.⁵ However, more than a quarter of delirium is preventable.⁶ Evidence-based guidelines for delirium uniformly call for multicomponent prevention strategies,⁷ and these are best delivered through collaborative models of care. In short, delirium impacts healthcare systems; therefore, interventions aimed at preventing delirium and its consequences ought to be systems-based.

Since the Institute of Medicine issued its 1999 report highlighting the critical role of medical errors in healthcare, healthcare systems have increasingly become team-based.⁸ “Medical care is inherently interdependent,”⁹ and this implies that delirium prevention rests not only on individuals but also on broader systems of care. Although nonpharmacological interventions are efficacious at preventing delirium,¹⁰ previous reviews have focused on specific interventions or multiple interventions rather than the systems of care needed to deliver them. Indeed, teams and the quality of their teamwork impact outcomes.¹¹

Herein, we provide a systematic review and meta-analysis of integrated models of care designed to prevent delirium. What distinguishes this review from previous reviews of nonpharmacological interventions to prevent delirium is our focus on discrete models of care that involve collaboration among clinicians. Our goal is to identify the most promising models that deserve further development, investigation, and dissemination. Viewing delirium prevention through a collaborative care lens is consistent with efforts to achieve value-based care and may encourage drawing from the expanding literature outlining the benefits of mental healthcare integration.^12,13 Specifically, a systems perspective highlights the potential for system-wide benefits such as reducing readmissions^14,15 and cost savings.¹⁶

This systematic review and meta-analysis follows PRISMA guidelines. A search of OVID, MEDLINE, CINAHL, Cochrane Database of Systematic Reviews, EMBASE, and PsycINFO was completed by a medical librarian for clinical studies in which models of care were implemented to prevent delirium using PICO (P patient, problem or population; I, intervention; C, comparison, control or comparator; O, outcome) inquiries. Search terms included delirium, acute confusional state, altered mental status, prevention, and control (“delirium”/exp OR “acute confusion”/exp OR “altered mental status”/exp) AND “prevention and control”/exp AND [English]/lim AND [embase]/lim).

One researcher (AK) screened articles by title for relevance. Relevant articles were then divided among four authors (AK, MO, NF, and OB), and the abstracts were screened for eligibility. The authors reviewed the full texts of any potentially eligible studies. Each full text was assigned to two authors for full review. Discrepancies were adjudicated by conference among all authors. In addition, references within all full-text publications were scanned for potential additional articles.

The inclusion criteria for review of full-text articles required English-language description of a model of care with multiple interventions, delirium reported as an outcome, and presence of a comparator group.

“Model of care” was defined by the Cochrane Effective Practice and Organization of Care Review Group as follows: (1) revision of professional roles, including shifting of professional roles or expansion of roles to new tasks; (2) creation of clinical multidisciplinary teams or addition of new members to the team who collaborate inpatient care; (3) delivery of multiple interventions across multiple domains (ie, studies involving a single intervention such as physical therapy or targeting a single domain such as sleep were excluded); and (4) formal integration of services whereby teams work together in collaboration with existing services to enhance care.¹⁷ For this review, we required that studies include a comparator group so that effectiveness of the intervention could be assessed. Quality improvement studies that lacked a comparator group were excluded.

Delirium incidence was the primary outcome and was evaluated by meta-analysis. Heterogeneity was assessed using I² and visual inspection of forest plots. I² values of 25%, 50%, and 75% represent low, moderate, and high heterogeneity, respectively. The studies were pooled according to study type as follows: randomized controlled trials, pre–post design, and other nonrandomized prospective studies. Random effects models were used to calculate estimates using the Comprehensive Meta-Analysis software (Version 3, Biostat, Englewood, New Jersey), which also generated forest plots.

Risk of bias was assessed using criteria established by the Cochrane Collaborative Review Criteria, which lists six categories of potential bias: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, and selective reporting.¹⁷ Each study was assessed by two authors (either MO and AK or MO-P. and OB) for bias and a numerical value was assigned to each of the six categories as follows: 1 = low risk, 2 = unknown/moderate risk, and 3 = high risk. Where scorers disagreed, all authors jointly conferred, and a consensus score was given. The values for each of these six categories were added to create a composite risk-of-bias score for each study, with 6 being the lowest possible score and 18 the highest. Overall risk was classified as follows: <9 = low risk, 9-12 = moderate risk, and >12 = high risk.

Study Selection Process

An initial literature search identified 352 articles. After reviewing the titles, 308 articles were excluded for irrelevance, and 44 abstracts were screened for eligibility. We excluded 27 articles upon abstract review, and the full texts of 17 were obtained for detailed review. In addition, we identified another 10 potentially eligible articles through review of references and obtained full texts of these as well. Of the 27 full-text articles reviewed, 15 were included in this systematic review, 10 of which were suitable for meta-analysis. The Figure shows the PRISMA flow chart.

Study Characteristics

The 15 studies that met the inclusion criteria are summarized in the Table.^18-32 Delirium prevention was among the primary outcomes of 13 studies; delirium outcomes were reported in the other two studies as well, which were primarily designed to assess feasibility.^26,27 Six studies were conducted in the United States, three in Sweden, two in Spain, two in the United Kingdom, and one each conducted in Korea and Canada. Healthcare settings among the included studies involved the intensive care unit (six studies), medical floors (four studies), surgical floors (three studies), a long-term care unit (one study), and an inpatient palliative care service (one study). We categorized the studies according to design and intent as follows: randomized controlled studies (three), pilot feasibility studies (two), pre–post design (six), and other nonrandomized prospective studies (four; Table).

Outcomes Reported

All but one of the studies reported delirium incidence. The most commonly used delirium screening instrument was the Confusion Assessment Method (CAM) or its modified version, the CAM-ICU (11 studies).^33,34 Other methods used to assess mentation included the Richmond Agitation Sedation Scale,³⁵ the Organic Brain Syndrome scale,³⁶ the revised Delirium Rating Scale,³⁷ the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition,³⁸ and the Confusion Rating Scale.³⁹ (Details regarding delirium screening tools can be found in the systematic review by De and Wand.⁴⁰) Researchers performed delirium assessment in nine studies, whereas assessments were performed by clinical staff in the remaining studies. Other outcomes reported included length of stay (LOS), mortality, number of days ventilated, and functional decline. None of the included studies reported cost effectiveness.

Risk of Bias Assessment

Risk of bias assessment identified only two studies—both randomized controlled trials—as low risk (Table). The remaining studies had moderate (four studies) or high risk (nine studies).

Results from Individual Studies

Of the 15 studies, nine reported a statistically significant reduction in delirium incidence, and another two reported a statistically insignificant reduction. In addition, seven of the eight studies that assessed delirium duration found reduced duration in the intervention cohort, and two of the three studies that reported delirium severity found a reduction in the intervention group.

Results of Meta-Analysis

Random effects models were created to meta-analyze groups of studies based on design as follows: randomized controlled trials (three studies^18,19,25), pre–post intervention studies (four of six studies included^28-31), and other nonrandomized studies (three of four studies included^21-23). Meta-analysis was not completed for the two feasibility studies^26,27 because delirium outcome data were limited due to the feasibility study design. The study of Dale et al.³² was excluded from the meta-analysis because the rates of CAM-ICU completion differed substantially between control and intervention groups (0.35 vs 1.49 per 24 hours, respectively), leading to imbalanced between-group sensitivity in delirium detection and Needham et al.²⁰ was also excluded because it reported only days of delirium, not delirium incidence. The study by Lundström et al.²⁴ was also excluded from the meta-analysis because delirium incidence was measured on days 1, 3, and 5, whereas the other studies reported delirium daily.

Meta-analysis of the three randomized controlled trials revealed a pooled odds ratio of 0.56 (95% CI: 0.37-0.85; P = .006) for delirium incidence among intervention group subjects relative to those in comparator groups. The heterogeneity across studies was low (I² = 29%). Pooling data from four pre–post studies found that the odds ratio for delirium incidence was 0.63 (95% CI: 0.37-1.07; P = .09). The heterogeneity across these studies was moderate (I² = 65%). Results from the three eligible, nonrandomized prospective studies were also pooled. The odds ratio for developing delirium among study subjects was 0.79 (95% CI: 0.46-1.37; P = .40), and the heterogeneity among these studies was high (I² = 85%).

We provide a systematic review and meta-analysis of delirium preventive models of care. Meta-analysis of the three randomized controlled trials found that these models of care led to a statistically significant reduction in delirium incidence; study subjects had an 11.5% reduction in absolute delirium incidence. The pooled odds ratios for both of the other sets of nonrandomized studies favored the intervention group but were not significant, each because of one included study. The pre–post meta-analysis failed to reach significance as one of the included studies found a trend toward higher delirium incidence; however, interestingly, in that same study, the overall delirium-free days were significantly reduced overall (24 vs 27, P = .002). Similarly, meta-analysis of the three additional nonrandomized prospective studies failed to reach significance because the largest included study found higher rates of delirium among intervention group subjects. Despite considerable risk of bias in several of these studies, their findings were broadly consistent; all but one study (Gagnon 2012²¹) reported a trend or a significant reduction in delirium incidence, duration, severity, or number of delirium episodes. Moreover, the value of such models of care extended beyond preventing delirium; for instance, other positive outcomes included reduced LOS and fewer medical complications.

Models of care ranged widely with respect to specific interventions, though several common elements highlighted their relevance for delirium care and as potential delirium prevention strategies in future studies. For example, two of the randomized controlled trials^18,19 employed early mobilization, enhanced nutrition, sleep hygiene, early reduction of invasive procedures (eg, urinary catheterization), and pain control in their multicomponent models. Five additional studies also incorporated early mobilization,^{20,22,23,31,32} and three sought to improve sleep quality.^22,28,30 Among other important strategies were delirium screening,^{18,20,22,30,31} monitoring medication,^{18,20,22,26,28,30,32} orientation,^18,21,23,28 addressing vision and hearing impairment,^18,22,23,32 hydration,^18,22,23 avoiding hypoxia,^18,20,30 and staff, patient, and caretaker education.^{19,21,23,27-30} Unique strategies were implemented in certain studies. For instance, one study used massage therapy,²⁸ preventing delays in transfer logistics in another,³⁰ and a third addressed psychosocial problems.²⁵ Overall, the selection of strategies depended on the patient setting; thus, no one care bundle should be expected to emerge as a universal model for delirium prevention. Rather, these results should be interpreted within their specific care contexts and judged on the quality of evidence (eg, effect size and statistically significant findings, low risk of bias, sound experimental design). The one study that failed to find any positive effect on delirium, that of Gagnon et al.,²¹ was conducted on an inpatient palliative care service in Canada, and its negative finding may reflect the unique delirium risk factors in patients who are nearing end of life.

This current review differs from previous delirium prevention reviews in operationally defining a “model of care.” We identified a great deal of variation in specific models and team composition. For example, some interventions were carried out by nurses^18-20,31 and physicians,^20,21,25,32 whereas others involved physical therapists,^20,22,28 medical residents,²³ geriatricians,^22,23,25 pharmacists,²⁶ researchers,¹⁸ and trained volunteers.²² In all cases, the staff roles were expanded to include new tasks, and the clinical team worked collaboratively to administer interventions across multiple domains. Team-related considerations are critical because modern medical care is inherently interdependent.⁹ These broad differences in team composition across studies demonstrate the number of potential options for team structure and function. They also highlight the number of “moving parts” to be considered when designing and implementing delirium care bundles.

Most of the delirium prevention studies implementing models of care are characterized by a substantial risk of bias. We evaluated risk of bias along six categories of potential sources, including random assignment to groups, ability to foresee future group allocation, blinding of participants and personnel to group assignment, blinding of outcome assessment, completeness of outcome data, selective reporting, and other potential sources of bias.¹⁷ Two of the three studies that used randomization had a low risk of bias, and four additional studies had a moderate risk of bias. Allocation concealment was accomplished only in randomized controlled trials, whereas blinding of both subjects and study personnel was not implemented in any of the studies. Although some studies relied on data analysis by research personnel blinded to group membership or the nature of the intervention, others failed to do so or failed to describe data analysis in sufficient detail. Studies also failed to report the percentage of unscorable or otherwise omitted delirium assessments necessary to calculate attrition rates or to understand the comprehensiveness of outcome assessment in a systematic manner. Other potential sources of bias included systematic differences between the intervention and control groups (such as differences in gender composition, age, or delirium risk) at study outset.

A primary limitation of this review is the heterogeneity of settings, interventions, and models of care across included studies. We excluded several studies from this review for being delivered by a single individual or service line (eg, introduction of a geriatric consult service, physical therapy, or volunteers), for providing a single intervention (eg, early ambulation alone), or for multiple interventions targeting a single domain (eg, sleep). We did so because the future of value-based care lies in collaboration of providers and services, and in a way the complexity across and within these studies ultimately reflects the complexity of medical settings as well as the multifactorial nature of delirium. The broader message is a call for increasing the integration of delirium-related care services. As discussed earlier, the high risk of bias across these studies is a limitation of our findings; high-quality evidence on the value of delirium prevention models of care remains limited. Thus, although our review suggests that there are multicomponent models of care that hold promise in mitigating delirium and its outcomes, additional randomized studies are required to confirm the efficacy of such models of care and to test which services, interventions, and clinical domains deserve priority.

To our knowledge, this is the first systematic review and meta-analysis of delirium preventive models of care. Models of care, as defined here, necessarily included a multidisciplinary team in which traditional staff roles had been revised to implement a multicomponent, multidomain intervention. Other recent reviews are available for multicomponent pharmacological and nonpharmacological interventions to prevent and manage delirium,^41-49 but just as important as which interventions are being delivered is the team that delivers them. Care delivery in a complex medical system is more than handing a patient a medication or facilitating ambulation; it requires a choreographed dance of teamwork and integration across services. This review identifies promising models of care that deserve further recognition, refinement, and ultimately widespread implementation.

Acknowledgments

The authors comprise a writing group created through the Delirium Boot Camp, an annual meeting originally sponsored by the Center of Excellence for Delirium in Aging: Research, Training, and Educational Enhancement (CEDARTREE, Boston, Massachusetts); it is currently supported by the Network for Investigation of Delirium: Unifying Scientists (NIDUS, Boston, Massachusetts). The authors would like to thank medical librarian Rita Mitchell (Aurora Health Care, Milwaukee, Wisconsin) for the literature search, senior scientific writer and editor Joe Grundle (Aurora Research Institute, Milwaukee, Wisconsin) for editorial assistance, and graphics specialist Brian Miller (Aurora Research Institute, Milwaukee, Wisconsin) for help with the figures.

Disclosures

The authors report no relevant conflicts of interest.

Funding

No funding was dedicated to the conduct of this review.

Delirium presents as an acute change in mentation characterized by reduced attention, clouding of awareness, and typically an altered level of arousal. It can be caused by a host of medical conditions, medications, or other psychoactive substances and is therefore encountered primarily in acute and postacute medical settings.¹ More than a quarter of all hospitalized patients develop delirium,² with rates up to 80% in the critically ill.³ Similarly, delirium occurs in more than one-third of patients who transition to postacute care.⁴ These high prevalence rates are alarming, especially because delirium is a risk factor for mortality, prolonged hospitalization, institutionalization, and overall higher cost of care.⁵ However, more than a quarter of delirium is preventable.⁶ Evidence-based guidelines for delirium uniformly call for multicomponent prevention strategies,⁷ and these are best delivered through collaborative models of care. In short, delirium impacts healthcare systems; therefore, interventions aimed at preventing delirium and its consequences ought to be systems-based.

Since the Institute of Medicine issued its 1999 report highlighting the critical role of medical errors in healthcare, healthcare systems have increasingly become team-based.⁸ “Medical care is inherently interdependent,”⁹ and this implies that delirium prevention rests not only on individuals but also on broader systems of care. Although nonpharmacological interventions are efficacious at preventing delirium,¹⁰ previous reviews have focused on specific interventions or multiple interventions rather than the systems of care needed to deliver them. Indeed, teams and the quality of their teamwork impact outcomes.¹¹

Herein, we provide a systematic review and meta-analysis of integrated models of care designed to prevent delirium. What distinguishes this review from previous reviews of nonpharmacological interventions to prevent delirium is our focus on discrete models of care that involve collaboration among clinicians. Our goal is to identify the most promising models that deserve further development, investigation, and dissemination. Viewing delirium prevention through a collaborative care lens is consistent with efforts to achieve value-based care and may encourage drawing from the expanding literature outlining the benefits of mental healthcare integration.^12,13 Specifically, a systems perspective highlights the potential for system-wide benefits such as reducing readmissions^14,15 and cost savings.¹⁶

This systematic review and meta-analysis follows PRISMA guidelines. A search of OVID, MEDLINE, CINAHL, Cochrane Database of Systematic Reviews, EMBASE, and PsycINFO was completed by a medical librarian for clinical studies in which models of care were implemented to prevent delirium using PICO (P patient, problem or population; I, intervention; C, comparison, control or comparator; O, outcome) inquiries. Search terms included delirium, acute confusional state, altered mental status, prevention, and control (“delirium”/exp OR “acute confusion”/exp OR “altered mental status”/exp) AND “prevention and control”/exp AND [English]/lim AND [embase]/lim).

One researcher (AK) screened articles by title for relevance. Relevant articles were then divided among four authors (AK, MO, NF, and OB), and the abstracts were screened for eligibility. The authors reviewed the full texts of any potentially eligible studies. Each full text was assigned to two authors for full review. Discrepancies were adjudicated by conference among all authors. In addition, references within all full-text publications were scanned for potential additional articles.

The inclusion criteria for review of full-text articles required English-language description of a model of care with multiple interventions, delirium reported as an outcome, and presence of a comparator group.

“Model of care” was defined by the Cochrane Effective Practice and Organization of Care Review Group as follows: (1) revision of professional roles, including shifting of professional roles or expansion of roles to new tasks; (2) creation of clinical multidisciplinary teams or addition of new members to the team who collaborate inpatient care; (3) delivery of multiple interventions across multiple domains (ie, studies involving a single intervention such as physical therapy or targeting a single domain such as sleep were excluded); and (4) formal integration of services whereby teams work together in collaboration with existing services to enhance care.¹⁷ For this review, we required that studies include a comparator group so that effectiveness of the intervention could be assessed. Quality improvement studies that lacked a comparator group were excluded.

Delirium incidence was the primary outcome and was evaluated by meta-analysis. Heterogeneity was assessed using I² and visual inspection of forest plots. I² values of 25%, 50%, and 75% represent low, moderate, and high heterogeneity, respectively. The studies were pooled according to study type as follows: randomized controlled trials, pre–post design, and other nonrandomized prospective studies. Random effects models were used to calculate estimates using the Comprehensive Meta-Analysis software (Version 3, Biostat, Englewood, New Jersey), which also generated forest plots.

Risk of bias was assessed using criteria established by the Cochrane Collaborative Review Criteria, which lists six categories of potential bias: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, and selective reporting.¹⁷ Each study was assessed by two authors (either MO and AK or MO-P. and OB) for bias and a numerical value was assigned to each of the six categories as follows: 1 = low risk, 2 = unknown/moderate risk, and 3 = high risk. Where scorers disagreed, all authors jointly conferred, and a consensus score was given. The values for each of these six categories were added to create a composite risk-of-bias score for each study, with 6 being the lowest possible score and 18 the highest. Overall risk was classified as follows: <9 = low risk, 9-12 = moderate risk, and >12 = high risk.

Study Selection Process

An initial literature search identified 352 articles. After reviewing the titles, 308 articles were excluded for irrelevance, and 44 abstracts were screened for eligibility. We excluded 27 articles upon abstract review, and the full texts of 17 were obtained for detailed review. In addition, we identified another 10 potentially eligible articles through review of references and obtained full texts of these as well. Of the 27 full-text articles reviewed, 15 were included in this systematic review, 10 of which were suitable for meta-analysis. The Figure shows the PRISMA flow chart.

Study Characteristics

The 15 studies that met the inclusion criteria are summarized in the Table.^18-32 Delirium prevention was among the primary outcomes of 13 studies; delirium outcomes were reported in the other two studies as well, which were primarily designed to assess feasibility.^26,27 Six studies were conducted in the United States, three in Sweden, two in Spain, two in the United Kingdom, and one each conducted in Korea and Canada. Healthcare settings among the included studies involved the intensive care unit (six studies), medical floors (four studies), surgical floors (three studies), a long-term care unit (one study), and an inpatient palliative care service (one study). We categorized the studies according to design and intent as follows: randomized controlled studies (three), pilot feasibility studies (two), pre–post design (six), and other nonrandomized prospective studies (four; Table).

Outcomes Reported

All but one of the studies reported delirium incidence. The most commonly used delirium screening instrument was the Confusion Assessment Method (CAM) or its modified version, the CAM-ICU (11 studies).^33,34 Other methods used to assess mentation included the Richmond Agitation Sedation Scale,³⁵ the Organic Brain Syndrome scale,³⁶ the revised Delirium Rating Scale,³⁷ the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition,³⁸ and the Confusion Rating Scale.³⁹ (Details regarding delirium screening tools can be found in the systematic review by De and Wand.⁴⁰) Researchers performed delirium assessment in nine studies, whereas assessments were performed by clinical staff in the remaining studies. Other outcomes reported included length of stay (LOS), mortality, number of days ventilated, and functional decline. None of the included studies reported cost effectiveness.

Risk of Bias Assessment

Risk of bias assessment identified only two studies—both randomized controlled trials—as low risk (Table). The remaining studies had moderate (four studies) or high risk (nine studies).

Results from Individual Studies

Of the 15 studies, nine reported a statistically significant reduction in delirium incidence, and another two reported a statistically insignificant reduction. In addition, seven of the eight studies that assessed delirium duration found reduced duration in the intervention cohort, and two of the three studies that reported delirium severity found a reduction in the intervention group.

Results of Meta-Analysis

Random effects models were created to meta-analyze groups of studies based on design as follows: randomized controlled trials (three studies^18,19,25), pre–post intervention studies (four of six studies included^28-31), and other nonrandomized studies (three of four studies included^21-23). Meta-analysis was not completed for the two feasibility studies^26,27 because delirium outcome data were limited due to the feasibility study design. The study of Dale et al.³² was excluded from the meta-analysis because the rates of CAM-ICU completion differed substantially between control and intervention groups (0.35 vs 1.49 per 24 hours, respectively), leading to imbalanced between-group sensitivity in delirium detection and Needham et al.²⁰ was also excluded because it reported only days of delirium, not delirium incidence. The study by Lundström et al.²⁴ was also excluded from the meta-analysis because delirium incidence was measured on days 1, 3, and 5, whereas the other studies reported delirium daily.

Meta-analysis of the three randomized controlled trials revealed a pooled odds ratio of 0.56 (95% CI: 0.37-0.85; P = .006) for delirium incidence among intervention group subjects relative to those in comparator groups. The heterogeneity across studies was low (I² = 29%). Pooling data from four pre–post studies found that the odds ratio for delirium incidence was 0.63 (95% CI: 0.37-1.07; P = .09). The heterogeneity across these studies was moderate (I² = 65%). Results from the three eligible, nonrandomized prospective studies were also pooled. The odds ratio for developing delirium among study subjects was 0.79 (95% CI: 0.46-1.37; P = .40), and the heterogeneity among these studies was high (I² = 85%).

We provide a systematic review and meta-analysis of delirium preventive models of care. Meta-analysis of the three randomized controlled trials found that these models of care led to a statistically significant reduction in delirium incidence; study subjects had an 11.5% reduction in absolute delirium incidence. The pooled odds ratios for both of the other sets of nonrandomized studies favored the intervention group but were not significant, each because of one included study. The pre–post meta-analysis failed to reach significance as one of the included studies found a trend toward higher delirium incidence; however, interestingly, in that same study, the overall delirium-free days were significantly reduced overall (24 vs 27, P = .002). Similarly, meta-analysis of the three additional nonrandomized prospective studies failed to reach significance because the largest included study found higher rates of delirium among intervention group subjects. Despite considerable risk of bias in several of these studies, their findings were broadly consistent; all but one study (Gagnon 2012²¹) reported a trend or a significant reduction in delirium incidence, duration, severity, or number of delirium episodes. Moreover, the value of such models of care extended beyond preventing delirium; for instance, other positive outcomes included reduced LOS and fewer medical complications.

Models of care ranged widely with respect to specific interventions, though several common elements highlighted their relevance for delirium care and as potential delirium prevention strategies in future studies. For example, two of the randomized controlled trials^18,19 employed early mobilization, enhanced nutrition, sleep hygiene, early reduction of invasive procedures (eg, urinary catheterization), and pain control in their multicomponent models. Five additional studies also incorporated early mobilization,^{20,22,23,31,32} and three sought to improve sleep quality.^22,28,30 Among other important strategies were delirium screening,^{18,20,22,30,31} monitoring medication,^{18,20,22,26,28,30,32} orientation,^18,21,23,28 addressing vision and hearing impairment,^18,22,23,32 hydration,^18,22,23 avoiding hypoxia,^18,20,30 and staff, patient, and caretaker education.^{19,21,23,27-30} Unique strategies were implemented in certain studies. For instance, one study used massage therapy,²⁸ preventing delays in transfer logistics in another,³⁰ and a third addressed psychosocial problems.²⁵ Overall, the selection of strategies depended on the patient setting; thus, no one care bundle should be expected to emerge as a universal model for delirium prevention. Rather, these results should be interpreted within their specific care contexts and judged on the quality of evidence (eg, effect size and statistically significant findings, low risk of bias, sound experimental design). The one study that failed to find any positive effect on delirium, that of Gagnon et al.,²¹ was conducted on an inpatient palliative care service in Canada, and its negative finding may reflect the unique delirium risk factors in patients who are nearing end of life.

This current review differs from previous delirium prevention reviews in operationally defining a “model of care.” We identified a great deal of variation in specific models and team composition. For example, some interventions were carried out by nurses^18-20,31 and physicians,^20,21,25,32 whereas others involved physical therapists,^20,22,28 medical residents,²³ geriatricians,^22,23,25 pharmacists,²⁶ researchers,¹⁸ and trained volunteers.²² In all cases, the staff roles were expanded to include new tasks, and the clinical team worked collaboratively to administer interventions across multiple domains. Team-related considerations are critical because modern medical care is inherently interdependent.⁹ These broad differences in team composition across studies demonstrate the number of potential options for team structure and function. They also highlight the number of “moving parts” to be considered when designing and implementing delirium care bundles.

Most of the delirium prevention studies implementing models of care are characterized by a substantial risk of bias. We evaluated risk of bias along six categories of potential sources, including random assignment to groups, ability to foresee future group allocation, blinding of participants and personnel to group assignment, blinding of outcome assessment, completeness of outcome data, selective reporting, and other potential sources of bias.¹⁷ Two of the three studies that used randomization had a low risk of bias, and four additional studies had a moderate risk of bias. Allocation concealment was accomplished only in randomized controlled trials, whereas blinding of both subjects and study personnel was not implemented in any of the studies. Although some studies relied on data analysis by research personnel blinded to group membership or the nature of the intervention, others failed to do so or failed to describe data analysis in sufficient detail. Studies also failed to report the percentage of unscorable or otherwise omitted delirium assessments necessary to calculate attrition rates or to understand the comprehensiveness of outcome assessment in a systematic manner. Other potential sources of bias included systematic differences between the intervention and control groups (such as differences in gender composition, age, or delirium risk) at study outset.

A primary limitation of this review is the heterogeneity of settings, interventions, and models of care across included studies. We excluded several studies from this review for being delivered by a single individual or service line (eg, introduction of a geriatric consult service, physical therapy, or volunteers), for providing a single intervention (eg, early ambulation alone), or for multiple interventions targeting a single domain (eg, sleep). We did so because the future of value-based care lies in collaboration of providers and services, and in a way the complexity across and within these studies ultimately reflects the complexity of medical settings as well as the multifactorial nature of delirium. The broader message is a call for increasing the integration of delirium-related care services. As discussed earlier, the high risk of bias across these studies is a limitation of our findings; high-quality evidence on the value of delirium prevention models of care remains limited. Thus, although our review suggests that there are multicomponent models of care that hold promise in mitigating delirium and its outcomes, additional randomized studies are required to confirm the efficacy of such models of care and to test which services, interventions, and clinical domains deserve priority.

To our knowledge, this is the first systematic review and meta-analysis of delirium preventive models of care. Models of care, as defined here, necessarily included a multidisciplinary team in which traditional staff roles had been revised to implement a multicomponent, multidomain intervention. Other recent reviews are available for multicomponent pharmacological and nonpharmacological interventions to prevent and manage delirium,^41-49 but just as important as which interventions are being delivered is the team that delivers them. Care delivery in a complex medical system is more than handing a patient a medication or facilitating ambulation; it requires a choreographed dance of teamwork and integration across services. This review identifies promising models of care that deserve further recognition, refinement, and ultimately widespread implementation.

Acknowledgments

The authors comprise a writing group created through the Delirium Boot Camp, an annual meeting originally sponsored by the Center of Excellence for Delirium in Aging: Research, Training, and Educational Enhancement (CEDARTREE, Boston, Massachusetts); it is currently supported by the Network for Investigation of Delirium: Unifying Scientists (NIDUS, Boston, Massachusetts). The authors would like to thank medical librarian Rita Mitchell (Aurora Health Care, Milwaukee, Wisconsin) for the literature search, senior scientific writer and editor Joe Grundle (Aurora Research Institute, Milwaukee, Wisconsin) for editorial assistance, and graphics specialist Brian Miller (Aurora Research Institute, Milwaukee, Wisconsin) for help with the figures.

Disclosures

The authors report no relevant conflicts of interest.

Funding

No funding was dedicated to the conduct of this review.

Delirium presents as an acute change in mentation characterized by reduced attention, clouding of awareness, and typically an altered level of arousal. It can be caused by a host of medical conditions, medications, or other psychoactive substances and is therefore encountered primarily in acute and postacute medical settings.¹ More than a quarter of all hospitalized patients develop delirium,² with rates up to 80% in the critically ill.³ Similarly, delirium occurs in more than one-third of patients who transition to postacute care.⁴ These high prevalence rates are alarming, especially because delirium is a risk factor for mortality, prolonged hospitalization, institutionalization, and overall higher cost of care.⁵ However, more than a quarter of delirium is preventable.⁶ Evidence-based guidelines for delirium uniformly call for multicomponent prevention strategies,⁷ and these are best delivered through collaborative models of care. In short, delirium impacts healthcare systems; therefore, interventions aimed at preventing delirium and its consequences ought to be systems-based.

Since the Institute of Medicine issued its 1999 report highlighting the critical role of medical errors in healthcare, healthcare systems have increasingly become team-based.⁸ “Medical care is inherently interdependent,”⁹ and this implies that delirium prevention rests not only on individuals but also on broader systems of care. Although nonpharmacological interventions are efficacious at preventing delirium,¹⁰ previous reviews have focused on specific interventions or multiple interventions rather than the systems of care needed to deliver them. Indeed, teams and the quality of their teamwork impact outcomes.¹¹

Herein, we provide a systematic review and meta-analysis of integrated models of care designed to prevent delirium. What distinguishes this review from previous reviews of nonpharmacological interventions to prevent delirium is our focus on discrete models of care that involve collaboration among clinicians. Our goal is to identify the most promising models that deserve further development, investigation, and dissemination. Viewing delirium prevention through a collaborative care lens is consistent with efforts to achieve value-based care and may encourage drawing from the expanding literature outlining the benefits of mental healthcare integration.^12,13 Specifically, a systems perspective highlights the potential for system-wide benefits such as reducing readmissions^14,15 and cost savings.¹⁶

This systematic review and meta-analysis follows PRISMA guidelines. A search of OVID, MEDLINE, CINAHL, Cochrane Database of Systematic Reviews, EMBASE, and PsycINFO was completed by a medical librarian for clinical studies in which models of care were implemented to prevent delirium using PICO (P patient, problem or population; I, intervention; C, comparison, control or comparator; O, outcome) inquiries. Search terms included delirium, acute confusional state, altered mental status, prevention, and control (“delirium”/exp OR “acute confusion”/exp OR “altered mental status”/exp) AND “prevention and control”/exp AND [English]/lim AND [embase]/lim).

One researcher (AK) screened articles by title for relevance. Relevant articles were then divided among four authors (AK, MO, NF, and OB), and the abstracts were screened for eligibility. The authors reviewed the full texts of any potentially eligible studies. Each full text was assigned to two authors for full review. Discrepancies were adjudicated by conference among all authors. In addition, references within all full-text publications were scanned for potential additional articles.

The inclusion criteria for review of full-text articles required English-language description of a model of care with multiple interventions, delirium reported as an outcome, and presence of a comparator group.

