Hospital Medicine

A 51-year-old man with end-stage liver disease from alcohol abuse presented with worsening dyspnea on exertion. He had a history of ascites requiring diuretic therapy and intermittent paracentesis, as well as symptomatic hepatic hydrothorax requiring thoracentesis. Chest radiography showed a large right hydrothorax (Figure 1).

See related commentary

The patient underwent high-volume thoracentesis, and 3.2 L of clear fluid was removed. Chest radiography after the procedure revealed a right-sided pneumothorax (Figure 2, arrow). The patient was mildly short of breath and was treated with high-flow oxygen. Later the same day, his shortness of breath worsened, and repeat chest radiography showed an unchanged pneumothorax that was now complicated by reexpansion pulmonary edema after thoracentesis (Figure 3, star). The reexpansion pulmonary edema resolved by the following day, and the pneumothorax resolved after placement of a pig-tail catheter into the pleural space (Figure 4).

Iatrogenic pneumothorax after thoracentesis occurs in 6% of cases.¹ Iatrogenic reexpansion pulmonary edema after thoracentesis occurs in fewer than 1% of cases.^2,3 Simultaneous pneumothorax and reexpansion pulmonary edema arising from the same procedure appears to be extremely rare.

A 51-year-old man with end-stage liver disease from alcohol abuse presented with worsening dyspnea on exertion. He had a history of ascites requiring diuretic therapy and intermittent paracentesis, as well as symptomatic hepatic hydrothorax requiring thoracentesis. Chest radiography showed a large right hydrothorax (Figure 1).

See related commentary

The patient underwent high-volume thoracentesis, and 3.2 L of clear fluid was removed. Chest radiography after the procedure revealed a right-sided pneumothorax (Figure 2, arrow). The patient was mildly short of breath and was treated with high-flow oxygen. Later the same day, his shortness of breath worsened, and repeat chest radiography showed an unchanged pneumothorax that was now complicated by reexpansion pulmonary edema after thoracentesis (Figure 3, star). The reexpansion pulmonary edema resolved by the following day, and the pneumothorax resolved after placement of a pig-tail catheter into the pleural space (Figure 4).

Iatrogenic pneumothorax after thoracentesis occurs in 6% of cases.¹ Iatrogenic reexpansion pulmonary edema after thoracentesis occurs in fewer than 1% of cases.^2,3 Simultaneous pneumothorax and reexpansion pulmonary edema arising from the same procedure appears to be extremely rare.

A 51-year-old man with end-stage liver disease from alcohol abuse presented with worsening dyspnea on exertion. He had a history of ascites requiring diuretic therapy and intermittent paracentesis, as well as symptomatic hepatic hydrothorax requiring thoracentesis. Chest radiography showed a large right hydrothorax (Figure 1).

See related commentary

The patient underwent high-volume thoracentesis, and 3.2 L of clear fluid was removed. Chest radiography after the procedure revealed a right-sided pneumothorax (Figure 2, arrow). The patient was mildly short of breath and was treated with high-flow oxygen. Later the same day, his shortness of breath worsened, and repeat chest radiography showed an unchanged pneumothorax that was now complicated by reexpansion pulmonary edema after thoracentesis (Figure 3, star). The reexpansion pulmonary edema resolved by the following day, and the pneumothorax resolved after placement of a pig-tail catheter into the pleural space (Figure 4).

Iatrogenic pneumothorax after thoracentesis occurs in 6% of cases.¹ Iatrogenic reexpansion pulmonary edema after thoracentesis occurs in fewer than 1% of cases.^2,3 Simultaneous pneumothorax and reexpansion pulmonary edema arising from the same procedure appears to be extremely rare.

Large-volume thoracentesis is defined as the drainage of more than 1 L of fluid. Inherent in this procedure is the removal of a large amount of fluid from a cavity with a rigid wall, which leads to changes in pleural pressure and to expansion of the lung. Two specific complications occur, pneumothorax and reexpansion pulmonary edema. The images submitted for the Clinical Picture article by Drs. Apter and Aronowitz in this issue of the Journal highlight these complications.

See related article

Retrospective studies have found an association between the amount of fluid drained and the incidence of pneumothorax.^1,2 Although technical issues may account for it (eg, needle injury to the lung that leads to postprocedural pneumothorax), the available evidence suggests that it has more to do with the drainage of larger volumes than the lung can expand to fill.^3,4 That is, the patient’s lung cannot expand,⁵ so drainage creates a vacuum, and air enters the pleural space³ through the lung parenchyma, or perhaps from around the drainage catheter.

In a series of patients who underwent therapeutic thoracentesis,³ 23 (8.7%) of 265 patients had pneumothorax. Interestingly, some patients had only symptoms, some had only excessively negative pressures (< 25 cm H₂O), some had both, and some had neither. Thus, there does not seem to be a reliable sign or symptom of an unexpanding lung, but pleural manometry may help increase its detection.⁶ This technique, however, is rarely used in clinical practice.

