Patients who are hospitalized for an acute event often have a range of prior outpatient experiences within the healthcare system, both before and after a hospitalization. In particular, continuity with a primary care provider can influence health outcomes.[1] In this issue of the Journal of Hospital Medicine, Boonyasai et al. found several characteristics of primary care physicians that were associated with whether their hospitalized patients were cared for by hospitalists.[2] Using Medicare claims data from the state of Texas during years 2001 to 2009, the authors calculated the percent of primary care physicians' hospitalized patients who were cared for by hospitalists. Hospitalist use increased overall during the time period, but primary care physicians differed in the rate and extent of hospitalist use. A minority of physicians were early adopters, with the majority of their hospitalized patients cared for by hospitalists during the entire time period. A sizeable group of primary care physicians mostly avoided using hospitalists. Moreover, there was a significant cluster of primary care physicians who, at some point during the study period, rapidly began using hospitalists within a relatively short time.

Several physician characteristics were associated with a greater adoption of the hospitalist model, including being female, in a family practice specialty, or in a rural practice setting. What this study lacks is the ability to explain why some physicians used hospitalists and others did not. It is probable that adoption (or not) of hospitalists is less an individual physician decision and instead reflects a choice of their clinical practice group. If an outpatient practice group or provider can influence whether or not their patients are cared for by hospitalists, it is also conceivable that they can affect hospital‐based outcomes as well. This finding reinforces the importance of examining the care and outcomes of patient care across the continuum of care, rather than focusing on the inpatient or outpatient setting.

As a result of the Affordable Care Act and rising healthcare costs, provider groups are beginning to form accountable care organizations (ACOs). An ACO is partnership between payers and providers to care for a population of patients across the continuum of care. In these arrangements, the providers often take on financial risk for the total cost of care for a population as well as for providing high‐quality care as monitored by specific metrics.[3] The population of patients for which ACOs take risks often include predominantly patients who receive primary care from the group.[4] For overall cost management, given that acute hospitalizations are disproportionately high cost, a primary focus of a majority of ACOs is to reduce unnecessary hospital days. Overall, ACOs that have been successful in the short term in managing costs have done so primarily by reducing overall hospital days.[5] ACOs have started to do so by creating intensive outpatient care management programs for high‐risk patients, by focusing on transitions of care to help decrease readmissions, by working with primary care clinics to transform into patient‐centered medical homes, where same‐day access to care is a priority, and developing other disease‐management tools to keep patients healthy.

To manage hospital utilization, many ACOs have developed plans to transform primary care and shift hospital care to outpatient care through enhanced outpatient case management for complex cases. As the way primary care is delivered changes, it will be very important to understand how this will modify the utilization and impact of hospitalist care on patients. The hope is that these modifications will work synergistically with hospitalist programs.

As the lines between outpatient and inpatient care become increasingly blurred, it may not be fair to attribute hospitalization outcome measures to hospitalists alone, particularly as ACOs are likely to move only the sickest or most difficult to manage patients to the inpatient setting. This may affect hospital‐based quality metrics such as readmissions and mortality. Seamless communication and transfer of information between outpatient and inpatient care will be vital to the success of ACOs.[6] In addition to improved communication, however, some systems may look to hospitalists to staff postdischarge clinics or act as extensivists or ambulatory intensivists to help manage the sickest in the population.[7]

Boonyasi et al. show that primary care physician characteristics as associated with whether or not patients' receive care from hospitalists.[2] As such, it reinforces the concept that providers in part of the continuum of care are integrally tied to care received by patients in different treatment settings. As our healthcare system rapidly transforms over the next few years, it will become more important to understand how outpatient and inpatient providers influence one another's care patterns and how these relationships influence care and cost‐related outcomes for patients.

Patients who are hospitalized for an acute event often have a range of prior outpatient experiences within the healthcare system, both before and after a hospitalization. In particular, continuity with a primary care provider can influence health outcomes.[1] In this issue of the Journal of Hospital Medicine, Boonyasai et al. found several characteristics of primary care physicians that were associated with whether their hospitalized patients were cared for by hospitalists.[2] Using Medicare claims data from the state of Texas during years 2001 to 2009, the authors calculated the percent of primary care physicians' hospitalized patients who were cared for by hospitalists. Hospitalist use increased overall during the time period, but primary care physicians differed in the rate and extent of hospitalist use. A minority of physicians were early adopters, with the majority of their hospitalized patients cared for by hospitalists during the entire time period. A sizeable group of primary care physicians mostly avoided using hospitalists. Moreover, there was a significant cluster of primary care physicians who, at some point during the study period, rapidly began using hospitalists within a relatively short time.

Several physician characteristics were associated with a greater adoption of the hospitalist model, including being female, in a family practice specialty, or in a rural practice setting. What this study lacks is the ability to explain why some physicians used hospitalists and others did not. It is probable that adoption (or not) of hospitalists is less an individual physician decision and instead reflects a choice of their clinical practice group. If an outpatient practice group or provider can influence whether or not their patients are cared for by hospitalists, it is also conceivable that they can affect hospital‐based outcomes as well. This finding reinforces the importance of examining the care and outcomes of patient care across the continuum of care, rather than focusing on the inpatient or outpatient setting.

As a result of the Affordable Care Act and rising healthcare costs, provider groups are beginning to form accountable care organizations (ACOs). An ACO is partnership between payers and providers to care for a population of patients across the continuum of care. In these arrangements, the providers often take on financial risk for the total cost of care for a population as well as for providing high‐quality care as monitored by specific metrics.[3] The population of patients for which ACOs take risks often include predominantly patients who receive primary care from the group.[4] For overall cost management, given that acute hospitalizations are disproportionately high cost, a primary focus of a majority of ACOs is to reduce unnecessary hospital days. Overall, ACOs that have been successful in the short term in managing costs have done so primarily by reducing overall hospital days.[5] ACOs have started to do so by creating intensive outpatient care management programs for high‐risk patients, by focusing on transitions of care to help decrease readmissions, by working with primary care clinics to transform into patient‐centered medical homes, where same‐day access to care is a priority, and developing other disease‐management tools to keep patients healthy.

To manage hospital utilization, many ACOs have developed plans to transform primary care and shift hospital care to outpatient care through enhanced outpatient case management for complex cases. As the way primary care is delivered changes, it will be very important to understand how this will modify the utilization and impact of hospitalist care on patients. The hope is that these modifications will work synergistically with hospitalist programs.

As the lines between outpatient and inpatient care become increasingly blurred, it may not be fair to attribute hospitalization outcome measures to hospitalists alone, particularly as ACOs are likely to move only the sickest or most difficult to manage patients to the inpatient setting. This may affect hospital‐based quality metrics such as readmissions and mortality. Seamless communication and transfer of information between outpatient and inpatient care will be vital to the success of ACOs.[6] In addition to improved communication, however, some systems may look to hospitalists to staff postdischarge clinics or act as extensivists or ambulatory intensivists to help manage the sickest in the population.[7]

Boonyasi et al. show that primary care physician characteristics as associated with whether or not patients' receive care from hospitalists.[2] As such, it reinforces the concept that providers in part of the continuum of care are integrally tied to care received by patients in different treatment settings. As our healthcare system rapidly transforms over the next few years, it will become more important to understand how outpatient and inpatient providers influence one another's care patterns and how these relationships influence care and cost‐related outcomes for patients.

Patients who are hospitalized for an acute event often have a range of prior outpatient experiences within the healthcare system, both before and after a hospitalization. In particular, continuity with a primary care provider can influence health outcomes.[1] In this issue of the Journal of Hospital Medicine, Boonyasai et al. found several characteristics of primary care physicians that were associated with whether their hospitalized patients were cared for by hospitalists.[2] Using Medicare claims data from the state of Texas during years 2001 to 2009, the authors calculated the percent of primary care physicians' hospitalized patients who were cared for by hospitalists. Hospitalist use increased overall during the time period, but primary care physicians differed in the rate and extent of hospitalist use. A minority of physicians were early adopters, with the majority of their hospitalized patients cared for by hospitalists during the entire time period. A sizeable group of primary care physicians mostly avoided using hospitalists. Moreover, there was a significant cluster of primary care physicians who, at some point during the study period, rapidly began using hospitalists within a relatively short time.

Several physician characteristics were associated with a greater adoption of the hospitalist model, including being female, in a family practice specialty, or in a rural practice setting. What this study lacks is the ability to explain why some physicians used hospitalists and others did not. It is probable that adoption (or not) of hospitalists is less an individual physician decision and instead reflects a choice of their clinical practice group. If an outpatient practice group or provider can influence whether or not their patients are cared for by hospitalists, it is also conceivable that they can affect hospital‐based outcomes as well. This finding reinforces the importance of examining the care and outcomes of patient care across the continuum of care, rather than focusing on the inpatient or outpatient setting.

As a result of the Affordable Care Act and rising healthcare costs, provider groups are beginning to form accountable care organizations (ACOs). An ACO is partnership between payers and providers to care for a population of patients across the continuum of care. In these arrangements, the providers often take on financial risk for the total cost of care for a population as well as for providing high‐quality care as monitored by specific metrics.[3] The population of patients for which ACOs take risks often include predominantly patients who receive primary care from the group.[4] For overall cost management, given that acute hospitalizations are disproportionately high cost, a primary focus of a majority of ACOs is to reduce unnecessary hospital days. Overall, ACOs that have been successful in the short term in managing costs have done so primarily by reducing overall hospital days.[5] ACOs have started to do so by creating intensive outpatient care management programs for high‐risk patients, by focusing on transitions of care to help decrease readmissions, by working with primary care clinics to transform into patient‐centered medical homes, where same‐day access to care is a priority, and developing other disease‐management tools to keep patients healthy.

To manage hospital utilization, many ACOs have developed plans to transform primary care and shift hospital care to outpatient care through enhanced outpatient case management for complex cases. As the way primary care is delivered changes, it will be very important to understand how this will modify the utilization and impact of hospitalist care on patients. The hope is that these modifications will work synergistically with hospitalist programs.

As the lines between outpatient and inpatient care become increasingly blurred, it may not be fair to attribute hospitalization outcome measures to hospitalists alone, particularly as ACOs are likely to move only the sickest or most difficult to manage patients to the inpatient setting. This may affect hospital‐based quality metrics such as readmissions and mortality. Seamless communication and transfer of information between outpatient and inpatient care will be vital to the success of ACOs.[6] In addition to improved communication, however, some systems may look to hospitalists to staff postdischarge clinics or act as extensivists or ambulatory intensivists to help manage the sickest in the population.[7]

Boonyasi et al. show that primary care physician characteristics as associated with whether or not patients' receive care from hospitalists.[2] As such, it reinforces the concept that providers in part of the continuum of care are integrally tied to care received by patients in different treatment settings. As our healthcare system rapidly transforms over the next few years, it will become more important to understand how outpatient and inpatient providers influence one another's care patterns and how these relationships influence care and cost‐related outcomes for patients.