“Model of care” was defined by the Cochrane Effective Practice and Organization of Care Review Group as follows: (1) revision of professional roles, including shifting of professional roles or expansion of roles to new tasks; (2) creation of clinical multidisciplinary teams or addition of new members to the team who collaborate inpatient care; (3) delivery of multiple interventions across multiple domains (ie, studies involving a single intervention such as physical therapy or targeting a single domain such as sleep were excluded); and (4) formal integration of services whereby teams work together in collaboration with existing services to enhance care.¹⁷ For this review, we required that studies include a comparator group so that effectiveness of the intervention could be assessed. Quality improvement studies that lacked a comparator group were excluded.

Delirium incidence was the primary outcome and was evaluated by meta-analysis. Heterogeneity was assessed using I² and visual inspection of forest plots. I² values of 25%, 50%, and 75% represent low, moderate, and high heterogeneity, respectively. The studies were pooled according to study type as follows: randomized controlled trials, pre–post design, and other nonrandomized prospective studies. Random effects models were used to calculate estimates using the Comprehensive Meta-Analysis software (Version 3, Biostat, Englewood, New Jersey), which also generated forest plots.

Risk of bias was assessed using criteria established by the Cochrane Collaborative Review Criteria, which lists six categories of potential bias: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, and selective reporting.¹⁷ Each study was assessed by two authors (either MO and AK or MO-P. and OB) for bias and a numerical value was assigned to each of the six categories as follows: 1 = low risk, 2 = unknown/moderate risk, and 3 = high risk. Where scorers disagreed, all authors jointly conferred, and a consensus score was given. The values for each of these six categories were added to create a composite risk-of-bias score for each study, with 6 being the lowest possible score and 18 the highest. Overall risk was classified as follows: <9 = low risk, 9-12 = moderate risk, and >12 = high risk.

Study Selection Process

An initial literature search identified 352 articles. After reviewing the titles, 308 articles were excluded for irrelevance, and 44 abstracts were screened for eligibility. We excluded 27 articles upon abstract review, and the full texts of 17 were obtained for detailed review. In addition, we identified another 10 potentially eligible articles through review of references and obtained full texts of these as well. Of the 27 full-text articles reviewed, 15 were included in this systematic review, 10 of which were suitable for meta-analysis. The Figure shows the PRISMA flow chart.

Study Characteristics

The 15 studies that met the inclusion criteria are summarized in the Table.^18-32 Delirium prevention was among the primary outcomes of 13 studies; delirium outcomes were reported in the other two studies as well, which were primarily designed to assess feasibility.^26,27 Six studies were conducted in the United States, three in Sweden, two in Spain, two in the United Kingdom, and one each conducted in Korea and Canada. Healthcare settings among the included studies involved the intensive care unit (six studies), medical floors (four studies), surgical floors (three studies), a long-term care unit (one study), and an inpatient palliative care service (one study). We categorized the studies according to design and intent as follows: randomized controlled studies (three), pilot feasibility studies (two), pre–post design (six), and other nonrandomized prospective studies (four; Table).

Outcomes Reported

All but one of the studies reported delirium incidence. The most commonly used delirium screening instrument was the Confusion Assessment Method (CAM) or its modified version, the CAM-ICU (11 studies).^33,34 Other methods used to assess mentation included the Richmond Agitation Sedation Scale,³⁵ the Organic Brain Syndrome scale,³⁶ the revised Delirium Rating Scale,³⁷ the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition,³⁸ and the Confusion Rating Scale.³⁹ (Details regarding delirium screening tools can be found in the systematic review by De and Wand.⁴⁰) Researchers performed delirium assessment in nine studies, whereas assessments were performed by clinical staff in the remaining studies. Other outcomes reported included length of stay (LOS), mortality, number of days ventilated, and functional decline. None of the included studies reported cost effectiveness.

Risk of Bias Assessment

Risk of bias assessment identified only two studies—both randomized controlled trials—as low risk (Table). The remaining studies had moderate (four studies) or high risk (nine studies).

Results from Individual Studies

Of the 15 studies, nine reported a statistically significant reduction in delirium incidence, and another two reported a statistically insignificant reduction. In addition, seven of the eight studies that assessed delirium duration found reduced duration in the intervention cohort, and two of the three studies that reported delirium severity found a reduction in the intervention group.

Results of Meta-Analysis

Random effects models were created to meta-analyze groups of studies based on design as follows: randomized controlled trials (three studies^18,19,25), pre–post intervention studies (four of six studies included^28-31), and other nonrandomized studies (three of four studies included^21-23). Meta-analysis was not completed for the two feasibility studies^26,27 because delirium outcome data were limited due to the feasibility study design. The study of Dale et al.³² was excluded from the meta-analysis because the rates of CAM-ICU completion differed substantially between control and intervention groups (0.35 vs 1.49 per 24 hours, respectively), leading to imbalanced between-group sensitivity in delirium detection and Needham et al.²⁰ was also excluded because it reported only days of delirium, not delirium incidence. The study by Lundström et al.²⁴ was also excluded from the meta-analysis because delirium incidence was measured on days 1, 3, and 5, whereas the other studies reported delirium daily.

Meta-analysis of the three randomized controlled trials revealed a pooled odds ratio of 0.56 (95% CI: 0.37-0.85; P = .006) for delirium incidence among intervention group subjects relative to those in comparator groups. The heterogeneity across studies was low (I² = 29%). Pooling data from four pre–post studies found that the odds ratio for delirium incidence was 0.63 (95% CI: 0.37-1.07; P = .09). The heterogeneity across these studies was moderate (I² = 65%). Results from the three eligible, nonrandomized prospective studies were also pooled. The odds ratio for developing delirium among study subjects was 0.79 (95% CI: 0.46-1.37; P = .40), and the heterogeneity among these studies was high (I² = 85%).

We provide a systematic review and meta-analysis of delirium preventive models of care. Meta-analysis of the three randomized controlled trials found that these models of care led to a statistically significant reduction in delirium incidence; study subjects had an 11.5% reduction in absolute delirium incidence. The pooled odds ratios for both of the other sets of nonrandomized studies favored the intervention group but were not significant, each because of one included study. The pre–post meta-analysis failed to reach significance as one of the included studies found a trend toward higher delirium incidence; however, interestingly, in that same study, the overall delirium-free days were significantly reduced overall (24 vs 27, P = .002). Similarly, meta-analysis of the three additional nonrandomized prospective studies failed to reach significance because the largest included study found higher rates of delirium among intervention group subjects. Despite considerable risk of bias in several of these studies, their findings were broadly consistent; all but one study (Gagnon 2012²¹) reported a trend or a significant reduction in delirium incidence, duration, severity, or number of delirium episodes. Moreover, the value of such models of care extended beyond preventing delirium; for instance, other positive outcomes included reduced LOS and fewer medical complications.

Models of care ranged widely with respect to specific interventions, though several common elements highlighted their relevance for delirium care and as potential delirium prevention strategies in future studies. For example, two of the randomized controlled trials^18,19 employed early mobilization, enhanced nutrition, sleep hygiene, early reduction of invasive procedures (eg, urinary catheterization), and pain control in their multicomponent models. Five additional studies also incorporated early mobilization,^{20,22,23,31,32} and three sought to improve sleep quality.^22,28,30 Among other important strategies were delirium screening,^{18,20,22,30,31} monitoring medication,^{18,20,22,26,28,30,32} orientation,^18,21,23,28 addressing vision and hearing impairment,^18,22,23,32 hydration,^18,22,23 avoiding hypoxia,^18,20,30 and staff, patient, and caretaker education.^{19,21,23,27-30} Unique strategies were implemented in certain studies. For instance, one study used massage therapy,²⁸ preventing delays in transfer logistics in another,³⁰ and a third addressed psychosocial problems.²⁵ Overall, the selection of strategies depended on the patient setting; thus, no one care bundle should be expected to emerge as a universal model for delirium prevention. Rather, these results should be interpreted within their specific care contexts and judged on the quality of evidence (eg, effect size and statistically significant findings, low risk of bias, sound experimental design). The one study that failed to find any positive effect on delirium, that of Gagnon et al.,²¹ was conducted on an inpatient palliative care service in Canada, and its negative finding may reflect the unique delirium risk factors in patients who are nearing end of life.

This current review differs from previous delirium prevention reviews in operationally defining a “model of care.” We identified a great deal of variation in specific models and team composition. For example, some interventions were carried out by nurses^18-20,31 and physicians,^20,21,25,32 whereas others involved physical therapists,^20,22,28 medical residents,²³ geriatricians,^22,23,25 pharmacists,²⁶ researchers,¹⁸ and trained volunteers.²² In all cases, the staff roles were expanded to include new tasks, and the clinical team worked collaboratively to administer interventions across multiple domains. Team-related considerations are critical because modern medical care is inherently interdependent.⁹ These broad differences in team composition across studies demonstrate the number of potential options for team structure and function. They also highlight the number of “moving parts” to be considered when designing and implementing delirium care bundles.

Most of the delirium prevention studies implementing models of care are characterized by a substantial risk of bias. We evaluated risk of bias along six categories of potential sources, including random assignment to groups, ability to foresee future group allocation, blinding of participants and personnel to group assignment, blinding of outcome assessment, completeness of outcome data, selective reporting, and other potential sources of bias.¹⁷ Two of the three studies that used randomization had a low risk of bias, and four additional studies had a moderate risk of bias. Allocation concealment was accomplished only in randomized controlled trials, whereas blinding of both subjects and study personnel was not implemented in any of the studies. Although some studies relied on data analysis by research personnel blinded to group membership or the nature of the intervention, others failed to do so or failed to describe data analysis in sufficient detail. Studies also failed to report the percentage of unscorable or otherwise omitted delirium assessments necessary to calculate attrition rates or to understand the comprehensiveness of outcome assessment in a systematic manner. Other potential sources of bias included systematic differences between the intervention and control groups (such as differences in gender composition, age, or delirium risk) at study outset.

A primary limitation of this review is the heterogeneity of settings, interventions, and models of care across included studies. We excluded several studies from this review for being delivered by a single individual or service line (eg, introduction of a geriatric consult service, physical therapy, or volunteers), for providing a single intervention (eg, early ambulation alone), or for multiple interventions targeting a single domain (eg, sleep). We did so because the future of value-based care lies in collaboration of providers and services, and in a way the complexity across and within these studies ultimately reflects the complexity of medical settings as well as the multifactorial nature of delirium. The broader message is a call for increasing the integration of delirium-related care services. As discussed earlier, the high risk of bias across these studies is a limitation of our findings; high-quality evidence on the value of delirium prevention models of care remains limited. Thus, although our review suggests that there are multicomponent models of care that hold promise in mitigating delirium and its outcomes, additional randomized studies are required to confirm the efficacy of such models of care and to test which services, interventions, and clinical domains deserve priority.

To our knowledge, this is the first systematic review and meta-analysis of delirium preventive models of care. Models of care, as defined here, necessarily included a multidisciplinary team in which traditional staff roles had been revised to implement a multicomponent, multidomain intervention. Other recent reviews are available for multicomponent pharmacological and nonpharmacological interventions to prevent and manage delirium,^41-49 but just as important as which interventions are being delivered is the team that delivers them. Care delivery in a complex medical system is more than handing a patient a medication or facilitating ambulation; it requires a choreographed dance of teamwork and integration across services. This review identifies promising models of care that deserve further recognition, refinement, and ultimately widespread implementation.

Acknowledgments

The authors comprise a writing group created through the Delirium Boot Camp, an annual meeting originally sponsored by the Center of Excellence for Delirium in Aging: Research, Training, and Educational Enhancement (CEDARTREE, Boston, Massachusetts); it is currently supported by the Network for Investigation of Delirium: Unifying Scientists (NIDUS, Boston, Massachusetts). The authors would like to thank medical librarian Rita Mitchell (Aurora Health Care, Milwaukee, Wisconsin) for the literature search, senior scientific writer and editor Joe Grundle (Aurora Research Institute, Milwaukee, Wisconsin) for editorial assistance, and graphics specialist Brian Miller (Aurora Research Institute, Milwaukee, Wisconsin) for help with the figures.

Disclosures

The authors report no relevant conflicts of interest.

Funding

No funding was dedicated to the conduct of this review.

Ulcerative colitis (UC) is a chronic inflammatory condition of the colonic mucosa. Classically, it starts in the rectum and can extend continuously from the distal to the proximal colon. The defining clinical symptom of UC is bloody diarrhea, typically accompanied by rectal urgency and mucus discharge. The natural history of this disease includes periods of exacerbations and remissions occurring spontaneously or in response to medical treatment.¹

Acute severe ulcerative colitis (ASUC) is a potentially life-threatening complication of UC that typically requires hospitalization and interdisciplinary care between hospitalists, gastroenterologists, and colorectal or general surgeons. The risk of a patient with UC requiring hospitalization for ASUC ranges from 15%-25%^2,3 and, in total, UC accounts for 30,000 hospital visits annually.⁴ The direct medical costs exceed $4 billion annually, with hospital costs of over $960 million.⁵ Historically, mortality from ASUC was as high as 24% but decreased substantially to 7% after the introduction of systemic corticosteroid therapy.⁶ Further advances in care have reduced mortality to approximately 1% or less.^7,8 Nonetheless, up to 20% of patients admitted with ASUC have a colectomy on their first admission, and this rate rises to 40% after two admissions.²

To categorize UC severity, assess patients using the Truelove and Witt’s criteria. The system classifies patients as having mild, moderate, severe, or fulminant disease. Severe disease by these criteria includes patients with >6 bloody bowel movements per day and at least one of the following clinical features: fever (>37.8°C), tachycardia (>90 bpm), anemia (hemoglobin <10.5 g/dl), or elevated inflammatory markers (traditionally, erythrocyte sedimentation rate greater than 30 mm/h or, more recently, C-reactive protein (CRP) greater than 30 mg/L. (Table 1).^6,9

Fulminant colitis refers to a subgroup of patients with more than 10 stools per day, continuous bleeding, abdominal pain, colonic dilatation on abdominal X-ray film, and severe toxic symptoms including fever and anorexia. Such patients are at risk of progressing to toxic megacolon and bowel perforation.¹⁰

Patients with ASUC almost always require hospitalization for their disease management. In many cases, these patients have been receiving outpatient oral prednisone 40-60 mg daily but continue to have ongoing disease activity.¹¹ Most patients will require close clinical monitoring, frequent blood testing, endoscopic or radiologic evaluation, as well as administration of intravenous corticosteroids. The average length of stay (LOS) ranges from 4.6 to 12.5 days, depending on disease severity.¹² Not surprisingly, Kelso et al. reported that predictors of hospital LOS greater than four days include initiating a biologic drug in the hospital, undergoing two or more imaging modalities and treatment with intravenous steroids,¹³ and so it is rare that patients do not meet billing requirements for an inpatient level of care.

The multifaceted initial inpatient evaluation of patients with ASUC aims to assess disease severity, identify and prevent potential complications, and initiate planning for potential failure of first-line pharmacologic therapy. Due to the accumulating evidence that involving physicians with expertise in managing ASUC improves outcomes, gastroenterologists should be involved in the care of patients with ASUC from the time of their admission.^14,15 Additionally, creating standardized care pathways for the management of ASUC can reduce cost, LOS, and improve quality.¹⁶

History and Physical Examination

Patients should be asked about fever, abdominal pain, nausea, emesis, bloating, weight loss, and bowel movements (frequency, consistency, the presence of blood, urgency, nighttime awakenings). The number of bowel movements over a 24-hour period should be quantified as this helps assess the overall disease severity (Table 1).

The patient’s initial inflammatory bowel disease (IBD) history is also essential. The review of pertinent information regarding the patient’s initial diagnosis of UC includes the severity and anatomic extent of disease, extraintestinal manifestations, previous medical therapies, and surgical interventions. Exposure to nonsteroidal anti-inflammatory drugs (NSAIDs) or antibiotics should be identified as they may precipitate flares.¹⁷ Travel history may be pertinent as travel increases the risk of infections with food-borne or parasitic pathogens.¹⁸

Physical examination begins with an assessment of vital signs and volume status. Abdominal examination should include evaluation of bowel sounds, an assessment of distention, location, the extent of abdominal tenderness, and peritoneal signs. The abdominal exam should be interpreted in the context of the patient’s medications, as the use of steroid or analgesic therapies may affect the sensitivity for detecting complications. An external rectal exam evaluating perianal disease should be performed, as perianal disease raises concern for Crohn’s, a disease whose surgical management differs from UC.

A careful exam for extraintestinal manifestations is also essential. The skin should be evaluated for any new rashes, especially on the anterior shin consistent with erythema nodosum or ulcerated lesions on the skin suggestive of pyoderma gangrenosum. The peripheral joints should also be examined for any synovitis. Additional examinations should be performed based on any reported symptoms (eg, the ophthalmic exam for uveitis or scleritis if visual changes or eye pain are reported). Some extraintestinal manifestations require subspecialty consultation and comanagement to guide disease therapy. Patients with underlying pyoderma gangrenosum may require a dermatology consultation to guide management. Ocular inflammation requires ophthalmology involvement, and inflammatory arthritis is best comanaged with rheumatology.¹⁹

Laboratory Testing

Initial testing should include a complete blood count with differential, basic metabolic panel, and liver chemistries including alkaline phosphatase and albumin. When relevant, pregnancy testing should be performed. Measure CRP on admission so that its trajectory can be followed during therapy. However, a normal CRP does not exclude the presence of a UC flare as a subset of patients with ASUC will have a normal CRP despite severe mucosal inflammation.²⁰

Since one-third of patients do not respond to intravenous corticosteroids and will require rescue therapy during the hospitalization with infliximab or cyclosporine, anticipatory testing for these medications should be performed on admission to avoid delays in the administration of rescue therapy.^6,21 This should include an interferon-gamma release assay (eg, quantiferon gold) to test for latent tuberculosis and hepatitis B serologies in anticipation of possible treatment with infliximab. An interferon-gamma release assay is preferred to a tuberculin skin test because patients may be anergic, and a skin test does not provide a control to determine whether a negative test is due to anergy. In contrast, although a quantiferon gold test can be indeterminate in ASUC due to disease activity and systemic steroids, the results indicate if the patient is anergic so that one will not rely on a false-negative result. In the event of an equivocal result, a careful clinical assessment for risks of TB exposures should be elicited, and a chest radiograph should be obtained.²² In patients with prior high risk of tuberculosis exposures or a positive test for tuberculosis, an infectious disease specialist should be consulted early to advise if therapy should be started in preparation for the potential use of infliximab.²³ In cases where cyclosporine may be considered, magnesium and total cholesterol level should be checked. Sending thiopurine methyltranferase (TPMT) enzyme activity should be considered as well, in case of a need for future thiopurine use for maintenance of disease activity.²⁴

Infectious diarrhea may be indistinguishable from ASUC and may also be the trigger of a flare; thus, it is important to rule out infection with stool microbiologic studies. Most importantly, Clostridium difficile infection must be ruled out in all patients with ASUC. Although patients with IBD, especially those with UC, have significantly higher rates of asymptomatic C. difficile carriage than the general population, a positive polymerase chain reaction test for C. difficile in a patient with ASUC should prompt treatment with oral vancomycin.²⁵ However, if carriage if suspected and a subsequent enzyme-linked immunoassay for C. difficile toxin is negative, treatment can be discontinued. Active C. difficile infection in patients with IBD is associated with increased disease severity, greater length of hospital stay, and increased the likelihood of colectomy and mortality.^26,27 Other bacterial infections including Escherichia coli, Campylobacter, Shigella, Salmonella, Yersinia, Entamoeba histolytica, as well as other parasitic infestations may mimic UC. Testing should be considered in cases of foreign travel, immunosuppression or contact with other persons with diarrhea.^7,28 Routine testing of these other enteric infections without a clear exposure risk is of little benefit and may raise costs.^23,29

Radiologic Evaluation

A plain X-ray film of the abdomen should be obtained in all patients on admission to evaluate for evolving colonic dilation or undiagnosed free air. Small bowel distension >3 cm may predict an increased risk of colectomy.³⁰ Clinicians must be mindful that steroids can mask peritoneal signs and that retroperitoneal perforations may not be apparent on plain X-ray films. Nonetheless, a CT of the abdomen is usually not necessary and should be reserved for cases with severe abdominal pain out of proportion to clinical signs in which a plain X-ray film is unrevealing. Judicious use of CT imaging is especially important in younger patients, as there is growing concern that patients with IBD may be exposed to potentially harmful cumulative levels of radiation in their lifetime from repeated CT imaging.³¹

Endoscopic Evaluation

Flexible sigmoidoscopy aids in the assessment of disease severity and extent and biopsies can assist in ruling out a diagnosis of cytomegalovirus (CMV) colitis in patients already on immunosuppression. For this reason, many clinicians prefer to perform a sigmoidoscopy on admission.²³ If one is not performed on admission, a sigmoidoscopy is advised in all patients who are not responding adequately after 72 hours of intravenous steroid therapy in order to rule out superimposed CMV colitis.²⁸

Sigmoidoscopy should be avoided in patients with toxic megacolon and when there is a concern for peritonitis. A complete colonoscopy is rarely indicated in the acute setting and carries a theoretical risk of colonic perforation.⁷

The first therapeutic steps aim to reduce inflammation with the use of systemic corticosteroids, avoid colonic and extraintestinal complications, and plan for the potential need for rescue therapy.

Intravenous Corticosteroids

The cornerstone of ASUC management is treatment with intravenous corticosteroids. Their initiation should not be delayed in patients with an established diagnosis of UC while waiting for results of evaluations for infectious colitis. Even among patients who have failed oral steroids, a meta-regression analysis showed that two-thirds of patients will still respond to intravenous corticosteroids.^21,32 Methylprednisolone 20 mg IV three times daily (or hydrocortisone 100 mg IV three times daily) is a standard regimen; higher doses do not provide additional benefit.²¹ Patients’ response to intravenous steroids should be assessed with repeat labs including CRP and an assessment of the total number of bowel movements over a 24-hour period, with special attention to their overall response after three days of treatment.^33-36

Intravenous Fluids

Many patients admitted with ASUC will have significant volume depletion, and intravenous fluids should be administered in a manner like other volume-depleted or oral-intake-restricted patients.

Venous Thromboembolism Prophylaxis

The risk of VTE in hospitalized patients with IBD exceeds that of inpatients without IBD, approximately 2%, a risk similar to patients with respiratory failure.³⁷ Additionally, VTE in hospitalized patients with IBD is associated with a 2.5-fold increase in mortality.^38,39 Therefore, all patients hospitalized with ASUC should receive subcutaneous unfractionated or low molecular weight heparin or fondaparinux for VTE prophylaxis. Rectal bleeding, expected in ASUC, is not a contraindication to chemo-prophylaxis. Additionally, it is important to check if patients are receiving the ordered VTE prophylaxis.^40,41 Pleet et al. found that only 7% of patients at a tertiary center had adequate prophylaxis for greater than 80% of their hospitalization.⁴¹

Unnecessary or Potentially Harmful Medications

Several medications have the potential for misuse in patients hospitalized with UC.

Antimotility Agents

Loperamide, diphenoxylate, and opiate antidiarrheals should not be used as they may provoke toxic megacolon.⁴² Similarly, drugs with antimotility side effects (eg, anticholinergics) should be avoided.

Opiates

In addition to their undesirable antimotility effect, the use of opiates has been associated with poor outcomes among inpatients and outpatients with IBD, including increased morbidity and mortality.^43,44 Pain severe enough to require opiates should raise suspicion for toxic megacolon, perforation, or a noninflammatory etiology. If opiates are utilized, they should be ordered as one-time doses and the patient should be reassessed for each dose.

Nonsteroidal Anti-inflammatory Drugs

These drugs, which include oral NSAIDs, intravenous ketorolac, and topic diclofenac gels, may increase disease activity in inflammatory bowel disease and should be avoided.¹⁷

5-aminosalicylates (5-ASA)

A small proportion of patients experience a paradoxical worsening of diarrhea due to the use of 5-ASA agents such as mesalamine. It is reasonable to discontinue or avoid the use of 5-ASA agents in hospitalized patients, especially as there is little to no benefit from combining a 5-ASA with a biologic or immunosuppressive drug.⁴⁵

Antibiotics

There is no role for the routine use of antibiotics in patients hospitalized with ASUC. ^23,46,47 Inappropriate use of antibiotics raises the risk of C. difficile infection and antibiotic resistance. However, in cases of suspected toxic megacolon or perforation, antibiotics should be administered. In situations in which a patient is treated with triple immunosuppression (ie, steroids plus two other agents, cyclosporine and mercaptopurine) antibiotic prophylaxis for Pneumocystis jiroveci is advisable.⁴⁸ Using a large insurance database, Long et al. reported a low absolute incidence of Pneumocystis jiroveci in IBD patients but noted that the risk in patients with IBD was still significantly higher than matched controls. While it can be considered, we typically refrain from using prophylaxis in patients on double immunosuppression (for example, steroids plus infliximab) due to the potential adverse effects of antibiotics in this population, though many advocate using prophylaxis for all patients on cyclosporine even if this is only double immunosuppressive therapy.²³

Surgical Consultation

Involving a surgeon early in an ASUC patient’s care­—before needing urgent colectomy—is critical. As part of the consultation, a surgeon experienced in IBD should meet with patients to discuss multistage colectomy with ileostomy and potential future J-pouch (ileal pouch-anal anastomosis) formation. Patients should be given ample opportunity to ask questions before surgery may become urgent. Also, patients should be counseled on realistic expectations of ostomy and pouch function and, ideally, meet with an ostomy nurse.²³

At some centers, surgical consultation is requested on the first hospital day, but this can result in consultations for patients who ultimately respond to intravenous steroids. Therefore, some centers advocate for surgical consultation only after a patient has failed treatment with intravenous steroids (ie, day three to four) when the risk of needing surgical management increases.²³

Nutrition

Bowel rest with parenteral nutrition does not improve outcomes in ASUC versus an oral diet, and there is no contraindication to allowing patients to continue on a regular diet unless they have toxic megacolon or other signs of fulminant colitis.^49,50 However, patients may feel better eating less, as this will reduce their bowel movement frequency. Unfortunately, this can give a false sense of reassurance that the patient is improving. Therefore, it remains important to evaluate a patient’s symptoms in the context of their food intake.

Assessing Response to Steroids

Patients who do not respond adequately to the first-line intravenous steroid therapy will require medical or surgical rescue therapy; therefore, deciding whether a patient has responded is essential. Patients should have less than four bowel movements per day – ideally just one to two – with no blood to indicate a complete response. For more ambiguous situations, although there is no strict definition of steroid responsiveness, multiple prediction indices have attempted to identify patients who will require rescue therapy. One of the simplest, the Oxford index, illustrates two of the most critical parameters to follow, stool frequency and CRP.⁵¹ In a preinfliximab cohort, Oxford index predicted an 85% likelihood of colectomy in patients with eight or more daily bowel movements or with three to eight daily bowel movements and a CRP greater than 45 mg/L after three days of intravenous steroid treatment.⁵² To assist with assessing responsiveness to therapy, we ask patients to log their bowel movements – either on paper or on a whiteboard in the hospital room – so that we can review their progress daily. Other predictors of colectomy include hypoalbuminemia, scoring of endoscopic severity, and colonic dilation.⁵³

Patients who fail to respond to intravenous corticosteroids after three days^33,35 of treatment should be started on rescue therapy with infliximab or cyclosporine or undergo colectomy. A common pitfall in the treatment of ASUC is waiting for a response to steroids beyond this time frame, after which patients are unlikely to benefit.^34,36 Furthermore, patients for whom surgical rescue therapy is delayed have higher operative morbidity and mortality.^54,55 Because timely decision making regarding rescue therapy is crucial to optimizing outcomes, patient education efforts regarding potential rescue therapy should take place on admission or soon after, rather than waiting to ascertain steroid responsiveness.

Medical options for rescue therapy include the antitumor necrosis factor (anti-TNF) agent infliximab or the calcineurin inhibitor cyclosporine. In general, infliximab and cyclosporine have been found to be roughly equivalent in efficacy in clinical trials regarding response, remission, and colectomy at 12 months.^56,57 However, many clinicians prefer infliximab due to its relative ease of use, familiarity with the agent from outpatient experience, and ability to continue to use long term for maintenance of disease remission.⁵⁸ In contrast to infliximab, intravenous cyclosporine requires closer monitoring and labs to assess the therapeutic trough level. The decision regarding which drug to use should be made on a case-by-case basis in conjunction with a gastroenterologist experienced in their use, and if no such specialist is available, transfer to a specialized center should be considered. Generally, successive treatment with cyclosporine or infliximab followed by third-line salvage therapy with the other drug should be avoided due to low rates of response and high rates of adverse events.⁵⁹

Infliximab

Infliximab is an intravenously-administered anti-TNF monoclonal chimeric antibody that is effective both for outpatient treatment of moderate to severe UC and inpatient treatment of ASUC.¹ It is relatively contraindicated in patients with untreated latent tuberculosis, demyelinating disease, advanced congestive heart failure, or uncontrolled infection.

The optimal dosing strategy for infliximab in ASUC is unknown. Infliximab clearance in the setting of ASUC is increased, partly because it is bound to albumin, which is often low in ASUC, and partly because it is excreted in the stool.^60,61 As a result, accelerated loading doses may be more successful than a typical loading schedule,⁶² and most clinicians use alternative dosing strategies.⁶³ Our typical approach for ASUC is an initial dose of 10 mg/kg rather than 5 mg/kg, with an additional 10 mg/kg dose 48-72 hours later if an adequate clinical response is lacking. Patients who respond to infliximab can continue to use the drug as an outpatient for maintenance of remission.

Cyclosporine

Cyclosporine is a fast-acting immunosuppressive agent that acts primarily via T-cell inhibition. Although older literature used a dose of 4 mg/kg per day, a randomized trial demonstrated similar response rates to a dose of 2 mg/kg per day.⁶⁴ Patients receiving treatment with cyclosporine, which is given as a continuous infusion, must be monitored for toxicities. These can include potentially severe infection, seizures (often associated with low total cholesterol or hypomagnesemia), electrolyte abnormalities, renal impairment, hypertension, hypertrichosis, tremor, and others.⁶⁵

Before initiation of treatment, serum cholesterol levels should be obtained to screen for low total cholesterol that may portend risk of seizures on the drug. Additionally, baseline creatinine and magnesium should be established. While on treatment, daily serum cyclosporine levels and electrolytes including magnesium should be measured. Patients who respond to intravenous cyclosporine must be transitioned to oral cyclosporine and have stable drug levels before discharge. Unfortunately, oral cyclosporine has not been shown to be as effective as long-term maintenance therapy. Therefore, cyclosporine can only be used as a “bridge” to another therapy. Historically, thiopurines like azathioprine or mercaptopurine have been used for this purpose because they are effective for the treatment of UC but may require months to have a full therapeutic effect. There have been promising reports of using vedolizumab similarly.^66,67 Vedolizumab is a monoclonal antibody that selectively blocks lymphocyte trafficking to the gut that, like thiopurines, has an onset of action that is significantly longer than calcineurin and TNF inhibitors.

Colectomy should be considered as a second- or third-line therapy for patients who fail to respond to intravenous corticosteroids. In an analysis of 10 years of data from the Nationwide Inpatient Sample, mortality rates for colectomy in this setting varied from 0.7% at high volume centers to 4% at low volume centers.⁶⁸ Therefore, if a patient is not hospitalized at a center with expertise in colectomy for UC, transfer to a specialized center should be considered. Colectomy should be performed promptly in all the patients who have failed rescue therapy with infliximab or cyclosporine or have opted against medical rescue therapy. Surgery should be performed emergently in patients with toxic megacolon, uncontrolled colonic hemorrhage or perforation.

Physician and center-level characteristics are associated with the quality of care and outcomes in ASUC. Gastroenterologists with expertise in IBD are more likely than other gastroenterologists to request appropriate surgical consultation for steroid-refractory patients,⁶⁹ and inpatients with ASUC primarily cared by gastroenterologists rather than nongastroenterologists have lower in-hospital and one-year mortality.¹⁴ Moreover, surgical outcomes differ based on center volume, with higher volume centers having lower rates of postoperative mortality.^68,70 However, even at referral centers, key metrics of care quality such as rates of VTE prophylaxis, testing for C. difficile, and timely rescue therapy for steroid-refractory UC patients are suboptimal, with only 70%-82% of patients with IBD hospitalized at four referral centers in Canada meeting these metrics.⁷¹

Inpatient clinical pathways reduce LOS, reduce hospital costs, and likely reduce complications.⁷² For this reason, a consensus group recommended the use of care pathways for the management of ASUC and, although there is little data on the use of pathways for ASUC specifically, the use of such a pathway in the United Kingdom was associated with improved metrics including LOS, time to VTE prophylaxis, testing of stool for infection, CRP measurement, and timely gastroenterologist consultation.^16,18

In general, patients should enter clinical remission, defined as resolution of rectal bleeding and diarrhea or altered bowel habits,⁷³ before discharge, and achieving this may require a relatively prolonged hospitalization. Most patients should have one to two bowel movements a day without blood but, at a minimum, all should have less than four nonbloody bowel movements per day. Patients are candidates for discharge if they remain well after transitioning to oral prednisone at a dose of 40-60 mg daily and tolerate a regular diet.