Another consequence of therapeutic thoracentesis is reexpansion pulmonary edema. This rare condition occurs only after large-volume thoracentesis or evacuation of a moderate to large pneumothorax.⁷ The pathophysiology behind this is controversial.⁸ As with pneumothorax, a large case series did not find a correlation between volume removed or pleural pressures and reexpansion pulmonary edema.⁷ Experimental data and analysis of case series^8–10 suggest that the duration of lung collapse and the speed of drainage and negative pressure applied contribute to the development of edema. Vacuum bottles are often used to speed drainage and to contain the large amount of fluid drained. These bottles have an initial negative pressure of about −723 mm Hg (personal communication with Baxter Healthcare Product information line), which may lead to rapid changes in lung volume and perhaps to higher negative pleural pressures.

Given the risks discussed above, we believe it is appropriate to avoid vacuum bottles and instead to use the syringe and one-way valve supplied in most thoracentesis kits. Further, pleural manometry to detect changes in pressure that suggest an unexpandable lung may lead to the appropriate early termination of a planned large-volume thoracentesis.³ The complications reported by Drs. Apter and Aronowitz are relatively rare and, at this point, unpredictable; therefore, generating high-quality evidence for prediction or management will be difficult. In the meantime, understanding the physiologic changes in the lung and the pleural space when draining large effusions from the chest may help avoid double trouble.

Large-volume thoracentesis is defined as the drainage of more than 1 L of fluid. Inherent in this procedure is the removal of a large amount of fluid from a cavity with a rigid wall, which leads to changes in pleural pressure and to expansion of the lung. Two specific complications occur, pneumothorax and reexpansion pulmonary edema. The images submitted for the Clinical Picture article by Drs. Apter and Aronowitz in this issue of the Journal highlight these complications.

See related article

Retrospective studies have found an association between the amount of fluid drained and the incidence of pneumothorax.^1,2 Although technical issues may account for it (eg, needle injury to the lung that leads to postprocedural pneumothorax), the available evidence suggests that it has more to do with the drainage of larger volumes than the lung can expand to fill.^3,4 That is, the patient’s lung cannot expand,⁵ so drainage creates a vacuum, and air enters the pleural space³ through the lung parenchyma, or perhaps from around the drainage catheter.

In a series of patients who underwent therapeutic thoracentesis,³ 23 (8.7%) of 265 patients had pneumothorax. Interestingly, some patients had only symptoms, some had only excessively negative pressures (< 25 cm H₂O), some had both, and some had neither. Thus, there does not seem to be a reliable sign or symptom of an unexpanding lung, but pleural manometry may help increase its detection.⁶ This technique, however, is rarely used in clinical practice.

Another consequence of therapeutic thoracentesis is reexpansion pulmonary edema. This rare condition occurs only after large-volume thoracentesis or evacuation of a moderate to large pneumothorax.⁷ The pathophysiology behind this is controversial.⁸ As with pneumothorax, a large case series did not find a correlation between volume removed or pleural pressures and reexpansion pulmonary edema.⁷ Experimental data and analysis of case series^8–10 suggest that the duration of lung collapse and the speed of drainage and negative pressure applied contribute to the development of edema. Vacuum bottles are often used to speed drainage and to contain the large amount of fluid drained. These bottles have an initial negative pressure of about −723 mm Hg (personal communication with Baxter Healthcare Product information line), which may lead to rapid changes in lung volume and perhaps to higher negative pleural pressures.

Given the risks discussed above, we believe it is appropriate to avoid vacuum bottles and instead to use the syringe and one-way valve supplied in most thoracentesis kits. Further, pleural manometry to detect changes in pressure that suggest an unexpandable lung may lead to the appropriate early termination of a planned large-volume thoracentesis.³ The complications reported by Drs. Apter and Aronowitz are relatively rare and, at this point, unpredictable; therefore, generating high-quality evidence for prediction or management will be difficult. In the meantime, understanding the physiologic changes in the lung and the pleural space when draining large effusions from the chest may help avoid double trouble.

Large-volume thoracentesis is defined as the drainage of more than 1 L of fluid. Inherent in this procedure is the removal of a large amount of fluid from a cavity with a rigid wall, which leads to changes in pleural pressure and to expansion of the lung. Two specific complications occur, pneumothorax and reexpansion pulmonary edema. The images submitted for the Clinical Picture article by Drs. Apter and Aronowitz in this issue of the Journal highlight these complications.

See related article

Retrospective studies have found an association between the amount of fluid drained and the incidence of pneumothorax.^1,2 Although technical issues may account for it (eg, needle injury to the lung that leads to postprocedural pneumothorax), the available evidence suggests that it has more to do with the drainage of larger volumes than the lung can expand to fill.^3,4 That is, the patient’s lung cannot expand,⁵ so drainage creates a vacuum, and air enters the pleural space³ through the lung parenchyma, or perhaps from around the drainage catheter.