The Affordable Care Act explicitly outlines improving the value of healthcare by increasing quality and decreasing costs. It emphasizes value‐based purchasing, the transparency of performance metrics, and the use of payment incentives to reward quality.[1, 2] Venous thromboembolism (VTE) prophylaxis is one of these publicly reported performance measures. The National Quality Forum recommends that each patient be evaluated on hospital admission and during their hospitalization for VTE risk level and for appropriate thromboprophylaxis to be used, if required.[3] Similarly, the Joint Commission includes appropriate VTE prophylaxis in its Core Measures.[4] Patient experience and performance metrics, including VTE prophylaxis, constitute the hospital value‐based purchasing (VBP) component of healthcare reform.[5] For a hypothetical 327‐bed hospital, an estimated $1.7 million of a hospital's inpatient payments from Medicare will be at risk from VBP alone.[2]

VTE prophylaxis is a common target of quality improvement projects. Effective, safe, and cost‐effective measures to prevent VTE exist, including pharmacologic and mechanical prophylaxis.[6, 7] Despite these measures, compliance rates are often below 50%.[8] Different interventions have been pursued to ensure appropriate VTE prophylaxis, including computerized provider order entry (CPOE), electronic alerts, mandatory VTE risk assessment and prophylaxis, and provider education campaigns.[9] Recent studies show that CPOE systems with mandatory fields can increase VTE prophylaxis rates to above 80%, yet the goal of a high reliability health system is for 100% of patients to receive recommended therapy.[10, 11, 12, 13, 14, 15] Interventions to improve prophylaxis rates that have included multiple strategies, such as computerized order sets, feedback, and education, have been the most effective, increasing compliance to above 90%.[9, 11, 16] These systems can be enhanced with additional interventions such as providing individualized provider education and feedback, understanding of work flow, and ensuring patients receive the prescribed therapies.[12] For example, a physician dashboard could be employed to provide a snapshot and historical trend of key performance indicators using graphical displays and indicators.[17]

Dashboards and pay‐for‐performance programs have been increasingly used to increase the visibility of these metrics, provide feedback, visually display benchmarks and goals, and proactively monitor for achievements and setbacks.[18] Although these strategies are often addressed at departmental (or greater) levels, applying them at the level of the individual provider may assist hospitals in reducing preventable harm and achieving safety and quality goals, especially at higher benchmarks. With their expanding role, hospitalists provide a key opportunity to lead improvement efforts and to study the impact of dashboards and pay‐for performance at the provider level to achieve VTE prophylaxis performance targets. Hospitalists are often the front‐line provider for inpatients and deliver up to 70% of inpatient general medical services.[19] The objective of our study was to evaluate the impact of providing individual provider feedback and employing a pay‐for‐performance program on baseline performance of VTE prophylaxis among hospitalists. We hypothesized that performance feedback through the use of a dashboard would increase appropriate VTE prophylaxis, and this effect would be further augmented by incorporation of a pay‐for‐performance program.

METHODS

Hospitalist Dashboard

In 2010, hospitalist program leaders met with hospital administrators to create a hospitalist dashboard that would provide regularly updated summaries of performance measures for individual hospitalists. The final set of metrics identified included appropriate VTE prophylaxis, length of stay, patients discharged per day, discharges before 3 _pm, depth of coding, patient satisfaction, readmissions, communication with the primary care provider, and time to signature for discharge summaries (Figure 1A). The dashboard was introduced at a general hospitalist meeting during which its purpose, methodology, and accessibility were described; it was subsequently implemented in January 2011.

Figure 1

Benchmarks were established for each metric, standardized to establish a scale ranging from 1 through 9, and incorporated into the dashboard (Figure 1A). Higher scores (creating a larger geometric shape) were desirable. For the VTE prophylaxis measure, scores of 1 through 9 corresponded to <60%, 60% to 64.9%, 65% to 69.9%, 70% to 74.9%, 75% to 79.9%, 80% to 84.9%, 85% to 89.9%, 90% to 94.9%, and 95% American College of Chest Physicians (ACCP)‐compliant VTE prophylaxis, respectively.[12, 20] Each provider was able to access the aggregated dashboard (showing the group mean) and his/her individualized dashboard using an individualized login and password for the institutional portal. This portal is used during the provider's workflow, including medical record review and order entry. Both a polygonal summary graphic (Figure 1A) and trend (Figure 1B) view of the dashboard were available to the provider. A comparison of the individual provider to the hospitalist group average was displayed (Figure 1A). At monthly program meetings, the dashboard, group results, and trends were discussed.

Venous Thromboembolism Prophylaxis Compliance

Our study was performed in a tertiary academic medical center with an approximately 20‐member hospitalist group (the precise membership varied over time), whose responsibilities include, among other clinical duties, staffing a 17‐bed general medicine unit with telemetry. The scope of diagnoses and acuity of patients admitted to the hospitalist service is similar to the housestaff services. Some hospitalist faculty serve both as hospitalist and nonhospitalist general medicine service team attendings, but the comparison groups were staffed by hospitalists for <20% of the time. For admissions, all hospitalists use a standardized general medicine admission order set that is integrated into the CPOE system (Sunrise Clinical Manager; Allscripts, Chicago, IL) and completed for all admitted patients. A mandatory VTE risk screen, which includes an assessment of VTE risk factors and pharmacological prophylaxis contraindications, must be completed by the ordering physician as part of this order set (Figure 2A). The system then prompts the provider with a risk‐appropriate VTE prophylaxis recommendation that the provider may subsequently order, including mechanical prophylaxis (Figure 2B). Based on ACCP VTE prevention guidelines, risk‐appropriate prophylaxis was determined using an electronic algorithm that categorized patients into risk categories based on the presence of major VTE risk factors (Figure 2A).[12, 15, 20] If none of these were present, the provider selected No major risk factors known. Both an assessment of current use of anticoagulation and a clinically high risk of bleeding were also included (Figure 2A). If none of these were present, the provider selected No contraindications known. This algorithm is published in detail elsewhere and has been shown to not increase major bleeding episodes.[12, 15] The VTE risk assessment, but not the VTE order itself, was a mandatory field. This allowed the physician discretion to choose among various pharmacological agents and mechanical mechanisms based on patient and physician preferences.

Figure 2

Compliance of risk‐appropriate VTE prophylaxis was determined 24 hours after the admission order set was completed using an automated electronic query of the CPOE system. Low molecular‐weight heparin prescription was included in the compliance algorithm as acceptable prophylaxis. Prescription of pharmacological VTE prophylaxis when a contraindication was present was considered noncompliant. The metric was assigned to the attending physician who billed for the first inpatient encounter.

RESULTS

Venous Thromboembolism Prophylaxis Compliance

We analyzed 3144 inpatient admissions by 38 hospitalists from 2009 to 2012. The 5 most frequent coded diagnoses were heart failure, acute kidney failure, syncope, pneumonia, and chest pain. Patients had a median length of stay of 3 days [interquartile range: 26]. During the dashboard‐only period, on average, providers improved in compliance by 4% (95% confidence interval [CI]: 35; P<0.001). With the addition of the pay‐for‐performance program, providers improved by an additional 4% (95% CI: 35; P<0.001). Group compliance significantly improved from 86% (95% CI: 8588) during the BASE period of the CPOE‐based VTE order set to 90% (95% CI: 8893) during the DASH period (P=0.01) and 94% (95% CI: 9396) during the subsequent P4P program (P=0.01) (Figure 3). Both inappropriate prophylaxis and lack of prophylaxis, when indicated, resulted in a non‐compliance rating. During the 3 periods, inappropriate prophylaxis decreased from 7.9% to 6.2% to 2.6% during the BASE, DASH, and subsequent P4P periods, respectively. Similarly, lack of prophylaxis when indicated decreased from 6.1% to 3.2% to 3.1% during the BASE, DASH, and subsequent P4P periods, respectively.

Figure 3

The average compliance of the 4 non‐hospitalist general medicine service teams was initially higher than that of the hospitalist service during the CPOE‐based VTE order set (90%) and DASH (92%) periods, but subsequently plateaued and was exceeded by the hospitalist service during the combined P4P (92%) period (Figure 3). However, there was no statistically significant difference between the general medicine service teams and hospitalist service during the DASH (P=0.15) and subsequent P4P (P=0.76) periods.

We also analyzed the rate of VTE prophylaxis compliance improvement (slope) with cut points at each time period transition (Figure 3). Risk‐appropriate VTE prophylaxis during the BASE period did not exhibit significant improvement as indicated by the slope (P=0.23) (Figure 3). In contrast, during the DASH period, VTE prophylaxis compliance significantly increased by 1.58% per month (95% CI: 0.41‐2.76; P=0.01). The addition of the P4P program, however, did not further significantly increase the rate of compliance (P=0.78).

A subgroup analysis restricted to the 19 providers present during all 3 periods was performed to assess for potential confounding from physician turnover. The percent compliance increased in a similar fashion: BASE period of CPOE‐based VTE order set, 85% (95% CI: 8386); DASH, 90% (95% CI: 8893); and P4P, 94% (95% CI: 9296).

DISCUSSION

A key component of healthcare reform has been value‐based purchasing, which emphasizes extrinsic motivation through the transparency of performance metrics and use of payment incentives to reward quality. Our study evaluates the impact of both extrinsic (payments) and intrinsic (professionalism and peer norms) motivation. It specifically attributed an individual performance metric, VTE prophylaxis, to an attending physician, provided both individualized and group feedback using an electronic dashboard, and incorporated a pay‐for‐performance program. Prescription of risk‐appropriate VTE prophylaxis significantly increased with the implementation of the dashboard and subsequent pay‐for performance program. The fastest rate of improvement occurred after the addition of the dashboard. Sensitivity analyses for provider turnover and comparisons to the general medicine services showed our results to be independent of a general trend of improvement, both at the provider and institutional levels.

Our prior studies demonstrated that order sets significantly improve performance, from a baseline compliance of risk‐appropriate VTE prophylaxis of 66% to 84%.[13, 15, 25] In the current study, compliance was relatively flat during the BASE period, which included these order sets. The greatest rate of continued improvement in compliance occurred during the DASH period, emphasizing both the importance of provider feedback and receptivity and adaptability in the prescribing behavior of hospitalists. Because the goal of a high‐reliability health system is for 100% of patients to receive recommended therapy, multiple approaches are necessary for success.

Nationally, benchmarks for performance measures continue to be raised, with the highest performers achieving above 95%.[26] Additional interventions, such as dashboards and pay‐for‐performance programs, supplement CPOE systems to achieve high reliability. In our study, the compliance rate during the baseline period, which included a CPOE‐based, clinical support‐enabled VTE order set, was 86%. Initially the compliance of the general medicine teams with residents exceeded that of the hospitalist attending teams, which may reflect a greater willingness of resident teams to comply with order sets and automated recommendations. This emphasizes the importance of continuous individual feedback and provider education at the attending physician level to enhance both guideline compliance and decrease provider care variation. Ultimately, with the addition of the dashboard and subsequent pay‐for‐performance program, compliance was increased to 90% and 94%, respectively. Although the major mechanism used by policymakers to improve quality of care is extrinsic motivation, this study demonstrates that intrinsic motivation through peer norms can enhance extrinsic efforts and may be more influential. Both of these programs, dashboards and pay‐for‐performance, may ultimately assist institutions in changing provider behavior and achieving these harder‐to‐achieve higher benchmarks.