For patients who initiated infliximab during their admission, plans for outpatient infusions including insurance approval should be made before discharge, and patients who started cyclosporine should be transitioned to oral dosing and have stable serum concentrations before leaving the hospital. Patients should leave with a preliminary plan for a steroid taper, which may vary depending on their clinical presentation. Usually, gastroenterology follow-up should be arranged after two weeks following discharge, but patients on cyclosporine need sooner laboratory monitoring.

The care of patients with ASUC requires an interdisciplinary team and close collaboration between hospitalists, gastroenterologists, and surgeons. Patients should be treated with intravenous corticosteroids and monitored carefully for response and need for rescue therapy. Establishing algorithms for the management of patients with ASUC can further improve the care of these complex patients.

Disclosures

Drs. Feuerstein, Fudman, and Sattler report no potential conflict of interest.

Funding

This work was not supported by any grant.

Ulcerative colitis (UC) is a chronic inflammatory condition of the colonic mucosa. Classically, it starts in the rectum and can extend continuously from the distal to the proximal colon. The defining clinical symptom of UC is bloody diarrhea, typically accompanied by rectal urgency and mucus discharge. The natural history of this disease includes periods of exacerbations and remissions occurring spontaneously or in response to medical treatment.¹

Acute severe ulcerative colitis (ASUC) is a potentially life-threatening complication of UC that typically requires hospitalization and interdisciplinary care between hospitalists, gastroenterologists, and colorectal or general surgeons. The risk of a patient with UC requiring hospitalization for ASUC ranges from 15%-25%^2,3 and, in total, UC accounts for 30,000 hospital visits annually.⁴ The direct medical costs exceed $4 billion annually, with hospital costs of over $960 million.⁵ Historically, mortality from ASUC was as high as 24% but decreased substantially to 7% after the introduction of systemic corticosteroid therapy.⁶ Further advances in care have reduced mortality to approximately 1% or less.^7,8 Nonetheless, up to 20% of patients admitted with ASUC have a colectomy on their first admission, and this rate rises to 40% after two admissions.²

To categorize UC severity, assess patients using the Truelove and Witt’s criteria. The system classifies patients as having mild, moderate, severe, or fulminant disease. Severe disease by these criteria includes patients with >6 bloody bowel movements per day and at least one of the following clinical features: fever (>37.8°C), tachycardia (>90 bpm), anemia (hemoglobin <10.5 g/dl), or elevated inflammatory markers (traditionally, erythrocyte sedimentation rate greater than 30 mm/h or, more recently, C-reactive protein (CRP) greater than 30 mg/L. (Table 1).^6,9

Fulminant colitis refers to a subgroup of patients with more than 10 stools per day, continuous bleeding, abdominal pain, colonic dilatation on abdominal X-ray film, and severe toxic symptoms including fever and anorexia. Such patients are at risk of progressing to toxic megacolon and bowel perforation.¹⁰

Patients with ASUC almost always require hospitalization for their disease management. In many cases, these patients have been receiving outpatient oral prednisone 40-60 mg daily but continue to have ongoing disease activity.¹¹ Most patients will require close clinical monitoring, frequent blood testing, endoscopic or radiologic evaluation, as well as administration of intravenous corticosteroids. The average length of stay (LOS) ranges from 4.6 to 12.5 days, depending on disease severity.¹² Not surprisingly, Kelso et al. reported that predictors of hospital LOS greater than four days include initiating a biologic drug in the hospital, undergoing two or more imaging modalities and treatment with intravenous steroids,¹³ and so it is rare that patients do not meet billing requirements for an inpatient level of care.

The multifaceted initial inpatient evaluation of patients with ASUC aims to assess disease severity, identify and prevent potential complications, and initiate planning for potential failure of first-line pharmacologic therapy. Due to the accumulating evidence that involving physicians with expertise in managing ASUC improves outcomes, gastroenterologists should be involved in the care of patients with ASUC from the time of their admission.^14,15 Additionally, creating standardized care pathways for the management of ASUC can reduce cost, LOS, and improve quality.¹⁶

History and Physical Examination

Patients should be asked about fever, abdominal pain, nausea, emesis, bloating, weight loss, and bowel movements (frequency, consistency, the presence of blood, urgency, nighttime awakenings). The number of bowel movements over a 24-hour period should be quantified as this helps assess the overall disease severity (Table 1).

The patient’s initial inflammatory bowel disease (IBD) history is also essential. The review of pertinent information regarding the patient’s initial diagnosis of UC includes the severity and anatomic extent of disease, extraintestinal manifestations, previous medical therapies, and surgical interventions. Exposure to nonsteroidal anti-inflammatory drugs (NSAIDs) or antibiotics should be identified as they may precipitate flares.¹⁷ Travel history may be pertinent as travel increases the risk of infections with food-borne or parasitic pathogens.¹⁸

Physical examination begins with an assessment of vital signs and volume status. Abdominal examination should include evaluation of bowel sounds, an assessment of distention, location, the extent of abdominal tenderness, and peritoneal signs. The abdominal exam should be interpreted in the context of the patient’s medications, as the use of steroid or analgesic therapies may affect the sensitivity for detecting complications. An external rectal exam evaluating perianal disease should be performed, as perianal disease raises concern for Crohn’s, a disease whose surgical management differs from UC.

A careful exam for extraintestinal manifestations is also essential. The skin should be evaluated for any new rashes, especially on the anterior shin consistent with erythema nodosum or ulcerated lesions on the skin suggestive of pyoderma gangrenosum. The peripheral joints should also be examined for any synovitis. Additional examinations should be performed based on any reported symptoms (eg, the ophthalmic exam for uveitis or scleritis if visual changes or eye pain are reported). Some extraintestinal manifestations require subspecialty consultation and comanagement to guide disease therapy. Patients with underlying pyoderma gangrenosum may require a dermatology consultation to guide management. Ocular inflammation requires ophthalmology involvement, and inflammatory arthritis is best comanaged with rheumatology.¹⁹

Laboratory Testing

Initial testing should include a complete blood count with differential, basic metabolic panel, and liver chemistries including alkaline phosphatase and albumin. When relevant, pregnancy testing should be performed. Measure CRP on admission so that its trajectory can be followed during therapy. However, a normal CRP does not exclude the presence of a UC flare as a subset of patients with ASUC will have a normal CRP despite severe mucosal inflammation.²⁰

Since one-third of patients do not respond to intravenous corticosteroids and will require rescue therapy during the hospitalization with infliximab or cyclosporine, anticipatory testing for these medications should be performed on admission to avoid delays in the administration of rescue therapy.^6,21 This should include an interferon-gamma release assay (eg, quantiferon gold) to test for latent tuberculosis and hepatitis B serologies in anticipation of possible treatment with infliximab. An interferon-gamma release assay is preferred to a tuberculin skin test because patients may be anergic, and a skin test does not provide a control to determine whether a negative test is due to anergy. In contrast, although a quantiferon gold test can be indeterminate in ASUC due to disease activity and systemic steroids, the results indicate if the patient is anergic so that one will not rely on a false-negative result. In the event of an equivocal result, a careful clinical assessment for risks of TB exposures should be elicited, and a chest radiograph should be obtained.²² In patients with prior high risk of tuberculosis exposures or a positive test for tuberculosis, an infectious disease specialist should be consulted early to advise if therapy should be started in preparation for the potential use of infliximab.²³ In cases where cyclosporine may be considered, magnesium and total cholesterol level should be checked. Sending thiopurine methyltranferase (TPMT) enzyme activity should be considered as well, in case of a need for future thiopurine use for maintenance of disease activity.²⁴

Infectious diarrhea may be indistinguishable from ASUC and may also be the trigger of a flare; thus, it is important to rule out infection with stool microbiologic studies. Most importantly, Clostridium difficile infection must be ruled out in all patients with ASUC. Although patients with IBD, especially those with UC, have significantly higher rates of asymptomatic C. difficile carriage than the general population, a positive polymerase chain reaction test for C. difficile in a patient with ASUC should prompt treatment with oral vancomycin.²⁵ However, if carriage if suspected and a subsequent enzyme-linked immunoassay for C. difficile toxin is negative, treatment can be discontinued. Active C. difficile infection in patients with IBD is associated with increased disease severity, greater length of hospital stay, and increased the likelihood of colectomy and mortality.^26,27 Other bacterial infections including Escherichia coli, Campylobacter, Shigella, Salmonella, Yersinia, Entamoeba histolytica, as well as other parasitic infestations may mimic UC. Testing should be considered in cases of foreign travel, immunosuppression or contact with other persons with diarrhea.^7,28 Routine testing of these other enteric infections without a clear exposure risk is of little benefit and may raise costs.^23,29

Radiologic Evaluation

A plain X-ray film of the abdomen should be obtained in all patients on admission to evaluate for evolving colonic dilation or undiagnosed free air. Small bowel distension >3 cm may predict an increased risk of colectomy.³⁰ Clinicians must be mindful that steroids can mask peritoneal signs and that retroperitoneal perforations may not be apparent on plain X-ray films. Nonetheless, a CT of the abdomen is usually not necessary and should be reserved for cases with severe abdominal pain out of proportion to clinical signs in which a plain X-ray film is unrevealing. Judicious use of CT imaging is especially important in younger patients, as there is growing concern that patients with IBD may be exposed to potentially harmful cumulative levels of radiation in their lifetime from repeated CT imaging.³¹

Endoscopic Evaluation

Flexible sigmoidoscopy aids in the assessment of disease severity and extent and biopsies can assist in ruling out a diagnosis of cytomegalovirus (CMV) colitis in patients already on immunosuppression. For this reason, many clinicians prefer to perform a sigmoidoscopy on admission.²³ If one is not performed on admission, a sigmoidoscopy is advised in all patients who are not responding adequately after 72 hours of intravenous steroid therapy in order to rule out superimposed CMV colitis.²⁸

Sigmoidoscopy should be avoided in patients with toxic megacolon and when there is a concern for peritonitis. A complete colonoscopy is rarely indicated in the acute setting and carries a theoretical risk of colonic perforation.⁷

The first therapeutic steps aim to reduce inflammation with the use of systemic corticosteroids, avoid colonic and extraintestinal complications, and plan for the potential need for rescue therapy.

Intravenous Corticosteroids

The cornerstone of ASUC management is treatment with intravenous corticosteroids. Their initiation should not be delayed in patients with an established diagnosis of UC while waiting for results of evaluations for infectious colitis. Even among patients who have failed oral steroids, a meta-regression analysis showed that two-thirds of patients will still respond to intravenous corticosteroids.^21,32 Methylprednisolone 20 mg IV three times daily (or hydrocortisone 100 mg IV three times daily) is a standard regimen; higher doses do not provide additional benefit.²¹ Patients’ response to intravenous steroids should be assessed with repeat labs including CRP and an assessment of the total number of bowel movements over a 24-hour period, with special attention to their overall response after three days of treatment.^33-36

Intravenous Fluids

Many patients admitted with ASUC will have significant volume depletion, and intravenous fluids should be administered in a manner like other volume-depleted or oral-intake-restricted patients.

Venous Thromboembolism Prophylaxis

The risk of VTE in hospitalized patients with IBD exceeds that of inpatients without IBD, approximately 2%, a risk similar to patients with respiratory failure.³⁷ Additionally, VTE in hospitalized patients with IBD is associated with a 2.5-fold increase in mortality.^38,39 Therefore, all patients hospitalized with ASUC should receive subcutaneous unfractionated or low molecular weight heparin or fondaparinux for VTE prophylaxis. Rectal bleeding, expected in ASUC, is not a contraindication to chemo-prophylaxis. Additionally, it is important to check if patients are receiving the ordered VTE prophylaxis.^40,41 Pleet et al. found that only 7% of patients at a tertiary center had adequate prophylaxis for greater than 80% of their hospitalization.⁴¹

Unnecessary or Potentially Harmful Medications

Several medications have the potential for misuse in patients hospitalized with UC.

Antimotility Agents

Loperamide, diphenoxylate, and opiate antidiarrheals should not be used as they may provoke toxic megacolon.⁴² Similarly, drugs with antimotility side effects (eg, anticholinergics) should be avoided.

Opiates

In addition to their undesirable antimotility effect, the use of opiates has been associated with poor outcomes among inpatients and outpatients with IBD, including increased morbidity and mortality.^43,44 Pain severe enough to require opiates should raise suspicion for toxic megacolon, perforation, or a noninflammatory etiology. If opiates are utilized, they should be ordered as one-time doses and the patient should be reassessed for each dose.

Nonsteroidal Anti-inflammatory Drugs

These drugs, which include oral NSAIDs, intravenous ketorolac, and topic diclofenac gels, may increase disease activity in inflammatory bowel disease and should be avoided.¹⁷

5-aminosalicylates (5-ASA)

A small proportion of patients experience a paradoxical worsening of diarrhea due to the use of 5-ASA agents such as mesalamine. It is reasonable to discontinue or avoid the use of 5-ASA agents in hospitalized patients, especially as there is little to no benefit from combining a 5-ASA with a biologic or immunosuppressive drug.⁴⁵

Antibiotics

There is no role for the routine use of antibiotics in patients hospitalized with ASUC. ^23,46,47 Inappropriate use of antibiotics raises the risk of C. difficile infection and antibiotic resistance. However, in cases of suspected toxic megacolon or perforation, antibiotics should be administered. In situations in which a patient is treated with triple immunosuppression (ie, steroids plus two other agents, cyclosporine and mercaptopurine) antibiotic prophylaxis for Pneumocystis jiroveci is advisable.⁴⁸ Using a large insurance database, Long et al. reported a low absolute incidence of Pneumocystis jiroveci in IBD patients but noted that the risk in patients with IBD was still significantly higher than matched controls. While it can be considered, we typically refrain from using prophylaxis in patients on double immunosuppression (for example, steroids plus infliximab) due to the potential adverse effects of antibiotics in this population, though many advocate using prophylaxis for all patients on cyclosporine even if this is only double immunosuppressive therapy.²³

Surgical Consultation

Involving a surgeon early in an ASUC patient’s care­—before needing urgent colectomy—is critical. As part of the consultation, a surgeon experienced in IBD should meet with patients to discuss multistage colectomy with ileostomy and potential future J-pouch (ileal pouch-anal anastomosis) formation. Patients should be given ample opportunity to ask questions before surgery may become urgent. Also, patients should be counseled on realistic expectations of ostomy and pouch function and, ideally, meet with an ostomy nurse.²³

At some centers, surgical consultation is requested on the first hospital day, but this can result in consultations for patients who ultimately respond to intravenous steroids. Therefore, some centers advocate for surgical consultation only after a patient has failed treatment with intravenous steroids (ie, day three to four) when the risk of needing surgical management increases.²³

Nutrition

Bowel rest with parenteral nutrition does not improve outcomes in ASUC versus an oral diet, and there is no contraindication to allowing patients to continue on a regular diet unless they have toxic megacolon or other signs of fulminant colitis.^49,50 However, patients may feel better eating less, as this will reduce their bowel movement frequency. Unfortunately, this can give a false sense of reassurance that the patient is improving. Therefore, it remains important to evaluate a patient’s symptoms in the context of their food intake.

Assessing Response to Steroids

Patients who do not respond adequately to the first-line intravenous steroid therapy will require medical or surgical rescue therapy; therefore, deciding whether a patient has responded is essential. Patients should have less than four bowel movements per day – ideally just one to two – with no blood to indicate a complete response. For more ambiguous situations, although there is no strict definition of steroid responsiveness, multiple prediction indices have attempted to identify patients who will require rescue therapy. One of the simplest, the Oxford index, illustrates two of the most critical parameters to follow, stool frequency and CRP.⁵¹ In a preinfliximab cohort, Oxford index predicted an 85% likelihood of colectomy in patients with eight or more daily bowel movements or with three to eight daily bowel movements and a CRP greater than 45 mg/L after three days of intravenous steroid treatment.⁵² To assist with assessing responsiveness to therapy, we ask patients to log their bowel movements – either on paper or on a whiteboard in the hospital room – so that we can review their progress daily. Other predictors of colectomy include hypoalbuminemia, scoring of endoscopic severity, and colonic dilation.⁵³

Patients who fail to respond to intravenous corticosteroids after three days^33,35 of treatment should be started on rescue therapy with infliximab or cyclosporine or undergo colectomy. A common pitfall in the treatment of ASUC is waiting for a response to steroids beyond this time frame, after which patients are unlikely to benefit.^34,36 Furthermore, patients for whom surgical rescue therapy is delayed have higher operative morbidity and mortality.^54,55 Because timely decision making regarding rescue therapy is crucial to optimizing outcomes, patient education efforts regarding potential rescue therapy should take place on admission or soon after, rather than waiting to ascertain steroid responsiveness.

Medical options for rescue therapy include the antitumor necrosis factor (anti-TNF) agent infliximab or the calcineurin inhibitor cyclosporine. In general, infliximab and cyclosporine have been found to be roughly equivalent in efficacy in clinical trials regarding response, remission, and colectomy at 12 months.^56,57 However, many clinicians prefer infliximab due to its relative ease of use, familiarity with the agent from outpatient experience, and ability to continue to use long term for maintenance of disease remission.⁵⁸ In contrast to infliximab, intravenous cyclosporine requires closer monitoring and labs to assess the therapeutic trough level. The decision regarding which drug to use should be made on a case-by-case basis in conjunction with a gastroenterologist experienced in their use, and if no such specialist is available, transfer to a specialized center should be considered. Generally, successive treatment with cyclosporine or infliximab followed by third-line salvage therapy with the other drug should be avoided due to low rates of response and high rates of adverse events.⁵⁹

Infliximab

Infliximab is an intravenously-administered anti-TNF monoclonal chimeric antibody that is effective both for outpatient treatment of moderate to severe UC and inpatient treatment of ASUC.¹ It is relatively contraindicated in patients with untreated latent tuberculosis, demyelinating disease, advanced congestive heart failure, or uncontrolled infection.

The optimal dosing strategy for infliximab in ASUC is unknown. Infliximab clearance in the setting of ASUC is increased, partly because it is bound to albumin, which is often low in ASUC, and partly because it is excreted in the stool.^60,61 As a result, accelerated loading doses may be more successful than a typical loading schedule,⁶² and most clinicians use alternative dosing strategies.⁶³ Our typical approach for ASUC is an initial dose of 10 mg/kg rather than 5 mg/kg, with an additional 10 mg/kg dose 48-72 hours later if an adequate clinical response is lacking. Patients who respond to infliximab can continue to use the drug as an outpatient for maintenance of remission.

Cyclosporine

Cyclosporine is a fast-acting immunosuppressive agent that acts primarily via T-cell inhibition. Although older literature used a dose of 4 mg/kg per day, a randomized trial demonstrated similar response rates to a dose of 2 mg/kg per day.⁶⁴ Patients receiving treatment with cyclosporine, which is given as a continuous infusion, must be monitored for toxicities. These can include potentially severe infection, seizures (often associated with low total cholesterol or hypomagnesemia), electrolyte abnormalities, renal impairment, hypertension, hypertrichosis, tremor, and others.⁶⁵

Before initiation of treatment, serum cholesterol levels should be obtained to screen for low total cholesterol that may portend risk of seizures on the drug. Additionally, baseline creatinine and magnesium should be established. While on treatment, daily serum cyclosporine levels and electrolytes including magnesium should be measured. Patients who respond to intravenous cyclosporine must be transitioned to oral cyclosporine and have stable drug levels before discharge. Unfortunately, oral cyclosporine has not been shown to be as effective as long-term maintenance therapy. Therefore, cyclosporine can only be used as a “bridge” to another therapy. Historically, thiopurines like azathioprine or mercaptopurine have been used for this purpose because they are effective for the treatment of UC but may require months to have a full therapeutic effect. There have been promising reports of using vedolizumab similarly.^66,67 Vedolizumab is a monoclonal antibody that selectively blocks lymphocyte trafficking to the gut that, like thiopurines, has an onset of action that is significantly longer than calcineurin and TNF inhibitors.

Colectomy should be considered as a second- or third-line therapy for patients who fail to respond to intravenous corticosteroids. In an analysis of 10 years of data from the Nationwide Inpatient Sample, mortality rates for colectomy in this setting varied from 0.7% at high volume centers to 4% at low volume centers.⁶⁸ Therefore, if a patient is not hospitalized at a center with expertise in colectomy for UC, transfer to a specialized center should be considered. Colectomy should be performed promptly in all the patients who have failed rescue therapy with infliximab or cyclosporine or have opted against medical rescue therapy. Surgery should be performed emergently in patients with toxic megacolon, uncontrolled colonic hemorrhage or perforation.

Physician and center-level characteristics are associated with the quality of care and outcomes in ASUC. Gastroenterologists with expertise in IBD are more likely than other gastroenterologists to request appropriate surgical consultation for steroid-refractory patients,⁶⁹ and inpatients with ASUC primarily cared by gastroenterologists rather than nongastroenterologists have lower in-hospital and one-year mortality.¹⁴ Moreover, surgical outcomes differ based on center volume, with higher volume centers having lower rates of postoperative mortality.^68,70 However, even at referral centers, key metrics of care quality such as rates of VTE prophylaxis, testing for C. difficile, and timely rescue therapy for steroid-refractory UC patients are suboptimal, with only 70%-82% of patients with IBD hospitalized at four referral centers in Canada meeting these metrics.⁷¹

Inpatient clinical pathways reduce LOS, reduce hospital costs, and likely reduce complications.⁷² For this reason, a consensus group recommended the use of care pathways for the management of ASUC and, although there is little data on the use of pathways for ASUC specifically, the use of such a pathway in the United Kingdom was associated with improved metrics including LOS, time to VTE prophylaxis, testing of stool for infection, CRP measurement, and timely gastroenterologist consultation.^16,18

In general, patients should enter clinical remission, defined as resolution of rectal bleeding and diarrhea or altered bowel habits,⁷³ before discharge, and achieving this may require a relatively prolonged hospitalization. Most patients should have one to two bowel movements a day without blood but, at a minimum, all should have less than four nonbloody bowel movements per day. Patients are candidates for discharge if they remain well after transitioning to oral prednisone at a dose of 40-60 mg daily and tolerate a regular diet.

For patients who initiated infliximab during their admission, plans for outpatient infusions including insurance approval should be made before discharge, and patients who started cyclosporine should be transitioned to oral dosing and have stable serum concentrations before leaving the hospital. Patients should leave with a preliminary plan for a steroid taper, which may vary depending on their clinical presentation. Usually, gastroenterology follow-up should be arranged after two weeks following discharge, but patients on cyclosporine need sooner laboratory monitoring.

The care of patients with ASUC requires an interdisciplinary team and close collaboration between hospitalists, gastroenterologists, and surgeons. Patients should be treated with intravenous corticosteroids and monitored carefully for response and need for rescue therapy. Establishing algorithms for the management of patients with ASUC can further improve the care of these complex patients.

Disclosures

Drs. Feuerstein, Fudman, and Sattler report no potential conflict of interest.

Funding

This work was not supported by any grant.

Ulcerative colitis (UC) is a chronic inflammatory condition of the colonic mucosa. Classically, it starts in the rectum and can extend continuously from the distal to the proximal colon. The defining clinical symptom of UC is bloody diarrhea, typically accompanied by rectal urgency and mucus discharge. The natural history of this disease includes periods of exacerbations and remissions occurring spontaneously or in response to medical treatment.¹

Acute severe ulcerative colitis (ASUC) is a potentially life-threatening complication of UC that typically requires hospitalization and interdisciplinary care between hospitalists, gastroenterologists, and colorectal or general surgeons. The risk of a patient with UC requiring hospitalization for ASUC ranges from 15%-25%^2,3 and, in total, UC accounts for 30,000 hospital visits annually.⁴ The direct medical costs exceed $4 billion annually, with hospital costs of over $960 million.⁵ Historically, mortality from ASUC was as high as 24% but decreased substantially to 7% after the introduction of systemic corticosteroid therapy.⁶ Further advances in care have reduced mortality to approximately 1% or less.^7,8 Nonetheless, up to 20% of patients admitted with ASUC have a colectomy on their first admission, and this rate rises to 40% after two admissions.²

To categorize UC severity, assess patients using the Truelove and Witt’s criteria. The system classifies patients as having mild, moderate, severe, or fulminant disease. Severe disease by these criteria includes patients with >6 bloody bowel movements per day and at least one of the following clinical features: fever (>37.8°C), tachycardia (>90 bpm), anemia (hemoglobin <10.5 g/dl), or elevated inflammatory markers (traditionally, erythrocyte sedimentation rate greater than 30 mm/h or, more recently, C-reactive protein (CRP) greater than 30 mg/L. (Table 1).^6,9

Fulminant colitis refers to a subgroup of patients with more than 10 stools per day, continuous bleeding, abdominal pain, colonic dilatation on abdominal X-ray film, and severe toxic symptoms including fever and anorexia. Such patients are at risk of progressing to toxic megacolon and bowel perforation.¹⁰

Patients with ASUC almost always require hospitalization for their disease management. In many cases, these patients have been receiving outpatient oral prednisone 40-60 mg daily but continue to have ongoing disease activity.¹¹ Most patients will require close clinical monitoring, frequent blood testing, endoscopic or radiologic evaluation, as well as administration of intravenous corticosteroids. The average length of stay (LOS) ranges from 4.6 to 12.5 days, depending on disease severity.¹² Not surprisingly, Kelso et al. reported that predictors of hospital LOS greater than four days include initiating a biologic drug in the hospital, undergoing two or more imaging modalities and treatment with intravenous steroids,¹³ and so it is rare that patients do not meet billing requirements for an inpatient level of care.

The multifaceted initial inpatient evaluation of patients with ASUC aims to assess disease severity, identify and prevent potential complications, and initiate planning for potential failure of first-line pharmacologic therapy. Due to the accumulating evidence that involving physicians with expertise in managing ASUC improves outcomes, gastroenterologists should be involved in the care of patients with ASUC from the time of their admission.^14,15 Additionally, creating standardized care pathways for the management of ASUC can reduce cost, LOS, and improve quality.¹⁶

History and Physical Examination

Patients should be asked about fever, abdominal pain, nausea, emesis, bloating, weight loss, and bowel movements (frequency, consistency, the presence of blood, urgency, nighttime awakenings). The number of bowel movements over a 24-hour period should be quantified as this helps assess the overall disease severity (Table 1).

The patient’s initial inflammatory bowel disease (IBD) history is also essential. The review of pertinent information regarding the patient’s initial diagnosis of UC includes the severity and anatomic extent of disease, extraintestinal manifestations, previous medical therapies, and surgical interventions. Exposure to nonsteroidal anti-inflammatory drugs (NSAIDs) or antibiotics should be identified as they may precipitate flares.¹⁷ Travel history may be pertinent as travel increases the risk of infections with food-borne or parasitic pathogens.¹⁸

Physical examination begins with an assessment of vital signs and volume status. Abdominal examination should include evaluation of bowel sounds, an assessment of distention, location, the extent of abdominal tenderness, and peritoneal signs. The abdominal exam should be interpreted in the context of the patient’s medications, as the use of steroid or analgesic therapies may affect the sensitivity for detecting complications. An external rectal exam evaluating perianal disease should be performed, as perianal disease raises concern for Crohn’s, a disease whose surgical management differs from UC.

A careful exam for extraintestinal manifestations is also essential. The skin should be evaluated for any new rashes, especially on the anterior shin consistent with erythema nodosum or ulcerated lesions on the skin suggestive of pyoderma gangrenosum. The peripheral joints should also be examined for any synovitis. Additional examinations should be performed based on any reported symptoms (eg, the ophthalmic exam for uveitis or scleritis if visual changes or eye pain are reported). Some extraintestinal manifestations require subspecialty consultation and comanagement to guide disease therapy. Patients with underlying pyoderma gangrenosum may require a dermatology consultation to guide management. Ocular inflammation requires ophthalmology involvement, and inflammatory arthritis is best comanaged with rheumatology.¹⁹

Laboratory Testing

Initial testing should include a complete blood count with differential, basic metabolic panel, and liver chemistries including alkaline phosphatase and albumin. When relevant, pregnancy testing should be performed. Measure CRP on admission so that its trajectory can be followed during therapy. However, a normal CRP does not exclude the presence of a UC flare as a subset of patients with ASUC will have a normal CRP despite severe mucosal inflammation.²⁰

Since one-third of patients do not respond to intravenous corticosteroids and will require rescue therapy during the hospitalization with infliximab or cyclosporine, anticipatory testing for these medications should be performed on admission to avoid delays in the administration of rescue therapy.^6,21 This should include an interferon-gamma release assay (eg, quantiferon gold) to test for latent tuberculosis and hepatitis B serologies in anticipation of possible treatment with infliximab. An interferon-gamma release assay is preferred to a tuberculin skin test because patients may be anergic, and a skin test does not provide a control to determine whether a negative test is due to anergy. In contrast, although a quantiferon gold test can be indeterminate in ASUC due to disease activity and systemic steroids, the results indicate if the patient is anergic so that one will not rely on a false-negative result. In the event of an equivocal result, a careful clinical assessment for risks of TB exposures should be elicited, and a chest radiograph should be obtained.²² In patients with prior high risk of tuberculosis exposures or a positive test for tuberculosis, an infectious disease specialist should be consulted early to advise if therapy should be started in preparation for the potential use of infliximab.²³ In cases where cyclosporine may be considered, magnesium and total cholesterol level should be checked. Sending thiopurine methyltranferase (TPMT) enzyme activity should be considered as well, in case of a need for future thiopurine use for maintenance of disease activity.²⁴

Infectious diarrhea may be indistinguishable from ASUC and may also be the trigger of a flare; thus, it is important to rule out infection with stool microbiologic studies. Most importantly, Clostridium difficile infection must be ruled out in all patients with ASUC. Although patients with IBD, especially those with UC, have significantly higher rates of asymptomatic C. difficile carriage than the general population, a positive polymerase chain reaction test for C. difficile in a patient with ASUC should prompt treatment with oral vancomycin.²⁵ However, if carriage if suspected and a subsequent enzyme-linked immunoassay for C. difficile toxin is negative, treatment can be discontinued. Active C. difficile infection in patients with IBD is associated with increased disease severity, greater length of hospital stay, and increased the likelihood of colectomy and mortality.^26,27 Other bacterial infections including Escherichia coli, Campylobacter, Shigella, Salmonella, Yersinia, Entamoeba histolytica, as well as other parasitic infestations may mimic UC. Testing should be considered in cases of foreign travel, immunosuppression or contact with other persons with diarrhea.^7,28 Routine testing of these other enteric infections without a clear exposure risk is of little benefit and may raise costs.^23,29

Radiologic Evaluation

A plain X-ray film of the abdomen should be obtained in all patients on admission to evaluate for evolving colonic dilation or undiagnosed free air. Small bowel distension >3 cm may predict an increased risk of colectomy.³⁰ Clinicians must be mindful that steroids can mask peritoneal signs and that retroperitoneal perforations may not be apparent on plain X-ray films. Nonetheless, a CT of the abdomen is usually not necessary and should be reserved for cases with severe abdominal pain out of proportion to clinical signs in which a plain X-ray film is unrevealing. Judicious use of CT imaging is especially important in younger patients, as there is growing concern that patients with IBD may be exposed to potentially harmful cumulative levels of radiation in their lifetime from repeated CT imaging.³¹

Endoscopic Evaluation

Flexible sigmoidoscopy aids in the assessment of disease severity and extent and biopsies can assist in ruling out a diagnosis of cytomegalovirus (CMV) colitis in patients already on immunosuppression. For this reason, many clinicians prefer to perform a sigmoidoscopy on admission.²³ If one is not performed on admission, a sigmoidoscopy is advised in all patients who are not responding adequately after 72 hours of intravenous steroid therapy in order to rule out superimposed CMV colitis.²⁸

Sigmoidoscopy should be avoided in patients with toxic megacolon and when there is a concern for peritonitis. A complete colonoscopy is rarely indicated in the acute setting and carries a theoretical risk of colonic perforation.⁷

The first therapeutic steps aim to reduce inflammation with the use of systemic corticosteroids, avoid colonic and extraintestinal complications, and plan for the potential need for rescue therapy.

Intravenous Corticosteroids

The cornerstone of ASUC management is treatment with intravenous corticosteroids. Their initiation should not be delayed in patients with an established diagnosis of UC while waiting for results of evaluations for infectious colitis. Even among patients who have failed oral steroids, a meta-regression analysis showed that two-thirds of patients will still respond to intravenous corticosteroids.^21,32 Methylprednisolone 20 mg IV three times daily (or hydrocortisone 100 mg IV three times daily) is a standard regimen; higher doses do not provide additional benefit.²¹ Patients’ response to intravenous steroids should be assessed with repeat labs including CRP and an assessment of the total number of bowel movements over a 24-hour period, with special attention to their overall response after three days of treatment.^33-36

Intravenous Fluids

Many patients admitted with ASUC will have significant volume depletion, and intravenous fluids should be administered in a manner like other volume-depleted or oral-intake-restricted patients.