In a series of patients who underwent therapeutic thoracentesis,³ 23 (8.7%) of 265 patients had pneumothorax. Interestingly, some patients had only symptoms, some had only excessively negative pressures (< 25 cm H₂O), some had both, and some had neither. Thus, there does not seem to be a reliable sign or symptom of an unexpanding lung, but pleural manometry may help increase its detection.⁶ This technique, however, is rarely used in clinical practice.

Another consequence of therapeutic thoracentesis is reexpansion pulmonary edema. This rare condition occurs only after large-volume thoracentesis or evacuation of a moderate to large pneumothorax.⁷ The pathophysiology behind this is controversial.⁸ As with pneumothorax, a large case series did not find a correlation between volume removed or pleural pressures and reexpansion pulmonary edema.⁷ Experimental data and analysis of case series^8–10 suggest that the duration of lung collapse and the speed of drainage and negative pressure applied contribute to the development of edema. Vacuum bottles are often used to speed drainage and to contain the large amount of fluid drained. These bottles have an initial negative pressure of about −723 mm Hg (personal communication with Baxter Healthcare Product information line), which may lead to rapid changes in lung volume and perhaps to higher negative pleural pressures.

Given the risks discussed above, we believe it is appropriate to avoid vacuum bottles and instead to use the syringe and one-way valve supplied in most thoracentesis kits. Further, pleural manometry to detect changes in pressure that suggest an unexpandable lung may lead to the appropriate early termination of a planned large-volume thoracentesis.³ The complications reported by Drs. Apter and Aronowitz are relatively rare and, at this point, unpredictable; therefore, generating high-quality evidence for prediction or management will be difficult. In the meantime, understanding the physiologic changes in the lung and the pleural space when draining large effusions from the chest may help avoid double trouble.

To the Editor: We appreciated the thoughtful 1-Minute Consult by Drs. Dobri and Lansang, “How should we manage insulin therapy before surgery?”¹ We agree with them in regard to the benefits of perioperative control of blood glucose levels. However, we disagree in general with their assertion that the full dose of the patient’s home dose of basal insulin be administered while the patient is nil per os (NPO) before surgery, with a reduction to 75% of the home dose only if the patient has a history of hypoglycemia, a recommendation that did not differentiate between patients with type 1 and type 2 diabetes mellitus.

The RABBIT 2 Surgery trial,² which showed superiority of basal-bolus insulin over sliding scale insulin in surgical patients with type 2 diabetes mellitus, also showed a surprisingly high rate of hypoglycemia—24 (23.1%) of 104 patients had blood glucose levels lower than 70 mg/dL, compared with a similar trial in nonsurgical patients in which 2 (3.1%) of 65 patients had a blood glucose level less than 60 mg/dL.³ The authors of the two studies explained² that “differences in hypoglycemic events between the two trials could be in part explained by reduced nutritional intake in surgical patients…”

Although patients with well-controlled type 1 diabetes mellitus may tolerate their full dose of basal insulin while NPO, we contend that patients with type 2 diabetes mellitus should be prescribed a reduced dose of basal insulin while NPO, regardless of the dose distribution or the patient’s overall glycemic control. It is routine practice on our consult service to reduce the basal insulin dose in such patients by roughly half.

To the Editor: We appreciated the thoughtful 1-Minute Consult by Drs. Dobri and Lansang, “How should we manage insulin therapy before surgery?”¹ We agree with them in regard to the benefits of perioperative control of blood glucose levels. However, we disagree in general with their assertion that the full dose of the patient’s home dose of basal insulin be administered while the patient is nil per os (NPO) before surgery, with a reduction to 75% of the home dose only if the patient has a history of hypoglycemia, a recommendation that did not differentiate between patients with type 1 and type 2 diabetes mellitus.

The RABBIT 2 Surgery trial,² which showed superiority of basal-bolus insulin over sliding scale insulin in surgical patients with type 2 diabetes mellitus, also showed a surprisingly high rate of hypoglycemia—24 (23.1%) of 104 patients had blood glucose levels lower than 70 mg/dL, compared with a similar trial in nonsurgical patients in which 2 (3.1%) of 65 patients had a blood glucose level less than 60 mg/dL.³ The authors of the two studies explained² that “differences in hypoglycemic events between the two trials could be in part explained by reduced nutritional intake in surgical patients…”

Although patients with well-controlled type 1 diabetes mellitus may tolerate their full dose of basal insulin while NPO, we contend that patients with type 2 diabetes mellitus should be prescribed a reduced dose of basal insulin while NPO, regardless of the dose distribution or the patient’s overall glycemic control. It is routine practice on our consult service to reduce the basal insulin dose in such patients by roughly half.