We recognize that there are several limitations to our study. First, this is a single‐site program limited to an attending‐physician‐only service. There was strong data support and a defined CPOE algorithm for this initiative. Multi‐site studies will need to overcome the additional challenges of varying service structures and electronic medical record and provider order entry systems. Second, it is difficult to show actual changes in VTE events over time with appropriate prophylaxis. Although VTE prophylaxis is recommended for patients with VTE risk factors, there are conflicting findings about whether prophylaxis prevents VTE events in lower‐risk patients, and current studies suggest that most patients with VTE events are severely ill and develop VTE despite receiving prophylaxis.[27, 28, 29] Our study was underpowered to detect these potential differences in VTE rates, and although the algorithm has been shown to not increase bleeding rates, we did not measure bleeding rates during this study.[12, 15] Our institutional experience suggests that the majority of VTE events occur despite appropriate prophylaxis.[30] Also, VTE prophylaxis may be ordered, but intervening events, such as procedures and changes in risk status or patient refusal, may prevent patients from receiving appropriate prophylaxis.[31, 32] Similarly, hospitals with higher quality scores have higher VTE prophylaxis rates but worse risk‐adjusted VTE rates, which may result from increased surveillance for VTE, suggesting surveillance bias limits the usefulness of the VTE quality measure.[33, 34] Nevertheless, VTE prophylaxis remains a publicly reported Core Measure tied to financial incentives.[4, 5] Third, there may be an unmeasured factor specific to the hospitalist program, which could potentially account for an overall improvement in quality of care. Although the rate of increase in appropriate prophylaxis was not statistically significant during the baseline period, there did appear to be some improvement in prophylaxis toward the end of the period. However, there were no other VTE‐related provider feedback programs being simultaneously pursued during this study. VTE prophylaxis for the non‐hospitalist services showed a relatively stable, non‐increasing compliance rate for the general medical services. Although it was possible for successful residents to age into the hospitalist service, thereby improving rates of prophylaxis based on changes in group makeup, our subgroup analysis of the providers present throughout all phases of the study showed our results to be robust. Similarly, there may have been a cross‐contamination effect of hospitalist faculty who attended on both hospitalist and non‐hospitalist general medicine service teams. This, however, would attenuate any impact of the programs, and thus the effects may in fact be greater than reported. Fourth, establishment of both the dashboard and pay‐for‐performance program required significant institutional and program leadership and resources. To be successful, the dashboard must be in the provider's workflow, transparent, minimize reporter burden, use existing systems, and be actively fed back to providers, ideally those directly entering orders. Our greatest rate of improvement occurred during the feedback‐only phase of this study, emphasizing the importance of physician feedback, provider‐level accountability, and engagement. We suspect that the relatively modest pay‐for‐performance incentive served mainly as a means of engaging providers in self‐monitoring, rather than as a means to change behavior through true incentivization. Although we did not track individual physician views of the dashboard, we reinforced trends, deviations, and expectations at regularly scheduled meetings and provided feedback and patient‐level data to individual providers. Fifth, the design of the pay‐for‐performance program may have also influenced its effectiveness. These types of programs may be more effective when they provide frequent visible, small payments rather than one large payment, and when the payment is framed as a loss rather than a gain.[35] Finally, physician champions and consistent feedback through departmental meetings or visual displays may be required for program success. The initial resources to create the dashboard, continued maintenance and monitoring of performance, and payment of financial incentives all require institutional commitment. A partnership of physicians, program leaders, and institutional administrators is necessary for both initial and continued success.

To achieve performance goals and benchmarks, multiple strategies that combine extrinsic and intrinsic motivation are necessary. As shown by our study, the use of a dashboard and pay‐for‐performance can be tailored to an institution's goals, in line with national standards. The specific goal (risk‐appropriate VTE prophylaxis) and benchmarks (80%, 85%, 90%, 95%) can be individualized to a particular institution. For example, if readmission rates are above target, readmissions could be added as a dashboard metric. The specific benchmark would be determined by historical trends and administrative targets. Similarly, the overall financial incentives could be adjusted based on the financial resources available. Other process measures, such as influenza vaccination screening and administration, could also be targeted. For all of these objectives, continued provider feedback and engagement are critical for progressive success, especially to decrease variability in care at the attending physician level. Incorporating the value‐based purchasing philosophy from the Affordable Care Act, our study suggests that the combination of standardized order sets, real‐time dashboards, and physician‐level incentives may assist hospitals in achieving quality and safety benchmarks, especially at higher targets.

The Affordable Care Act explicitly outlines improving the value of healthcare by increasing quality and decreasing costs. It emphasizes value‐based purchasing, the transparency of performance metrics, and the use of payment incentives to reward quality.[1, 2] Venous thromboembolism (VTE) prophylaxis is one of these publicly reported performance measures. The National Quality Forum recommends that each patient be evaluated on hospital admission and during their hospitalization for VTE risk level and for appropriate thromboprophylaxis to be used, if required.[3] Similarly, the Joint Commission includes appropriate VTE prophylaxis in its Core Measures.[4] Patient experience and performance metrics, including VTE prophylaxis, constitute the hospital value‐based purchasing (VBP) component of healthcare reform.[5] For a hypothetical 327‐bed hospital, an estimated $1.7 million of a hospital's inpatient payments from Medicare will be at risk from VBP alone.[2]

VTE prophylaxis is a common target of quality improvement projects. Effective, safe, and cost‐effective measures to prevent VTE exist, including pharmacologic and mechanical prophylaxis.[6, 7] Despite these measures, compliance rates are often below 50%.[8] Different interventions have been pursued to ensure appropriate VTE prophylaxis, including computerized provider order entry (CPOE), electronic alerts, mandatory VTE risk assessment and prophylaxis, and provider education campaigns.[9] Recent studies show that CPOE systems with mandatory fields can increase VTE prophylaxis rates to above 80%, yet the goal of a high reliability health system is for 100% of patients to receive recommended therapy.[10, 11, 12, 13, 14, 15] Interventions to improve prophylaxis rates that have included multiple strategies, such as computerized order sets, feedback, and education, have been the most effective, increasing compliance to above 90%.[9, 11, 16] These systems can be enhanced with additional interventions such as providing individualized provider education and feedback, understanding of work flow, and ensuring patients receive the prescribed therapies.[12] For example, a physician dashboard could be employed to provide a snapshot and historical trend of key performance indicators using graphical displays and indicators.[17]

Dashboards and pay‐for‐performance programs have been increasingly used to increase the visibility of these metrics, provide feedback, visually display benchmarks and goals, and proactively monitor for achievements and setbacks.[18] Although these strategies are often addressed at departmental (or greater) levels, applying them at the level of the individual provider may assist hospitals in reducing preventable harm and achieving safety and quality goals, especially at higher benchmarks. With their expanding role, hospitalists provide a key opportunity to lead improvement efforts and to study the impact of dashboards and pay‐for performance at the provider level to achieve VTE prophylaxis performance targets. Hospitalists are often the front‐line provider for inpatients and deliver up to 70% of inpatient general medical services.[19] The objective of our study was to evaluate the impact of providing individual provider feedback and employing a pay‐for‐performance program on baseline performance of VTE prophylaxis among hospitalists. We hypothesized that performance feedback through the use of a dashboard would increase appropriate VTE prophylaxis, and this effect would be further augmented by incorporation of a pay‐for‐performance program.

METHODS

Hospitalist Dashboard

In 2010, hospitalist program leaders met with hospital administrators to create a hospitalist dashboard that would provide regularly updated summaries of performance measures for individual hospitalists. The final set of metrics identified included appropriate VTE prophylaxis, length of stay, patients discharged per day, discharges before 3 _pm, depth of coding, patient satisfaction, readmissions, communication with the primary care provider, and time to signature for discharge summaries (Figure 1A). The dashboard was introduced at a general hospitalist meeting during which its purpose, methodology, and accessibility were described; it was subsequently implemented in January 2011.

Figure 1

Benchmarks were established for each metric, standardized to establish a scale ranging from 1 through 9, and incorporated into the dashboard (Figure 1A). Higher scores (creating a larger geometric shape) were desirable. For the VTE prophylaxis measure, scores of 1 through 9 corresponded to <60%, 60% to 64.9%, 65% to 69.9%, 70% to 74.9%, 75% to 79.9%, 80% to 84.9%, 85% to 89.9%, 90% to 94.9%, and 95% American College of Chest Physicians (ACCP)‐compliant VTE prophylaxis, respectively.[12, 20] Each provider was able to access the aggregated dashboard (showing the group mean) and his/her individualized dashboard using an individualized login and password for the institutional portal. This portal is used during the provider's workflow, including medical record review and order entry. Both a polygonal summary graphic (Figure 1A) and trend (Figure 1B) view of the dashboard were available to the provider. A comparison of the individual provider to the hospitalist group average was displayed (Figure 1A). At monthly program meetings, the dashboard, group results, and trends were discussed.

Venous Thromboembolism Prophylaxis Compliance

Our study was performed in a tertiary academic medical center with an approximately 20‐member hospitalist group (the precise membership varied over time), whose responsibilities include, among other clinical duties, staffing a 17‐bed general medicine unit with telemetry. The scope of diagnoses and acuity of patients admitted to the hospitalist service is similar to the housestaff services. Some hospitalist faculty serve both as hospitalist and nonhospitalist general medicine service team attendings, but the comparison groups were staffed by hospitalists for <20% of the time. For admissions, all hospitalists use a standardized general medicine admission order set that is integrated into the CPOE system (Sunrise Clinical Manager; Allscripts, Chicago, IL) and completed for all admitted patients. A mandatory VTE risk screen, which includes an assessment of VTE risk factors and pharmacological prophylaxis contraindications, must be completed by the ordering physician as part of this order set (Figure 2A). The system then prompts the provider with a risk‐appropriate VTE prophylaxis recommendation that the provider may subsequently order, including mechanical prophylaxis (Figure 2B). Based on ACCP VTE prevention guidelines, risk‐appropriate prophylaxis was determined using an electronic algorithm that categorized patients into risk categories based on the presence of major VTE risk factors (Figure 2A).[12, 15, 20] If none of these were present, the provider selected No major risk factors known. Both an assessment of current use of anticoagulation and a clinically high risk of bleeding were also included (Figure 2A). If none of these were present, the provider selected No contraindications known. This algorithm is published in detail elsewhere and has been shown to not increase major bleeding episodes.[12, 15] The VTE risk assessment, but not the VTE order itself, was a mandatory field. This allowed the physician discretion to choose among various pharmacological agents and mechanical mechanisms based on patient and physician preferences.

Figure 2

Compliance of risk‐appropriate VTE prophylaxis was determined 24 hours after the admission order set was completed using an automated electronic query of the CPOE system. Low molecular‐weight heparin prescription was included in the compliance algorithm as acceptable prophylaxis. Prescription of pharmacological VTE prophylaxis when a contraindication was present was considered noncompliant. The metric was assigned to the attending physician who billed for the first inpatient encounter.