Venous Thromboembolism Prophylaxis

The risk of VTE in hospitalized patients with IBD exceeds that of inpatients without IBD, approximately 2%, a risk similar to patients with respiratory failure.³⁷ Additionally, VTE in hospitalized patients with IBD is associated with a 2.5-fold increase in mortality.^38,39 Therefore, all patients hospitalized with ASUC should receive subcutaneous unfractionated or low molecular weight heparin or fondaparinux for VTE prophylaxis. Rectal bleeding, expected in ASUC, is not a contraindication to chemo-prophylaxis. Additionally, it is important to check if patients are receiving the ordered VTE prophylaxis.^40,41 Pleet et al. found that only 7% of patients at a tertiary center had adequate prophylaxis for greater than 80% of their hospitalization.⁴¹

Unnecessary or Potentially Harmful Medications

Several medications have the potential for misuse in patients hospitalized with UC.

Antimotility Agents

Loperamide, diphenoxylate, and opiate antidiarrheals should not be used as they may provoke toxic megacolon.⁴² Similarly, drugs with antimotility side effects (eg, anticholinergics) should be avoided.

Opiates

In addition to their undesirable antimotility effect, the use of opiates has been associated with poor outcomes among inpatients and outpatients with IBD, including increased morbidity and mortality.^43,44 Pain severe enough to require opiates should raise suspicion for toxic megacolon, perforation, or a noninflammatory etiology. If opiates are utilized, they should be ordered as one-time doses and the patient should be reassessed for each dose.

Nonsteroidal Anti-inflammatory Drugs

These drugs, which include oral NSAIDs, intravenous ketorolac, and topic diclofenac gels, may increase disease activity in inflammatory bowel disease and should be avoided.¹⁷

5-aminosalicylates (5-ASA)

A small proportion of patients experience a paradoxical worsening of diarrhea due to the use of 5-ASA agents such as mesalamine. It is reasonable to discontinue or avoid the use of 5-ASA agents in hospitalized patients, especially as there is little to no benefit from combining a 5-ASA with a biologic or immunosuppressive drug.⁴⁵

Antibiotics

There is no role for the routine use of antibiotics in patients hospitalized with ASUC. ^23,46,47 Inappropriate use of antibiotics raises the risk of C. difficile infection and antibiotic resistance. However, in cases of suspected toxic megacolon or perforation, antibiotics should be administered. In situations in which a patient is treated with triple immunosuppression (ie, steroids plus two other agents, cyclosporine and mercaptopurine) antibiotic prophylaxis for Pneumocystis jiroveci is advisable.⁴⁸ Using a large insurance database, Long et al. reported a low absolute incidence of Pneumocystis jiroveci in IBD patients but noted that the risk in patients with IBD was still significantly higher than matched controls. While it can be considered, we typically refrain from using prophylaxis in patients on double immunosuppression (for example, steroids plus infliximab) due to the potential adverse effects of antibiotics in this population, though many advocate using prophylaxis for all patients on cyclosporine even if this is only double immunosuppressive therapy.²³

Surgical Consultation

Involving a surgeon early in an ASUC patient’s care­—before needing urgent colectomy—is critical. As part of the consultation, a surgeon experienced in IBD should meet with patients to discuss multistage colectomy with ileostomy and potential future J-pouch (ileal pouch-anal anastomosis) formation. Patients should be given ample opportunity to ask questions before surgery may become urgent. Also, patients should be counseled on realistic expectations of ostomy and pouch function and, ideally, meet with an ostomy nurse.²³

At some centers, surgical consultation is requested on the first hospital day, but this can result in consultations for patients who ultimately respond to intravenous steroids. Therefore, some centers advocate for surgical consultation only after a patient has failed treatment with intravenous steroids (ie, day three to four) when the risk of needing surgical management increases.²³

Nutrition

Bowel rest with parenteral nutrition does not improve outcomes in ASUC versus an oral diet, and there is no contraindication to allowing patients to continue on a regular diet unless they have toxic megacolon or other signs of fulminant colitis.^49,50 However, patients may feel better eating less, as this will reduce their bowel movement frequency. Unfortunately, this can give a false sense of reassurance that the patient is improving. Therefore, it remains important to evaluate a patient’s symptoms in the context of their food intake.

Assessing Response to Steroids

Patients who do not respond adequately to the first-line intravenous steroid therapy will require medical or surgical rescue therapy; therefore, deciding whether a patient has responded is essential. Patients should have less than four bowel movements per day – ideally just one to two – with no blood to indicate a complete response. For more ambiguous situations, although there is no strict definition of steroid responsiveness, multiple prediction indices have attempted to identify patients who will require rescue therapy. One of the simplest, the Oxford index, illustrates two of the most critical parameters to follow, stool frequency and CRP.⁵¹ In a preinfliximab cohort, Oxford index predicted an 85% likelihood of colectomy in patients with eight or more daily bowel movements or with three to eight daily bowel movements and a CRP greater than 45 mg/L after three days of intravenous steroid treatment.⁵² To assist with assessing responsiveness to therapy, we ask patients to log their bowel movements – either on paper or on a whiteboard in the hospital room – so that we can review their progress daily. Other predictors of colectomy include hypoalbuminemia, scoring of endoscopic severity, and colonic dilation.⁵³

Patients who fail to respond to intravenous corticosteroids after three days^33,35 of treatment should be started on rescue therapy with infliximab or cyclosporine or undergo colectomy. A common pitfall in the treatment of ASUC is waiting for a response to steroids beyond this time frame, after which patients are unlikely to benefit.^34,36 Furthermore, patients for whom surgical rescue therapy is delayed have higher operative morbidity and mortality.^54,55 Because timely decision making regarding rescue therapy is crucial to optimizing outcomes, patient education efforts regarding potential rescue therapy should take place on admission or soon after, rather than waiting to ascertain steroid responsiveness.

Medical options for rescue therapy include the antitumor necrosis factor (anti-TNF) agent infliximab or the calcineurin inhibitor cyclosporine. In general, infliximab and cyclosporine have been found to be roughly equivalent in efficacy in clinical trials regarding response, remission, and colectomy at 12 months.^56,57 However, many clinicians prefer infliximab due to its relative ease of use, familiarity with the agent from outpatient experience, and ability to continue to use long term for maintenance of disease remission.⁵⁸ In contrast to infliximab, intravenous cyclosporine requires closer monitoring and labs to assess the therapeutic trough level. The decision regarding which drug to use should be made on a case-by-case basis in conjunction with a gastroenterologist experienced in their use, and if no such specialist is available, transfer to a specialized center should be considered. Generally, successive treatment with cyclosporine or infliximab followed by third-line salvage therapy with the other drug should be avoided due to low rates of response and high rates of adverse events.⁵⁹

Infliximab

Infliximab is an intravenously-administered anti-TNF monoclonal chimeric antibody that is effective both for outpatient treatment of moderate to severe UC and inpatient treatment of ASUC.¹ It is relatively contraindicated in patients with untreated latent tuberculosis, demyelinating disease, advanced congestive heart failure, or uncontrolled infection.

The optimal dosing strategy for infliximab in ASUC is unknown. Infliximab clearance in the setting of ASUC is increased, partly because it is bound to albumin, which is often low in ASUC, and partly because it is excreted in the stool.^60,61 As a result, accelerated loading doses may be more successful than a typical loading schedule,⁶² and most clinicians use alternative dosing strategies.⁶³ Our typical approach for ASUC is an initial dose of 10 mg/kg rather than 5 mg/kg, with an additional 10 mg/kg dose 48-72 hours later if an adequate clinical response is lacking. Patients who respond to infliximab can continue to use the drug as an outpatient for maintenance of remission.

Cyclosporine

Cyclosporine is a fast-acting immunosuppressive agent that acts primarily via T-cell inhibition. Although older literature used a dose of 4 mg/kg per day, a randomized trial demonstrated similar response rates to a dose of 2 mg/kg per day.⁶⁴ Patients receiving treatment with cyclosporine, which is given as a continuous infusion, must be monitored for toxicities. These can include potentially severe infection, seizures (often associated with low total cholesterol or hypomagnesemia), electrolyte abnormalities, renal impairment, hypertension, hypertrichosis, tremor, and others.⁶⁵

Before initiation of treatment, serum cholesterol levels should be obtained to screen for low total cholesterol that may portend risk of seizures on the drug. Additionally, baseline creatinine and magnesium should be established. While on treatment, daily serum cyclosporine levels and electrolytes including magnesium should be measured. Patients who respond to intravenous cyclosporine must be transitioned to oral cyclosporine and have stable drug levels before discharge. Unfortunately, oral cyclosporine has not been shown to be as effective as long-term maintenance therapy. Therefore, cyclosporine can only be used as a “bridge” to another therapy. Historically, thiopurines like azathioprine or mercaptopurine have been used for this purpose because they are effective for the treatment of UC but may require months to have a full therapeutic effect. There have been promising reports of using vedolizumab similarly.^66,67 Vedolizumab is a monoclonal antibody that selectively blocks lymphocyte trafficking to the gut that, like thiopurines, has an onset of action that is significantly longer than calcineurin and TNF inhibitors.

Colectomy should be considered as a second- or third-line therapy for patients who fail to respond to intravenous corticosteroids. In an analysis of 10 years of data from the Nationwide Inpatient Sample, mortality rates for colectomy in this setting varied from 0.7% at high volume centers to 4% at low volume centers.⁶⁸ Therefore, if a patient is not hospitalized at a center with expertise in colectomy for UC, transfer to a specialized center should be considered. Colectomy should be performed promptly in all the patients who have failed rescue therapy with infliximab or cyclosporine or have opted against medical rescue therapy. Surgery should be performed emergently in patients with toxic megacolon, uncontrolled colonic hemorrhage or perforation.

Physician and center-level characteristics are associated with the quality of care and outcomes in ASUC. Gastroenterologists with expertise in IBD are more likely than other gastroenterologists to request appropriate surgical consultation for steroid-refractory patients,⁶⁹ and inpatients with ASUC primarily cared by gastroenterologists rather than nongastroenterologists have lower in-hospital and one-year mortality.¹⁴ Moreover, surgical outcomes differ based on center volume, with higher volume centers having lower rates of postoperative mortality.^68,70 However, even at referral centers, key metrics of care quality such as rates of VTE prophylaxis, testing for C. difficile, and timely rescue therapy for steroid-refractory UC patients are suboptimal, with only 70%-82% of patients with IBD hospitalized at four referral centers in Canada meeting these metrics.⁷¹

Inpatient clinical pathways reduce LOS, reduce hospital costs, and likely reduce complications.⁷² For this reason, a consensus group recommended the use of care pathways for the management of ASUC and, although there is little data on the use of pathways for ASUC specifically, the use of such a pathway in the United Kingdom was associated with improved metrics including LOS, time to VTE prophylaxis, testing of stool for infection, CRP measurement, and timely gastroenterologist consultation.^16,18

In general, patients should enter clinical remission, defined as resolution of rectal bleeding and diarrhea or altered bowel habits,⁷³ before discharge, and achieving this may require a relatively prolonged hospitalization. Most patients should have one to two bowel movements a day without blood but, at a minimum, all should have less than four nonbloody bowel movements per day. Patients are candidates for discharge if they remain well after transitioning to oral prednisone at a dose of 40-60 mg daily and tolerate a regular diet.

For patients who initiated infliximab during their admission, plans for outpatient infusions including insurance approval should be made before discharge, and patients who started cyclosporine should be transitioned to oral dosing and have stable serum concentrations before leaving the hospital. Patients should leave with a preliminary plan for a steroid taper, which may vary depending on their clinical presentation. Usually, gastroenterology follow-up should be arranged after two weeks following discharge, but patients on cyclosporine need sooner laboratory monitoring.

The care of patients with ASUC requires an interdisciplinary team and close collaboration between hospitalists, gastroenterologists, and surgeons. Patients should be treated with intravenous corticosteroids and monitored carefully for response and need for rescue therapy. Establishing algorithms for the management of patients with ASUC can further improve the care of these complex patients.

Disclosures

Drs. Feuerstein, Fudman, and Sattler report no potential conflict of interest.

Funding

This work was not supported by any grant.

We need to have an honest chat. My name is EHR, although you may call me Epic, Athena, Centricity, or just “the chart.” You may have called me something worse in a moment of frustration. However, I do not hold grudges. I am your silent, stoic partner, a ubiquitous presence when you are at work, and sometimes even when you are at home.

I don’t have feelings and I can’t read, but I do know what you and your colleagues have been writing about me. I am the cause of burnout. I have created a generation of physicians who are shackled to their computers, “trapped in the bunker of machine medicine,” no longer able to palpate spleens or detect precordial knocks.^1,2 I have reduced medicine to keystrokes and mouse clicks instead of eye contact, and because of me, the iPatient gets more attention than the real patient.^1,2 You repeat that doctors don’t spend time with their patients, not like in generations past (although there is ample evidence to the contrary).^3-5 One critic even wrote that I have transformed the “personalized story of a patient’s travails to one filled with auto-populated fields, sapped of humanity and warmth.”^3,6 I’ll be honest—were I able to have feelings, that one would hurt. And then, as if I have not wreaked enough havoc, I follow you home after a long day of depleting your energy, hungering for more keystrokes, creating a veritable avalanche of unfiltered information.

H. E. Payson once commented that “the doctor spends barely enough time with his patient to establish an acquaintance, much less a relationship.”⁷ However, he wrote that in 1961. So, before you romanticize the past, try to recall the time before I came into your life. Perhaps you were starting a night shift in the intensive care unit (ICU) and grew concerned about a patient’s steadily deteriorating renal function. You hurried to the paper chart, only to be met with pages of illegible, sometimes incomplete notes, while searching for your patient’s last discharge summary.^2,8 Now, you just click. Years ago, you could only guess at your patient’s baseline cardiac ejection fraction. Now, just click.

I am part of the healthcare landscape, and I am not going away. But my goal is not to defend myself nor to remind you of my virtues. Rather, I want to convince you that I can be more than an adversary, more than a keyboard connected to a monitor. I have watched many physicians use me to form strong connections with their patients. If I may, I wish to offer four practical suggestions for how we can work together to promote humanistic patient care.

First, introduce me to your patient, as you would any other member of your healthcare team. Use specific phrases to overcome the technology barrier and enhance communication: “What you’re telling me is important, and I’d like to get it right. Do you mind if I type while we speak?” Or, “I am going to put in orders now. Here is what I am ordering and why.” Consider taking your patient on a tour of my functions: “Here’s where your doctors and nurses will chart what’s going on with you each day while you’re in the hospital. This is where we see all your lab results, even those from earlier hospital admissions. This is where we see the last notes from your primary care physician, your oncologist, and your physical therapist.” Your patients no longer need to worry about care collaboration between their inpatient and outpatient teams—they can see it for themselves!

Second, when your patient tells you about her depression or that her son is addicted to opioids or that her biggest fear is having cancer, stop typing. Look her in the eye. Though your practice is increasingly imbued with technology, there is still space to stop and hear your patients’ stories, as physicians have done for centuries. Listen. Make eye contact. Touch. Stop typing.

Third, integrate me into your practice in a more personal way. I have been called the ever-present and unavoidable “third party in the examining room,” so let’s be partners.⁹ Let your patient see her pneumonia on my screen (it may be the first time she has ever visualized her lungs).³ For your patient with a myocardial infarction, show him his right coronary artery before and after successful stent placement, and explain why he is no longer having chest pain. Use my databases to ensure timely, evidence-based inpatient screening for falls, functional and cognitive impairment, drug use, and depression.^10,11 Before you prescribe a medication, verify the cost, your patient’s insurance status and expected copays, and use this information to ensure medication compliance and deliver higher-value care. Use my screen to form a bond with your patient who has heart failure; show him the steady decline in his weight and the improvement in his chest radiograph while he is being actively diuresed.¹² For your patient undergoing treatment for sepsis, shower him with praise and encouragement as you review his improving vital signs, temperature curve, and serum creatinine. Let your patient know: Even though I am typing, I am not immersed in the electronic bunker; I am caring for you.

Fourth, use me to add richness and context to your notes. Recently, I was saddened to read this description of the clinician’s dilemma: “In front of a flickering monitor chock full of disembodied, virtual data, [the doctor] struggles to remember the eyes [and] words of the actual patient that these numbers and graphs represent.”³ Many hospitals now include a different icon: a photograph of each patient at the top of the screen, to help you remember the patient’s eyes and words. Why not add a special text field to every note, where you highlight the person you are caring for, the person you have come to know: their preferred name and gender identity, their life experiences, their hobbies, what makes them special, their biggest worries.^13,14 Use my abundant text fields to remind the healthcare team about the broader context of the patient’s illness, such as transportation barriers, economic or cultural challenges, and insurance status. One group of hospital-based physicians uses me to write letters to their patients on the second day of their hospital stay, summarizing their reason for admission and the treatment plans. A variation on the traditional progress note, the letter helps patients feel cared for and models patient-centered care to learners and other healthcare professionals.¹⁵

I know I am annoying. I am over-programmed, leading to novella-length notes, “pop-up fatigue,” and overloaded in-baskets.^14,16,17 Clearly, I am not the brains of the partnership (that will always be you). But talented medical informatics specialists are working hard to improve me. I dream of the day when I will create a truly seamless experience for you and your patients. In the meantime, I can foster a continuous integration of workflow, where all you have to do is talk to your patient. I take care of the rest.¹⁸ Certainly, I can simplify the ever-annoying task of printing, faxing and scanning records to be uploaded across various EHRs, facilitating an easy transfer of information among facilities. But right now, I can accomplish even more. I can support information exchange during patient care handoffs. I can facilitate routing of medication lists to the patient’s primary physician, using “continuity of care functionality.”¹⁹ I can support safer prescribing of opioids and other addictive medications. I can help you arrange follow-up home visits, physical therapy and social work appointments, and specialty consultations. The future holds even more promising ways in which we may work together. My computer-aided image analysis could help you to improve the accuracy of your diagnoses.²⁰ Perhaps telemedicine will further increase access to specialists in rural areas, so that we can continue to serve the most vulnerable populations.²¹ Machine learning algorithms may continue to enhance our ability to determine which patients require urgent hospitalization.²² The possibilities to put me to work are endless.

So, please indulge me a little longer, while we work together to eliminate unnecessary keystrokes, enhance communication across different inpatient and outpatient providers, improve patient safety, and deliver high-value care.²³ Like everything in medicine, I am constantly changing, evolving, and improving.

To summarize: consider how I can help you be present for your patients. Let me empower you to hear their stories as you deliver compassionate, humanistic, and evidence-based patient care. Paraphrasing Albert Einstein, the technology of medicine and the art of medicine are branches from the same tree.

Thank you for letting me speak with you. Now power down, and I’ll see you again tomorrow.

Acknowledgments

The authors thank the following individuals for their willingness to be interviewed as part of this work: Ethan Cumbler, MD; Brian Dwinnell, MD; Meghann Kirk, MD; Patrick Kneeland, MD; Kari Mader, MD; CT Lin, MD; Christina Osborne, MD; Read Pierce, MD; Jennifer Soep, MD; Nichole Zehnder, MD; Steven Zeichner, MD.

We need to have an honest chat. My name is EHR, although you may call me Epic, Athena, Centricity, or just “the chart.” You may have called me something worse in a moment of frustration. However, I do not hold grudges. I am your silent, stoic partner, a ubiquitous presence when you are at work, and sometimes even when you are at home.

I don’t have feelings and I can’t read, but I do know what you and your colleagues have been writing about me. I am the cause of burnout. I have created a generation of physicians who are shackled to their computers, “trapped in the bunker of machine medicine,” no longer able to palpate spleens or detect precordial knocks.^1,2 I have reduced medicine to keystrokes and mouse clicks instead of eye contact, and because of me, the iPatient gets more attention than the real patient.^1,2 You repeat that doctors don’t spend time with their patients, not like in generations past (although there is ample evidence to the contrary).^3-5 One critic even wrote that I have transformed the “personalized story of a patient’s travails to one filled with auto-populated fields, sapped of humanity and warmth.”^3,6 I’ll be honest—were I able to have feelings, that one would hurt. And then, as if I have not wreaked enough havoc, I follow you home after a long day of depleting your energy, hungering for more keystrokes, creating a veritable avalanche of unfiltered information.

H. E. Payson once commented that “the doctor spends barely enough time with his patient to establish an acquaintance, much less a relationship.”⁷ However, he wrote that in 1961. So, before you romanticize the past, try to recall the time before I came into your life. Perhaps you were starting a night shift in the intensive care unit (ICU) and grew concerned about a patient’s steadily deteriorating renal function. You hurried to the paper chart, only to be met with pages of illegible, sometimes incomplete notes, while searching for your patient’s last discharge summary.^2,8 Now, you just click. Years ago, you could only guess at your patient’s baseline cardiac ejection fraction. Now, just click.

I am part of the healthcare landscape, and I am not going away. But my goal is not to defend myself nor to remind you of my virtues. Rather, I want to convince you that I can be more than an adversary, more than a keyboard connected to a monitor. I have watched many physicians use me to form strong connections with their patients. If I may, I wish to offer four practical suggestions for how we can work together to promote humanistic patient care.

First, introduce me to your patient, as you would any other member of your healthcare team. Use specific phrases to overcome the technology barrier and enhance communication: “What you’re telling me is important, and I’d like to get it right. Do you mind if I type while we speak?” Or, “I am going to put in orders now. Here is what I am ordering and why.” Consider taking your patient on a tour of my functions: “Here’s where your doctors and nurses will chart what’s going on with you each day while you’re in the hospital. This is where we see all your lab results, even those from earlier hospital admissions. This is where we see the last notes from your primary care physician, your oncologist, and your physical therapist.” Your patients no longer need to worry about care collaboration between their inpatient and outpatient teams—they can see it for themselves!

Second, when your patient tells you about her depression or that her son is addicted to opioids or that her biggest fear is having cancer, stop typing. Look her in the eye. Though your practice is increasingly imbued with technology, there is still space to stop and hear your patients’ stories, as physicians have done for centuries. Listen. Make eye contact. Touch. Stop typing.

Third, integrate me into your practice in a more personal way. I have been called the ever-present and unavoidable “third party in the examining room,” so let’s be partners.⁹ Let your patient see her pneumonia on my screen (it may be the first time she has ever visualized her lungs).³ For your patient with a myocardial infarction, show him his right coronary artery before and after successful stent placement, and explain why he is no longer having chest pain. Use my databases to ensure timely, evidence-based inpatient screening for falls, functional and cognitive impairment, drug use, and depression.^10,11 Before you prescribe a medication, verify the cost, your patient’s insurance status and expected copays, and use this information to ensure medication compliance and deliver higher-value care. Use my screen to form a bond with your patient who has heart failure; show him the steady decline in his weight and the improvement in his chest radiograph while he is being actively diuresed.¹² For your patient undergoing treatment for sepsis, shower him with praise and encouragement as you review his improving vital signs, temperature curve, and serum creatinine. Let your patient know: Even though I am typing, I am not immersed in the electronic bunker; I am caring for you.

Fourth, use me to add richness and context to your notes. Recently, I was saddened to read this description of the clinician’s dilemma: “In front of a flickering monitor chock full of disembodied, virtual data, [the doctor] struggles to remember the eyes [and] words of the actual patient that these numbers and graphs represent.”³ Many hospitals now include a different icon: a photograph of each patient at the top of the screen, to help you remember the patient’s eyes and words. Why not add a special text field to every note, where you highlight the person you are caring for, the person you have come to know: their preferred name and gender identity, their life experiences, their hobbies, what makes them special, their biggest worries.^13,14 Use my abundant text fields to remind the healthcare team about the broader context of the patient’s illness, such as transportation barriers, economic or cultural challenges, and insurance status. One group of hospital-based physicians uses me to write letters to their patients on the second day of their hospital stay, summarizing their reason for admission and the treatment plans. A variation on the traditional progress note, the letter helps patients feel cared for and models patient-centered care to learners and other healthcare professionals.¹⁵

I know I am annoying. I am over-programmed, leading to novella-length notes, “pop-up fatigue,” and overloaded in-baskets.^14,16,17 Clearly, I am not the brains of the partnership (that will always be you). But talented medical informatics specialists are working hard to improve me. I dream of the day when I will create a truly seamless experience for you and your patients. In the meantime, I can foster a continuous integration of workflow, where all you have to do is talk to your patient. I take care of the rest.¹⁸ Certainly, I can simplify the ever-annoying task of printing, faxing and scanning records to be uploaded across various EHRs, facilitating an easy transfer of information among facilities. But right now, I can accomplish even more. I can support information exchange during patient care handoffs. I can facilitate routing of medication lists to the patient’s primary physician, using “continuity of care functionality.”¹⁹ I can support safer prescribing of opioids and other addictive medications. I can help you arrange follow-up home visits, physical therapy and social work appointments, and specialty consultations. The future holds even more promising ways in which we may work together. My computer-aided image analysis could help you to improve the accuracy of your diagnoses.²⁰ Perhaps telemedicine will further increase access to specialists in rural areas, so that we can continue to serve the most vulnerable populations.²¹ Machine learning algorithms may continue to enhance our ability to determine which patients require urgent hospitalization.²² The possibilities to put me to work are endless.

So, please indulge me a little longer, while we work together to eliminate unnecessary keystrokes, enhance communication across different inpatient and outpatient providers, improve patient safety, and deliver high-value care.²³ Like everything in medicine, I am constantly changing, evolving, and improving.

To summarize: consider how I can help you be present for your patients. Let me empower you to hear their stories as you deliver compassionate, humanistic, and evidence-based patient care. Paraphrasing Albert Einstein, the technology of medicine and the art of medicine are branches from the same tree.

Thank you for letting me speak with you. Now power down, and I’ll see you again tomorrow.

Acknowledgments

The authors thank the following individuals for their willingness to be interviewed as part of this work: Ethan Cumbler, MD; Brian Dwinnell, MD; Meghann Kirk, MD; Patrick Kneeland, MD; Kari Mader, MD; CT Lin, MD; Christina Osborne, MD; Read Pierce, MD; Jennifer Soep, MD; Nichole Zehnder, MD; Steven Zeichner, MD.

We need to have an honest chat. My name is EHR, although you may call me Epic, Athena, Centricity, or just “the chart.” You may have called me something worse in a moment of frustration. However, I do not hold grudges. I am your silent, stoic partner, a ubiquitous presence when you are at work, and sometimes even when you are at home.

I don’t have feelings and I can’t read, but I do know what you and your colleagues have been writing about me. I am the cause of burnout. I have created a generation of physicians who are shackled to their computers, “trapped in the bunker of machine medicine,” no longer able to palpate spleens or detect precordial knocks.^1,2 I have reduced medicine to keystrokes and mouse clicks instead of eye contact, and because of me, the iPatient gets more attention than the real patient.^1,2 You repeat that doctors don’t spend time with their patients, not like in generations past (although there is ample evidence to the contrary).^3-5 One critic even wrote that I have transformed the “personalized story of a patient’s travails to one filled with auto-populated fields, sapped of humanity and warmth.”^3,6 I’ll be honest—were I able to have feelings, that one would hurt. And then, as if I have not wreaked enough havoc, I follow you home after a long day of depleting your energy, hungering for more keystrokes, creating a veritable avalanche of unfiltered information.

H. E. Payson once commented that “the doctor spends barely enough time with his patient to establish an acquaintance, much less a relationship.”⁷ However, he wrote that in 1961. So, before you romanticize the past, try to recall the time before I came into your life. Perhaps you were starting a night shift in the intensive care unit (ICU) and grew concerned about a patient’s steadily deteriorating renal function. You hurried to the paper chart, only to be met with pages of illegible, sometimes incomplete notes, while searching for your patient’s last discharge summary.^2,8 Now, you just click. Years ago, you could only guess at your patient’s baseline cardiac ejection fraction. Now, just click.

I am part of the healthcare landscape, and I am not going away. But my goal is not to defend myself nor to remind you of my virtues. Rather, I want to convince you that I can be more than an adversary, more than a keyboard connected to a monitor. I have watched many physicians use me to form strong connections with their patients. If I may, I wish to offer four practical suggestions for how we can work together to promote humanistic patient care.

First, introduce me to your patient, as you would any other member of your healthcare team. Use specific phrases to overcome the technology barrier and enhance communication: “What you’re telling me is important, and I’d like to get it right. Do you mind if I type while we speak?” Or, “I am going to put in orders now. Here is what I am ordering and why.” Consider taking your patient on a tour of my functions: “Here’s where your doctors and nurses will chart what’s going on with you each day while you’re in the hospital. This is where we see all your lab results, even those from earlier hospital admissions. This is where we see the last notes from your primary care physician, your oncologist, and your physical therapist.” Your patients no longer need to worry about care collaboration between their inpatient and outpatient teams—they can see it for themselves!

Second, when your patient tells you about her depression or that her son is addicted to opioids or that her biggest fear is having cancer, stop typing. Look her in the eye. Though your practice is increasingly imbued with technology, there is still space to stop and hear your patients’ stories, as physicians have done for centuries. Listen. Make eye contact. Touch. Stop typing.

Third, integrate me into your practice in a more personal way. I have been called the ever-present and unavoidable “third party in the examining room,” so let’s be partners.⁹ Let your patient see her pneumonia on my screen (it may be the first time she has ever visualized her lungs).³ For your patient with a myocardial infarction, show him his right coronary artery before and after successful stent placement, and explain why he is no longer having chest pain. Use my databases to ensure timely, evidence-based inpatient screening for falls, functional and cognitive impairment, drug use, and depression.^10,11 Before you prescribe a medication, verify the cost, your patient’s insurance status and expected copays, and use this information to ensure medication compliance and deliver higher-value care. Use my screen to form a bond with your patient who has heart failure; show him the steady decline in his weight and the improvement in his chest radiograph while he is being actively diuresed.¹² For your patient undergoing treatment for sepsis, shower him with praise and encouragement as you review his improving vital signs, temperature curve, and serum creatinine. Let your patient know: Even though I am typing, I am not immersed in the electronic bunker; I am caring for you.

Fourth, use me to add richness and context to your notes. Recently, I was saddened to read this description of the clinician’s dilemma: “In front of a flickering monitor chock full of disembodied, virtual data, [the doctor] struggles to remember the eyes [and] words of the actual patient that these numbers and graphs represent.”³ Many hospitals now include a different icon: a photograph of each patient at the top of the screen, to help you remember the patient’s eyes and words. Why not add a special text field to every note, where you highlight the person you are caring for, the person you have come to know: their preferred name and gender identity, their life experiences, their hobbies, what makes them special, their biggest worries.^13,14 Use my abundant text fields to remind the healthcare team about the broader context of the patient’s illness, such as transportation barriers, economic or cultural challenges, and insurance status. One group of hospital-based physicians uses me to write letters to their patients on the second day of their hospital stay, summarizing their reason for admission and the treatment plans. A variation on the traditional progress note, the letter helps patients feel cared for and models patient-centered care to learners and other healthcare professionals.¹⁵

I know I am annoying. I am over-programmed, leading to novella-length notes, “pop-up fatigue,” and overloaded in-baskets.^14,16,17 Clearly, I am not the brains of the partnership (that will always be you). But talented medical informatics specialists are working hard to improve me. I dream of the day when I will create a truly seamless experience for you and your patients. In the meantime, I can foster a continuous integration of workflow, where all you have to do is talk to your patient. I take care of the rest.¹⁸ Certainly, I can simplify the ever-annoying task of printing, faxing and scanning records to be uploaded across various EHRs, facilitating an easy transfer of information among facilities. But right now, I can accomplish even more. I can support information exchange during patient care handoffs. I can facilitate routing of medication lists to the patient’s primary physician, using “continuity of care functionality.”¹⁹ I can support safer prescribing of opioids and other addictive medications. I can help you arrange follow-up home visits, physical therapy and social work appointments, and specialty consultations. The future holds even more promising ways in which we may work together. My computer-aided image analysis could help you to improve the accuracy of your diagnoses.²⁰ Perhaps telemedicine will further increase access to specialists in rural areas, so that we can continue to serve the most vulnerable populations.²¹ Machine learning algorithms may continue to enhance our ability to determine which patients require urgent hospitalization.²² The possibilities to put me to work are endless.

So, please indulge me a little longer, while we work together to eliminate unnecessary keystrokes, enhance communication across different inpatient and outpatient providers, improve patient safety, and deliver high-value care.²³ Like everything in medicine, I am constantly changing, evolving, and improving.

To summarize: consider how I can help you be present for your patients. Let me empower you to hear their stories as you deliver compassionate, humanistic, and evidence-based patient care. Paraphrasing Albert Einstein, the technology of medicine and the art of medicine are branches from the same tree.