To the Editor: We appreciated the thoughtful 1-Minute Consult by Drs. Dobri and Lansang, “How should we manage insulin therapy before surgery?”¹ We agree with them in regard to the benefits of perioperative control of blood glucose levels. However, we disagree in general with their assertion that the full dose of the patient’s home dose of basal insulin be administered while the patient is nil per os (NPO) before surgery, with a reduction to 75% of the home dose only if the patient has a history of hypoglycemia, a recommendation that did not differentiate between patients with type 1 and type 2 diabetes mellitus.

The RABBIT 2 Surgery trial,² which showed superiority of basal-bolus insulin over sliding scale insulin in surgical patients with type 2 diabetes mellitus, also showed a surprisingly high rate of hypoglycemia—24 (23.1%) of 104 patients had blood glucose levels lower than 70 mg/dL, compared with a similar trial in nonsurgical patients in which 2 (3.1%) of 65 patients had a blood glucose level less than 60 mg/dL.³ The authors of the two studies explained² that “differences in hypoglycemic events between the two trials could be in part explained by reduced nutritional intake in surgical patients…”

Although patients with well-controlled type 1 diabetes mellitus may tolerate their full dose of basal insulin while NPO, we contend that patients with type 2 diabetes mellitus should be prescribed a reduced dose of basal insulin while NPO, regardless of the dose distribution or the patient’s overall glycemic control. It is routine practice on our consult service to reduce the basal insulin dose in such patients by roughly half.

In Reply: We appreciate the kind words of Drs. Ditch and Moore, as well as their opinion.

Our article was intentionally brief—a 1-Minute Consult—and so could not cover all specific situations we encounter in clinical practice. We meant only to provide a general approach in this matter.

Quite often before surgery, patients receive less basal insulin than needed, or none at all, rather than too much. It has to be borne in mind that perioperative hyperglycemia—not just hypoglycemia—is linked with poor outcomes in cardiac¹ and noncardiac surgery.^2,3

Through our scenarios and suggestions, we have taken steps to err on the side of preventing hypoglycemia while averting hyperglycemia, at the same time making it easy to calculate the dose. In a scenario in which the basal insulin dose is about the same as the total of the prandial boluses, we have not yet seen evidence that raises concern for hypoglycemia, maybe because many of the patients with type 2 diabetes seen in our institution for surgery take, in addition to insulin, oral agents or noninsulin injections (which are appropriately withheld before surgery), and have suboptimal glycemic control on their home regimen. But if a physician has concerns for hypoglycemia, a dose reduction should be made.

There were some differences between the RABBIT 2 trial in medical patients⁴ and the RABBIT 2 Surgery trial⁵ that would make the results not completely comparable. In RABBIT 2, the medical patients included were on diet alone or any combination of oral antidiabetic agents (not on insulin), and they were started on a total daily dose of insulin of either 0.4 or 0.5 U/kg/day, depending on the glucose level. In RABBIT 2 Surgery, patients who were on insulin at home with a total daily dose of 0.4 U/kg or less were also included, and the starting daily dose of insulin was 0.5 U/kg (unless they were older or had a high serum creatinine).

In view of all the above, we agree with Drs. Ditch and Moore that if there is concern for hypoglycemia, the clinician should reduce the insulin dose in the manner that evidence from the local practice suggests, without causing undue hyperglycemia and postsurgical complications.

In Reply: We appreciate the kind words of Drs. Ditch and Moore, as well as their opinion.

Our article was intentionally brief—a 1-Minute Consult—and so could not cover all specific situations we encounter in clinical practice. We meant only to provide a general approach in this matter.

Quite often before surgery, patients receive less basal insulin than needed, or none at all, rather than too much. It has to be borne in mind that perioperative hyperglycemia—not just hypoglycemia—is linked with poor outcomes in cardiac¹ and noncardiac surgery.^2,3

Through our scenarios and suggestions, we have taken steps to err on the side of preventing hypoglycemia while averting hyperglycemia, at the same time making it easy to calculate the dose. In a scenario in which the basal insulin dose is about the same as the total of the prandial boluses, we have not yet seen evidence that raises concern for hypoglycemia, maybe because many of the patients with type 2 diabetes seen in our institution for surgery take, in addition to insulin, oral agents or noninsulin injections (which are appropriately withheld before surgery), and have suboptimal glycemic control on their home regimen. But if a physician has concerns for hypoglycemia, a dose reduction should be made.

There were some differences between the RABBIT 2 trial in medical patients⁴ and the RABBIT 2 Surgery trial⁵ that would make the results not completely comparable. In RABBIT 2, the medical patients included were on diet alone or any combination of oral antidiabetic agents (not on insulin), and they were started on a total daily dose of insulin of either 0.4 or 0.5 U/kg/day, depending on the glucose level. In RABBIT 2 Surgery, patients who were on insulin at home with a total daily dose of 0.4 U/kg or less were also included, and the starting daily dose of insulin was 0.5 U/kg (unless they were older or had a high serum creatinine).