RESULTS

Venous Thromboembolism Prophylaxis Compliance

We analyzed 3144 inpatient admissions by 38 hospitalists from 2009 to 2012. The 5 most frequent coded diagnoses were heart failure, acute kidney failure, syncope, pneumonia, and chest pain. Patients had a median length of stay of 3 days [interquartile range: 26]. During the dashboard‐only period, on average, providers improved in compliance by 4% (95% confidence interval [CI]: 35; P<0.001). With the addition of the pay‐for‐performance program, providers improved by an additional 4% (95% CI: 35; P<0.001). Group compliance significantly improved from 86% (95% CI: 8588) during the BASE period of the CPOE‐based VTE order set to 90% (95% CI: 8893) during the DASH period (P=0.01) and 94% (95% CI: 9396) during the subsequent P4P program (P=0.01) (Figure 3). Both inappropriate prophylaxis and lack of prophylaxis, when indicated, resulted in a non‐compliance rating. During the 3 periods, inappropriate prophylaxis decreased from 7.9% to 6.2% to 2.6% during the BASE, DASH, and subsequent P4P periods, respectively. Similarly, lack of prophylaxis when indicated decreased from 6.1% to 3.2% to 3.1% during the BASE, DASH, and subsequent P4P periods, respectively.

Figure 3

The average compliance of the 4 non‐hospitalist general medicine service teams was initially higher than that of the hospitalist service during the CPOE‐based VTE order set (90%) and DASH (92%) periods, but subsequently plateaued and was exceeded by the hospitalist service during the combined P4P (92%) period (Figure 3). However, there was no statistically significant difference between the general medicine service teams and hospitalist service during the DASH (P=0.15) and subsequent P4P (P=0.76) periods.

We also analyzed the rate of VTE prophylaxis compliance improvement (slope) with cut points at each time period transition (Figure 3). Risk‐appropriate VTE prophylaxis during the BASE period did not exhibit significant improvement as indicated by the slope (P=0.23) (Figure 3). In contrast, during the DASH period, VTE prophylaxis compliance significantly increased by 1.58% per month (95% CI: 0.41‐2.76; P=0.01). The addition of the P4P program, however, did not further significantly increase the rate of compliance (P=0.78).

A subgroup analysis restricted to the 19 providers present during all 3 periods was performed to assess for potential confounding from physician turnover. The percent compliance increased in a similar fashion: BASE period of CPOE‐based VTE order set, 85% (95% CI: 8386); DASH, 90% (95% CI: 8893); and P4P, 94% (95% CI: 9296).

DISCUSSION

A key component of healthcare reform has been value‐based purchasing, which emphasizes extrinsic motivation through the transparency of performance metrics and use of payment incentives to reward quality. Our study evaluates the impact of both extrinsic (payments) and intrinsic (professionalism and peer norms) motivation. It specifically attributed an individual performance metric, VTE prophylaxis, to an attending physician, provided both individualized and group feedback using an electronic dashboard, and incorporated a pay‐for‐performance program. Prescription of risk‐appropriate VTE prophylaxis significantly increased with the implementation of the dashboard and subsequent pay‐for performance program. The fastest rate of improvement occurred after the addition of the dashboard. Sensitivity analyses for provider turnover and comparisons to the general medicine services showed our results to be independent of a general trend of improvement, both at the provider and institutional levels.

Our prior studies demonstrated that order sets significantly improve performance, from a baseline compliance of risk‐appropriate VTE prophylaxis of 66% to 84%.[13, 15, 25] In the current study, compliance was relatively flat during the BASE period, which included these order sets. The greatest rate of continued improvement in compliance occurred during the DASH period, emphasizing both the importance of provider feedback and receptivity and adaptability in the prescribing behavior of hospitalists. Because the goal of a high‐reliability health system is for 100% of patients to receive recommended therapy, multiple approaches are necessary for success.

Nationally, benchmarks for performance measures continue to be raised, with the highest performers achieving above 95%.[26] Additional interventions, such as dashboards and pay‐for‐performance programs, supplement CPOE systems to achieve high reliability. In our study, the compliance rate during the baseline period, which included a CPOE‐based, clinical support‐enabled VTE order set, was 86%. Initially the compliance of the general medicine teams with residents exceeded that of the hospitalist attending teams, which may reflect a greater willingness of resident teams to comply with order sets and automated recommendations. This emphasizes the importance of continuous individual feedback and provider education at the attending physician level to enhance both guideline compliance and decrease provider care variation. Ultimately, with the addition of the dashboard and subsequent pay‐for‐performance program, compliance was increased to 90% and 94%, respectively. Although the major mechanism used by policymakers to improve quality of care is extrinsic motivation, this study demonstrates that intrinsic motivation through peer norms can enhance extrinsic efforts and may be more influential. Both of these programs, dashboards and pay‐for‐performance, may ultimately assist institutions in changing provider behavior and achieving these harder‐to‐achieve higher benchmarks.

We recognize that there are several limitations to our study. First, this is a single‐site program limited to an attending‐physician‐only service. There was strong data support and a defined CPOE algorithm for this initiative. Multi‐site studies will need to overcome the additional challenges of varying service structures and electronic medical record and provider order entry systems. Second, it is difficult to show actual changes in VTE events over time with appropriate prophylaxis. Although VTE prophylaxis is recommended for patients with VTE risk factors, there are conflicting findings about whether prophylaxis prevents VTE events in lower‐risk patients, and current studies suggest that most patients with VTE events are severely ill and develop VTE despite receiving prophylaxis.[27, 28, 29] Our study was underpowered to detect these potential differences in VTE rates, and although the algorithm has been shown to not increase bleeding rates, we did not measure bleeding rates during this study.[12, 15] Our institutional experience suggests that the majority of VTE events occur despite appropriate prophylaxis.[30] Also, VTE prophylaxis may be ordered, but intervening events, such as procedures and changes in risk status or patient refusal, may prevent patients from receiving appropriate prophylaxis.[31, 32] Similarly, hospitals with higher quality scores have higher VTE prophylaxis rates but worse risk‐adjusted VTE rates, which may result from increased surveillance for VTE, suggesting surveillance bias limits the usefulness of the VTE quality measure.[33, 34] Nevertheless, VTE prophylaxis remains a publicly reported Core Measure tied to financial incentives.[4, 5] Third, there may be an unmeasured factor specific to the hospitalist program, which could potentially account for an overall improvement in quality of care. Although the rate of increase in appropriate prophylaxis was not statistically significant during the baseline period, there did appear to be some improvement in prophylaxis toward the end of the period. However, there were no other VTE‐related provider feedback programs being simultaneously pursued during this study. VTE prophylaxis for the non‐hospitalist services showed a relatively stable, non‐increasing compliance rate for the general medical services. Although it was possible for successful residents to age into the hospitalist service, thereby improving rates of prophylaxis based on changes in group makeup, our subgroup analysis of the providers present throughout all phases of the study showed our results to be robust. Similarly, there may have been a cross‐contamination effect of hospitalist faculty who attended on both hospitalist and non‐hospitalist general medicine service teams. This, however, would attenuate any impact of the programs, and thus the effects may in fact be greater than reported. Fourth, establishment of both the dashboard and pay‐for‐performance program required significant institutional and program leadership and resources. To be successful, the dashboard must be in the provider's workflow, transparent, minimize reporter burden, use existing systems, and be actively fed back to providers, ideally those directly entering orders. Our greatest rate of improvement occurred during the feedback‐only phase of this study, emphasizing the importance of physician feedback, provider‐level accountability, and engagement. We suspect that the relatively modest pay‐for‐performance incentive served mainly as a means of engaging providers in self‐monitoring, rather than as a means to change behavior through true incentivization. Although we did not track individual physician views of the dashboard, we reinforced trends, deviations, and expectations at regularly scheduled meetings and provided feedback and patient‐level data to individual providers. Fifth, the design of the pay‐for‐performance program may have also influenced its effectiveness. These types of programs may be more effective when they provide frequent visible, small payments rather than one large payment, and when the payment is framed as a loss rather than a gain.[35] Finally, physician champions and consistent feedback through departmental meetings or visual displays may be required for program success. The initial resources to create the dashboard, continued maintenance and monitoring of performance, and payment of financial incentives all require institutional commitment. A partnership of physicians, program leaders, and institutional administrators is necessary for both initial and continued success.

To achieve performance goals and benchmarks, multiple strategies that combine extrinsic and intrinsic motivation are necessary. As shown by our study, the use of a dashboard and pay‐for‐performance can be tailored to an institution's goals, in line with national standards. The specific goal (risk‐appropriate VTE prophylaxis) and benchmarks (80%, 85%, 90%, 95%) can be individualized to a particular institution. For example, if readmission rates are above target, readmissions could be added as a dashboard metric. The specific benchmark would be determined by historical trends and administrative targets. Similarly, the overall financial incentives could be adjusted based on the financial resources available. Other process measures, such as influenza vaccination screening and administration, could also be targeted. For all of these objectives, continued provider feedback and engagement are critical for progressive success, especially to decrease variability in care at the attending physician level. Incorporating the value‐based purchasing philosophy from the Affordable Care Act, our study suggests that the combination of standardized order sets, real‐time dashboards, and physician‐level incentives may assist hospitals in achieving quality and safety benchmarks, especially at higher targets.

The Affordable Care Act explicitly outlines improving the value of healthcare by increasing quality and decreasing costs. It emphasizes value‐based purchasing, the transparency of performance metrics, and the use of payment incentives to reward quality.[1, 2] Venous thromboembolism (VTE) prophylaxis is one of these publicly reported performance measures. The National Quality Forum recommends that each patient be evaluated on hospital admission and during their hospitalization for VTE risk level and for appropriate thromboprophylaxis to be used, if required.[3] Similarly, the Joint Commission includes appropriate VTE prophylaxis in its Core Measures.[4] Patient experience and performance metrics, including VTE prophylaxis, constitute the hospital value‐based purchasing (VBP) component of healthcare reform.[5] For a hypothetical 327‐bed hospital, an estimated $1.7 million of a hospital's inpatient payments from Medicare will be at risk from VBP alone.[2]

VTE prophylaxis is a common target of quality improvement projects. Effective, safe, and cost‐effective measures to prevent VTE exist, including pharmacologic and mechanical prophylaxis.[6, 7] Despite these measures, compliance rates are often below 50%.[8] Different interventions have been pursued to ensure appropriate VTE prophylaxis, including computerized provider order entry (CPOE), electronic alerts, mandatory VTE risk assessment and prophylaxis, and provider education campaigns.[9] Recent studies show that CPOE systems with mandatory fields can increase VTE prophylaxis rates to above 80%, yet the goal of a high reliability health system is for 100% of patients to receive recommended therapy.[10, 11, 12, 13, 14, 15] Interventions to improve prophylaxis rates that have included multiple strategies, such as computerized order sets, feedback, and education, have been the most effective, increasing compliance to above 90%.[9, 11, 16] These systems can be enhanced with additional interventions such as providing individualized provider education and feedback, understanding of work flow, and ensuring patients receive the prescribed therapies.[12] For example, a physician dashboard could be employed to provide a snapshot and historical trend of key performance indicators using graphical displays and indicators.[17]

Dashboards and pay‐for‐performance programs have been increasingly used to increase the visibility of these metrics, provide feedback, visually display benchmarks and goals, and proactively monitor for achievements and setbacks.[18] Although these strategies are often addressed at departmental (or greater) levels, applying them at the level of the individual provider may assist hospitals in reducing preventable harm and achieving safety and quality goals, especially at higher benchmarks. With their expanding role, hospitalists provide a key opportunity to lead improvement efforts and to study the impact of dashboards and pay‐for performance at the provider level to achieve VTE prophylaxis performance targets. Hospitalists are often the front‐line provider for inpatients and deliver up to 70% of inpatient general medical services.[19] The objective of our study was to evaluate the impact of providing individual provider feedback and employing a pay‐for‐performance program on baseline performance of VTE prophylaxis among hospitalists. We hypothesized that performance feedback through the use of a dashboard would increase appropriate VTE prophylaxis, and this effect would be further augmented by incorporation of a pay‐for‐performance program.