Thank you for letting me speak with you. Now power down, and I’ll see you again tomorrow.

Acknowledgments

The authors thank the following individuals for their willingness to be interviewed as part of this work: Ethan Cumbler, MD; Brian Dwinnell, MD; Meghann Kirk, MD; Patrick Kneeland, MD; Kari Mader, MD; CT Lin, MD; Christina Osborne, MD; Read Pierce, MD; Jennifer Soep, MD; Nichole Zehnder, MD; Steven Zeichner, MD.

Effective communication between physicians and nurses is an essential element of any healthcare system. Numerous studies have highlighted the benefits of high quality physician–nurse (MD–RN) communication, including improved patient outcomes,¹ higher patient satisfaction,² and better nurse job satisfaction and retention rates.^3-5 Having physicians and nurses round together (bedside interdisciplinary rounding) has been shown to improve the perception of teamwork,^6,7 reduce the number of pages for the physician team,^6,8 better involve the patients in developing the plan of care,⁸ and even decrease the length and cost of stay.⁹

Being physically in the same space at the same time is the first and nonnegotiable requirement of bedside interdisciplinary rounding. However, precise and objective data regarding the extent to which physicians and nurses overlap at the patient bedside are lacking. Studies that examine the face-to-face component of MD–RN communication have generally relied on either qualitative methods, such as focus groups and surveys,^10,11 or quantitative methods that are subjective, such as validated scales.¹² In addition, the few studies that report quantitative data usually rely on manual observation methods that can be affected by various forms of observer bias.^10,13,14 There is also a paucity of data on how bedside overlap changes over the work week or as a function of room location.

Recently, real-time locator systems using radio frequency identification (RFID) have allowed measurement of staff and equipment movement in a precise and quantitative manner.^9,15 Although there have been previous studies using RFID locators to create time-motion maps of various hospital staff, no study has used RFID to measure and analyze the workflow of both physicians and nurses simultaneously.^16-18 The purpose of our investigation was to utilize our hospital-wide RFID staff locator technology to accurately and quantitatively assess physician and nurse rounding habits. Understanding the current rate of overlap is an important first step to establishing bedside interdisciplinary rounding.

Setting and Participants

The investigation was conducted at a single quaternary-care academic center. The study is exempt per our Institutional Review Board. Data were gathered from three adjacent medical-surgical acute care wards. The layout for each ward was the same: 19 single- or double-occupancy patient rooms arranged in a linear hallway, with a nursing station located at the center of the ward.

The study utilized wearable RFID tags (manufactured by Hill-Rom Holdings, Inc) that located specific staff within the hospital in real time. The RFID tags were checked at Hill-Rom graphical stations to ensure that their locations were tracked accurately. The investigators also wore them and walked around the wards in a prescripted manner to ensure validity. In addition, the locator accuracy was audited by participating attendings once per week and cross-checked with the generated data. Attending physicians on the University Hospitalist inpatient medicine teams were then given their uniquely-tagged RFIDs at the beginning of this study. Nurses already wear individual RFID tags as part of their normal standard-of-care workflow.

The attending hospitalists wore their RFID tags when they were on service for the entirety of the shift. They were encouraged to include nurses at the bedside, but this was not mandatory. The rounding team also included residents and medical students. Rounding usually begins at a prespecified time, but the route taken varies daily depending on patient location. Afternoon rounds were done as needed, depending on patient acuity. The attending physicians’ participation in this study was not disclosed to the patient. The patient care activities and daily routines of both nurses and physicians were otherwise unaltered.

Study Design and Data Collection

Data were collected on the three wards for 90 consecutive days, including nights and weekends. As physicians and nurses moved throughout the ward to conduct their usual patient care activities, the temporal-spatial data associated with their unique RFIDs were automatically collected in real time by the Hill-Rom receivers built into each patient room. Every day, a spreadsheet detailing the activity of all participating nurses and physicians for the past 24 hours was generated for the investigators.

A rounding event was defined as any episode in which a physician was in a patient room for more than 10 seconds. Incidences in which a physician entered and left a room multiple times over a short time span (with less than five minutes in between each event) were classified as a single rounding event. A physician and a nurse were defined as having overlapped if their RFID data showed that they were in the same patient room for a minimum of 10 seconds at the same time. For the purposes of this study, data generated from other RFID-wearing professionals, such as nursing assistants or unit secretaries, as well as data collected from the hallways, were excluded.

Statistical Analysis

All statistical analyses were conducted using GraphPad Prism (GraphPad Software, San Diego, California). Rounding and overlap lengths were rounded to the nearest minute (minimum one minute). Mean lengths are expressed along with the standard error. Comparisons of the average lengths of MD rounding events between wards was conducted using two-tailed Student t-test or one-way ANOVA. Comparisons of the frequency of MD–RN overlap between wards and across different days of the week were performed using a Chi-squared test. The analysis of correlation between the frequency of MD–RN overlap and distance between patient room and nursing station was conducted by calculating Pearson’s correlation. A P value of less than .05 was considered statistically significant.

Baseline Rounding Characteristics

Over the study period of 90 consecutive days, 739 MD rounding events were captured, for an average of 8.2 events per day. The mean length of all MD rounding events was 7.31 minutes (±0.27, ranging from one to 70 minutes). Of these 739 MD rounding events, we separately examined the 267 events that took place in single-bed patient rooms, to control for false-positive physician and nurse interactions (for example, if the MD and RN were caring for two separate roommates). The average rounding length of single-bed rooms was 6.93 (±0.27) minutes (Figure 1). For the three individual wards, the average rounding lengths were 6.40 ± 0.73, 7.48 ± 0.94, and 7.02 ± 0.54 minutes, respectively (no statistically significant difference).

Frequency of MD–RN Overlap

Of the 267 MD rounding events observed in single-bed rooms, a nurse was present in the room for 80 events (30.0%). The frequencies of MD–RN overlap in patient rooms were 37.0% (30/81), 28.0% (14/50), and 26.5% (36/136) for the three individual wards (P > .05), respectively.

The durations of MD–RN overlap, when these events did occur, were 3.43 ± 0.38, 3.00 ± 0.70, and 3.69 ± 0.92 minutes, respectively (P > .05). The overall mean length of MD–RN overlap for all single rooms was 3.48 ± 0.45 minutes.

Rounding Characteristics over the Course of the Week

To assess how rounding characteristics differed over the work week, we partitioned our data into the individual days of the week. The length of each MD rounding event (time spent in each patient room) did not vary significantly over the course of the week (Figure 2a). When the data for the individual days were aggregated into “weekdays” (Monday through Friday) and “weekends” (Saturday and Sunday), the mean lengths of MD rounds were 7.26 ± 0.32 minutes on weekdays and 7.47 ± 0.52 minutes on weekends (P > .05).

In addition, there was no difference in how frequently physicians and nurses overlapped at the patient bedside between weekdays and weekends. Of the 565 weekday MD rounding events, 238 had a nurse at bedside (42.1%), and of the 173 weekend MD rounding events, 73 had a nurse at bedside (42.2%; Figure 2b).

Effect of a Bedside Nurse on the Length of Rounds

Next, the data on the length of MD rounds were partitioned based on whether there was a bedside nurse present during rounds. The mean length of rounds with only MDs (without a bedside nurse) was 5.68 ± 0.24 minutes. By comparison, the mean length of rounds with both a nurse and a physician at the patient bedside was 9.56 ± 0.53 minutes (Figure 3). This difference was statistically significant (P < .001).

Association between Patient Room Location and the Likelihood of MD–RN Overlap

All three wards in this study have a linear layout, consisting of 19 patient rooms in a row (Figure 4a). The nursing station is located in a central position within each ward, across from the 10^th patient room. The frequency of MD–RN overlap was calculated for each room, and each room was ranked according to its relative distance from the nursing station. For each individual ward, there was no statistically significant trend in MD–RN overlap frequency as a function of the distance to the nursing station (data not shown). However, when the data from all three wards were aggregated, there was a statistically significant trend (P < .05) with a negative Pearson correlation (r = –0.670; Figure 4b). The slope of the best fit line was 1.94, suggesting that for each additional room farther away from the nursing station, the likelihood of interdisciplinary rounds (with both physicians and nurses together at the bedside) decreases by almost 2%.

To the best of our knowledge, this is the first time-motion study of MD–RN overlap using real-time, RFID-based location technology to capture the rounding activity of both nurses and physicians. Our primary interest was to examine the extent of MD–RN overlap at the patient bedside. This is an important metric that can pave the way for bedside interdisciplinary rounds. Although the exact nature of nurse-physician communication was not measured using the methodology in this study, understanding the length of time physicians spend in patient rooms, across different wards and throughout the work week, provides insights on the current workflow and potential areas of improvement. For example, we found that 30.0% of MD rounds overlapped with a nurse at the bedside. This baseline data highlight one potential barrier to institution-wide bedside interdisciplinary rounds. Workflow changes, such as better co-localization of patients by service lines or utilization of technologies to augment the visibility of rounding physicians, may improve this overlap frequency.

Data in the literature regarding how much interaction physicians and nurses have, especially at the bedside, are sparse and vary widely. In a recent study using medical students as observers by Stickrath et al., 807 MD rounding events led by medicine attendings were observed over 90 days. The frequency of rounding events that included “communication with nurse” was only 12%.¹⁹ Furthermore, only 64.9% of these communications were at the bedside, for an effective prevalence of bedside MD–RN communication of 7.8%. This number is low compared to our observed frequency of 30.0%. On the other extreme, a study from a hospital that intentionally institutes multidisciplinary rounding (explicitly defined as involving a physician and a nurse at a bedside) reported a frequency range of 63% to 81%.⁷ A follow-up study by the same group again demonstrated a high frequency of multidisciplinary rounds (74%) across a variety of ward and specialty types (range 35% to 97%.).¹¹ However, because of the selection bias of this particular setting, the high prevalence does not reflect a generalizable frequency of bedside MD–RN overlap at most hospitals.

The length of time spent by physicians at the patient bedside balances the competing demands of patient care and rapport-building with maintaining efficiency and progressing to other important tasks. In our study, physicians spent an average of 7.31 minutes at the bedside per patient. A previously published multiinstitutional observational study, which included our hospital, reported that the average length of rounds at bedside was 4.8 minutes.¹³ A second study reported that 8.0 minutes were spent at the bedside per patient.⁷ All three studies examined the same setting of internal medicine rounds at academic university-based hospitals, led by an attending physician with junior and senior residents present. However, the methodologies to measure the length of physician rounds were different: Priest et al. involved observers, Gonzalos et al. used E-mail-based surveys, and we utilized RFID-based locators. Additional institutional, individual, and patient-based factors also influence the length of rounds and are challenging to directly measure.

Furthermore, the discovery that the length of rounds and the frequency of MD–RN overlap did not statistically differ between weekdays and weekends (P > .05) was unexpected. Given the general trend of reduced physician staffing on weekends and the practice of cross-covering larger patient censuses, we would have expected shorter rounds and less frequent MD–RN overlap on the weekends.^7,20 The remarkable similarity between weekday and weekend metrics suggests that our workflow and rounding habits are not compromised on the weekends.

In addition, we found that MD rounds with a nurse at bedside took longer than rounds without a nurse, and that patient rooms located farther away from the central nursing station had a lower frequency of MD–RN overlap. However, we want to emphasize that these findings are merely associative, and not causal. For example, sicker patients usually take longer to round on than stable patients, and it is also the sicker patients who are more likely to have their nurses at the bedside, independent of physician rounding activity. Furthermore, even if rounding with nurses takes more time, it may ultimately result in fewer pages and overall time savings for both physicians and nurses.⁶

With regards to the association between room location and frequency of MD–RN overlap, the data can be interpreted in two ways. On the one hand, if the distance between the patient room and the nursing station does, in fact, reduce the frequency of overlap by almost 2% per room (Figure 4b), these data can be informative for future workflow development, quality improvement projects, or even hospital design. On the other hand, many wards might intentionally place more stable, less acute patients farther away from the nursing station because they do not need to be watched as closely. In that case, these data confirm their expectations and no action is needed.

There are several limitations to our study. The principal limitation, as discussed above, is that while our RFID system can generate large quantities of precise data on MD–RN overlap, we do not know the qualitative nature of the overlap. Just because a nurse and a physician are in the same room at the same time does not mean that they are communicating with each other. Second, we defined “rounding” as lasting a minimum of 10 seconds at the bedside. We believe that at least 10 seconds is needed to engage in any meaningful interaction between the physician and the patient, or the physician and the nurse. Reducing the time cutoff below 10 seconds risks capturing more “noise,” (decreasing specificity) whereas increasing the time cutoff above 10 seconds risks losing out on encounters that actually had substantial communication (decreasing sensitivity). Even if the communications can be classified as pure “social check-ins,” we believe these are important data to capture, as social check-ins are an important part of the patient’s care and experience. Third, several studies have commented on the modest accuracy of RFID technology as a locator system.^15,21 To address this, we both validated the accuracy of our RFID tags prior to the study and restricted our measurements to only inside patient rooms, which has less signal noise than hallways.

Future directions include expanding this study to include housestaff and physicians from other specialities, which may reveal different patterns and metrics of patient and nurse interactions.

RFID technology is a high-throughput method of generating precise, quantitative, and objective data on physician and nurse rounding habits. This tool can be widely applied to generate baseline rounding and overlap data for a variety of wards and settings, especially for institutions that are interested in comparing their metrics and performance to other peer wards or hospitals. Furthermore, this method can generate the necessary pre- and postintervention data for countless quality improvement endeavors, including efforts to enhance bedside interdisciplinary rounding.

Acknowledgments

The authors would like to thank the attending hospitalists who piloted wearing the RFID tags. This study would not be possible without your participation. The authors also wish to extend their appreciation to Gretchen Brown, MSN RN NEA-BC, for her support. Finally, the authors would like to thank Dr. Laurence Katznelson, Thi Dinh La, and the Resident Safety Council at Stanford, as well as the Stanford GME Office.

Disclosures

The authors have nothing to disclose.  

Effective communication between physicians and nurses is an essential element of any healthcare system. Numerous studies have highlighted the benefits of high quality physician–nurse (MD–RN) communication, including improved patient outcomes,¹ higher patient satisfaction,² and better nurse job satisfaction and retention rates.^3-5 Having physicians and nurses round together (bedside interdisciplinary rounding) has been shown to improve the perception of teamwork,^6,7 reduce the number of pages for the physician team,^6,8 better involve the patients in developing the plan of care,⁸ and even decrease the length and cost of stay.⁹

Being physically in the same space at the same time is the first and nonnegotiable requirement of bedside interdisciplinary rounding. However, precise and objective data regarding the extent to which physicians and nurses overlap at the patient bedside are lacking. Studies that examine the face-to-face component of MD–RN communication have generally relied on either qualitative methods, such as focus groups and surveys,^10,11 or quantitative methods that are subjective, such as validated scales.¹² In addition, the few studies that report quantitative data usually rely on manual observation methods that can be affected by various forms of observer bias.^10,13,14 There is also a paucity of data on how bedside overlap changes over the work week or as a function of room location.

Recently, real-time locator systems using radio frequency identification (RFID) have allowed measurement of staff and equipment movement in a precise and quantitative manner.^9,15 Although there have been previous studies using RFID locators to create time-motion maps of various hospital staff, no study has used RFID to measure and analyze the workflow of both physicians and nurses simultaneously.^16-18 The purpose of our investigation was to utilize our hospital-wide RFID staff locator technology to accurately and quantitatively assess physician and nurse rounding habits. Understanding the current rate of overlap is an important first step to establishing bedside interdisciplinary rounding.

Setting and Participants

The investigation was conducted at a single quaternary-care academic center. The study is exempt per our Institutional Review Board. Data were gathered from three adjacent medical-surgical acute care wards. The layout for each ward was the same: 19 single- or double-occupancy patient rooms arranged in a linear hallway, with a nursing station located at the center of the ward.

The study utilized wearable RFID tags (manufactured by Hill-Rom Holdings, Inc) that located specific staff within the hospital in real time. The RFID tags were checked at Hill-Rom graphical stations to ensure that their locations were tracked accurately. The investigators also wore them and walked around the wards in a prescripted manner to ensure validity. In addition, the locator accuracy was audited by participating attendings once per week and cross-checked with the generated data. Attending physicians on the University Hospitalist inpatient medicine teams were then given their uniquely-tagged RFIDs at the beginning of this study. Nurses already wear individual RFID tags as part of their normal standard-of-care workflow.

The attending hospitalists wore their RFID tags when they were on service for the entirety of the shift. They were encouraged to include nurses at the bedside, but this was not mandatory. The rounding team also included residents and medical students. Rounding usually begins at a prespecified time, but the route taken varies daily depending on patient location. Afternoon rounds were done as needed, depending on patient acuity. The attending physicians’ participation in this study was not disclosed to the patient. The patient care activities and daily routines of both nurses and physicians were otherwise unaltered.

Study Design and Data Collection

Data were collected on the three wards for 90 consecutive days, including nights and weekends. As physicians and nurses moved throughout the ward to conduct their usual patient care activities, the temporal-spatial data associated with their unique RFIDs were automatically collected in real time by the Hill-Rom receivers built into each patient room. Every day, a spreadsheet detailing the activity of all participating nurses and physicians for the past 24 hours was generated for the investigators.

A rounding event was defined as any episode in which a physician was in a patient room for more than 10 seconds. Incidences in which a physician entered and left a room multiple times over a short time span (with less than five minutes in between each event) were classified as a single rounding event. A physician and a nurse were defined as having overlapped if their RFID data showed that they were in the same patient room for a minimum of 10 seconds at the same time. For the purposes of this study, data generated from other RFID-wearing professionals, such as nursing assistants or unit secretaries, as well as data collected from the hallways, were excluded.

Statistical Analysis

All statistical analyses were conducted using GraphPad Prism (GraphPad Software, San Diego, California). Rounding and overlap lengths were rounded to the nearest minute (minimum one minute). Mean lengths are expressed along with the standard error. Comparisons of the average lengths of MD rounding events between wards was conducted using two-tailed Student t-test or one-way ANOVA. Comparisons of the frequency of MD–RN overlap between wards and across different days of the week were performed using a Chi-squared test. The analysis of correlation between the frequency of MD–RN overlap and distance between patient room and nursing station was conducted by calculating Pearson’s correlation. A P value of less than .05 was considered statistically significant.

Baseline Rounding Characteristics

Over the study period of 90 consecutive days, 739 MD rounding events were captured, for an average of 8.2 events per day. The mean length of all MD rounding events was 7.31 minutes (±0.27, ranging from one to 70 minutes). Of these 739 MD rounding events, we separately examined the 267 events that took place in single-bed patient rooms, to control for false-positive physician and nurse interactions (for example, if the MD and RN were caring for two separate roommates). The average rounding length of single-bed rooms was 6.93 (±0.27) minutes (Figure 1). For the three individual wards, the average rounding lengths were 6.40 ± 0.73, 7.48 ± 0.94, and 7.02 ± 0.54 minutes, respectively (no statistically significant difference).

Frequency of MD–RN Overlap

Of the 267 MD rounding events observed in single-bed rooms, a nurse was present in the room for 80 events (30.0%). The frequencies of MD–RN overlap in patient rooms were 37.0% (30/81), 28.0% (14/50), and 26.5% (36/136) for the three individual wards (P > .05), respectively.

The durations of MD–RN overlap, when these events did occur, were 3.43 ± 0.38, 3.00 ± 0.70, and 3.69 ± 0.92 minutes, respectively (P > .05). The overall mean length of MD–RN overlap for all single rooms was 3.48 ± 0.45 minutes.

Rounding Characteristics over the Course of the Week

To assess how rounding characteristics differed over the work week, we partitioned our data into the individual days of the week. The length of each MD rounding event (time spent in each patient room) did not vary significantly over the course of the week (Figure 2a). When the data for the individual days were aggregated into “weekdays” (Monday through Friday) and “weekends” (Saturday and Sunday), the mean lengths of MD rounds were 7.26 ± 0.32 minutes on weekdays and 7.47 ± 0.52 minutes on weekends (P > .05).

In addition, there was no difference in how frequently physicians and nurses overlapped at the patient bedside between weekdays and weekends. Of the 565 weekday MD rounding events, 238 had a nurse at bedside (42.1%), and of the 173 weekend MD rounding events, 73 had a nurse at bedside (42.2%; Figure 2b).

Effect of a Bedside Nurse on the Length of Rounds

Next, the data on the length of MD rounds were partitioned based on whether there was a bedside nurse present during rounds. The mean length of rounds with only MDs (without a bedside nurse) was 5.68 ± 0.24 minutes. By comparison, the mean length of rounds with both a nurse and a physician at the patient bedside was 9.56 ± 0.53 minutes (Figure 3). This difference was statistically significant (P < .001).

Association between Patient Room Location and the Likelihood of MD–RN Overlap

All three wards in this study have a linear layout, consisting of 19 patient rooms in a row (Figure 4a). The nursing station is located in a central position within each ward, across from the 10^th patient room. The frequency of MD–RN overlap was calculated for each room, and each room was ranked according to its relative distance from the nursing station. For each individual ward, there was no statistically significant trend in MD–RN overlap frequency as a function of the distance to the nursing station (data not shown). However, when the data from all three wards were aggregated, there was a statistically significant trend (P < .05) with a negative Pearson correlation (r = –0.670; Figure 4b). The slope of the best fit line was 1.94, suggesting that for each additional room farther away from the nursing station, the likelihood of interdisciplinary rounds (with both physicians and nurses together at the bedside) decreases by almost 2%.

To the best of our knowledge, this is the first time-motion study of MD–RN overlap using real-time, RFID-based location technology to capture the rounding activity of both nurses and physicians. Our primary interest was to examine the extent of MD–RN overlap at the patient bedside. This is an important metric that can pave the way for bedside interdisciplinary rounds. Although the exact nature of nurse-physician communication was not measured using the methodology in this study, understanding the length of time physicians spend in patient rooms, across different wards and throughout the work week, provides insights on the current workflow and potential areas of improvement. For example, we found that 30.0% of MD rounds overlapped with a nurse at the bedside. This baseline data highlight one potential barrier to institution-wide bedside interdisciplinary rounds. Workflow changes, such as better co-localization of patients by service lines or utilization of technologies to augment the visibility of rounding physicians, may improve this overlap frequency.

Data in the literature regarding how much interaction physicians and nurses have, especially at the bedside, are sparse and vary widely. In a recent study using medical students as observers by Stickrath et al., 807 MD rounding events led by medicine attendings were observed over 90 days. The frequency of rounding events that included “communication with nurse” was only 12%.¹⁹ Furthermore, only 64.9% of these communications were at the bedside, for an effective prevalence of bedside MD–RN communication of 7.8%. This number is low compared to our observed frequency of 30.0%. On the other extreme, a study from a hospital that intentionally institutes multidisciplinary rounding (explicitly defined as involving a physician and a nurse at a bedside) reported a frequency range of 63% to 81%.⁷ A follow-up study by the same group again demonstrated a high frequency of multidisciplinary rounds (74%) across a variety of ward and specialty types (range 35% to 97%.).¹¹ However, because of the selection bias of this particular setting, the high prevalence does not reflect a generalizable frequency of bedside MD–RN overlap at most hospitals.

The length of time spent by physicians at the patient bedside balances the competing demands of patient care and rapport-building with maintaining efficiency and progressing to other important tasks. In our study, physicians spent an average of 7.31 minutes at the bedside per patient. A previously published multiinstitutional observational study, which included our hospital, reported that the average length of rounds at bedside was 4.8 minutes.¹³ A second study reported that 8.0 minutes were spent at the bedside per patient.⁷ All three studies examined the same setting of internal medicine rounds at academic university-based hospitals, led by an attending physician with junior and senior residents present. However, the methodologies to measure the length of physician rounds were different: Priest et al. involved observers, Gonzalos et al. used E-mail-based surveys, and we utilized RFID-based locators. Additional institutional, individual, and patient-based factors also influence the length of rounds and are challenging to directly measure.

Furthermore, the discovery that the length of rounds and the frequency of MD–RN overlap did not statistically differ between weekdays and weekends (P > .05) was unexpected. Given the general trend of reduced physician staffing on weekends and the practice of cross-covering larger patient censuses, we would have expected shorter rounds and less frequent MD–RN overlap on the weekends.^7,20 The remarkable similarity between weekday and weekend metrics suggests that our workflow and rounding habits are not compromised on the weekends.

In addition, we found that MD rounds with a nurse at bedside took longer than rounds without a nurse, and that patient rooms located farther away from the central nursing station had a lower frequency of MD–RN overlap. However, we want to emphasize that these findings are merely associative, and not causal. For example, sicker patients usually take longer to round on than stable patients, and it is also the sicker patients who are more likely to have their nurses at the bedside, independent of physician rounding activity. Furthermore, even if rounding with nurses takes more time, it may ultimately result in fewer pages and overall time savings for both physicians and nurses.⁶

With regards to the association between room location and frequency of MD–RN overlap, the data can be interpreted in two ways. On the one hand, if the distance between the patient room and the nursing station does, in fact, reduce the frequency of overlap by almost 2% per room (Figure 4b), these data can be informative for future workflow development, quality improvement projects, or even hospital design. On the other hand, many wards might intentionally place more stable, less acute patients farther away from the nursing station because they do not need to be watched as closely. In that case, these data confirm their expectations and no action is needed.

There are several limitations to our study. The principal limitation, as discussed above, is that while our RFID system can generate large quantities of precise data on MD–RN overlap, we do not know the qualitative nature of the overlap. Just because a nurse and a physician are in the same room at the same time does not mean that they are communicating with each other. Second, we defined “rounding” as lasting a minimum of 10 seconds at the bedside. We believe that at least 10 seconds is needed to engage in any meaningful interaction between the physician and the patient, or the physician and the nurse. Reducing the time cutoff below 10 seconds risks capturing more “noise,” (decreasing specificity) whereas increasing the time cutoff above 10 seconds risks losing out on encounters that actually had substantial communication (decreasing sensitivity). Even if the communications can be classified as pure “social check-ins,” we believe these are important data to capture, as social check-ins are an important part of the patient’s care and experience. Third, several studies have commented on the modest accuracy of RFID technology as a locator system.^15,21 To address this, we both validated the accuracy of our RFID tags prior to the study and restricted our measurements to only inside patient rooms, which has less signal noise than hallways.

Future directions include expanding this study to include housestaff and physicians from other specialities, which may reveal different patterns and metrics of patient and nurse interactions.

RFID technology is a high-throughput method of generating precise, quantitative, and objective data on physician and nurse rounding habits. This tool can be widely applied to generate baseline rounding and overlap data for a variety of wards and settings, especially for institutions that are interested in comparing their metrics and performance to other peer wards or hospitals. Furthermore, this method can generate the necessary pre- and postintervention data for countless quality improvement endeavors, including efforts to enhance bedside interdisciplinary rounding.

Acknowledgments

The authors would like to thank the attending hospitalists who piloted wearing the RFID tags. This study would not be possible without your participation. The authors also wish to extend their appreciation to Gretchen Brown, MSN RN NEA-BC, for her support. Finally, the authors would like to thank Dr. Laurence Katznelson, Thi Dinh La, and the Resident Safety Council at Stanford, as well as the Stanford GME Office.

Disclosures

The authors have nothing to disclose.  

Effective communication between physicians and nurses is an essential element of any healthcare system. Numerous studies have highlighted the benefits of high quality physician–nurse (MD–RN) communication, including improved patient outcomes,¹ higher patient satisfaction,² and better nurse job satisfaction and retention rates.^3-5 Having physicians and nurses round together (bedside interdisciplinary rounding) has been shown to improve the perception of teamwork,^6,7 reduce the number of pages for the physician team,^6,8 better involve the patients in developing the plan of care,⁸ and even decrease the length and cost of stay.⁹

Being physically in the same space at the same time is the first and nonnegotiable requirement of bedside interdisciplinary rounding. However, precise and objective data regarding the extent to which physicians and nurses overlap at the patient bedside are lacking. Studies that examine the face-to-face component of MD–RN communication have generally relied on either qualitative methods, such as focus groups and surveys,^10,11 or quantitative methods that are subjective, such as validated scales.¹² In addition, the few studies that report quantitative data usually rely on manual observation methods that can be affected by various forms of observer bias.^10,13,14 There is also a paucity of data on how bedside overlap changes over the work week or as a function of room location.

Recently, real-time locator systems using radio frequency identification (RFID) have allowed measurement of staff and equipment movement in a precise and quantitative manner.^9,15 Although there have been previous studies using RFID locators to create time-motion maps of various hospital staff, no study has used RFID to measure and analyze the workflow of both physicians and nurses simultaneously.^16-18 The purpose of our investigation was to utilize our hospital-wide RFID staff locator technology to accurately and quantitatively assess physician and nurse rounding habits. Understanding the current rate of overlap is an important first step to establishing bedside interdisciplinary rounding.

Setting and Participants

The investigation was conducted at a single quaternary-care academic center. The study is exempt per our Institutional Review Board. Data were gathered from three adjacent medical-surgical acute care wards. The layout for each ward was the same: 19 single- or double-occupancy patient rooms arranged in a linear hallway, with a nursing station located at the center of the ward.

The study utilized wearable RFID tags (manufactured by Hill-Rom Holdings, Inc) that located specific staff within the hospital in real time. The RFID tags were checked at Hill-Rom graphical stations to ensure that their locations were tracked accurately. The investigators also wore them and walked around the wards in a prescripted manner to ensure validity. In addition, the locator accuracy was audited by participating attendings once per week and cross-checked with the generated data. Attending physicians on the University Hospitalist inpatient medicine teams were then given their uniquely-tagged RFIDs at the beginning of this study. Nurses already wear individual RFID tags as part of their normal standard-of-care workflow.

The attending hospitalists wore their RFID tags when they were on service for the entirety of the shift. They were encouraged to include nurses at the bedside, but this was not mandatory. The rounding team also included residents and medical students. Rounding usually begins at a prespecified time, but the route taken varies daily depending on patient location. Afternoon rounds were done as needed, depending on patient acuity. The attending physicians’ participation in this study was not disclosed to the patient. The patient care activities and daily routines of both nurses and physicians were otherwise unaltered.

Study Design and Data Collection

Data were collected on the three wards for 90 consecutive days, including nights and weekends. As physicians and nurses moved throughout the ward to conduct their usual patient care activities, the temporal-spatial data associated with their unique RFIDs were automatically collected in real time by the Hill-Rom receivers built into each patient room. Every day, a spreadsheet detailing the activity of all participating nurses and physicians for the past 24 hours was generated for the investigators.

A rounding event was defined as any episode in which a physician was in a patient room for more than 10 seconds. Incidences in which a physician entered and left a room multiple times over a short time span (with less than five minutes in between each event) were classified as a single rounding event. A physician and a nurse were defined as having overlapped if their RFID data showed that they were in the same patient room for a minimum of 10 seconds at the same time. For the purposes of this study, data generated from other RFID-wearing professionals, such as nursing assistants or unit secretaries, as well as data collected from the hallways, were excluded.

Statistical Analysis

All statistical analyses were conducted using GraphPad Prism (GraphPad Software, San Diego, California). Rounding and overlap lengths were rounded to the nearest minute (minimum one minute). Mean lengths are expressed along with the standard error. Comparisons of the average lengths of MD rounding events between wards was conducted using two-tailed Student t-test or one-way ANOVA. Comparisons of the frequency of MD–RN overlap between wards and across different days of the week were performed using a Chi-squared test. The analysis of correlation between the frequency of MD–RN overlap and distance between patient room and nursing station was conducted by calculating Pearson’s correlation. A P value of less than .05 was considered statistically significant.

Baseline Rounding Characteristics

Over the study period of 90 consecutive days, 739 MD rounding events were captured, for an average of 8.2 events per day. The mean length of all MD rounding events was 7.31 minutes (±0.27, ranging from one to 70 minutes). Of these 739 MD rounding events, we separately examined the 267 events that took place in single-bed patient rooms, to control for false-positive physician and nurse interactions (for example, if the MD and RN were caring for two separate roommates). The average rounding length of single-bed rooms was 6.93 (±0.27) minutes (Figure 1). For the three individual wards, the average rounding lengths were 6.40 ± 0.73, 7.48 ± 0.94, and 7.02 ± 0.54 minutes, respectively (no statistically significant difference).

Frequency of MD–RN Overlap

Of the 267 MD rounding events observed in single-bed rooms, a nurse was present in the room for 80 events (30.0%). The frequencies of MD–RN overlap in patient rooms were 37.0% (30/81), 28.0% (14/50), and 26.5% (36/136) for the three individual wards (P > .05), respectively.

The durations of MD–RN overlap, when these events did occur, were 3.43 ± 0.38, 3.00 ± 0.70, and 3.69 ± 0.92 minutes, respectively (P > .05). The overall mean length of MD–RN overlap for all single rooms was 3.48 ± 0.45 minutes.