In view of all the above, we agree with Drs. Ditch and Moore that if there is concern for hypoglycemia, the clinician should reduce the insulin dose in the manner that evidence from the local practice suggests, without causing undue hyperglycemia and postsurgical complications.

In Reply: We appreciate the kind words of Drs. Ditch and Moore, as well as their opinion.

Our article was intentionally brief—a 1-Minute Consult—and so could not cover all specific situations we encounter in clinical practice. We meant only to provide a general approach in this matter.

Quite often before surgery, patients receive less basal insulin than needed, or none at all, rather than too much. It has to be borne in mind that perioperative hyperglycemia—not just hypoglycemia—is linked with poor outcomes in cardiac¹ and noncardiac surgery.^2,3

Through our scenarios and suggestions, we have taken steps to err on the side of preventing hypoglycemia while averting hyperglycemia, at the same time making it easy to calculate the dose. In a scenario in which the basal insulin dose is about the same as the total of the prandial boluses, we have not yet seen evidence that raises concern for hypoglycemia, maybe because many of the patients with type 2 diabetes seen in our institution for surgery take, in addition to insulin, oral agents or noninsulin injections (which are appropriately withheld before surgery), and have suboptimal glycemic control on their home regimen. But if a physician has concerns for hypoglycemia, a dose reduction should be made.

There were some differences between the RABBIT 2 trial in medical patients⁴ and the RABBIT 2 Surgery trial⁵ that would make the results not completely comparable. In RABBIT 2, the medical patients included were on diet alone or any combination of oral antidiabetic agents (not on insulin), and they were started on a total daily dose of insulin of either 0.4 or 0.5 U/kg/day, depending on the glucose level. In RABBIT 2 Surgery, patients who were on insulin at home with a total daily dose of 0.4 U/kg or less were also included, and the starting daily dose of insulin was 0.5 U/kg (unless they were older or had a high serum creatinine).

In view of all the above, we agree with Drs. Ditch and Moore that if there is concern for hypoglycemia, the clinician should reduce the insulin dose in the manner that evidence from the local practice suggests, without causing undue hyperglycemia and postsurgical complications.

To the Editor: In the January 2014 issue, Eisa et al¹ suggested that patients who require prolonged mechanical ventilatory support, ie, for more than 48 hours, should receive stress ulcer prophylaxis. This recommendation came from a study by Cook et al² in 1994, which found a significant increase in the risk of gastrointestinal blood loss in this group of patients. Other studies have shown a different result. Zandstra et al³ found an extremely low rate of stress ulcer-related bleeding in this group in the absence of stress ulcer prophylaxis. Another study⁴ in critically ill patients also found no relationship between stress ulcer incidence and prolonged mechanical ventilatory support. Interestingly, that study found that prolonged use of a nasogastric tube is the major risk factor for developing a stress ulcer.⁴ The explanation for why newer studies did not demonstrate the relationship between mechanical ventilation and stress ulcer development may lie in the result of a meta-analysis by Marik et al,⁵ which showed that stress ulcer prophylaxis may not be required in a patient who receives early enteral nutrition. That practice was not common in the past, including at the time the original study was conducted.

According to current evidence, mechanical ventilation for more than 48 hours does not seem to increase the risk of stress ulcer. The medical community should start questioning the routine practice of stress ulcer prophylaxis in this group of patients. In addition, more studies have identified the adverse effects of acid-suppression therapy in this group of patients, and these effects likely make the harms outweigh the benefits. This notion was confirmed in the most recent meta-analysis by Krag et al.⁶ In summary, the practice of routine stress ulcer prophylaxis in all mechanically ventilated patients will likely change in the future, with more focus on patients who are at higher risk.

To the Editor: In the January 2014 issue, Eisa et al¹ suggested that patients who require prolonged mechanical ventilatory support, ie, for more than 48 hours, should receive stress ulcer prophylaxis. This recommendation came from a study by Cook et al² in 1994, which found a significant increase in the risk of gastrointestinal blood loss in this group of patients. Other studies have shown a different result. Zandstra et al³ found an extremely low rate of stress ulcer-related bleeding in this group in the absence of stress ulcer prophylaxis. Another study⁴ in critically ill patients also found no relationship between stress ulcer incidence and prolonged mechanical ventilatory support. Interestingly, that study found that prolonged use of a nasogastric tube is the major risk factor for developing a stress ulcer.⁴ The explanation for why newer studies did not demonstrate the relationship between mechanical ventilation and stress ulcer development may lie in the result of a meta-analysis by Marik et al,⁵ which showed that stress ulcer prophylaxis may not be required in a patient who receives early enteral nutrition. That practice was not common in the past, including at the time the original study was conducted.