METHODS

Hospitalist Dashboard

In 2010, hospitalist program leaders met with hospital administrators to create a hospitalist dashboard that would provide regularly updated summaries of performance measures for individual hospitalists. The final set of metrics identified included appropriate VTE prophylaxis, length of stay, patients discharged per day, discharges before 3 _pm, depth of coding, patient satisfaction, readmissions, communication with the primary care provider, and time to signature for discharge summaries (Figure 1A). The dashboard was introduced at a general hospitalist meeting during which its purpose, methodology, and accessibility were described; it was subsequently implemented in January 2011.

Figure 1

Benchmarks were established for each metric, standardized to establish a scale ranging from 1 through 9, and incorporated into the dashboard (Figure 1A). Higher scores (creating a larger geometric shape) were desirable. For the VTE prophylaxis measure, scores of 1 through 9 corresponded to <60%, 60% to 64.9%, 65% to 69.9%, 70% to 74.9%, 75% to 79.9%, 80% to 84.9%, 85% to 89.9%, 90% to 94.9%, and 95% American College of Chest Physicians (ACCP)‐compliant VTE prophylaxis, respectively.[12, 20] Each provider was able to access the aggregated dashboard (showing the group mean) and his/her individualized dashboard using an individualized login and password for the institutional portal. This portal is used during the provider's workflow, including medical record review and order entry. Both a polygonal summary graphic (Figure 1A) and trend (Figure 1B) view of the dashboard were available to the provider. A comparison of the individual provider to the hospitalist group average was displayed (Figure 1A). At monthly program meetings, the dashboard, group results, and trends were discussed.

Venous Thromboembolism Prophylaxis Compliance

Our study was performed in a tertiary academic medical center with an approximately 20‐member hospitalist group (the precise membership varied over time), whose responsibilities include, among other clinical duties, staffing a 17‐bed general medicine unit with telemetry. The scope of diagnoses and acuity of patients admitted to the hospitalist service is similar to the housestaff services. Some hospitalist faculty serve both as hospitalist and nonhospitalist general medicine service team attendings, but the comparison groups were staffed by hospitalists for <20% of the time. For admissions, all hospitalists use a standardized general medicine admission order set that is integrated into the CPOE system (Sunrise Clinical Manager; Allscripts, Chicago, IL) and completed for all admitted patients. A mandatory VTE risk screen, which includes an assessment of VTE risk factors and pharmacological prophylaxis contraindications, must be completed by the ordering physician as part of this order set (Figure 2A). The system then prompts the provider with a risk‐appropriate VTE prophylaxis recommendation that the provider may subsequently order, including mechanical prophylaxis (Figure 2B). Based on ACCP VTE prevention guidelines, risk‐appropriate prophylaxis was determined using an electronic algorithm that categorized patients into risk categories based on the presence of major VTE risk factors (Figure 2A).[12, 15, 20] If none of these were present, the provider selected No major risk factors known. Both an assessment of current use of anticoagulation and a clinically high risk of bleeding were also included (Figure 2A). If none of these were present, the provider selected No contraindications known. This algorithm is published in detail elsewhere and has been shown to not increase major bleeding episodes.[12, 15] The VTE risk assessment, but not the VTE order itself, was a mandatory field. This allowed the physician discretion to choose among various pharmacological agents and mechanical mechanisms based on patient and physician preferences.

Figure 2

Compliance of risk‐appropriate VTE prophylaxis was determined 24 hours after the admission order set was completed using an automated electronic query of the CPOE system. Low molecular‐weight heparin prescription was included in the compliance algorithm as acceptable prophylaxis. Prescription of pharmacological VTE prophylaxis when a contraindication was present was considered noncompliant. The metric was assigned to the attending physician who billed for the first inpatient encounter.

RESULTS

Venous Thromboembolism Prophylaxis Compliance

We analyzed 3144 inpatient admissions by 38 hospitalists from 2009 to 2012. The 5 most frequent coded diagnoses were heart failure, acute kidney failure, syncope, pneumonia, and chest pain. Patients had a median length of stay of 3 days [interquartile range: 26]. During the dashboard‐only period, on average, providers improved in compliance by 4% (95% confidence interval [CI]: 35; P<0.001). With the addition of the pay‐for‐performance program, providers improved by an additional 4% (95% CI: 35; P<0.001). Group compliance significantly improved from 86% (95% CI: 8588) during the BASE period of the CPOE‐based VTE order set to 90% (95% CI: 8893) during the DASH period (P=0.01) and 94% (95% CI: 9396) during the subsequent P4P program (P=0.01) (Figure 3). Both inappropriate prophylaxis and lack of prophylaxis, when indicated, resulted in a non‐compliance rating. During the 3 periods, inappropriate prophylaxis decreased from 7.9% to 6.2% to 2.6% during the BASE, DASH, and subsequent P4P periods, respectively. Similarly, lack of prophylaxis when indicated decreased from 6.1% to 3.2% to 3.1% during the BASE, DASH, and subsequent P4P periods, respectively.

Figure 3

The average compliance of the 4 non‐hospitalist general medicine service teams was initially higher than that of the hospitalist service during the CPOE‐based VTE order set (90%) and DASH (92%) periods, but subsequently plateaued and was exceeded by the hospitalist service during the combined P4P (92%) period (Figure 3). However, there was no statistically significant difference between the general medicine service teams and hospitalist service during the DASH (P=0.15) and subsequent P4P (P=0.76) periods.

We also analyzed the rate of VTE prophylaxis compliance improvement (slope) with cut points at each time period transition (Figure 3). Risk‐appropriate VTE prophylaxis during the BASE period did not exhibit significant improvement as indicated by the slope (P=0.23) (Figure 3). In contrast, during the DASH period, VTE prophylaxis compliance significantly increased by 1.58% per month (95% CI: 0.41‐2.76; P=0.01). The addition of the P4P program, however, did not further significantly increase the rate of compliance (P=0.78).

A subgroup analysis restricted to the 19 providers present during all 3 periods was performed to assess for potential confounding from physician turnover. The percent compliance increased in a similar fashion: BASE period of CPOE‐based VTE order set, 85% (95% CI: 8386); DASH, 90% (95% CI: 8893); and P4P, 94% (95% CI: 9296).

DISCUSSION

A key component of healthcare reform has been value‐based purchasing, which emphasizes extrinsic motivation through the transparency of performance metrics and use of payment incentives to reward quality. Our study evaluates the impact of both extrinsic (payments) and intrinsic (professionalism and peer norms) motivation. It specifically attributed an individual performance metric, VTE prophylaxis, to an attending physician, provided both individualized and group feedback using an electronic dashboard, and incorporated a pay‐for‐performance program. Prescription of risk‐appropriate VTE prophylaxis significantly increased with the implementation of the dashboard and subsequent pay‐for performance program. The fastest rate of improvement occurred after the addition of the dashboard. Sensitivity analyses for provider turnover and comparisons to the general medicine services showed our results to be independent of a general trend of improvement, both at the provider and institutional levels.

Our prior studies demonstrated that order sets significantly improve performance, from a baseline compliance of risk‐appropriate VTE prophylaxis of 66% to 84%.[13, 15, 25] In the current study, compliance was relatively flat during the BASE period, which included these order sets. The greatest rate of continued improvement in compliance occurred during the DASH period, emphasizing both the importance of provider feedback and receptivity and adaptability in the prescribing behavior of hospitalists. Because the goal of a high‐reliability health system is for 100% of patients to receive recommended therapy, multiple approaches are necessary for success.

Nationally, benchmarks for performance measures continue to be raised, with the highest performers achieving above 95%.[26] Additional interventions, such as dashboards and pay‐for‐performance programs, supplement CPOE systems to achieve high reliability. In our study, the compliance rate during the baseline period, which included a CPOE‐based, clinical support‐enabled VTE order set, was 86%. Initially the compliance of the general medicine teams with residents exceeded that of the hospitalist attending teams, which may reflect a greater willingness of resident teams to comply with order sets and automated recommendations. This emphasizes the importance of continuous individual feedback and provider education at the attending physician level to enhance both guideline compliance and decrease provider care variation. Ultimately, with the addition of the dashboard and subsequent pay‐for‐performance program, compliance was increased to 90% and 94%, respectively. Although the major mechanism used by policymakers to improve quality of care is extrinsic motivation, this study demonstrates that intrinsic motivation through peer norms can enhance extrinsic efforts and may be more influential. Both of these programs, dashboards and pay‐for‐performance, may ultimately assist institutions in changing provider behavior and achieving these harder‐to‐achieve higher benchmarks.