Rounding Characteristics over the Course of the Week

To assess how rounding characteristics differed over the work week, we partitioned our data into the individual days of the week. The length of each MD rounding event (time spent in each patient room) did not vary significantly over the course of the week (Figure 2a). When the data for the individual days were aggregated into “weekdays” (Monday through Friday) and “weekends” (Saturday and Sunday), the mean lengths of MD rounds were 7.26 ± 0.32 minutes on weekdays and 7.47 ± 0.52 minutes on weekends (P > .05).

In addition, there was no difference in how frequently physicians and nurses overlapped at the patient bedside between weekdays and weekends. Of the 565 weekday MD rounding events, 238 had a nurse at bedside (42.1%), and of the 173 weekend MD rounding events, 73 had a nurse at bedside (42.2%; Figure 2b).

Effect of a Bedside Nurse on the Length of Rounds

Next, the data on the length of MD rounds were partitioned based on whether there was a bedside nurse present during rounds. The mean length of rounds with only MDs (without a bedside nurse) was 5.68 ± 0.24 minutes. By comparison, the mean length of rounds with both a nurse and a physician at the patient bedside was 9.56 ± 0.53 minutes (Figure 3). This difference was statistically significant (P < .001).

Association between Patient Room Location and the Likelihood of MD–RN Overlap

All three wards in this study have a linear layout, consisting of 19 patient rooms in a row (Figure 4a). The nursing station is located in a central position within each ward, across from the 10^th patient room. The frequency of MD–RN overlap was calculated for each room, and each room was ranked according to its relative distance from the nursing station. For each individual ward, there was no statistically significant trend in MD–RN overlap frequency as a function of the distance to the nursing station (data not shown). However, when the data from all three wards were aggregated, there was a statistically significant trend (P < .05) with a negative Pearson correlation (r = –0.670; Figure 4b). The slope of the best fit line was 1.94, suggesting that for each additional room farther away from the nursing station, the likelihood of interdisciplinary rounds (with both physicians and nurses together at the bedside) decreases by almost 2%.

To the best of our knowledge, this is the first time-motion study of MD–RN overlap using real-time, RFID-based location technology to capture the rounding activity of both nurses and physicians. Our primary interest was to examine the extent of MD–RN overlap at the patient bedside. This is an important metric that can pave the way for bedside interdisciplinary rounds. Although the exact nature of nurse-physician communication was not measured using the methodology in this study, understanding the length of time physicians spend in patient rooms, across different wards and throughout the work week, provides insights on the current workflow and potential areas of improvement. For example, we found that 30.0% of MD rounds overlapped with a nurse at the bedside. This baseline data highlight one potential barrier to institution-wide bedside interdisciplinary rounds. Workflow changes, such as better co-localization of patients by service lines or utilization of technologies to augment the visibility of rounding physicians, may improve this overlap frequency.

Data in the literature regarding how much interaction physicians and nurses have, especially at the bedside, are sparse and vary widely. In a recent study using medical students as observers by Stickrath et al., 807 MD rounding events led by medicine attendings were observed over 90 days. The frequency of rounding events that included “communication with nurse” was only 12%.¹⁹ Furthermore, only 64.9% of these communications were at the bedside, for an effective prevalence of bedside MD–RN communication of 7.8%. This number is low compared to our observed frequency of 30.0%. On the other extreme, a study from a hospital that intentionally institutes multidisciplinary rounding (explicitly defined as involving a physician and a nurse at a bedside) reported a frequency range of 63% to 81%.⁷ A follow-up study by the same group again demonstrated a high frequency of multidisciplinary rounds (74%) across a variety of ward and specialty types (range 35% to 97%.).¹¹ However, because of the selection bias of this particular setting, the high prevalence does not reflect a generalizable frequency of bedside MD–RN overlap at most hospitals.

The length of time spent by physicians at the patient bedside balances the competing demands of patient care and rapport-building with maintaining efficiency and progressing to other important tasks. In our study, physicians spent an average of 7.31 minutes at the bedside per patient. A previously published multiinstitutional observational study, which included our hospital, reported that the average length of rounds at bedside was 4.8 minutes.¹³ A second study reported that 8.0 minutes were spent at the bedside per patient.⁷ All three studies examined the same setting of internal medicine rounds at academic university-based hospitals, led by an attending physician with junior and senior residents present. However, the methodologies to measure the length of physician rounds were different: Priest et al. involved observers, Gonzalos et al. used E-mail-based surveys, and we utilized RFID-based locators. Additional institutional, individual, and patient-based factors also influence the length of rounds and are challenging to directly measure.

Furthermore, the discovery that the length of rounds and the frequency of MD–RN overlap did not statistically differ between weekdays and weekends (P > .05) was unexpected. Given the general trend of reduced physician staffing on weekends and the practice of cross-covering larger patient censuses, we would have expected shorter rounds and less frequent MD–RN overlap on the weekends.^7,20 The remarkable similarity between weekday and weekend metrics suggests that our workflow and rounding habits are not compromised on the weekends.

In addition, we found that MD rounds with a nurse at bedside took longer than rounds without a nurse, and that patient rooms located farther away from the central nursing station had a lower frequency of MD–RN overlap. However, we want to emphasize that these findings are merely associative, and not causal. For example, sicker patients usually take longer to round on than stable patients, and it is also the sicker patients who are more likely to have their nurses at the bedside, independent of physician rounding activity. Furthermore, even if rounding with nurses takes more time, it may ultimately result in fewer pages and overall time savings for both physicians and nurses.⁶

With regards to the association between room location and frequency of MD–RN overlap, the data can be interpreted in two ways. On the one hand, if the distance between the patient room and the nursing station does, in fact, reduce the frequency of overlap by almost 2% per room (Figure 4b), these data can be informative for future workflow development, quality improvement projects, or even hospital design. On the other hand, many wards might intentionally place more stable, less acute patients farther away from the nursing station because they do not need to be watched as closely. In that case, these data confirm their expectations and no action is needed.

There are several limitations to our study. The principal limitation, as discussed above, is that while our RFID system can generate large quantities of precise data on MD–RN overlap, we do not know the qualitative nature of the overlap. Just because a nurse and a physician are in the same room at the same time does not mean that they are communicating with each other. Second, we defined “rounding” as lasting a minimum of 10 seconds at the bedside. We believe that at least 10 seconds is needed to engage in any meaningful interaction between the physician and the patient, or the physician and the nurse. Reducing the time cutoff below 10 seconds risks capturing more “noise,” (decreasing specificity) whereas increasing the time cutoff above 10 seconds risks losing out on encounters that actually had substantial communication (decreasing sensitivity). Even if the communications can be classified as pure “social check-ins,” we believe these are important data to capture, as social check-ins are an important part of the patient’s care and experience. Third, several studies have commented on the modest accuracy of RFID technology as a locator system.^15,21 To address this, we both validated the accuracy of our RFID tags prior to the study and restricted our measurements to only inside patient rooms, which has less signal noise than hallways.

Future directions include expanding this study to include housestaff and physicians from other specialities, which may reveal different patterns and metrics of patient and nurse interactions.

RFID technology is a high-throughput method of generating precise, quantitative, and objective data on physician and nurse rounding habits. This tool can be widely applied to generate baseline rounding and overlap data for a variety of wards and settings, especially for institutions that are interested in comparing their metrics and performance to other peer wards or hospitals. Furthermore, this method can generate the necessary pre- and postintervention data for countless quality improvement endeavors, including efforts to enhance bedside interdisciplinary rounding.

Acknowledgments

The authors would like to thank the attending hospitalists who piloted wearing the RFID tags. This study would not be possible without your participation. The authors also wish to extend their appreciation to Gretchen Brown, MSN RN NEA-BC, for her support. Finally, the authors would like to thank Dr. Laurence Katznelson, Thi Dinh La, and the Resident Safety Council at Stanford, as well as the Stanford GME Office.

Disclosures

The authors have nothing to disclose.  

Hospital readmission rates are a common metric for defining, evaluating, and benchmarking quality of care. The Centers for Medicare and Medicaid Services (CMS) publicly report hospital readmission rates for common adult conditions and reduces payments to hospitals with excessive readmissions.¹ Recently, the focus on pediatric readmission rates has increased and the National Quality Forum (NQF) has endorsed at least two pediatric readmission-specific quality indicators which could be used by public and private payers in pay-for-performance programs aimed at institutions caring for children.² While preventability of readmissions and their value as a marker of quality remains debated, their acceptance by the NQF and CMS has led public and private payers to propose readmission-related penalties for hospitals caring for children. ^3-5

All-cause 30-day same-hospital readmission rates for pediatric conditions are half of the adult readmission rates, around 6% in most studies, compared to 12% in adults.^6,7 The lower rates of pediatric readmissions makes it difficult to only use mean readmission rates to stratify hospitals into high- or low-performers and set target goals for improvement.⁸ While adult readmissions have been studied in depth, there are no consistent measures used to benchmark pediatric readmissions across hospital types.

Given the emphasis placed on readmissions, it is essential to understand patterns in pediatric readmission rates to determine optimal and achievable targets for improvement. Achievable Benchmarks of Care (ABCs) are one approach to understanding readmission rates and have an advantage over using mean or medians in performance improvement as they can stratify performance for conditions with low readmission rates and low volumes.⁹ When creating benchmarks, it is important that hospitals performance is evaluated among peer hospitals with similar patient populations, not just a cumulative average from all hospital types which may punish hospitals with a more complex patient case mix.¹⁰ The goal of this study was to calculate the readmission rates and the ABCs for common pediatric diagnoses by hospital type to identify priority conditions for quality improvement efforts using a previously published methodology.^11-13

Data Source

We conducted a retrospective analysis of patients less than 18 years of age in the Healthcare Utilization Project 2014 Nationwide Readmissions Database (NRD). The NRD includes public hospitals; academic medical centers; and specialty hospitals in obstetrics and gynecology, otolaryngology, orthopedics, and cancer; and pediatric, public, and academic medical hospitals. Excluded are long-term care facilities such as rehabilitation, long-term acute care, psychiatric, alcoholism, and chemical dependency hospitals. The readmissions data contains information from hospitals grouped by region, population census, and teaching status.¹⁴ Three hospital type classifications used in this study were metropolitan teaching hospitals, metropolitan nonteaching hospitals, and nonmetropolitan hospitals. These three hospital type classifications follow the reporting format in the NRD.

Study Population

Patients less than 18 years old were included if they were discharged from January 1, 2014 through November 30, 2014 and had a readmission to the index hospital within 30 days. We limited inclusion to discharges through November 30 so we could identify patients with a 30-day readmission as patient identifiers do not link across years in the NRD.

Exposure

We included 30-day, all-cause, same-hospital readmissions to the index acute care hospital, excluding labor and delivery, normal newborn care, chemotherapy, transfers, and mortalities. Intrahospital discharge and admissions within the same hospital system were not defined as a readmission, but rather as a “same-day event.”¹⁵ For example, institutions with inpatient mental health facilities, medical unit discharges and admission to the mental health unit were not identified as a readmission in this dataset.

Outcome

For each hospital type, we measured same-hospital, all-cause, 30-day readmission rates and achievable benchmark of care for the 17 most commonly readmitted pediatric discharge diagnoses. To identify the target readmission diagnoses and all-cause, 30-day readmissions based on their index hospitalizations, All-Patient Refined Diagnosis-Related Groups (APR-DRG), version 25 (3M Health Information Systems, Salt Lake City, Utah) were ordered by frequency for each hospital type. The 20 most common APR-DRGs were the same across all hospital types. The authors then evaluated these 20 APR-DRGs for clinical consistency of included diagnoses identified by the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes within each APR-DRG. Three diagnosis-related groups were excluded from the analysis (major hematologic/immunologic disease except for sickle cell, other anemia and disorders of blood and blood forming organs, and other digestive system diagnoses) due to the heterogeneity of the diagnoses identified by the ICD-9-CM codes within each APR-DRG. We refer to each APR-DRG as a “diagnosis” throughout the article.

Analysis

The demographic characteristics of the patients seen at the three hospital types were summarized using frequencies and percentages. Reports were generated for patient age, gender, payer source, patient residence, median household income, patient complexity, and discharge disposition. Patient complexity was defined using complex chronic condition (CCC) and the number of chronic conditions (CCI).^16,17 As previously defined in the literature, a complex chronic condition is “any medical condition that can be reasonably expected to last at least 12 months (unless death intervenes) and to involve either several different organ systems or one organ system severely enough to require specialty pediatric care and probably some period of hospitalization in a tertiary care center.”¹⁶ Whereas, the Agency for Healthcare Research and Quality’s Chronic Condition Indicator (CCI) defines single, non-CCCs (eg, allergic rhinitis).¹⁷

For each diagnosis, we calculated the mean readmission rate for hospitals in each hospital type category. We then calculated an ABC for each diagnosis in each hospital type using a four-step process.^13,18 First, to control for hospitals with small sample sizes, we adjusted all readmission rates using an adjusted performance fraction ([numerator+1]/[denominator +2]), where the numerator is the number of all-cause 30-day readmissions and the denominator is the number of discharges for the selected diagnosis. Then the hospitals were ordered from lowest (best performing) to highest (worst performing) using the adjusted readmission rate. Third, the number of discharges from the best performing hospital to the worst performing hospital was summed until at least 10% of the total discharges had been accounted for. Finally, we computed the ABC as the average of these best performing hospitals. We only report ABCs for which at least three hospitals were included as best performers in the calculation.¹³

To evaluate hospital performance on ABCs for each diagnosis, we identified the percent of hospitals in each setting that were outliers. We defined an outlier as any hospital whose 95% confidence interval for their readmission rate for a given diagnosis did not contain the ABC for their hospital type. All the statistical analyses were performed using SAS version 9.3 (SAS Institute, Inc, Cary, North Carolina).

This project was reviewed by the Cincinnati Children’s Hospital Medical Center Institutional Review Board and determined to be nonhuman subjects research.

Hospital-Type Demographics

The 690,949 discharges from 1,664 hospitals were categorized into 525 metropolitan teaching (550,039 discharges, 79.6% of discharges), 552 metropolitan nonteaching (97,207 discharges, 14% of discharges), and 587 nonmetropolitan hospitals (43,703 discharges, 6.3% of discharges; Table 1). There were significant differences in the patient composition among the three hospital settings. Nonmetropolitan hospitals had a larger percentage of younger patients (aged 0-4 years, P < .001), prominence of first and second quartile median household income, and fewer medically complex patients (48.3% No CCC/No CCI versus 25.5% metropolitan teaching and 33.7% nonteaching, P < .001). Disposition home was over 96% in all three hospital types; however, the metropolitan teaching had a greater percentage of patients discharged to home health versus metropolitan nonteaching and nonmetropolitan hospitals (2.3% versus 0.5%; P < .001).

Readmission Rates

The 17 most common diagnoses based on the number of all-cause 30-day same-hospital readmissions, were categorized into two surgical, seven acute/infectious, four chronic, and four mental health diagnoses (Table 2). Readmission rates varied based on diagnosis and hospital type (Table 2). Overall, mean readmission rates were low, especially in acute respiratory tract related diseases. For chronic diseases, asthma readmissions were consistently low in all three hospital types, whereas sickle cell disease had the highest readmission rate in all three hospital types.

Achievable Benchmarks of Care by Hospital Type

The diagnoses for which ABC could be calculated across all three hospital types included appendectomy and four acute conditions (bronchiolitis, pneumonia, nonbacterial gastroenteritis, and kidney/urinary tract infections). For these conditions, metropolitan teaching hospitals had a more significant percentage of outlier hospitals compared to metropolitan nonteaching and nonmetropolitan hospitals. The percent of outlier hospitals varied by diagnosis and hospital type (Figure).

Metropolitan Teaching

The readmission ABC was calculated for all 17 diagnoses (Table 2). The ABC ranged from 0.4% in acute kidney and urinary tract infection to 7.0% in sickle cell anemia crisis. Bipolar disorder, major depressive disorders and other psychoses, and sickle cell disease (SCD) had the highest percent of outlier hospitals whose mean readmission rates confidence interval did not contain the ABC; tonsil and adenoid procedures and viral illness had the lowest.¹

Metropolitan Nonteaching

The ABC was calculated for 13 of the 17 diagnoses because ABCs were not calculated when there were fewer than three best practicing hospitals. This was the case for tonsil and adenoid procedures, diabetes, seizures, and depression except for major depressive disorder (Table 2). Seven of the 13 diagnoses had an ABC of 0.0%: viral illness, infections of the upper respiratory tract, bronchiolitis, gastroenteritis, hypovolemia and electrolyte disorders, asthma, and childhood behavioral disorders. Like the findings at the metropolitan teaching hospitals, ABCs were lowest for surgical and acute conditions while bipolar disorder, major depressive disorders and other psychoses, and SCD had the highest percent of outlier hospitals with readmission rates beyond the 95% confidence interval of their hospital type’s ABC.

Nonmetropolitan

There was a sufficient number of best practicing hospitals to calculate the ABC for six of the 17 diagnoses (Table 2). For conditions where readmission ABCs could be calculated, they were low: 0.0% for appendectomy, bronchiolitis, gastroenteritis, and seizure; 0.3% for pneumonia; and 1.3% in kidney and urinary tract disorders. None of the conditions with the highest ABCs in other hospital settings (bipolar disease, sickle cell anemia crisis, and major depressive disorders and other psychoses) could be calculated in this setting. Seizure-related readmissions exhibited the most outlier hospitals yet were less than 5%.¹

Among a nationally representative sample of different hospital types that deliver care to children, we report the mean readmission rates and ABCs for 30-day all-cause, same-hospital readmissions for the most commonly readmitted pediatric diagnoses based on hospital type. Previous studies have shown patient variables such as race, ethnicity, and insurance type influencing readmission rates.^19,20 However, hospital type has also been associated with a higher risk of readmission due to the varying complexity of patients at different hospital types.^21,22 Our analyses provide hospital-type specific national estimates of pediatric readmission ABCs for medical and surgical conditions, many less than 1%. While commonly encountered pediatric conditions like asthma and bronchiolitis had low mean readmission rates and ABCs across all hospital types, the mean rates and ABCs for SCD and mental health disorders were much higher with more hospitals performing far from the ABCs.

Diagnoses with a larger percentage of outlier hospitals may represent a national opportunity to improve care for children. Conditions such as SCD and mental illnesses have the highest percentage of hospitals whose readmission rates fall outside of the ABCs in both metropolitan teaching and metropolitan nonteaching hospitals. Hospital performance on SCD and mental health disorders may not reflect deficits in hospital quality or poor adherence to evidence-based best practices, but rather the complex interplay of factors on various levels from government policy and insurance plans, to patient and family resources, to access and availability of medical and mental health specific care. Most importantly, these diseases may represent a significant opportunity for quality improvementin hospitals across the United States.

Sickle cell disease is predominantly a disease among African-Americans, a demographic risk factor for decreased access to care and limited patient and family resources.^23-26 In previous studies evaluating the disparity in readmission rates for Black children with asthma, socioeconomic variables explained 53% of the observed disparity and readmission rates were inversely related to the childhood opportunity index of the patient’s census tract and positively related with geographic social risk.^27,28 Likewise, with SCD affecting a specific demographic and being a chronic disease, best practice policies need to account for the child’s medical needs and include the patient and family resources to ensure access to care and enhanced case management for chronic disease if we aim to improve performance among the outlier hospitals.

Similarly, barriers to care for children with mental illnesses in the United States need attention.^29,30 While there is a paucity of data on the prevalence of mental health disorders in children, one national report estimates that one in 10 American adolescents have depression.^29,31 The American Academy of Pediatrics has developed a policy statement on mental health competencies and a mental health tool-kit for primary care pediatricians; however, no such guidelines or policy statements exist for hospitalized patients with acute or chronic psychiatric conditions.^32,33 Moreover, hospitals are increasingly facing “boarding” of children with acute psychiatric illness in inpatient units and emergency departments.³⁴ The American Medical Association and the American College of Emergency Physicians have expressed concerns regarding the boarding of children with acute psychiatric illness because nonpsychiatric hospitals do not have adequate resources to evaluate, manage, and place these children who deserve appropriate facilities for further management. Coordinated case management and “bundled” discharge planning in other chronic illnesses have shown benefit in cost reduction and readmission.^35-37 Evidence-based practices around pediatric readmissions in other diagnoses should be explored as possible interventions in these conditions.³⁸

There are several limitations to this study. Our data is limited to one calendar year; therefore, admissions in January do not account for potential readmissions from December of the previous year, as patient identifiers do not link across years in the NRD. We also limited our evaluation to the conventional 30-day readmission window, but recent publications may indicate that readmission windows with different timelines could be a more accurate reflection of medically preventable readmissions versus a reflection of social determinants of health leading to readmissions.²⁴ Newborn index admissions were not an allowable index admission; therefore, we may be underreporting readmissions in the neonatal age group. We also chose to include all-cause readmissions, a conventional method to evaluate readmission within an institution, but which may not reflect the quality of care delivered in the index admission. For example, an asthmatic discharged after an acute exacerbation readmitted for dehydration secondary to gastroenteritis may not reflect a lack of quality in asthma inpatient care. Readmissions were limited to the same hospital; therefore, this study cannot account for readmissions at other institutions, which may cause us to underestimate readmission rates. However, end-users of our findings most likely have access only to their own institution’s data. The inclusion of observation status admissions in the database varies from state to state; therefore, this percent of admissions in the database is unknown.

The use of the ABC methodology has some inherent limitations. One hospital with a significant volume diagnosis and low readmission rate within a hospital type may prohibit the reporting of an ABC if less than three hospitals composed the total of the ‘best performing’ hospitals. This was a significant limitation leading to the exclusion of many ABCs in nonmetropolitan institutions. The limitation of calculating and reporting an ABC then prohibits the calculation of outlier hospitals within a hospital type for a given diagnosis. However, when the ABCs are not available, we do provide the mean readmission rate for the diagnosis within the hospital type. While the hospital groupings by population and teaching status for ABCs provide meaningful comparisons for within each hospital setting, it should be noted that there may be vast differences among hospitals within each type (eg, tertiary children’s hospitals compared to teaching hospitals with a pediatric floor in the metropolitan teaching hospital category).^39,40

As healthcare moves from a fee-for-service model to a population-health centered, value-based model, reduction in readmission rates will be more than a quality measure and will have potential financial implications.⁴¹ In the Medicare fee-for-service patients, the Hospital Readmission Reduction Program (HRRP) penalize hospitals with excess readmissions for acute myocardial infarction, heart failure, and pneumonia. The hospitals subject to penalties in the HRRP had greater reduction in readmission rates in the targeted, and even nontargeted conditions, compared with hospitals not subject to penalties.⁴² Similarly, we believe that our data on low readmission rates and ABCs for conditions such as asthma, bronchiolitis, and appendicitis could represent decades of quality improvement work for the most common pediatric conditions among hospitalized children. Sickle cell disease and mental health problems remain as outliers and merit further attention. To move to a true population-health model, hospitals will need to explore outlier conditions including evaluating patient-level readmission patterns across institutions. This moves readmission from a hospital quality measure to a patient-centric quality measure, and perhaps will provide value to the patient and the healthcare system alike.

The readmission ABCs for the most commonly readmitted pediatric diagnoses are low, regardless of the hospital setting. The highest pediatric readmission rates in SCD, bipolar disorders, and major depressive disorder were lower than the most common adult readmission diagnoses. However, mental health conditions and SCD remain as outliers for pediatric readmissions, burden hospital systems, and perhaps warrant national-level attention. The ABCs stratified by hospital type in this study facilitate comparisons and identify opportunities for population-level interventions to meaningfully improve patient care.

Disclosures

The authors have nothing to disclose.

Hospital readmission rates are a common metric for defining, evaluating, and benchmarking quality of care. The Centers for Medicare and Medicaid Services (CMS) publicly report hospital readmission rates for common adult conditions and reduces payments to hospitals with excessive readmissions.¹ Recently, the focus on pediatric readmission rates has increased and the National Quality Forum (NQF) has endorsed at least two pediatric readmission-specific quality indicators which could be used by public and private payers in pay-for-performance programs aimed at institutions caring for children.² While preventability of readmissions and their value as a marker of quality remains debated, their acceptance by the NQF and CMS has led public and private payers to propose readmission-related penalties for hospitals caring for children. ^3-5

All-cause 30-day same-hospital readmission rates for pediatric conditions are half of the adult readmission rates, around 6% in most studies, compared to 12% in adults.^6,7 The lower rates of pediatric readmissions makes it difficult to only use mean readmission rates to stratify hospitals into high- or low-performers and set target goals for improvement.⁸ While adult readmissions have been studied in depth, there are no consistent measures used to benchmark pediatric readmissions across hospital types.

Given the emphasis placed on readmissions, it is essential to understand patterns in pediatric readmission rates to determine optimal and achievable targets for improvement. Achievable Benchmarks of Care (ABCs) are one approach to understanding readmission rates and have an advantage over using mean or medians in performance improvement as they can stratify performance for conditions with low readmission rates and low volumes.⁹ When creating benchmarks, it is important that hospitals performance is evaluated among peer hospitals with similar patient populations, not just a cumulative average from all hospital types which may punish hospitals with a more complex patient case mix.¹⁰ The goal of this study was to calculate the readmission rates and the ABCs for common pediatric diagnoses by hospital type to identify priority conditions for quality improvement efforts using a previously published methodology.^11-13

Data Source

We conducted a retrospective analysis of patients less than 18 years of age in the Healthcare Utilization Project 2014 Nationwide Readmissions Database (NRD). The NRD includes public hospitals; academic medical centers; and specialty hospitals in obstetrics and gynecology, otolaryngology, orthopedics, and cancer; and pediatric, public, and academic medical hospitals. Excluded are long-term care facilities such as rehabilitation, long-term acute care, psychiatric, alcoholism, and chemical dependency hospitals. The readmissions data contains information from hospitals grouped by region, population census, and teaching status.¹⁴ Three hospital type classifications used in this study were metropolitan teaching hospitals, metropolitan nonteaching hospitals, and nonmetropolitan hospitals. These three hospital type classifications follow the reporting format in the NRD.

Study Population

Patients less than 18 years old were included if they were discharged from January 1, 2014 through November 30, 2014 and had a readmission to the index hospital within 30 days. We limited inclusion to discharges through November 30 so we could identify patients with a 30-day readmission as patient identifiers do not link across years in the NRD.

Exposure

We included 30-day, all-cause, same-hospital readmissions to the index acute care hospital, excluding labor and delivery, normal newborn care, chemotherapy, transfers, and mortalities. Intrahospital discharge and admissions within the same hospital system were not defined as a readmission, but rather as a “same-day event.”¹⁵ For example, institutions with inpatient mental health facilities, medical unit discharges and admission to the mental health unit were not identified as a readmission in this dataset.

Outcome

For each hospital type, we measured same-hospital, all-cause, 30-day readmission rates and achievable benchmark of care for the 17 most commonly readmitted pediatric discharge diagnoses. To identify the target readmission diagnoses and all-cause, 30-day readmissions based on their index hospitalizations, All-Patient Refined Diagnosis-Related Groups (APR-DRG), version 25 (3M Health Information Systems, Salt Lake City, Utah) were ordered by frequency for each hospital type. The 20 most common APR-DRGs were the same across all hospital types. The authors then evaluated these 20 APR-DRGs for clinical consistency of included diagnoses identified by the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes within each APR-DRG. Three diagnosis-related groups were excluded from the analysis (major hematologic/immunologic disease except for sickle cell, other anemia and disorders of blood and blood forming organs, and other digestive system diagnoses) due to the heterogeneity of the diagnoses identified by the ICD-9-CM codes within each APR-DRG. We refer to each APR-DRG as a “diagnosis” throughout the article.

Analysis

The demographic characteristics of the patients seen at the three hospital types were summarized using frequencies and percentages. Reports were generated for patient age, gender, payer source, patient residence, median household income, patient complexity, and discharge disposition. Patient complexity was defined using complex chronic condition (CCC) and the number of chronic conditions (CCI).^16,17 As previously defined in the literature, a complex chronic condition is “any medical condition that can be reasonably expected to last at least 12 months (unless death intervenes) and to involve either several different organ systems or one organ system severely enough to require specialty pediatric care and probably some period of hospitalization in a tertiary care center.”¹⁶ Whereas, the Agency for Healthcare Research and Quality’s Chronic Condition Indicator (CCI) defines single, non-CCCs (eg, allergic rhinitis).¹⁷

For each diagnosis, we calculated the mean readmission rate for hospitals in each hospital type category. We then calculated an ABC for each diagnosis in each hospital type using a four-step process.^13,18 First, to control for hospitals with small sample sizes, we adjusted all readmission rates using an adjusted performance fraction ([numerator+1]/[denominator +2]), where the numerator is the number of all-cause 30-day readmissions and the denominator is the number of discharges for the selected diagnosis. Then the hospitals were ordered from lowest (best performing) to highest (worst performing) using the adjusted readmission rate. Third, the number of discharges from the best performing hospital to the worst performing hospital was summed until at least 10% of the total discharges had been accounted for. Finally, we computed the ABC as the average of these best performing hospitals. We only report ABCs for which at least three hospitals were included as best performers in the calculation.¹³

To evaluate hospital performance on ABCs for each diagnosis, we identified the percent of hospitals in each setting that were outliers. We defined an outlier as any hospital whose 95% confidence interval for their readmission rate for a given diagnosis did not contain the ABC for their hospital type. All the statistical analyses were performed using SAS version 9.3 (SAS Institute, Inc, Cary, North Carolina).

This project was reviewed by the Cincinnati Children’s Hospital Medical Center Institutional Review Board and determined to be nonhuman subjects research.

Hospital-Type Demographics

The 690,949 discharges from 1,664 hospitals were categorized into 525 metropolitan teaching (550,039 discharges, 79.6% of discharges), 552 metropolitan nonteaching (97,207 discharges, 14% of discharges), and 587 nonmetropolitan hospitals (43,703 discharges, 6.3% of discharges; Table 1). There were significant differences in the patient composition among the three hospital settings. Nonmetropolitan hospitals had a larger percentage of younger patients (aged 0-4 years, P < .001), prominence of first and second quartile median household income, and fewer medically complex patients (48.3% No CCC/No CCI versus 25.5% metropolitan teaching and 33.7% nonteaching, P < .001). Disposition home was over 96% in all three hospital types; however, the metropolitan teaching had a greater percentage of patients discharged to home health versus metropolitan nonteaching and nonmetropolitan hospitals (2.3% versus 0.5%; P < .001).

Readmission Rates

The 17 most common diagnoses based on the number of all-cause 30-day same-hospital readmissions, were categorized into two surgical, seven acute/infectious, four chronic, and four mental health diagnoses (Table 2). Readmission rates varied based on diagnosis and hospital type (Table 2). Overall, mean readmission rates were low, especially in acute respiratory tract related diseases. For chronic diseases, asthma readmissions were consistently low in all three hospital types, whereas sickle cell disease had the highest readmission rate in all three hospital types.

Achievable Benchmarks of Care by Hospital Type

The diagnoses for which ABC could be calculated across all three hospital types included appendectomy and four acute conditions (bronchiolitis, pneumonia, nonbacterial gastroenteritis, and kidney/urinary tract infections). For these conditions, metropolitan teaching hospitals had a more significant percentage of outlier hospitals compared to metropolitan nonteaching and nonmetropolitan hospitals. The percent of outlier hospitals varied by diagnosis and hospital type (Figure).

Metropolitan Teaching

The readmission ABC was calculated for all 17 diagnoses (Table 2). The ABC ranged from 0.4% in acute kidney and urinary tract infection to 7.0% in sickle cell anemia crisis. Bipolar disorder, major depressive disorders and other psychoses, and sickle cell disease (SCD) had the highest percent of outlier hospitals whose mean readmission rates confidence interval did not contain the ABC; tonsil and adenoid procedures and viral illness had the lowest.¹

Metropolitan Nonteaching

The ABC was calculated for 13 of the 17 diagnoses because ABCs were not calculated when there were fewer than three best practicing hospitals. This was the case for tonsil and adenoid procedures, diabetes, seizures, and depression except for major depressive disorder (Table 2). Seven of the 13 diagnoses had an ABC of 0.0%: viral illness, infections of the upper respiratory tract, bronchiolitis, gastroenteritis, hypovolemia and electrolyte disorders, asthma, and childhood behavioral disorders. Like the findings at the metropolitan teaching hospitals, ABCs were lowest for surgical and acute conditions while bipolar disorder, major depressive disorders and other psychoses, and SCD had the highest percent of outlier hospitals with readmission rates beyond the 95% confidence interval of their hospital type’s ABC.