According to current evidence, mechanical ventilation for more than 48 hours does not seem to increase the risk of stress ulcer. The medical community should start questioning the routine practice of stress ulcer prophylaxis in this group of patients. In addition, more studies have identified the adverse effects of acid-suppression therapy in this group of patients, and these effects likely make the harms outweigh the benefits. This notion was confirmed in the most recent meta-analysis by Krag et al.⁶ In summary, the practice of routine stress ulcer prophylaxis in all mechanically ventilated patients will likely change in the future, with more focus on patients who are at higher risk.

To the Editor: In the January 2014 issue, Eisa et al¹ suggested that patients who require prolonged mechanical ventilatory support, ie, for more than 48 hours, should receive stress ulcer prophylaxis. This recommendation came from a study by Cook et al² in 1994, which found a significant increase in the risk of gastrointestinal blood loss in this group of patients. Other studies have shown a different result. Zandstra et al³ found an extremely low rate of stress ulcer-related bleeding in this group in the absence of stress ulcer prophylaxis. Another study⁴ in critically ill patients also found no relationship between stress ulcer incidence and prolonged mechanical ventilatory support. Interestingly, that study found that prolonged use of a nasogastric tube is the major risk factor for developing a stress ulcer.⁴ The explanation for why newer studies did not demonstrate the relationship between mechanical ventilation and stress ulcer development may lie in the result of a meta-analysis by Marik et al,⁵ which showed that stress ulcer prophylaxis may not be required in a patient who receives early enteral nutrition. That practice was not common in the past, including at the time the original study was conducted.

According to current evidence, mechanical ventilation for more than 48 hours does not seem to increase the risk of stress ulcer. The medical community should start questioning the routine practice of stress ulcer prophylaxis in this group of patients. In addition, more studies have identified the adverse effects of acid-suppression therapy in this group of patients, and these effects likely make the harms outweigh the benefits. This notion was confirmed in the most recent meta-analysis by Krag et al.⁶ In summary, the practice of routine stress ulcer prophylaxis in all mechanically ventilated patients will likely change in the future, with more focus on patients who are at higher risk.

In Reply: We welcome the comments from Dr. Chongnarungsin on our article and the opportunity to further discuss our opinions.

In our paper, we discussed current recommendations for prophylaxis of stress ulcer-related bleeding in hospitalized patients and advocated against the blind administration of drugs without risk stratification.

The landmark trial that provides the most-cited definitions and the risk factors for clinically significant stress ulcer-related bleeding in critically ill patients was published in 1994 by Cook et al.¹ In their multicenter prospective cohort study of 2,252 patients, the authors reported that prolonged mechanical ventilation is an important risk factor for clinically significant stress ulcer-related bleeding.

Another major prospective cohort study observed an incidence rate of clinically significant stress ulcer-related bleeding of 3.5%.²

Dr. Chongnarungsin cites another prospective cohort study of 183 patients from the same era,³ wherein the authors defined stress ulcer-related bleeding as bleeding requiring transfusion of packed red blood cells, found on endoscopy or on postmortem evaluation. This was in contrast to the 1994 study of Cook et al,¹ who had a more rigorous and comprehensive definition for overt and clinically significant stress ulcer-related bleeding, applied by up to three independent adjudicators not involved in the patients’ care. Their definition not only entailed a more accurate transfusion-dependent bleeding criterion, but also included hemodynamic and laboratory criteria. As such, the “very low rate” of stress ulcer-related bleeding reported by Zandstra et al³ should be critically appraised. Of note, the authors in that study did not report the rates of patients who received early enteral feeding, and their patients received cefotaxime for digestive tract decontamination, an important confounder to the interpretation of the study results.

Indeed, the remarkable variation in estimates of the incidence of stress ulcer-related bleeding is probably related to the lack of a uniform definition. Even when rates of endoscopic and occult bleeding are set aside, agreement is lacking as to which category of bleeding is clinically significant.

Dr. Chongnarungsin also cites the study by Ellison et al⁴ of a cohort of 874 patients who had no previous gastrointestinal bleeding or peptic ulcer disease and who were enrolled in a multicenter randomized controlled trial of prophylactic intravenous immune globulin to prevent infections associated with an intensive care unit. In a secondary objective, the authors did not identify coagulopathy or prolonged mechanical ventilation as a principal risk factor for bleeding. The authors ascribed this discrepancy with previously published literature to their unique study population, which consisted predominantly of elderly men and rarely included trauma patients. In light of these unique peculiarities of their population, the lack of an association between prolonged mechanical ventilation and stress ulcer-related bleeding cannot be determined. Moreover, that study showed that prolonged nasogastric tube insertion was one of the risk factors for increased risk of gastrointestinal bleeding, and not the risk factor for development of stress ulcer as stated by Dr. Chongnarungsin.