We recognize that there are several limitations to our study. First, this is a single‐site program limited to an attending‐physician‐only service. There was strong data support and a defined CPOE algorithm for this initiative. Multi‐site studies will need to overcome the additional challenges of varying service structures and electronic medical record and provider order entry systems. Second, it is difficult to show actual changes in VTE events over time with appropriate prophylaxis. Although VTE prophylaxis is recommended for patients with VTE risk factors, there are conflicting findings about whether prophylaxis prevents VTE events in lower‐risk patients, and current studies suggest that most patients with VTE events are severely ill and develop VTE despite receiving prophylaxis.[27, 28, 29] Our study was underpowered to detect these potential differences in VTE rates, and although the algorithm has been shown to not increase bleeding rates, we did not measure bleeding rates during this study.[12, 15] Our institutional experience suggests that the majority of VTE events occur despite appropriate prophylaxis.[30] Also, VTE prophylaxis may be ordered, but intervening events, such as procedures and changes in risk status or patient refusal, may prevent patients from receiving appropriate prophylaxis.[31, 32] Similarly, hospitals with higher quality scores have higher VTE prophylaxis rates but worse risk‐adjusted VTE rates, which may result from increased surveillance for VTE, suggesting surveillance bias limits the usefulness of the VTE quality measure.[33, 34] Nevertheless, VTE prophylaxis remains a publicly reported Core Measure tied to financial incentives.[4, 5] Third, there may be an unmeasured factor specific to the hospitalist program, which could potentially account for an overall improvement in quality of care. Although the rate of increase in appropriate prophylaxis was not statistically significant during the baseline period, there did appear to be some improvement in prophylaxis toward the end of the period. However, there were no other VTE‐related provider feedback programs being simultaneously pursued during this study. VTE prophylaxis for the non‐hospitalist services showed a relatively stable, non‐increasing compliance rate for the general medical services. Although it was possible for successful residents to age into the hospitalist service, thereby improving rates of prophylaxis based on changes in group makeup, our subgroup analysis of the providers present throughout all phases of the study showed our results to be robust. Similarly, there may have been a cross‐contamination effect of hospitalist faculty who attended on both hospitalist and non‐hospitalist general medicine service teams. This, however, would attenuate any impact of the programs, and thus the effects may in fact be greater than reported. Fourth, establishment of both the dashboard and pay‐for‐performance program required significant institutional and program leadership and resources. To be successful, the dashboard must be in the provider's workflow, transparent, minimize reporter burden, use existing systems, and be actively fed back to providers, ideally those directly entering orders. Our greatest rate of improvement occurred during the feedback‐only phase of this study, emphasizing the importance of physician feedback, provider‐level accountability, and engagement. We suspect that the relatively modest pay‐for‐performance incentive served mainly as a means of engaging providers in self‐monitoring, rather than as a means to change behavior through true incentivization. Although we did not track individual physician views of the dashboard, we reinforced trends, deviations, and expectations at regularly scheduled meetings and provided feedback and patient‐level data to individual providers. Fifth, the design of the pay‐for‐performance program may have also influenced its effectiveness. These types of programs may be more effective when they provide frequent visible, small payments rather than one large payment, and when the payment is framed as a loss rather than a gain.[35] Finally, physician champions and consistent feedback through departmental meetings or visual displays may be required for program success. The initial resources to create the dashboard, continued maintenance and monitoring of performance, and payment of financial incentives all require institutional commitment. A partnership of physicians, program leaders, and institutional administrators is necessary for both initial and continued success.

To achieve performance goals and benchmarks, multiple strategies that combine extrinsic and intrinsic motivation are necessary. As shown by our study, the use of a dashboard and pay‐for‐performance can be tailored to an institution's goals, in line with national standards. The specific goal (risk‐appropriate VTE prophylaxis) and benchmarks (80%, 85%, 90%, 95%) can be individualized to a particular institution. For example, if readmission rates are above target, readmissions could be added as a dashboard metric. The specific benchmark would be determined by historical trends and administrative targets. Similarly, the overall financial incentives could be adjusted based on the financial resources available. Other process measures, such as influenza vaccination screening and administration, could also be targeted. For all of these objectives, continued provider feedback and engagement are critical for progressive success, especially to decrease variability in care at the attending physician level. Incorporating the value‐based purchasing philosophy from the Affordable Care Act, our study suggests that the combination of standardized order sets, real‐time dashboards, and physician‐level incentives may assist hospitals in achieving quality and safety benchmarks, especially at higher targets.

Hospitals typically reduce staffing levels and the availability of diagnostic, laboratory, and treatment services on weekends, and patients admitted on weekends exhibit poorer in‐hospital outcomes for several medical conditions.[1, 2, 3, 4, 5, 6, 7, 8, 9] Whether or not patients discharged on weekends have worse clinical outcomes has been less well studied.[10, 11, 12] Discharge rates on Saturday and Sunday are lower than for the other 5 days of the week,[12] but bed shortages and hospital overcrowding have increased the demand for maximizing 24/7 week‐round discharge efficiency. Given that the number of patients discharged on weekends is likely to continue to increase, it is important to assess the risk of weekend discharge on outcomes monitored as performance indicators by organizations such as the Centers for Medicare and Medicaid Services, the American Medical Association Physicians Consortium for Performance Improvement, the National Quality Forum, and the Joint Commission.

Thus, we designed this study to evaluate baseline characteristics, length of stay (LOS), and postdischarge outcomes for general internal medicine (GIM) patients in teaching hospitals discharged on weekends compared to weekdays. Our objective was to determine whether postdischarge outcomes differed for patients discharged on weekends versus weekdays.

METHODS

RESULTS

Weekday Versus Weekend Discharge

Although patients admitted on weekdays and weekends were very similar (data available upon request), patients discharged on weekends (compared to those discharged on weekdays) were younger, more likely to be discharged home without additional support, and had fewer comorbidities (Table 1, Figure 1). Patients discharged on weekends had shorter lengths of stay than those discharged on weekdays (5.6 days vs 7.9 days, P<0.0001). In adjusted linear regression analyses, this 2.3‐day difference remained statistically significant (adjusted P value <0.0001).

Figure 1

Patients discharged on a weekend exhibited lower unadjusted 30‐day rates of death or readmission than those discharged on a weekday (10.6% vs 13.2%), but these differences disappeared after multivariable adjustment that accounted for differences in risk profile (aOR: 0.94, 95% CI: 0.771.16 (Table 2). Results were similar in sensitivity analyses adjusting for LACE scores without LOS included (aOR: 0.88, 95% CI: 0.711.08) or adjusting for LACE scores using expected LOS rather than actual LOS (aOR: 0.90, 95% CI: 0.731.10). Restricting the analysis to only those patients deemed to be at high risk for events due to LACE scores of 10 or greater confirmed that weekend and weekday discharges had similar outcomes in the first 30 days after discharge (aOR: 1.09, 95% CI: 0.851.41, Table 2). Similar patterns were seen when we included ED visits as part of the composite endpoint (ie, death, unplanned readmission, or unplanned ED visit within 30 days of discharge) (Table 2).

Table 2.Postdischarge Outcomes After a General Internal Medicine Hospitalization in a Teaching Hospital

	Weekend Discharge, n/N (%)	Weekday Discharge, n/N (%)	Unadjusted P Value	aOR* (95% CI)	Adjusted P Value
NOTE: Abbreviations: aOR, adjusted odds ratio; CI, confidence interval; ED, emergency department; LACE, length of hospital stay, acuity of admission, comorbidity burden quantified using the Charlson Comorbidity Index, and emergency department visits in the 6 months prior to admission. *Multivariable models adjust for age, sex, hospital, and LACE score at time of discharge from index hospitalization. Weekday discharge is reference group for odds ratios.
Death/readmission within 30 days
All 7 teaching hospitals, all patients	121/1146 (10.6)	901/6845 (13.2)	0.01	0.94 (0.77‐1.16)	0.58
All 7 teaching hospitals, but only patients with LACE <10	37/647 (5.7)	225/2753 (8.2)	0.04	0.72 (0.50, 1.03)	0.07
All 7 teaching hospitals, but only patients with LACE 10	84/499 (16.8)	676/4092 (16.5)	0.86	1.09 (0.85‐1.41)	0.49
Death/readmission/ED visit within 30 days
All 7 teaching hospitals, all patients	218/1146 (19.0)	1445/6845 (21.1)	0.11	0.98 (0.83‐1.15)	0.79
All 7 teaching hospitals, but only patients with LACE <10	90/647 (13.9)	460/2753 (16.7)	0.08	0.83 (0.64‐1.06)	0.13
All 7 teaching hospitals, but only patients with LACE 10	128/499 (25.7)	985/4092 (24.1)	0.44	1.12 (0.90‐1.39)	0.31
Death within 30 days
All 7 teaching hospitals, all patients	24/1146 (2.1)	215/6845 (3.1)	0.05	0.97 (0.63‐1.51)	0.89
All 7 teaching hospitals, but only patients with LACE <10	4/647 (0.6)	23/2753 (0.8)	0.58	0.89 (0.30, 2.62)	0.83
All 7 teaching hospitals, but only patients with LACE 10	20/499 (4.0)	192/4092 (4.7)	0.49	0.99 (0.61‐1.61)	0.98
Readmission within 30 days
All 7 teaching hospitals, all patients	105/1146 (9.2)	751/6845 (11.0)	0.07	0.94 (0.76‐1.17)	0.59
All 7 teaching hospitals, but only patients with LACE <10	33/647 (5.1)	211/2753 (7.7)	0.02	0.68 (0.46‐0.99)	0.04
All 7 teaching hospitals, but only patients with LACE 10	72/499 (14.4)	540/4092 (13.2)	0.44	1.14 (0.87‐1.49)	0.34
ED visit within 30 days
All 7 teaching hospitals, all patients	182/1146 (15.9)	1118/6845 (16.3)	0.70	1.00 (0.84‐1.19)	0.99
All 7 teaching hospitals, but only patients with LACE <10	83/647 (12.8)	412/2753 (15.0)	0.17	0.84 (0.65, 1.09)	0.20
All 7 teaching hospitals, but only patients with LACE 10	99/499 (19.8)	706/4092 (17.3)	0.15	1.17 (0.92‐1.48)	0.20

DISCUSSION

Our data suggest that patients discharged from the GIM teaching wards we studied on weekends were appropriately triaged, as they did not exhibit a higher risk of adverse events postdischarge. Although patients discharged on weekends tended to be younger and had less comorbidities than those discharged during the week, we adjusted for baseline covariates in analyses, and we did not find an association between weekend discharge and increased postdischarge events even among the subset of patients deemed to be at high risk for postdischarge adverse events (based on high LACE scores). To our knowledge, although we previously examined this issue in patients with a most‐responsible diagnosis of heart failure,[10] examining weekend versus weekday discharges in the full gamut of general medical patients admitted to teaching hospitals has not previously been examined.

In our previous study[10] of over 24,000 heart failure patients discharged over 10 years (up to June 2009, therefore no overlap with any patients in this study), we also found that patients discharged on the weekends were younger, had fewer comorbidities, and shorter lengths of stay. Although postdischarge death/readmission rates were higher for weekend discharged patients in our earlier study (21.1% vs 19.5%, adjusted hazard ratio: 1.15, 95% CI: 1.061.25), it is worth noting that this was almost entirely driven by data from nonteaching hospitals and cardiology wards. Thus, it is important to reiterate that the findings in our current study are for GIM wards in teaching hospitals and may not be generalizable to less‐structured nonteaching settings.

Although we did not study physician decision making, our results suggest that physicians are incorporating discharge day into their discharge decision making. They may be selecting younger patients with less comorbidities for weekend discharges, or they may be delaying the discharges of older patients with more comorbidities for weekday discharges. Either is not surprising given the realities of weekend inpatient care: reduced staffing and frequent cross‐coverage (of physicians, nurses, physiotherapists, pharmacists, and occupational therapists), limited support services (such as laboratory services or diagnostic imaging), and decreased availability of community services (including home care and social support services).[17] For example, in 1 large US heart failure registry, patients discharged on a weekend received less complete discharge instructions than those discharged on weekdays.[11] Given that early follow‐up postdischarge is associated with better outcomes,[18, 19] future studies should also explore whether patterns of patient follow‐up differ after weekend versus weekday discharges.