Nonmetropolitan

There was a sufficient number of best practicing hospitals to calculate the ABC for six of the 17 diagnoses (Table 2). For conditions where readmission ABCs could be calculated, they were low: 0.0% for appendectomy, bronchiolitis, gastroenteritis, and seizure; 0.3% for pneumonia; and 1.3% in kidney and urinary tract disorders. None of the conditions with the highest ABCs in other hospital settings (bipolar disease, sickle cell anemia crisis, and major depressive disorders and other psychoses) could be calculated in this setting. Seizure-related readmissions exhibited the most outlier hospitals yet were less than 5%.¹

Among a nationally representative sample of different hospital types that deliver care to children, we report the mean readmission rates and ABCs for 30-day all-cause, same-hospital readmissions for the most commonly readmitted pediatric diagnoses based on hospital type. Previous studies have shown patient variables such as race, ethnicity, and insurance type influencing readmission rates.^19,20 However, hospital type has also been associated with a higher risk of readmission due to the varying complexity of patients at different hospital types.^21,22 Our analyses provide hospital-type specific national estimates of pediatric readmission ABCs for medical and surgical conditions, many less than 1%. While commonly encountered pediatric conditions like asthma and bronchiolitis had low mean readmission rates and ABCs across all hospital types, the mean rates and ABCs for SCD and mental health disorders were much higher with more hospitals performing far from the ABCs.

Diagnoses with a larger percentage of outlier hospitals may represent a national opportunity to improve care for children. Conditions such as SCD and mental illnesses have the highest percentage of hospitals whose readmission rates fall outside of the ABCs in both metropolitan teaching and metropolitan nonteaching hospitals. Hospital performance on SCD and mental health disorders may not reflect deficits in hospital quality or poor adherence to evidence-based best practices, but rather the complex interplay of factors on various levels from government policy and insurance plans, to patient and family resources, to access and availability of medical and mental health specific care. Most importantly, these diseases may represent a significant opportunity for quality improvementin hospitals across the United States.

Sickle cell disease is predominantly a disease among African-Americans, a demographic risk factor for decreased access to care and limited patient and family resources.^23-26 In previous studies evaluating the disparity in readmission rates for Black children with asthma, socioeconomic variables explained 53% of the observed disparity and readmission rates were inversely related to the childhood opportunity index of the patient’s census tract and positively related with geographic social risk.^27,28 Likewise, with SCD affecting a specific demographic and being a chronic disease, best practice policies need to account for the child’s medical needs and include the patient and family resources to ensure access to care and enhanced case management for chronic disease if we aim to improve performance among the outlier hospitals.

Similarly, barriers to care for children with mental illnesses in the United States need attention.^29,30 While there is a paucity of data on the prevalence of mental health disorders in children, one national report estimates that one in 10 American adolescents have depression.^29,31 The American Academy of Pediatrics has developed a policy statement on mental health competencies and a mental health tool-kit for primary care pediatricians; however, no such guidelines or policy statements exist for hospitalized patients with acute or chronic psychiatric conditions.^32,33 Moreover, hospitals are increasingly facing “boarding” of children with acute psychiatric illness in inpatient units and emergency departments.³⁴ The American Medical Association and the American College of Emergency Physicians have expressed concerns regarding the boarding of children with acute psychiatric illness because nonpsychiatric hospitals do not have adequate resources to evaluate, manage, and place these children who deserve appropriate facilities for further management. Coordinated case management and “bundled” discharge planning in other chronic illnesses have shown benefit in cost reduction and readmission.^35-37 Evidence-based practices around pediatric readmissions in other diagnoses should be explored as possible interventions in these conditions.³⁸

There are several limitations to this study. Our data is limited to one calendar year; therefore, admissions in January do not account for potential readmissions from December of the previous year, as patient identifiers do not link across years in the NRD. We also limited our evaluation to the conventional 30-day readmission window, but recent publications may indicate that readmission windows with different timelines could be a more accurate reflection of medically preventable readmissions versus a reflection of social determinants of health leading to readmissions.²⁴ Newborn index admissions were not an allowable index admission; therefore, we may be underreporting readmissions in the neonatal age group. We also chose to include all-cause readmissions, a conventional method to evaluate readmission within an institution, but which may not reflect the quality of care delivered in the index admission. For example, an asthmatic discharged after an acute exacerbation readmitted for dehydration secondary to gastroenteritis may not reflect a lack of quality in asthma inpatient care. Readmissions were limited to the same hospital; therefore, this study cannot account for readmissions at other institutions, which may cause us to underestimate readmission rates. However, end-users of our findings most likely have access only to their own institution’s data. The inclusion of observation status admissions in the database varies from state to state; therefore, this percent of admissions in the database is unknown.

The use of the ABC methodology has some inherent limitations. One hospital with a significant volume diagnosis and low readmission rate within a hospital type may prohibit the reporting of an ABC if less than three hospitals composed the total of the ‘best performing’ hospitals. This was a significant limitation leading to the exclusion of many ABCs in nonmetropolitan institutions. The limitation of calculating and reporting an ABC then prohibits the calculation of outlier hospitals within a hospital type for a given diagnosis. However, when the ABCs are not available, we do provide the mean readmission rate for the diagnosis within the hospital type. While the hospital groupings by population and teaching status for ABCs provide meaningful comparisons for within each hospital setting, it should be noted that there may be vast differences among hospitals within each type (eg, tertiary children’s hospitals compared to teaching hospitals with a pediatric floor in the metropolitan teaching hospital category).^39,40

As healthcare moves from a fee-for-service model to a population-health centered, value-based model, reduction in readmission rates will be more than a quality measure and will have potential financial implications.⁴¹ In the Medicare fee-for-service patients, the Hospital Readmission Reduction Program (HRRP) penalize hospitals with excess readmissions for acute myocardial infarction, heart failure, and pneumonia. The hospitals subject to penalties in the HRRP had greater reduction in readmission rates in the targeted, and even nontargeted conditions, compared with hospitals not subject to penalties.⁴² Similarly, we believe that our data on low readmission rates and ABCs for conditions such as asthma, bronchiolitis, and appendicitis could represent decades of quality improvement work for the most common pediatric conditions among hospitalized children. Sickle cell disease and mental health problems remain as outliers and merit further attention. To move to a true population-health model, hospitals will need to explore outlier conditions including evaluating patient-level readmission patterns across institutions. This moves readmission from a hospital quality measure to a patient-centric quality measure, and perhaps will provide value to the patient and the healthcare system alike.

The readmission ABCs for the most commonly readmitted pediatric diagnoses are low, regardless of the hospital setting. The highest pediatric readmission rates in SCD, bipolar disorders, and major depressive disorder were lower than the most common adult readmission diagnoses. However, mental health conditions and SCD remain as outliers for pediatric readmissions, burden hospital systems, and perhaps warrant national-level attention. The ABCs stratified by hospital type in this study facilitate comparisons and identify opportunities for population-level interventions to meaningfully improve patient care.

Disclosures

The authors have nothing to disclose.

Hospital readmission rates are a common metric for defining, evaluating, and benchmarking quality of care. The Centers for Medicare and Medicaid Services (CMS) publicly report hospital readmission rates for common adult conditions and reduces payments to hospitals with excessive readmissions.¹ Recently, the focus on pediatric readmission rates has increased and the National Quality Forum (NQF) has endorsed at least two pediatric readmission-specific quality indicators which could be used by public and private payers in pay-for-performance programs aimed at institutions caring for children.² While preventability of readmissions and their value as a marker of quality remains debated, their acceptance by the NQF and CMS has led public and private payers to propose readmission-related penalties for hospitals caring for children. ^3-5

All-cause 30-day same-hospital readmission rates for pediatric conditions are half of the adult readmission rates, around 6% in most studies, compared to 12% in adults.^6,7 The lower rates of pediatric readmissions makes it difficult to only use mean readmission rates to stratify hospitals into high- or low-performers and set target goals for improvement.⁸ While adult readmissions have been studied in depth, there are no consistent measures used to benchmark pediatric readmissions across hospital types.

Given the emphasis placed on readmissions, it is essential to understand patterns in pediatric readmission rates to determine optimal and achievable targets for improvement. Achievable Benchmarks of Care (ABCs) are one approach to understanding readmission rates and have an advantage over using mean or medians in performance improvement as they can stratify performance for conditions with low readmission rates and low volumes.⁹ When creating benchmarks, it is important that hospitals performance is evaluated among peer hospitals with similar patient populations, not just a cumulative average from all hospital types which may punish hospitals with a more complex patient case mix.¹⁰ The goal of this study was to calculate the readmission rates and the ABCs for common pediatric diagnoses by hospital type to identify priority conditions for quality improvement efforts using a previously published methodology.^11-13

Data Source

We conducted a retrospective analysis of patients less than 18 years of age in the Healthcare Utilization Project 2014 Nationwide Readmissions Database (NRD). The NRD includes public hospitals; academic medical centers; and specialty hospitals in obstetrics and gynecology, otolaryngology, orthopedics, and cancer; and pediatric, public, and academic medical hospitals. Excluded are long-term care facilities such as rehabilitation, long-term acute care, psychiatric, alcoholism, and chemical dependency hospitals. The readmissions data contains information from hospitals grouped by region, population census, and teaching status.¹⁴ Three hospital type classifications used in this study were metropolitan teaching hospitals, metropolitan nonteaching hospitals, and nonmetropolitan hospitals. These three hospital type classifications follow the reporting format in the NRD.

Study Population

Patients less than 18 years old were included if they were discharged from January 1, 2014 through November 30, 2014 and had a readmission to the index hospital within 30 days. We limited inclusion to discharges through November 30 so we could identify patients with a 30-day readmission as patient identifiers do not link across years in the NRD.

Exposure

We included 30-day, all-cause, same-hospital readmissions to the index acute care hospital, excluding labor and delivery, normal newborn care, chemotherapy, transfers, and mortalities. Intrahospital discharge and admissions within the same hospital system were not defined as a readmission, but rather as a “same-day event.”¹⁵ For example, institutions with inpatient mental health facilities, medical unit discharges and admission to the mental health unit were not identified as a readmission in this dataset.

Outcome

For each hospital type, we measured same-hospital, all-cause, 30-day readmission rates and achievable benchmark of care for the 17 most commonly readmitted pediatric discharge diagnoses. To identify the target readmission diagnoses and all-cause, 30-day readmissions based on their index hospitalizations, All-Patient Refined Diagnosis-Related Groups (APR-DRG), version 25 (3M Health Information Systems, Salt Lake City, Utah) were ordered by frequency for each hospital type. The 20 most common APR-DRGs were the same across all hospital types. The authors then evaluated these 20 APR-DRGs for clinical consistency of included diagnoses identified by the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes within each APR-DRG. Three diagnosis-related groups were excluded from the analysis (major hematologic/immunologic disease except for sickle cell, other anemia and disorders of blood and blood forming organs, and other digestive system diagnoses) due to the heterogeneity of the diagnoses identified by the ICD-9-CM codes within each APR-DRG. We refer to each APR-DRG as a “diagnosis” throughout the article.

Analysis

The demographic characteristics of the patients seen at the three hospital types were summarized using frequencies and percentages. Reports were generated for patient age, gender, payer source, patient residence, median household income, patient complexity, and discharge disposition. Patient complexity was defined using complex chronic condition (CCC) and the number of chronic conditions (CCI).^16,17 As previously defined in the literature, a complex chronic condition is “any medical condition that can be reasonably expected to last at least 12 months (unless death intervenes) and to involve either several different organ systems or one organ system severely enough to require specialty pediatric care and probably some period of hospitalization in a tertiary care center.”¹⁶ Whereas, the Agency for Healthcare Research and Quality’s Chronic Condition Indicator (CCI) defines single, non-CCCs (eg, allergic rhinitis).¹⁷

For each diagnosis, we calculated the mean readmission rate for hospitals in each hospital type category. We then calculated an ABC for each diagnosis in each hospital type using a four-step process.^13,18 First, to control for hospitals with small sample sizes, we adjusted all readmission rates using an adjusted performance fraction ([numerator+1]/[denominator +2]), where the numerator is the number of all-cause 30-day readmissions and the denominator is the number of discharges for the selected diagnosis. Then the hospitals were ordered from lowest (best performing) to highest (worst performing) using the adjusted readmission rate. Third, the number of discharges from the best performing hospital to the worst performing hospital was summed until at least 10% of the total discharges had been accounted for. Finally, we computed the ABC as the average of these best performing hospitals. We only report ABCs for which at least three hospitals were included as best performers in the calculation.¹³

To evaluate hospital performance on ABCs for each diagnosis, we identified the percent of hospitals in each setting that were outliers. We defined an outlier as any hospital whose 95% confidence interval for their readmission rate for a given diagnosis did not contain the ABC for their hospital type. All the statistical analyses were performed using SAS version 9.3 (SAS Institute, Inc, Cary, North Carolina).

This project was reviewed by the Cincinnati Children’s Hospital Medical Center Institutional Review Board and determined to be nonhuman subjects research.

Hospital-Type Demographics

The 690,949 discharges from 1,664 hospitals were categorized into 525 metropolitan teaching (550,039 discharges, 79.6% of discharges), 552 metropolitan nonteaching (97,207 discharges, 14% of discharges), and 587 nonmetropolitan hospitals (43,703 discharges, 6.3% of discharges; Table 1). There were significant differences in the patient composition among the three hospital settings. Nonmetropolitan hospitals had a larger percentage of younger patients (aged 0-4 years, P < .001), prominence of first and second quartile median household income, and fewer medically complex patients (48.3% No CCC/No CCI versus 25.5% metropolitan teaching and 33.7% nonteaching, P < .001). Disposition home was over 96% in all three hospital types; however, the metropolitan teaching had a greater percentage of patients discharged to home health versus metropolitan nonteaching and nonmetropolitan hospitals (2.3% versus 0.5%; P < .001).

Readmission Rates

The 17 most common diagnoses based on the number of all-cause 30-day same-hospital readmissions, were categorized into two surgical, seven acute/infectious, four chronic, and four mental health diagnoses (Table 2). Readmission rates varied based on diagnosis and hospital type (Table 2). Overall, mean readmission rates were low, especially in acute respiratory tract related diseases. For chronic diseases, asthma readmissions were consistently low in all three hospital types, whereas sickle cell disease had the highest readmission rate in all three hospital types.

Achievable Benchmarks of Care by Hospital Type

The diagnoses for which ABC could be calculated across all three hospital types included appendectomy and four acute conditions (bronchiolitis, pneumonia, nonbacterial gastroenteritis, and kidney/urinary tract infections). For these conditions, metropolitan teaching hospitals had a more significant percentage of outlier hospitals compared to metropolitan nonteaching and nonmetropolitan hospitals. The percent of outlier hospitals varied by diagnosis and hospital type (Figure).

Metropolitan Teaching

The readmission ABC was calculated for all 17 diagnoses (Table 2). The ABC ranged from 0.4% in acute kidney and urinary tract infection to 7.0% in sickle cell anemia crisis. Bipolar disorder, major depressive disorders and other psychoses, and sickle cell disease (SCD) had the highest percent of outlier hospitals whose mean readmission rates confidence interval did not contain the ABC; tonsil and adenoid procedures and viral illness had the lowest.¹

Metropolitan Nonteaching

The ABC was calculated for 13 of the 17 diagnoses because ABCs were not calculated when there were fewer than three best practicing hospitals. This was the case for tonsil and adenoid procedures, diabetes, seizures, and depression except for major depressive disorder (Table 2). Seven of the 13 diagnoses had an ABC of 0.0%: viral illness, infections of the upper respiratory tract, bronchiolitis, gastroenteritis, hypovolemia and electrolyte disorders, asthma, and childhood behavioral disorders. Like the findings at the metropolitan teaching hospitals, ABCs were lowest for surgical and acute conditions while bipolar disorder, major depressive disorders and other psychoses, and SCD had the highest percent of outlier hospitals with readmission rates beyond the 95% confidence interval of their hospital type’s ABC.

Nonmetropolitan

There was a sufficient number of best practicing hospitals to calculate the ABC for six of the 17 diagnoses (Table 2). For conditions where readmission ABCs could be calculated, they were low: 0.0% for appendectomy, bronchiolitis, gastroenteritis, and seizure; 0.3% for pneumonia; and 1.3% in kidney and urinary tract disorders. None of the conditions with the highest ABCs in other hospital settings (bipolar disease, sickle cell anemia crisis, and major depressive disorders and other psychoses) could be calculated in this setting. Seizure-related readmissions exhibited the most outlier hospitals yet were less than 5%.¹

Among a nationally representative sample of different hospital types that deliver care to children, we report the mean readmission rates and ABCs for 30-day all-cause, same-hospital readmissions for the most commonly readmitted pediatric diagnoses based on hospital type. Previous studies have shown patient variables such as race, ethnicity, and insurance type influencing readmission rates.^19,20 However, hospital type has also been associated with a higher risk of readmission due to the varying complexity of patients at different hospital types.^21,22 Our analyses provide hospital-type specific national estimates of pediatric readmission ABCs for medical and surgical conditions, many less than 1%. While commonly encountered pediatric conditions like asthma and bronchiolitis had low mean readmission rates and ABCs across all hospital types, the mean rates and ABCs for SCD and mental health disorders were much higher with more hospitals performing far from the ABCs.

Diagnoses with a larger percentage of outlier hospitals may represent a national opportunity to improve care for children. Conditions such as SCD and mental illnesses have the highest percentage of hospitals whose readmission rates fall outside of the ABCs in both metropolitan teaching and metropolitan nonteaching hospitals. Hospital performance on SCD and mental health disorders may not reflect deficits in hospital quality or poor adherence to evidence-based best practices, but rather the complex interplay of factors on various levels from government policy and insurance plans, to patient and family resources, to access and availability of medical and mental health specific care. Most importantly, these diseases may represent a significant opportunity for quality improvementin hospitals across the United States.

Sickle cell disease is predominantly a disease among African-Americans, a demographic risk factor for decreased access to care and limited patient and family resources.^23-26 In previous studies evaluating the disparity in readmission rates for Black children with asthma, socioeconomic variables explained 53% of the observed disparity and readmission rates were inversely related to the childhood opportunity index of the patient’s census tract and positively related with geographic social risk.^27,28 Likewise, with SCD affecting a specific demographic and being a chronic disease, best practice policies need to account for the child’s medical needs and include the patient and family resources to ensure access to care and enhanced case management for chronic disease if we aim to improve performance among the outlier hospitals.

Similarly, barriers to care for children with mental illnesses in the United States need attention.^29,30 While there is a paucity of data on the prevalence of mental health disorders in children, one national report estimates that one in 10 American adolescents have depression.^29,31 The American Academy of Pediatrics has developed a policy statement on mental health competencies and a mental health tool-kit for primary care pediatricians; however, no such guidelines or policy statements exist for hospitalized patients with acute or chronic psychiatric conditions.^32,33 Moreover, hospitals are increasingly facing “boarding” of children with acute psychiatric illness in inpatient units and emergency departments.³⁴ The American Medical Association and the American College of Emergency Physicians have expressed concerns regarding the boarding of children with acute psychiatric illness because nonpsychiatric hospitals do not have adequate resources to evaluate, manage, and place these children who deserve appropriate facilities for further management. Coordinated case management and “bundled” discharge planning in other chronic illnesses have shown benefit in cost reduction and readmission.^35-37 Evidence-based practices around pediatric readmissions in other diagnoses should be explored as possible interventions in these conditions.³⁸

There are several limitations to this study. Our data is limited to one calendar year; therefore, admissions in January do not account for potential readmissions from December of the previous year, as patient identifiers do not link across years in the NRD. We also limited our evaluation to the conventional 30-day readmission window, but recent publications may indicate that readmission windows with different timelines could be a more accurate reflection of medically preventable readmissions versus a reflection of social determinants of health leading to readmissions.²⁴ Newborn index admissions were not an allowable index admission; therefore, we may be underreporting readmissions in the neonatal age group. We also chose to include all-cause readmissions, a conventional method to evaluate readmission within an institution, but which may not reflect the quality of care delivered in the index admission. For example, an asthmatic discharged after an acute exacerbation readmitted for dehydration secondary to gastroenteritis may not reflect a lack of quality in asthma inpatient care. Readmissions were limited to the same hospital; therefore, this study cannot account for readmissions at other institutions, which may cause us to underestimate readmission rates. However, end-users of our findings most likely have access only to their own institution’s data. The inclusion of observation status admissions in the database varies from state to state; therefore, this percent of admissions in the database is unknown.

The use of the ABC methodology has some inherent limitations. One hospital with a significant volume diagnosis and low readmission rate within a hospital type may prohibit the reporting of an ABC if less than three hospitals composed the total of the ‘best performing’ hospitals. This was a significant limitation leading to the exclusion of many ABCs in nonmetropolitan institutions. The limitation of calculating and reporting an ABC then prohibits the calculation of outlier hospitals within a hospital type for a given diagnosis. However, when the ABCs are not available, we do provide the mean readmission rate for the diagnosis within the hospital type. While the hospital groupings by population and teaching status for ABCs provide meaningful comparisons for within each hospital setting, it should be noted that there may be vast differences among hospitals within each type (eg, tertiary children’s hospitals compared to teaching hospitals with a pediatric floor in the metropolitan teaching hospital category).^39,40

As healthcare moves from a fee-for-service model to a population-health centered, value-based model, reduction in readmission rates will be more than a quality measure and will have potential financial implications.⁴¹ In the Medicare fee-for-service patients, the Hospital Readmission Reduction Program (HRRP) penalize hospitals with excess readmissions for acute myocardial infarction, heart failure, and pneumonia. The hospitals subject to penalties in the HRRP had greater reduction in readmission rates in the targeted, and even nontargeted conditions, compared with hospitals not subject to penalties.⁴² Similarly, we believe that our data on low readmission rates and ABCs for conditions such as asthma, bronchiolitis, and appendicitis could represent decades of quality improvement work for the most common pediatric conditions among hospitalized children. Sickle cell disease and mental health problems remain as outliers and merit further attention. To move to a true population-health model, hospitals will need to explore outlier conditions including evaluating patient-level readmission patterns across institutions. This moves readmission from a hospital quality measure to a patient-centric quality measure, and perhaps will provide value to the patient and the healthcare system alike.

The readmission ABCs for the most commonly readmitted pediatric diagnoses are low, regardless of the hospital setting. The highest pediatric readmission rates in SCD, bipolar disorders, and major depressive disorder were lower than the most common adult readmission diagnoses. However, mental health conditions and SCD remain as outliers for pediatric readmissions, burden hospital systems, and perhaps warrant national-level attention. The ABCs stratified by hospital type in this study facilitate comparisons and identify opportunities for population-level interventions to meaningfully improve patient care.

Disclosures

The authors have nothing to disclose.

In 2016, 15.2% of older Americans were hospitalized compared with 7% of the overall population and their length of stay (LOS) was 0.7 days greater.¹ Geriatric hospitalizations frequently result in complications, functional decline, nursing home transfers, and increased cost.^2-4 This pattern of decline has been termed “hospitalitis” or dysfunctional syndrome.^5,6 Hospitals need data-driven approaches to improve outcomes for elders. The Acute Care for Elders (ACE) program, which has been in existence for roughly 25 years, is one such model. ACE features include an environment prepared for older adults, patient-centered care to prevent functional and cognitive decline, frequent medical review to prevent iatrogenic injury or new geriatric syndromes, and early discharge and rehabilitation planning to maximize the likelihood of return to the community.⁷ Although published data vary somewhat, ACE programs have robust evidence documenting improved safety, quality, and value.^8-15 A recent meta-analysis found that ACE programs decrease LOS, costs, new nursing home discharges, falls, delirium, and functional decline.¹⁶ However, of the 13 ACE trials reported to date, only five were published in the last decade. Recent rising pressure to decrease hospitalizations and reduce LOS has shifted some care to other settings and it is unclear whether the same results would persist in today’s rapid-paced hospitals.

ACE programs require enhanced resources and restructured care processes but there is a notable lack of data to guide patient selection. Admission criteria vary among the published reports, and information on whether comorbidity burden impacts the magnitude of benefit is scarce. One ACE investigator commented, “We were not able to identify a subgroup of patients who were most likely to benefit.”¹⁷ Not all hospitalized older adults can receive ACE care, and some units have closed due to financial and logistic pressures; thus, criteria to target this scarce resource are urgently needed. Our hospital implemented an ACE program in 2014 and we have measured and internally benchmarked important quality improvement metrics. Using this data, we conducted an exploratory analysis to generate hypotheses on the differential impact across the spectrum of cost, LOS, 30-day readmissions, and variations across quartiles of comorbidity severity.

Setting and Patients

In September 2014, our 716-bed teaching hospital in Springfield, Massachusetts launched an ACE program to improve care for older adults on a single medical unit. The program succeeded in engaging the senior leadership, and geriatrics was identified as a priority in Baystate’s 5-year strategic plan. ACE patients ≥70 years were admitted from the emergency department with inpatient status. Patients transferred from other units or with advanced dementia or nearing death were excluded. Core components of the ACE program were derived from published summaries (see supplementary material).^7,16

Interprofessional ”ACE Rounds”

Interprofessional ACE Rounds occurred every weekday. As one ACE analyst has noted, “the interdisciplinary team…ensures that the multifactorial nature of functional decline is met with a multicomponent plan to prevent it.”¹⁸ Rounds participants shifted over time but always included a geriatrics physician assistant (PA) or geriatrician (team leader), a pharmacist, staff nurses, and a chaplain. The nurse educator, dietician, research assistant, and patient advocate/volunteers attended intermittently. Before rounds, the PA reviewed the admission notes for new ACE patients. Initially, rounds were lengthy and included nurse coaching. Later, nurses’ presentations were structured by the SPICES tool (Sleep, Problems with eating/feeding, Incontinence, Confusion, Evidence of falls, Skin Breakdown)¹⁹ and tracking and reporting templates. Coaching and education, along with conversations that did not require the full team, were removed from rounds. Thus, the time required for rounds declined from about 75 minutes to 35 minutes, which allowed more patients to be discussed efficiently. This change was critical as the number of ACE patients rose following the shift to the larger unit. The pharmacist reviewed medications focusing on potentially inappropriate drugs. Following rounds, the nurses and pharmacist conveyed recommendations to the hospitalists.

Patient-Centered Activities to Prevent Functional and Cognitive Decline

Project leaders coached staff about the importance of mobility, sleep, and delirium prevention and identification. The nurses screened patients using the Confusion Assessment Method (CAM) and reported delirium promptly. Specific care sets for ACE patients were implemented (see supplementary material).

The project was enhanced by several palliative care components, ie tracking pain, noting psychiatric symptoms, and considering prognosis by posing the “Surprise Question” during rounds.²⁰ (“Would you be surprised if this patient died in the next year?”). As far as staffing and logistics allowed, the goals of care conversation were held by a geriatrics PA with patients/families who “screened in.”

Prepared Environment

The ACE program’s unit was remodeled to facilitate physical and cognitive functioning and promote sleep at night (quiet hours: 10 PM-6 AM).

In accordance with quality improvement processes, iterative shifts were implemented over time in terms of checklist, presentation format, timing, and team participation. In December 2016, the program relocated to a unit with 34 ACE beds and 5 end-of-life beds; this move markedly increased the number of eligible ACE patients.

Study Design, Data Source, and Patients

Since we were implementing and measuring our ACE program with a quality improvement lens, we chose a descriptive cross-sectional study design to generate hypotheses regarding our program’s impact compared to usual care. Using a hospital-wide billing database (McKesson Performance Analytics, v19, Alpharetta, Georgia) we sampled inpatients aged >70 years with a medical Diagnosis Related Group (DRGs) admitted through the emergency department and discharged from a medical unit from September 22, 2014 to August 31, 2017. These criteria mirrored those in the ACE unit. Older adults requiring specialized care (eg, those with myocardial infarct) were excluded, as were those with billing codes for mechanical ventilation, admission to critical care units, or discharge to hospice. Because one of our outcomes was readmission, we excluded patients who died during hospitalization. Patient characteristics collected included demographics and insurance category. To evaluate comorbidity burden, we collected ICD-9/ICD-10 diagnostic codes and generated a combined comorbidity score as described by Gagne, et al.²¹ This score was devised to predict mortality and 30-day readmissions and had better predictive ability in elders than the Elixhauser or Charlson scores. Scores ranged from −2 to 26, although values >20 are rare.

Exposure

Subjects were categorized as either discharged from the ACE or discharged from usual care. ACE discharges were tracked daily on a spreadsheet that was linked into our sample of eligible subjects.

Outcomes

Total cost of hospitalization (direct plus indirect costs), LOS, and all-cause 30-day readmissions were queried from the same billing database.

Statistical Analysis

As this study was a quality improvement project, analyses were descriptive and exploratory; no statistical hypothesis testing was conducted. We initially evaluated subject characteristics and comorbidities across study groups to determine group balance and comparability using means and standard deviations for continuous data and frequencies and percentages for categorical data. To analyze total cost and LOS, we utilized quantile regression with clustered standard errors to account for clustering by patient. We calculated the median difference between hospitalization cost and LOS for usual care versus ACE patients (with ACE as the referent group). To explore variations across the distributions of outcomes, we determined differences in cost and LOS and their 95% confidence intervals at the 25th, 50th, 75th, and 90th percentiles. Thirty-day readmission risk was estimated using a generalized estimating equation model with a logit link and binomial family. Readmission risk is presented along with 95% confidence intervals. For all models, we initially evaluated change over time (by quarter). After establishing the absence of time trends, we collapsed results into a comparison of usual care versus ACE care. Model estimates are presented both unadjusted and adjusted for age and comorbidity score. Following our initial analyses of cost, LOS, and 30-day readmission risk; we explored differences across quartiles of combined comorbidity scores. We used the same unadjusted models described above but incorporated an interaction term to generate estimates stratified by quartile of comorbidity score. We performed two additional analyses to evaluate the robustness of our findings. First, because hemiplegia prevalence was higher in the usual-care group than in the ACE group and can result in higher cost of care, we repeated the analysis after excluding those patients with hemiplegia. Second, because we were unable to control for functional capacity in the entire sample, we evaluated group differences in mobility for a subsample obtained prior to October 2015 using ICD-9 diagnostic codes, which can be considered surrogate markers for mobility.²² The results of our first analysis did not substantively change our main findings; in our second analysis, groups were balanced by mobility factors which suggested that confounding by functional capacity would be limited in our full sample. The results of these analyses are reported in the supplemental material.

Analysis was completed using Stata v15.1 (StataCorp, LP College Station, Texas). The Baystate Medical Center Institutional Review Board determined that the initiative was quality improvement and “not research.”

A total of 13,209 patients met the initial inclusion criteria; 1,621 were excluded, resulting in a sample of 11,588 patients. Over the 3-year study period, 1,429 (12.3%) were discharged from ACE and 10,159 (87.7%) were discharged from usual care. The groups were similar in age, sex, race and insurance status. Compared with the usual-care group, ACE patients had a higher median comorbidity score (3 vs 2 for usual care) and higher rates for anemia, dementia, fluid and electrolyte disorders, hypertension, and chronic obstructive pulmonary disease (COPD). However, ACE patients had lower rates of hemiplegia (0.9% vs 3%), arrhythmias, and pulmonary circulation disorders than those with usual care (Table 1).

The median cost per ACE patient was slightly lower at $6,258 (interquartile range [IQR] = $4,683-$8,547) versus $6,858 (IQR = $4,855-$10,478) in usual care. Across the cost distribution, the ACE program had lower costs than usual care; however, these differences became more pronounced at the higher end of the distribution. For example, compared with the ACE group, the usual-care group’s unadjusted cost difference was $171 higher at the 25th percentile, $600 higher at the median, $1,932 higher at the 75^th percentile, and $3,687 higher at the 90th percentile. The ACE median LOS was 3.7 days (IQR = 2.7-5.0) compared with 3.8 days (IQR = 2.7-6.0) for non-ACE patients. Similar to cost, LOS differences rose at higher percentiles of the distribution, with shorter stays for the ACE patients within each grouping. Compared with the ACE group, the unadjusted LOS difference for usual-care patients ranged from 0 days at the 25th percentile to 0.2 day longer at the median, 1.0 day longer at the 75th percentile, and 1.9 days longer at the 90th percentile. For both cost and LOS models, estimates remained stable after adjusting for age and combined comorbidity score (Table 2).