The decrease in the incidence of stress ulcer-related bleeding in critically ill patients over the years could be attributed to an era effect, from advances in critical care medicine and prophylactic methods.⁵ We agree with Dr. Chongnarungsin that the increased introduction of early enteral feeding may have also contributed to the reduced incidence of stress ulcer-related bleeding.⁶ However, we think the conclusion that “mechanical ventilation for more than 48 hours does not seem to increase the risk of stress ulcer” is overelaborated, and we believe that strong evidence demonstrates this association.^1,2

Alternatively, we recognize the lack of mortality-benefit evidence for stress ulcer prophylaxis. This notwithstanding, according to recent Surviving Sepsis Campaign guidelines, the use of stress ulcer prophylaxis is listed as a 1B recommendation (strong recommendation) for severely septic patients who require prolonged mechanical ventilation. In addition, the updated 2014 guidelines of the American Society of Health-System Pharmacists⁷ continue to recommend stress ulcer prophylaxis in the context of mechanical ventilation, with H2 receptor antagonists being the preferred first-line agents.⁸

It is important to acknowledge that these recommendations were endorsed despite the lack of obvious mortality benefit, and it is our opinion that large randomized controlled studies are needed to evaluate the risks and mortality benefit of these prophylaxis methods.

In Reply: We welcome the comments from Dr. Chongnarungsin on our article and the opportunity to further discuss our opinions.

In our paper, we discussed current recommendations for prophylaxis of stress ulcer-related bleeding in hospitalized patients and advocated against the blind administration of drugs without risk stratification.

The landmark trial that provides the most-cited definitions and the risk factors for clinically significant stress ulcer-related bleeding in critically ill patients was published in 1994 by Cook et al.¹ In their multicenter prospective cohort study of 2,252 patients, the authors reported that prolonged mechanical ventilation is an important risk factor for clinically significant stress ulcer-related bleeding.

Another major prospective cohort study observed an incidence rate of clinically significant stress ulcer-related bleeding of 3.5%.²

Dr. Chongnarungsin cites another prospective cohort study of 183 patients from the same era,³ wherein the authors defined stress ulcer-related bleeding as bleeding requiring transfusion of packed red blood cells, found on endoscopy or on postmortem evaluation. This was in contrast to the 1994 study of Cook et al,¹ who had a more rigorous and comprehensive definition for overt and clinically significant stress ulcer-related bleeding, applied by up to three independent adjudicators not involved in the patients’ care. Their definition not only entailed a more accurate transfusion-dependent bleeding criterion, but also included hemodynamic and laboratory criteria. As such, the “very low rate” of stress ulcer-related bleeding reported by Zandstra et al³ should be critically appraised. Of note, the authors in that study did not report the rates of patients who received early enteral feeding, and their patients received cefotaxime for digestive tract decontamination, an important confounder to the interpretation of the study results.

Indeed, the remarkable variation in estimates of the incidence of stress ulcer-related bleeding is probably related to the lack of a uniform definition. Even when rates of endoscopic and occult bleeding are set aside, agreement is lacking as to which category of bleeding is clinically significant.

Dr. Chongnarungsin also cites the study by Ellison et al⁴ of a cohort of 874 patients who had no previous gastrointestinal bleeding or peptic ulcer disease and who were enrolled in a multicenter randomized controlled trial of prophylactic intravenous immune globulin to prevent infections associated with an intensive care unit. In a secondary objective, the authors did not identify coagulopathy or prolonged mechanical ventilation as a principal risk factor for bleeding. The authors ascribed this discrepancy with previously published literature to their unique study population, which consisted predominantly of elderly men and rarely included trauma patients. In light of these unique peculiarities of their population, the lack of an association between prolonged mechanical ventilation and stress ulcer-related bleeding cannot be determined. Moreover, that study showed that prolonged nasogastric tube insertion was one of the risk factors for increased risk of gastrointestinal bleeding, and not the risk factor for development of stress ulcer as stated by Dr. Chongnarungsin.

The decrease in the incidence of stress ulcer-related bleeding in critically ill patients over the years could be attributed to an era effect, from advances in critical care medicine and prophylactic methods.⁵ We agree with Dr. Chongnarungsin that the increased introduction of early enteral feeding may have also contributed to the reduced incidence of stress ulcer-related bleeding.⁶ However, we think the conclusion that “mechanical ventilation for more than 48 hours does not seem to increase the risk of stress ulcer” is overelaborated, and we believe that strong evidence demonstrates this association.^1,2

Alternatively, we recognize the lack of mortality-benefit evidence for stress ulcer prophylaxis. This notwithstanding, according to recent Surviving Sepsis Campaign guidelines, the use of stress ulcer prophylaxis is listed as a 1B recommendation (strong recommendation) for severely septic patients who require prolonged mechanical ventilation. In addition, the updated 2014 guidelines of the American Society of Health-System Pharmacists⁷ continue to recommend stress ulcer prophylaxis in the context of mechanical ventilation, with H2 receptor antagonists being the preferred first-line agents.⁸

It is important to acknowledge that these recommendations were endorsed despite the lack of obvious mortality benefit, and it is our opinion that large randomized controlled studies are needed to evaluate the risks and mortality benefit of these prophylaxis methods.