Although we were able to capture all interactions with the healthcare system in a single payer system with universal access, there are some limitations to our study. First, we used administrative data, which preclude fully adjusting for severity of diagnoses or functional status, although we used proxies such as admission from/discharge to a long‐term care facility.[20, 21] Second, we did not have access to process of care measures such as diagnostic testing or prescribing data, and thus cannot determine whether quality of care or patient adherence differed by the day of the week they were discharged on, although this seems unlikely. Third, although postdischarge follow‐up may be associated with better outcomes,[18, 19] we were unable to adjust for patterns of outpatient follow‐up in this study. Fourth, we acknowledge that death or readmission soon after discharge does not necessarily mean that the quality of care during the preceding hospitalization was suboptimal or that these deaths or readmissions were even potentially preventable. Many factors influence postdischarge mortality and/or readmission, and quality of inpatient care is only one.[22, 23, 24, 25] Fifth, although some may express concern that LOS may be a mediator in the causal pathway between discharge decision and postdischarge events, and that adjusting for LOS in analyses could thus spuriously obscure a true association, it is worth pointing out that our 2 sensitivity analyses to explore this (the 1 in which we excluded LOS from the analyses and the 1 in which we included expected LOS rather than the actual LOS) revealed nearly identical point estimates and 95% CI as our main analysis. Finally, as our study is observational, we cannot definitively conclude causality, nor can we exclude an 18% excess risk for patients discharged on weekends (or a 22% lower risk either), given our 95% CI for postdischarge adverse outcomes.

CONCLUSION

We found that the proportion of patients discharged on weekends is lower than the proportion admitted on weekends. We also found that lower risk/less severely ill patients appear to be preferentially discharged on weekends, and as a result, postdischarge outcomes are similar between weekend and weekday discharges despite shorter LOS and less availability of outpatient resources for patients discharged on a weekend. The reasons why more complicated patients are not discharged on weekends deserves further study, as safely increasing weekend discharge rates would improve efficiency and safety (by reducing unnecessary exposure to in‐hospital adverse events such as falls, unnecessary urinary catheterizations, and healthcare‐acquired infections). Although hospital admission has become a 24/7 business, we believe that hospital discharge processes should strive for the same level of efficiency.

Hospitals typically reduce staffing levels and the availability of diagnostic, laboratory, and treatment services on weekends, and patients admitted on weekends exhibit poorer in‐hospital outcomes for several medical conditions.[1, 2, 3, 4, 5, 6, 7, 8, 9] Whether or not patients discharged on weekends have worse clinical outcomes has been less well studied.[10, 11, 12] Discharge rates on Saturday and Sunday are lower than for the other 5 days of the week,[12] but bed shortages and hospital overcrowding have increased the demand for maximizing 24/7 week‐round discharge efficiency. Given that the number of patients discharged on weekends is likely to continue to increase, it is important to assess the risk of weekend discharge on outcomes monitored as performance indicators by organizations such as the Centers for Medicare and Medicaid Services, the American Medical Association Physicians Consortium for Performance Improvement, the National Quality Forum, and the Joint Commission.

Thus, we designed this study to evaluate baseline characteristics, length of stay (LOS), and postdischarge outcomes for general internal medicine (GIM) patients in teaching hospitals discharged on weekends compared to weekdays. Our objective was to determine whether postdischarge outcomes differed for patients discharged on weekends versus weekdays.

METHODS

RESULTS

Weekday Versus Weekend Discharge

Although patients admitted on weekdays and weekends were very similar (data available upon request), patients discharged on weekends (compared to those discharged on weekdays) were younger, more likely to be discharged home without additional support, and had fewer comorbidities (Table 1, Figure 1). Patients discharged on weekends had shorter lengths of stay than those discharged on weekdays (5.6 days vs 7.9 days, P<0.0001). In adjusted linear regression analyses, this 2.3‐day difference remained statistically significant (adjusted P value <0.0001).

Figure 1

Patients discharged on a weekend exhibited lower unadjusted 30‐day rates of death or readmission than those discharged on a weekday (10.6% vs 13.2%), but these differences disappeared after multivariable adjustment that accounted for differences in risk profile (aOR: 0.94, 95% CI: 0.771.16 (Table 2). Results were similar in sensitivity analyses adjusting for LACE scores without LOS included (aOR: 0.88, 95% CI: 0.711.08) or adjusting for LACE scores using expected LOS rather than actual LOS (aOR: 0.90, 95% CI: 0.731.10). Restricting the analysis to only those patients deemed to be at high risk for events due to LACE scores of 10 or greater confirmed that weekend and weekday discharges had similar outcomes in the first 30 days after discharge (aOR: 1.09, 95% CI: 0.851.41, Table 2). Similar patterns were seen when we included ED visits as part of the composite endpoint (ie, death, unplanned readmission, or unplanned ED visit within 30 days of discharge) (Table 2).

Table 2.Postdischarge Outcomes After a General Internal Medicine Hospitalization in a Teaching Hospital

	Weekend Discharge, n/N (%)	Weekday Discharge, n/N (%)	Unadjusted P Value	aOR* (95% CI)	Adjusted P Value
NOTE: Abbreviations: aOR, adjusted odds ratio; CI, confidence interval; ED, emergency department; LACE, length of hospital stay, acuity of admission, comorbidity burden quantified using the Charlson Comorbidity Index, and emergency department visits in the 6 months prior to admission. *Multivariable models adjust for age, sex, hospital, and LACE score at time of discharge from index hospitalization. Weekday discharge is reference group for odds ratios.
Death/readmission within 30 days
All 7 teaching hospitals, all patients	121/1146 (10.6)	901/6845 (13.2)	0.01	0.94 (0.77‐1.16)	0.58
All 7 teaching hospitals, but only patients with LACE <10	37/647 (5.7)	225/2753 (8.2)	0.04	0.72 (0.50, 1.03)	0.07
All 7 teaching hospitals, but only patients with LACE 10	84/499 (16.8)	676/4092 (16.5)	0.86	1.09 (0.85‐1.41)	0.49
Death/readmission/ED visit within 30 days
All 7 teaching hospitals, all patients	218/1146 (19.0)	1445/6845 (21.1)	0.11	0.98 (0.83‐1.15)	0.79
All 7 teaching hospitals, but only patients with LACE <10	90/647 (13.9)	460/2753 (16.7)	0.08	0.83 (0.64‐1.06)	0.13
All 7 teaching hospitals, but only patients with LACE 10	128/499 (25.7)	985/4092 (24.1)	0.44	1.12 (0.90‐1.39)	0.31
Death within 30 days
All 7 teaching hospitals, all patients	24/1146 (2.1)	215/6845 (3.1)	0.05	0.97 (0.63‐1.51)	0.89
All 7 teaching hospitals, but only patients with LACE <10	4/647 (0.6)	23/2753 (0.8)	0.58	0.89 (0.30, 2.62)	0.83
All 7 teaching hospitals, but only patients with LACE 10	20/499 (4.0)	192/4092 (4.7)	0.49	0.99 (0.61‐1.61)	0.98
Readmission within 30 days
All 7 teaching hospitals, all patients	105/1146 (9.2)	751/6845 (11.0)	0.07	0.94 (0.76‐1.17)	0.59
All 7 teaching hospitals, but only patients with LACE <10	33/647 (5.1)	211/2753 (7.7)	0.02	0.68 (0.46‐0.99)	0.04
All 7 teaching hospitals, but only patients with LACE 10	72/499 (14.4)	540/4092 (13.2)	0.44	1.14 (0.87‐1.49)	0.34
ED visit within 30 days
All 7 teaching hospitals, all patients	182/1146 (15.9)	1118/6845 (16.3)	0.70	1.00 (0.84‐1.19)	0.99
All 7 teaching hospitals, but only patients with LACE <10	83/647 (12.8)	412/2753 (15.0)	0.17	0.84 (0.65, 1.09)	0.20
All 7 teaching hospitals, but only patients with LACE 10	99/499 (19.8)	706/4092 (17.3)	0.15	1.17 (0.92‐1.48)	0.20

DISCUSSION

Our data suggest that patients discharged from the GIM teaching wards we studied on weekends were appropriately triaged, as they did not exhibit a higher risk of adverse events postdischarge. Although patients discharged on weekends tended to be younger and had less comorbidities than those discharged during the week, we adjusted for baseline covariates in analyses, and we did not find an association between weekend discharge and increased postdischarge events even among the subset of patients deemed to be at high risk for postdischarge adverse events (based on high LACE scores). To our knowledge, although we previously examined this issue in patients with a most‐responsible diagnosis of heart failure,[10] examining weekend versus weekday discharges in the full gamut of general medical patients admitted to teaching hospitals has not previously been examined.

In our previous study[10] of over 24,000 heart failure patients discharged over 10 years (up to June 2009, therefore no overlap with any patients in this study), we also found that patients discharged on the weekends were younger, had fewer comorbidities, and shorter lengths of stay. Although postdischarge death/readmission rates were higher for weekend discharged patients in our earlier study (21.1% vs 19.5%, adjusted hazard ratio: 1.15, 95% CI: 1.061.25), it is worth noting that this was almost entirely driven by data from nonteaching hospitals and cardiology wards. Thus, it is important to reiterate that the findings in our current study are for GIM wards in teaching hospitals and may not be generalizable to less‐structured nonteaching settings.

Although we did not study physician decision making, our results suggest that physicians are incorporating discharge day into their discharge decision making. They may be selecting younger patients with less comorbidities for weekend discharges, or they may be delaying the discharges of older patients with more comorbidities for weekday discharges. Either is not surprising given the realities of weekend inpatient care: reduced staffing and frequent cross‐coverage (of physicians, nurses, physiotherapists, pharmacists, and occupational therapists), limited support services (such as laboratory services or diagnostic imaging), and decreased availability of community services (including home care and social support services).[17] For example, in 1 large US heart failure registry, patients discharged on a weekend received less complete discharge instructions than those discharged on weekdays.[11] Given that early follow‐up postdischarge is associated with better outcomes,[18, 19] future studies should also explore whether patterns of patient follow‐up differ after weekend versus weekday discharges.

Although we were able to capture all interactions with the healthcare system in a single payer system with universal access, there are some limitations to our study. First, we used administrative data, which preclude fully adjusting for severity of diagnoses or functional status, although we used proxies such as admission from/discharge to a long‐term care facility.[20, 21] Second, we did not have access to process of care measures such as diagnostic testing or prescribing data, and thus cannot determine whether quality of care or patient adherence differed by the day of the week they were discharged on, although this seems unlikely. Third, although postdischarge follow‐up may be associated with better outcomes,[18, 19] we were unable to adjust for patterns of outpatient follow‐up in this study. Fourth, we acknowledge that death or readmission soon after discharge does not necessarily mean that the quality of care during the preceding hospitalization was suboptimal or that these deaths or readmissions were even potentially preventable. Many factors influence postdischarge mortality and/or readmission, and quality of inpatient care is only one.[22, 23, 24, 25] Fifth, although some may express concern that LOS may be a mediator in the causal pathway between discharge decision and postdischarge events, and that adjusting for LOS in analyses could thus spuriously obscure a true association, it is worth pointing out that our 2 sensitivity analyses to explore this (the 1 in which we excluded LOS from the analyses and the 1 in which we included expected LOS rather than the actual LOS) revealed nearly identical point estimates and 95% CI as our main analysis. Finally, as our study is observational, we cannot definitively conclude causality, nor can we exclude an 18% excess risk for patients discharged on weekends (or a 22% lower risk either), given our 95% CI for postdischarge adverse outcomes.