This quality improvement initiative evaluated which ACE admissions yielded the greatest value and found the largest reductions in LOS and cost in patients with the greatest comorbidity scores (frequently referred to as “high need, high cost”).^23,24 Based on prior literature, we had anticipated that moderate risk patients would show the maximum benefit.^15,25 In contrast to our findings, a University of Alabama (UAB) ACE program subgroup analysis using the CMS Case Mix Index (CMI) found a cost reduction for patients with low or intermediate CMI scores but not for those with high scores.¹⁵ The Hospital Elder Life Program (HELP) has yielded maximal impact for patients at moderate risk for delirium.²⁶ Our results are supported by a University of Texas, Houston, study revealing lower LOS and cost for ACE patients, despite high CMI scores and endemic frailty, although it did not report outcomes across a range of comorbidities or costs.²⁷ Our results may be determined by the specific characteristics of the Baystate ACE initiative. Our emphasis on considering prognosis and encouraging advance care planning could have contributed to the improved metrics for more complicated patients. It is possible that patients with high comorbidity burden were more likely to screen in with the surprise question, leading to more frequent goals of care discussions by the hospitalists or geriatrics team, which, in turn, may have resulted in less aggressive care and consequently lower costs. The emphasis on prognosis and palliative care was not a feature of the UAB or Texas studies. Additional components, such as the delirium screening and the presence of volunteer advocates, could also have impacted the results. Our tiered approach during rounds with rapid reviews for most patients and longer discussions for those at highest risk may have further contributed to the findings. Finally, although we did not track the recommendation acceptance rate for the entire study period, in the first 9 months of the project, 9,325 recommendations were made with an acceptance rate of >85%. We previously reported a similar acceptance rate for medication recommendations.²⁸ Another factor contributing to our results may be the ways in which we categorized patients and calculated costs. We used the Gagne combined comorbidity score, which includes only prior conditions;²¹ the UAB study used CMI, which includes severity of presenting illness and complications, as well as baseline comorbidities. We also compared total cost, while UAB reported variable direct cost.

This study has a number of limitations. First, it was conducted at a single site and may not apply to other hospitals. Second, as a quality improvement program, its design, processes, and personnel evolved over time, and, as in any multicomponent initiative, the effect of individual factors on the outcomes is unknown. Third, this is an observational study with the aim of generating hypotheses for more rigorous studies in the future and residual confounding factors may exist despite efforts to adjust for variables present in an administrative database. Thus, we were unable to completely adjust for potentially important social factors, presence of delirium, or baseline functional status.

To our knowledge, this study is the first report on the differential impact of comorbidity scores and cost distribution on ACE total cost and LOS reductions. Despite its limitations, it contributes to the existing literature by suggesting that the Gagne comorbidity score can help identify which admissions will yield the greatest value. The Gagne score could be calculated at admission using the ePrognosis risk calculator or incorporated and automated in the EMR.²⁹ Many health systems are reluctant to designate beds for specific subpopulations since doing so decreases flexibility and complicates the admission process. A dynamic tension exists between increasing income streams now and generating future savings by supporting initiatives with upfront costs. Other successful acute care geriatrics programs, such as NICHE,³⁰ HELP,³¹ MACE,³² and consultation teams, exist.³³ Studies reporting the outcomes of combining ACE units with these other approaches in a “portfolio approach” will inform the design of the most efficient and impactful programs.³⁴ Scrupulous attention to symptom control and advance care planning are key features of our program, and, given the high prevalence of advanced serious illness in hospitalized older adults, this consideration may be critical for success.

As ACE units can only care for a small fraction of hospitalized older adults, determining which patients will maximally benefit from the structured, team-based care on ACE units is crucial. We found that the greatest impact on LOS and costs occurred in the subgroup with the highest comorbidity scores and overall cost. ACE care for the most vulnerable patients appeared to yield the greatest value for the system; thus, these older adults may need to be prioritized for admission. This improvement may enhance quality and value outcomes, maximize a scarce resource, and secure results needed to sustain the “clinician-led and data-driven” ACE model in the face of changing clinical and financial landscapes.³⁵

Acknowledgments

All those with significant contributions to this work are included as authors.

The authors express their deep appreciation to all their Baystate collaborators, particularly to Rebecca Starr, MD, the first geriatrics medical director of the program, Ms. Virginia Chipps, RN, the program’s first nurse manager, and Tasmiah Chowdhury, PharmD, the program’s first pharmacist. We are also deeply grateful to those persons who provided programmatic advice and input on model ACE programs elsewhere, including Kyle Allen, MD, Michael Malone MD, Robert Palmer MD, and, especially, Kellie Flood, MD.

Disclosures

None of the authors have any existing or potential personal or financial conflicts relevant to this paper to report.

Funding

This work was supported in part by a Geriatric Workforce Enhancement Program award (grant # U1QHP28702) from the Health Resources and Services Administration and by internal support from Baystate Health

In 2016, 15.2% of older Americans were hospitalized compared with 7% of the overall population and their length of stay (LOS) was 0.7 days greater.¹ Geriatric hospitalizations frequently result in complications, functional decline, nursing home transfers, and increased cost.^2-4 This pattern of decline has been termed “hospitalitis” or dysfunctional syndrome.^5,6 Hospitals need data-driven approaches to improve outcomes for elders. The Acute Care for Elders (ACE) program, which has been in existence for roughly 25 years, is one such model. ACE features include an environment prepared for older adults, patient-centered care to prevent functional and cognitive decline, frequent medical review to prevent iatrogenic injury or new geriatric syndromes, and early discharge and rehabilitation planning to maximize the likelihood of return to the community.⁷ Although published data vary somewhat, ACE programs have robust evidence documenting improved safety, quality, and value.^8-15 A recent meta-analysis found that ACE programs decrease LOS, costs, new nursing home discharges, falls, delirium, and functional decline.¹⁶ However, of the 13 ACE trials reported to date, only five were published in the last decade. Recent rising pressure to decrease hospitalizations and reduce LOS has shifted some care to other settings and it is unclear whether the same results would persist in today’s rapid-paced hospitals.

ACE programs require enhanced resources and restructured care processes but there is a notable lack of data to guide patient selection. Admission criteria vary among the published reports, and information on whether comorbidity burden impacts the magnitude of benefit is scarce. One ACE investigator commented, “We were not able to identify a subgroup of patients who were most likely to benefit.”¹⁷ Not all hospitalized older adults can receive ACE care, and some units have closed due to financial and logistic pressures; thus, criteria to target this scarce resource are urgently needed. Our hospital implemented an ACE program in 2014 and we have measured and internally benchmarked important quality improvement metrics. Using this data, we conducted an exploratory analysis to generate hypotheses on the differential impact across the spectrum of cost, LOS, 30-day readmissions, and variations across quartiles of comorbidity severity.

Setting and Patients

In September 2014, our 716-bed teaching hospital in Springfield, Massachusetts launched an ACE program to improve care for older adults on a single medical unit. The program succeeded in engaging the senior leadership, and geriatrics was identified as a priority in Baystate’s 5-year strategic plan. ACE patients ≥70 years were admitted from the emergency department with inpatient status. Patients transferred from other units or with advanced dementia or nearing death were excluded. Core components of the ACE program were derived from published summaries (see supplementary material).^7,16

Interprofessional ”ACE Rounds”

Interprofessional ACE Rounds occurred every weekday. As one ACE analyst has noted, “the interdisciplinary team…ensures that the multifactorial nature of functional decline is met with a multicomponent plan to prevent it.”¹⁸ Rounds participants shifted over time but always included a geriatrics physician assistant (PA) or geriatrician (team leader), a pharmacist, staff nurses, and a chaplain. The nurse educator, dietician, research assistant, and patient advocate/volunteers attended intermittently. Before rounds, the PA reviewed the admission notes for new ACE patients. Initially, rounds were lengthy and included nurse coaching. Later, nurses’ presentations were structured by the SPICES tool (Sleep, Problems with eating/feeding, Incontinence, Confusion, Evidence of falls, Skin Breakdown)¹⁹ and tracking and reporting templates. Coaching and education, along with conversations that did not require the full team, were removed from rounds. Thus, the time required for rounds declined from about 75 minutes to 35 minutes, which allowed more patients to be discussed efficiently. This change was critical as the number of ACE patients rose following the shift to the larger unit. The pharmacist reviewed medications focusing on potentially inappropriate drugs. Following rounds, the nurses and pharmacist conveyed recommendations to the hospitalists.

Patient-Centered Activities to Prevent Functional and Cognitive Decline

Project leaders coached staff about the importance of mobility, sleep, and delirium prevention and identification. The nurses screened patients using the Confusion Assessment Method (CAM) and reported delirium promptly. Specific care sets for ACE patients were implemented (see supplementary material).

The project was enhanced by several palliative care components, ie tracking pain, noting psychiatric symptoms, and considering prognosis by posing the “Surprise Question” during rounds.²⁰ (“Would you be surprised if this patient died in the next year?”). As far as staffing and logistics allowed, the goals of care conversation were held by a geriatrics PA with patients/families who “screened in.”

Prepared Environment

The ACE program’s unit was remodeled to facilitate physical and cognitive functioning and promote sleep at night (quiet hours: 10 PM-6 AM).

In accordance with quality improvement processes, iterative shifts were implemented over time in terms of checklist, presentation format, timing, and team participation. In December 2016, the program relocated to a unit with 34 ACE beds and 5 end-of-life beds; this move markedly increased the number of eligible ACE patients.

Study Design, Data Source, and Patients

Since we were implementing and measuring our ACE program with a quality improvement lens, we chose a descriptive cross-sectional study design to generate hypotheses regarding our program’s impact compared to usual care. Using a hospital-wide billing database (McKesson Performance Analytics, v19, Alpharetta, Georgia) we sampled inpatients aged >70 years with a medical Diagnosis Related Group (DRGs) admitted through the emergency department and discharged from a medical unit from September 22, 2014 to August 31, 2017. These criteria mirrored those in the ACE unit. Older adults requiring specialized care (eg, those with myocardial infarct) were excluded, as were those with billing codes for mechanical ventilation, admission to critical care units, or discharge to hospice. Because one of our outcomes was readmission, we excluded patients who died during hospitalization. Patient characteristics collected included demographics and insurance category. To evaluate comorbidity burden, we collected ICD-9/ICD-10 diagnostic codes and generated a combined comorbidity score as described by Gagne, et al.²¹ This score was devised to predict mortality and 30-day readmissions and had better predictive ability in elders than the Elixhauser or Charlson scores. Scores ranged from −2 to 26, although values >20 are rare.

Exposure

Subjects were categorized as either discharged from the ACE or discharged from usual care. ACE discharges were tracked daily on a spreadsheet that was linked into our sample of eligible subjects.

Outcomes

Total cost of hospitalization (direct plus indirect costs), LOS, and all-cause 30-day readmissions were queried from the same billing database.

Statistical Analysis

As this study was a quality improvement project, analyses were descriptive and exploratory; no statistical hypothesis testing was conducted. We initially evaluated subject characteristics and comorbidities across study groups to determine group balance and comparability using means and standard deviations for continuous data and frequencies and percentages for categorical data. To analyze total cost and LOS, we utilized quantile regression with clustered standard errors to account for clustering by patient. We calculated the median difference between hospitalization cost and LOS for usual care versus ACE patients (with ACE as the referent group). To explore variations across the distributions of outcomes, we determined differences in cost and LOS and their 95% confidence intervals at the 25th, 50th, 75th, and 90th percentiles. Thirty-day readmission risk was estimated using a generalized estimating equation model with a logit link and binomial family. Readmission risk is presented along with 95% confidence intervals. For all models, we initially evaluated change over time (by quarter). After establishing the absence of time trends, we collapsed results into a comparison of usual care versus ACE care. Model estimates are presented both unadjusted and adjusted for age and comorbidity score. Following our initial analyses of cost, LOS, and 30-day readmission risk; we explored differences across quartiles of combined comorbidity scores. We used the same unadjusted models described above but incorporated an interaction term to generate estimates stratified by quartile of comorbidity score. We performed two additional analyses to evaluate the robustness of our findings. First, because hemiplegia prevalence was higher in the usual-care group than in the ACE group and can result in higher cost of care, we repeated the analysis after excluding those patients with hemiplegia. Second, because we were unable to control for functional capacity in the entire sample, we evaluated group differences in mobility for a subsample obtained prior to October 2015 using ICD-9 diagnostic codes, which can be considered surrogate markers for mobility.²² The results of our first analysis did not substantively change our main findings; in our second analysis, groups were balanced by mobility factors which suggested that confounding by functional capacity would be limited in our full sample. The results of these analyses are reported in the supplemental material.

Analysis was completed using Stata v15.1 (StataCorp, LP College Station, Texas). The Baystate Medical Center Institutional Review Board determined that the initiative was quality improvement and “not research.”

A total of 13,209 patients met the initial inclusion criteria; 1,621 were excluded, resulting in a sample of 11,588 patients. Over the 3-year study period, 1,429 (12.3%) were discharged from ACE and 10,159 (87.7%) were discharged from usual care. The groups were similar in age, sex, race and insurance status. Compared with the usual-care group, ACE patients had a higher median comorbidity score (3 vs 2 for usual care) and higher rates for anemia, dementia, fluid and electrolyte disorders, hypertension, and chronic obstructive pulmonary disease (COPD). However, ACE patients had lower rates of hemiplegia (0.9% vs 3%), arrhythmias, and pulmonary circulation disorders than those with usual care (Table 1).

The median cost per ACE patient was slightly lower at $6,258 (interquartile range [IQR] = $4,683-$8,547) versus $6,858 (IQR = $4,855-$10,478) in usual care. Across the cost distribution, the ACE program had lower costs than usual care; however, these differences became more pronounced at the higher end of the distribution. For example, compared with the ACE group, the usual-care group’s unadjusted cost difference was $171 higher at the 25th percentile, $600 higher at the median, $1,932 higher at the 75^th percentile, and $3,687 higher at the 90th percentile. The ACE median LOS was 3.7 days (IQR = 2.7-5.0) compared with 3.8 days (IQR = 2.7-6.0) for non-ACE patients. Similar to cost, LOS differences rose at higher percentiles of the distribution, with shorter stays for the ACE patients within each grouping. Compared with the ACE group, the unadjusted LOS difference for usual-care patients ranged from 0 days at the 25th percentile to 0.2 day longer at the median, 1.0 day longer at the 75th percentile, and 1.9 days longer at the 90th percentile. For both cost and LOS models, estimates remained stable after adjusting for age and combined comorbidity score (Table 2).

This quality improvement initiative evaluated which ACE admissions yielded the greatest value and found the largest reductions in LOS and cost in patients with the greatest comorbidity scores (frequently referred to as “high need, high cost”).^23,24 Based on prior literature, we had anticipated that moderate risk patients would show the maximum benefit.^15,25 In contrast to our findings, a University of Alabama (UAB) ACE program subgroup analysis using the CMS Case Mix Index (CMI) found a cost reduction for patients with low or intermediate CMI scores but not for those with high scores.¹⁵ The Hospital Elder Life Program (HELP) has yielded maximal impact for patients at moderate risk for delirium.²⁶ Our results are supported by a University of Texas, Houston, study revealing lower LOS and cost for ACE patients, despite high CMI scores and endemic frailty, although it did not report outcomes across a range of comorbidities or costs.²⁷ Our results may be determined by the specific characteristics of the Baystate ACE initiative. Our emphasis on considering prognosis and encouraging advance care planning could have contributed to the improved metrics for more complicated patients. It is possible that patients with high comorbidity burden were more likely to screen in with the surprise question, leading to more frequent goals of care discussions by the hospitalists or geriatrics team, which, in turn, may have resulted in less aggressive care and consequently lower costs. The emphasis on prognosis and palliative care was not a feature of the UAB or Texas studies. Additional components, such as the delirium screening and the presence of volunteer advocates, could also have impacted the results. Our tiered approach during rounds with rapid reviews for most patients and longer discussions for those at highest risk may have further contributed to the findings. Finally, although we did not track the recommendation acceptance rate for the entire study period, in the first 9 months of the project, 9,325 recommendations were made with an acceptance rate of >85%. We previously reported a similar acceptance rate for medication recommendations.²⁸ Another factor contributing to our results may be the ways in which we categorized patients and calculated costs. We used the Gagne combined comorbidity score, which includes only prior conditions;²¹ the UAB study used CMI, which includes severity of presenting illness and complications, as well as baseline comorbidities. We also compared total cost, while UAB reported variable direct cost.

This study has a number of limitations. First, it was conducted at a single site and may not apply to other hospitals. Second, as a quality improvement program, its design, processes, and personnel evolved over time, and, as in any multicomponent initiative, the effect of individual factors on the outcomes is unknown. Third, this is an observational study with the aim of generating hypotheses for more rigorous studies in the future and residual confounding factors may exist despite efforts to adjust for variables present in an administrative database. Thus, we were unable to completely adjust for potentially important social factors, presence of delirium, or baseline functional status.

To our knowledge, this study is the first report on the differential impact of comorbidity scores and cost distribution on ACE total cost and LOS reductions. Despite its limitations, it contributes to the existing literature by suggesting that the Gagne comorbidity score can help identify which admissions will yield the greatest value. The Gagne score could be calculated at admission using the ePrognosis risk calculator or incorporated and automated in the EMR.²⁹ Many health systems are reluctant to designate beds for specific subpopulations since doing so decreases flexibility and complicates the admission process. A dynamic tension exists between increasing income streams now and generating future savings by supporting initiatives with upfront costs. Other successful acute care geriatrics programs, such as NICHE,³⁰ HELP,³¹ MACE,³² and consultation teams, exist.³³ Studies reporting the outcomes of combining ACE units with these other approaches in a “portfolio approach” will inform the design of the most efficient and impactful programs.³⁴ Scrupulous attention to symptom control and advance care planning are key features of our program, and, given the high prevalence of advanced serious illness in hospitalized older adults, this consideration may be critical for success.

As ACE units can only care for a small fraction of hospitalized older adults, determining which patients will maximally benefit from the structured, team-based care on ACE units is crucial. We found that the greatest impact on LOS and costs occurred in the subgroup with the highest comorbidity scores and overall cost. ACE care for the most vulnerable patients appeared to yield the greatest value for the system; thus, these older adults may need to be prioritized for admission. This improvement may enhance quality and value outcomes, maximize a scarce resource, and secure results needed to sustain the “clinician-led and data-driven” ACE model in the face of changing clinical and financial landscapes.³⁵

Acknowledgments

All those with significant contributions to this work are included as authors.

The authors express their deep appreciation to all their Baystate collaborators, particularly to Rebecca Starr, MD, the first geriatrics medical director of the program, Ms. Virginia Chipps, RN, the program’s first nurse manager, and Tasmiah Chowdhury, PharmD, the program’s first pharmacist. We are also deeply grateful to those persons who provided programmatic advice and input on model ACE programs elsewhere, including Kyle Allen, MD, Michael Malone MD, Robert Palmer MD, and, especially, Kellie Flood, MD.

Disclosures

None of the authors have any existing or potential personal or financial conflicts relevant to this paper to report.

Funding

This work was supported in part by a Geriatric Workforce Enhancement Program award (grant # U1QHP28702) from the Health Resources and Services Administration and by internal support from Baystate Health

In 2016, 15.2% of older Americans were hospitalized compared with 7% of the overall population and their length of stay (LOS) was 0.7 days greater.¹ Geriatric hospitalizations frequently result in complications, functional decline, nursing home transfers, and increased cost.^2-4 This pattern of decline has been termed “hospitalitis” or dysfunctional syndrome.^5,6 Hospitals need data-driven approaches to improve outcomes for elders. The Acute Care for Elders (ACE) program, which has been in existence for roughly 25 years, is one such model. ACE features include an environment prepared for older adults, patient-centered care to prevent functional and cognitive decline, frequent medical review to prevent iatrogenic injury or new geriatric syndromes, and early discharge and rehabilitation planning to maximize the likelihood of return to the community.⁷ Although published data vary somewhat, ACE programs have robust evidence documenting improved safety, quality, and value.^8-15 A recent meta-analysis found that ACE programs decrease LOS, costs, new nursing home discharges, falls, delirium, and functional decline.¹⁶ However, of the 13 ACE trials reported to date, only five were published in the last decade. Recent rising pressure to decrease hospitalizations and reduce LOS has shifted some care to other settings and it is unclear whether the same results would persist in today’s rapid-paced hospitals.

ACE programs require enhanced resources and restructured care processes but there is a notable lack of data to guide patient selection. Admission criteria vary among the published reports, and information on whether comorbidity burden impacts the magnitude of benefit is scarce. One ACE investigator commented, “We were not able to identify a subgroup of patients who were most likely to benefit.”¹⁷ Not all hospitalized older adults can receive ACE care, and some units have closed due to financial and logistic pressures; thus, criteria to target this scarce resource are urgently needed. Our hospital implemented an ACE program in 2014 and we have measured and internally benchmarked important quality improvement metrics. Using this data, we conducted an exploratory analysis to generate hypotheses on the differential impact across the spectrum of cost, LOS, 30-day readmissions, and variations across quartiles of comorbidity severity.

Setting and Patients

In September 2014, our 716-bed teaching hospital in Springfield, Massachusetts launched an ACE program to improve care for older adults on a single medical unit. The program succeeded in engaging the senior leadership, and geriatrics was identified as a priority in Baystate’s 5-year strategic plan. ACE patients ≥70 years were admitted from the emergency department with inpatient status. Patients transferred from other units or with advanced dementia or nearing death were excluded. Core components of the ACE program were derived from published summaries (see supplementary material).^7,16

Interprofessional ”ACE Rounds”

Interprofessional ACE Rounds occurred every weekday. As one ACE analyst has noted, “the interdisciplinary team…ensures that the multifactorial nature of functional decline is met with a multicomponent plan to prevent it.”¹⁸ Rounds participants shifted over time but always included a geriatrics physician assistant (PA) or geriatrician (team leader), a pharmacist, staff nurses, and a chaplain. The nurse educator, dietician, research assistant, and patient advocate/volunteers attended intermittently. Before rounds, the PA reviewed the admission notes for new ACE patients. Initially, rounds were lengthy and included nurse coaching. Later, nurses’ presentations were structured by the SPICES tool (Sleep, Problems with eating/feeding, Incontinence, Confusion, Evidence of falls, Skin Breakdown)¹⁹ and tracking and reporting templates. Coaching and education, along with conversations that did not require the full team, were removed from rounds. Thus, the time required for rounds declined from about 75 minutes to 35 minutes, which allowed more patients to be discussed efficiently. This change was critical as the number of ACE patients rose following the shift to the larger unit. The pharmacist reviewed medications focusing on potentially inappropriate drugs. Following rounds, the nurses and pharmacist conveyed recommendations to the hospitalists.

Patient-Centered Activities to Prevent Functional and Cognitive Decline

Project leaders coached staff about the importance of mobility, sleep, and delirium prevention and identification. The nurses screened patients using the Confusion Assessment Method (CAM) and reported delirium promptly. Specific care sets for ACE patients were implemented (see supplementary material).

The project was enhanced by several palliative care components, ie tracking pain, noting psychiatric symptoms, and considering prognosis by posing the “Surprise Question” during rounds.²⁰ (“Would you be surprised if this patient died in the next year?”). As far as staffing and logistics allowed, the goals of care conversation were held by a geriatrics PA with patients/families who “screened in.”

Prepared Environment

The ACE program’s unit was remodeled to facilitate physical and cognitive functioning and promote sleep at night (quiet hours: 10 PM-6 AM).

In accordance with quality improvement processes, iterative shifts were implemented over time in terms of checklist, presentation format, timing, and team participation. In December 2016, the program relocated to a unit with 34 ACE beds and 5 end-of-life beds; this move markedly increased the number of eligible ACE patients.

Study Design, Data Source, and Patients

Since we were implementing and measuring our ACE program with a quality improvement lens, we chose a descriptive cross-sectional study design to generate hypotheses regarding our program’s impact compared to usual care. Using a hospital-wide billing database (McKesson Performance Analytics, v19, Alpharetta, Georgia) we sampled inpatients aged >70 years with a medical Diagnosis Related Group (DRGs) admitted through the emergency department and discharged from a medical unit from September 22, 2014 to August 31, 2017. These criteria mirrored those in the ACE unit. Older adults requiring specialized care (eg, those with myocardial infarct) were excluded, as were those with billing codes for mechanical ventilation, admission to critical care units, or discharge to hospice. Because one of our outcomes was readmission, we excluded patients who died during hospitalization. Patient characteristics collected included demographics and insurance category. To evaluate comorbidity burden, we collected ICD-9/ICD-10 diagnostic codes and generated a combined comorbidity score as described by Gagne, et al.²¹ This score was devised to predict mortality and 30-day readmissions and had better predictive ability in elders than the Elixhauser or Charlson scores. Scores ranged from −2 to 26, although values >20 are rare.

Exposure

Subjects were categorized as either discharged from the ACE or discharged from usual care. ACE discharges were tracked daily on a spreadsheet that was linked into our sample of eligible subjects.

Outcomes

Total cost of hospitalization (direct plus indirect costs), LOS, and all-cause 30-day readmissions were queried from the same billing database.

Statistical Analysis

As this study was a quality improvement project, analyses were descriptive and exploratory; no statistical hypothesis testing was conducted. We initially evaluated subject characteristics and comorbidities across study groups to determine group balance and comparability using means and standard deviations for continuous data and frequencies and percentages for categorical data. To analyze total cost and LOS, we utilized quantile regression with clustered standard errors to account for clustering by patient. We calculated the median difference between hospitalization cost and LOS for usual care versus ACE patients (with ACE as the referent group). To explore variations across the distributions of outcomes, we determined differences in cost and LOS and their 95% confidence intervals at the 25th, 50th, 75th, and 90th percentiles. Thirty-day readmission risk was estimated using a generalized estimating equation model with a logit link and binomial family. Readmission risk is presented along with 95% confidence intervals. For all models, we initially evaluated change over time (by quarter). After establishing the absence of time trends, we collapsed results into a comparison of usual care versus ACE care. Model estimates are presented both unadjusted and adjusted for age and comorbidity score. Following our initial analyses of cost, LOS, and 30-day readmission risk; we explored differences across quartiles of combined comorbidity scores. We used the same unadjusted models described above but incorporated an interaction term to generate estimates stratified by quartile of comorbidity score. We performed two additional analyses to evaluate the robustness of our findings. First, because hemiplegia prevalence was higher in the usual-care group than in the ACE group and can result in higher cost of care, we repeated the analysis after excluding those patients with hemiplegia. Second, because we were unable to control for functional capacity in the entire sample, we evaluated group differences in mobility for a subsample obtained prior to October 2015 using ICD-9 diagnostic codes, which can be considered surrogate markers for mobility.²² The results of our first analysis did not substantively change our main findings; in our second analysis, groups were balanced by mobility factors which suggested that confounding by functional capacity would be limited in our full sample. The results of these analyses are reported in the supplemental material.

Analysis was completed using Stata v15.1 (StataCorp, LP College Station, Texas). The Baystate Medical Center Institutional Review Board determined that the initiative was quality improvement and “not research.”

A total of 13,209 patients met the initial inclusion criteria; 1,621 were excluded, resulting in a sample of 11,588 patients. Over the 3-year study period, 1,429 (12.3%) were discharged from ACE and 10,159 (87.7%) were discharged from usual care. The groups were similar in age, sex, race and insurance status. Compared with the usual-care group, ACE patients had a higher median comorbidity score (3 vs 2 for usual care) and higher rates for anemia, dementia, fluid and electrolyte disorders, hypertension, and chronic obstructive pulmonary disease (COPD). However, ACE patients had lower rates of hemiplegia (0.9% vs 3%), arrhythmias, and pulmonary circulation disorders than those with usual care (Table 1).

The median cost per ACE patient was slightly lower at $6,258 (interquartile range [IQR] = $4,683-$8,547) versus $6,858 (IQR = $4,855-$10,478) in usual care. Across the cost distribution, the ACE program had lower costs than usual care; however, these differences became more pronounced at the higher end of the distribution. For example, compared with the ACE group, the usual-care group’s unadjusted cost difference was $171 higher at the 25th percentile, $600 higher at the median, $1,932 higher at the 75^th percentile, and $3,687 higher at the 90th percentile. The ACE median LOS was 3.7 days (IQR = 2.7-5.0) compared with 3.8 days (IQR = 2.7-6.0) for non-ACE patients. Similar to cost, LOS differences rose at higher percentiles of the distribution, with shorter stays for the ACE patients within each grouping. Compared with the ACE group, the unadjusted LOS difference for usual-care patients ranged from 0 days at the 25th percentile to 0.2 day longer at the median, 1.0 day longer at the 75th percentile, and 1.9 days longer at the 90th percentile. For both cost and LOS models, estimates remained stable after adjusting for age and combined comorbidity score (Table 2).

This quality improvement initiative evaluated which ACE admissions yielded the greatest value and found the largest reductions in LOS and cost in patients with the greatest comorbidity scores (frequently referred to as “high need, high cost”).^23,24 Based on prior literature, we had anticipated that moderate risk patients would show the maximum benefit.^15,25 In contrast to our findings, a University of Alabama (UAB) ACE program subgroup analysis using the CMS Case Mix Index (CMI) found a cost reduction for patients with low or intermediate CMI scores but not for those with high scores.¹⁵ The Hospital Elder Life Program (HELP) has yielded maximal impact for patients at moderate risk for delirium.²⁶ Our results are supported by a University of Texas, Houston, study revealing lower LOS and cost for ACE patients, despite high CMI scores and endemic frailty, although it did not report outcomes across a range of comorbidities or costs.²⁷ Our results may be determined by the specific characteristics of the Baystate ACE initiative. Our emphasis on considering prognosis and encouraging advance care planning could have contributed to the improved metrics for more complicated patients. It is possible that patients with high comorbidity burden were more likely to screen in with the surprise question, leading to more frequent goals of care discussions by the hospitalists or geriatrics team, which, in turn, may have resulted in less aggressive care and consequently lower costs. The emphasis on prognosis and palliative care was not a feature of the UAB or Texas studies. Additional components, such as the delirium screening and the presence of volunteer advocates, could also have impacted the results. Our tiered approach during rounds with rapid reviews for most patients and longer discussions for those at highest risk may have further contributed to the findings. Finally, although we did not track the recommendation acceptance rate for the entire study period, in the first 9 months of the project, 9,325 recommendations were made with an acceptance rate of >85%. We previously reported a similar acceptance rate for medication recommendations.²⁸ Another factor contributing to our results may be the ways in which we categorized patients and calculated costs. We used the Gagne combined comorbidity score, which includes only prior conditions;²¹ the UAB study used CMI, which includes severity of presenting illness and complications, as well as baseline comorbidities. We also compared total cost, while UAB reported variable direct cost.

This study has a number of limitations. First, it was conducted at a single site and may not apply to other hospitals. Second, as a quality improvement program, its design, processes, and personnel evolved over time, and, as in any multicomponent initiative, the effect of individual factors on the outcomes is unknown. Third, this is an observational study with the aim of generating hypotheses for more rigorous studies in the future and residual confounding factors may exist despite efforts to adjust for variables present in an administrative database. Thus, we were unable to completely adjust for potentially important social factors, presence of delirium, or baseline functional status.

To our knowledge, this study is the first report on the differential impact of comorbidity scores and cost distribution on ACE total cost and LOS reductions. Despite its limitations, it contributes to the existing literature by suggesting that the Gagne comorbidity score can help identify which admissions will yield the greatest value. The Gagne score could be calculated at admission using the ePrognosis risk calculator or incorporated and automated in the EMR.²⁹ Many health systems are reluctant to designate beds for specific subpopulations since doing so decreases flexibility and complicates the admission process. A dynamic tension exists between increasing income streams now and generating future savings by supporting initiatives with upfront costs. Other successful acute care geriatrics programs, such as NICHE,³⁰ HELP,³¹ MACE,³² and consultation teams, exist.³³ Studies reporting the outcomes of combining ACE units with these other approaches in a “portfolio approach” will inform the design of the most efficient and impactful programs.³⁴ Scrupulous attention to symptom control and advance care planning are key features of our program, and, given the high prevalence of advanced serious illness in hospitalized older adults, this consideration may be critical for success.

As ACE units can only care for a small fraction of hospitalized older adults, determining which patients will maximally benefit from the structured, team-based care on ACE units is crucial. We found that the greatest impact on LOS and costs occurred in the subgroup with the highest comorbidity scores and overall cost. ACE care for the most vulnerable patients appeared to yield the greatest value for the system; thus, these older adults may need to be prioritized for admission. This improvement may enhance quality and value outcomes, maximize a scarce resource, and secure results needed to sustain the “clinician-led and data-driven” ACE model in the face of changing clinical and financial landscapes.³⁵

Acknowledgments

All those with significant contributions to this work are included as authors.

The authors express their deep appreciation to all their Baystate collaborators, particularly to Rebecca Starr, MD, the first geriatrics medical director of the program, Ms. Virginia Chipps, RN, the program’s first nurse manager, and Tasmiah Chowdhury, PharmD, the program’s first pharmacist. We are also deeply grateful to those persons who provided programmatic advice and input on model ACE programs elsewhere, including Kyle Allen, MD, Michael Malone MD, Robert Palmer MD, and, especially, Kellie Flood, MD.

Disclosures

None of the authors have any existing or potential personal or financial conflicts relevant to this paper to report.

Funding

This work was supported in part by a Geriatric Workforce Enhancement Program award (grant # U1QHP28702) from the Health Resources and Services Administration and by internal support from Baystate Health