CONCLUSION

We found that the proportion of patients discharged on weekends is lower than the proportion admitted on weekends. We also found that lower risk/less severely ill patients appear to be preferentially discharged on weekends, and as a result, postdischarge outcomes are similar between weekend and weekday discharges despite shorter LOS and less availability of outpatient resources for patients discharged on a weekend. The reasons why more complicated patients are not discharged on weekends deserves further study, as safely increasing weekend discharge rates would improve efficiency and safety (by reducing unnecessary exposure to in‐hospital adverse events such as falls, unnecessary urinary catheterizations, and healthcare‐acquired infections). Although hospital admission has become a 24/7 business, we believe that hospital discharge processes should strive for the same level of efficiency.

Hospitals typically reduce staffing levels and the availability of diagnostic, laboratory, and treatment services on weekends, and patients admitted on weekends exhibit poorer in‐hospital outcomes for several medical conditions.[1, 2, 3, 4, 5, 6, 7, 8, 9] Whether or not patients discharged on weekends have worse clinical outcomes has been less well studied.[10, 11, 12] Discharge rates on Saturday and Sunday are lower than for the other 5 days of the week,[12] but bed shortages and hospital overcrowding have increased the demand for maximizing 24/7 week‐round discharge efficiency. Given that the number of patients discharged on weekends is likely to continue to increase, it is important to assess the risk of weekend discharge on outcomes monitored as performance indicators by organizations such as the Centers for Medicare and Medicaid Services, the American Medical Association Physicians Consortium for Performance Improvement, the National Quality Forum, and the Joint Commission.

Thus, we designed this study to evaluate baseline characteristics, length of stay (LOS), and postdischarge outcomes for general internal medicine (GIM) patients in teaching hospitals discharged on weekends compared to weekdays. Our objective was to determine whether postdischarge outcomes differed for patients discharged on weekends versus weekdays.

METHODS

RESULTS

Weekday Versus Weekend Discharge

Although patients admitted on weekdays and weekends were very similar (data available upon request), patients discharged on weekends (compared to those discharged on weekdays) were younger, more likely to be discharged home without additional support, and had fewer comorbidities (Table 1, Figure 1). Patients discharged on weekends had shorter lengths of stay than those discharged on weekdays (5.6 days vs 7.9 days, P<0.0001). In adjusted linear regression analyses, this 2.3‐day difference remained statistically significant (adjusted P value <0.0001).

Figure 1

Patients discharged on a weekend exhibited lower unadjusted 30‐day rates of death or readmission than those discharged on a weekday (10.6% vs 13.2%), but these differences disappeared after multivariable adjustment that accounted for differences in risk profile (aOR: 0.94, 95% CI: 0.771.16 (Table 2). Results were similar in sensitivity analyses adjusting for LACE scores without LOS included (aOR: 0.88, 95% CI: 0.711.08) or adjusting for LACE scores using expected LOS rather than actual LOS (aOR: 0.90, 95% CI: 0.731.10). Restricting the analysis to only those patients deemed to be at high risk for events due to LACE scores of 10 or greater confirmed that weekend and weekday discharges had similar outcomes in the first 30 days after discharge (aOR: 1.09, 95% CI: 0.851.41, Table 2). Similar patterns were seen when we included ED visits as part of the composite endpoint (ie, death, unplanned readmission, or unplanned ED visit within 30 days of discharge) (Table 2).

Table 2.Postdischarge Outcomes After a General Internal Medicine Hospitalization in a Teaching Hospital

	Weekend Discharge, n/N (%)	Weekday Discharge, n/N (%)	Unadjusted P Value	aOR* (95% CI)	Adjusted P Value
NOTE: Abbreviations: aOR, adjusted odds ratio; CI, confidence interval; ED, emergency department; LACE, length of hospital stay, acuity of admission, comorbidity burden quantified using the Charlson Comorbidity Index, and emergency department visits in the 6 months prior to admission. *Multivariable models adjust for age, sex, hospital, and LACE score at time of discharge from index hospitalization. Weekday discharge is reference group for odds ratios.
Death/readmission within 30 days
All 7 teaching hospitals, all patients	121/1146 (10.6)	901/6845 (13.2)	0.01	0.94 (0.77‐1.16)	0.58
All 7 teaching hospitals, but only patients with LACE <10	37/647 (5.7)	225/2753 (8.2)	0.04	0.72 (0.50, 1.03)	0.07
All 7 teaching hospitals, but only patients with LACE 10	84/499 (16.8)	676/4092 (16.5)	0.86	1.09 (0.85‐1.41)	0.49
Death/readmission/ED visit within 30 days
All 7 teaching hospitals, all patients	218/1146 (19.0)	1445/6845 (21.1)	0.11	0.98 (0.83‐1.15)	0.79
All 7 teaching hospitals, but only patients with LACE <10	90/647 (13.9)	460/2753 (16.7)	0.08	0.83 (0.64‐1.06)	0.13
All 7 teaching hospitals, but only patients with LACE 10	128/499 (25.7)	985/4092 (24.1)	0.44	1.12 (0.90‐1.39)	0.31
Death within 30 days
All 7 teaching hospitals, all patients	24/1146 (2.1)	215/6845 (3.1)	0.05	0.97 (0.63‐1.51)	0.89
All 7 teaching hospitals, but only patients with LACE <10	4/647 (0.6)	23/2753 (0.8)	0.58	0.89 (0.30, 2.62)	0.83
All 7 teaching hospitals, but only patients with LACE 10	20/499 (4.0)	192/4092 (4.7)	0.49	0.99 (0.61‐1.61)	0.98
Readmission within 30 days
All 7 teaching hospitals, all patients	105/1146 (9.2)	751/6845 (11.0)	0.07	0.94 (0.76‐1.17)	0.59
All 7 teaching hospitals, but only patients with LACE <10	33/647 (5.1)	211/2753 (7.7)	0.02	0.68 (0.46‐0.99)	0.04
All 7 teaching hospitals, but only patients with LACE 10	72/499 (14.4)	540/4092 (13.2)	0.44	1.14 (0.87‐1.49)	0.34
ED visit within 30 days
All 7 teaching hospitals, all patients	182/1146 (15.9)	1118/6845 (16.3)	0.70	1.00 (0.84‐1.19)	0.99
All 7 teaching hospitals, but only patients with LACE <10	83/647 (12.8)	412/2753 (15.0)	0.17	0.84 (0.65, 1.09)	0.20
All 7 teaching hospitals, but only patients with LACE 10	99/499 (19.8)	706/4092 (17.3)	0.15	1.17 (0.92‐1.48)	0.20

DISCUSSION

Our data suggest that patients discharged from the GIM teaching wards we studied on weekends were appropriately triaged, as they did not exhibit a higher risk of adverse events postdischarge. Although patients discharged on weekends tended to be younger and had less comorbidities than those discharged during the week, we adjusted for baseline covariates in analyses, and we did not find an association between weekend discharge and increased postdischarge events even among the subset of patients deemed to be at high risk for postdischarge adverse events (based on high LACE scores). To our knowledge, although we previously examined this issue in patients with a most‐responsible diagnosis of heart failure,[10] examining weekend versus weekday discharges in the full gamut of general medical patients admitted to teaching hospitals has not previously been examined.

In our previous study[10] of over 24,000 heart failure patients discharged over 10 years (up to June 2009, therefore no overlap with any patients in this study), we also found that patients discharged on the weekends were younger, had fewer comorbidities, and shorter lengths of stay. Although postdischarge death/readmission rates were higher for weekend discharged patients in our earlier study (21.1% vs 19.5%, adjusted hazard ratio: 1.15, 95% CI: 1.061.25), it is worth noting that this was almost entirely driven by data from nonteaching hospitals and cardiology wards. Thus, it is important to reiterate that the findings in our current study are for GIM wards in teaching hospitals and may not be generalizable to less‐structured nonteaching settings.

Although we did not study physician decision making, our results suggest that physicians are incorporating discharge day into their discharge decision making. They may be selecting younger patients with less comorbidities for weekend discharges, or they may be delaying the discharges of older patients with more comorbidities for weekday discharges. Either is not surprising given the realities of weekend inpatient care: reduced staffing and frequent cross‐coverage (of physicians, nurses, physiotherapists, pharmacists, and occupational therapists), limited support services (such as laboratory services or diagnostic imaging), and decreased availability of community services (including home care and social support services).[17] For example, in 1 large US heart failure registry, patients discharged on a weekend received less complete discharge instructions than those discharged on weekdays.[11] Given that early follow‐up postdischarge is associated with better outcomes,[18, 19] future studies should also explore whether patterns of patient follow‐up differ after weekend versus weekday discharges.

Although we were able to capture all interactions with the healthcare system in a single payer system with universal access, there are some limitations to our study. First, we used administrative data, which preclude fully adjusting for severity of diagnoses or functional status, although we used proxies such as admission from/discharge to a long‐term care facility.[20, 21] Second, we did not have access to process of care measures such as diagnostic testing or prescribing data, and thus cannot determine whether quality of care or patient adherence differed by the day of the week they were discharged on, although this seems unlikely. Third, although postdischarge follow‐up may be associated with better outcomes,[18, 19] we were unable to adjust for patterns of outpatient follow‐up in this study. Fourth, we acknowledge that death or readmission soon after discharge does not necessarily mean that the quality of care during the preceding hospitalization was suboptimal or that these deaths or readmissions were even potentially preventable. Many factors influence postdischarge mortality and/or readmission, and quality of inpatient care is only one.[22, 23, 24, 25] Fifth, although some may express concern that LOS may be a mediator in the causal pathway between discharge decision and postdischarge events, and that adjusting for LOS in analyses could thus spuriously obscure a true association, it is worth pointing out that our 2 sensitivity analyses to explore this (the 1 in which we excluded LOS from the analyses and the 1 in which we included expected LOS rather than the actual LOS) revealed nearly identical point estimates and 95% CI as our main analysis. Finally, as our study is observational, we cannot definitively conclude causality, nor can we exclude an 18% excess risk for patients discharged on weekends (or a 22% lower risk either), given our 95% CI for postdischarge adverse outcomes.

CONCLUSION

We found that the proportion of patients discharged on weekends is lower than the proportion admitted on weekends. We also found that lower risk/less severely ill patients appear to be preferentially discharged on weekends, and as a result, postdischarge outcomes are similar between weekend and weekday discharges despite shorter LOS and less availability of outpatient resources for patients discharged on a weekend. The reasons why more complicated patients are not discharged on weekends deserves further study, as safely increasing weekend discharge rates would improve efficiency and safety (by reducing unnecessary exposure to in‐hospital adverse events such as falls, unnecessary urinary catheterizations, and healthcare‐acquired infections). Although hospital admission has become a 24/7 business, we believe that hospital discharge processes should strive for the same level of efficiency.