In Reply: We welcome the comments from Dr. Chongnarungsin on our article and the opportunity to further discuss our opinions.

In our paper, we discussed current recommendations for prophylaxis of stress ulcer-related bleeding in hospitalized patients and advocated against the blind administration of drugs without risk stratification.

The landmark trial that provides the most-cited definitions and the risk factors for clinically significant stress ulcer-related bleeding in critically ill patients was published in 1994 by Cook et al.¹ In their multicenter prospective cohort study of 2,252 patients, the authors reported that prolonged mechanical ventilation is an important risk factor for clinically significant stress ulcer-related bleeding.

Another major prospective cohort study observed an incidence rate of clinically significant stress ulcer-related bleeding of 3.5%.²

Dr. Chongnarungsin cites another prospective cohort study of 183 patients from the same era,³ wherein the authors defined stress ulcer-related bleeding as bleeding requiring transfusion of packed red blood cells, found on endoscopy or on postmortem evaluation. This was in contrast to the 1994 study of Cook et al,¹ who had a more rigorous and comprehensive definition for overt and clinically significant stress ulcer-related bleeding, applied by up to three independent adjudicators not involved in the patients’ care. Their definition not only entailed a more accurate transfusion-dependent bleeding criterion, but also included hemodynamic and laboratory criteria. As such, the “very low rate” of stress ulcer-related bleeding reported by Zandstra et al³ should be critically appraised. Of note, the authors in that study did not report the rates of patients who received early enteral feeding, and their patients received cefotaxime for digestive tract decontamination, an important confounder to the interpretation of the study results.

Indeed, the remarkable variation in estimates of the incidence of stress ulcer-related bleeding is probably related to the lack of a uniform definition. Even when rates of endoscopic and occult bleeding are set aside, agreement is lacking as to which category of bleeding is clinically significant.

Dr. Chongnarungsin also cites the study by Ellison et al⁴ of a cohort of 874 patients who had no previous gastrointestinal bleeding or peptic ulcer disease and who were enrolled in a multicenter randomized controlled trial of prophylactic intravenous immune globulin to prevent infections associated with an intensive care unit. In a secondary objective, the authors did not identify coagulopathy or prolonged mechanical ventilation as a principal risk factor for bleeding. The authors ascribed this discrepancy with previously published literature to their unique study population, which consisted predominantly of elderly men and rarely included trauma patients. In light of these unique peculiarities of their population, the lack of an association between prolonged mechanical ventilation and stress ulcer-related bleeding cannot be determined. Moreover, that study showed that prolonged nasogastric tube insertion was one of the risk factors for increased risk of gastrointestinal bleeding, and not the risk factor for development of stress ulcer as stated by Dr. Chongnarungsin.

The decrease in the incidence of stress ulcer-related bleeding in critically ill patients over the years could be attributed to an era effect, from advances in critical care medicine and prophylactic methods.⁵ We agree with Dr. Chongnarungsin that the increased introduction of early enteral feeding may have also contributed to the reduced incidence of stress ulcer-related bleeding.⁶ However, we think the conclusion that “mechanical ventilation for more than 48 hours does not seem to increase the risk of stress ulcer” is overelaborated, and we believe that strong evidence demonstrates this association.^1,2

Alternatively, we recognize the lack of mortality-benefit evidence for stress ulcer prophylaxis. This notwithstanding, according to recent Surviving Sepsis Campaign guidelines, the use of stress ulcer prophylaxis is listed as a 1B recommendation (strong recommendation) for severely septic patients who require prolonged mechanical ventilation. In addition, the updated 2014 guidelines of the American Society of Health-System Pharmacists⁷ continue to recommend stress ulcer prophylaxis in the context of mechanical ventilation, with H2 receptor antagonists being the preferred first-line agents.⁸

It is important to acknowledge that these recommendations were endorsed despite the lack of obvious mortality benefit, and it is our opinion that large randomized controlled studies are needed to evaluate the risks and mortality benefit of these prophylaxis methods.

You have spent several days checking on a patient hospitalized for an acute exacerbation of heart failure. You have straightened out her medications and diet and discussed a plan for follow-up with the patient and a family member, and now she is being wheeled out the door. What happens to her next?

Too often, not your desired plan. If she is going home, maybe she understands what she needs to do, maybe not. Maybe she will get your prescriptions filled and take the medications as directed, maybe not. If she is going to a nursing home, maybe the physician covering the nursing home will get your plan, maybe not. There is a good chance she will be back in the emergency room soon, all because of a poor transition of care.

Transitions of care are changes in the level, location, or providers of care as patients move within the health care system. These can be critical junctures in patients’ lives, and if poorly executed can result in many adverse effects—including rehospitalization.¹

Although high rehospitalization rates gained national attention in 2009 after a analysis of Medicare data,² health care providers have known about the lack of coordinated care transitions for more than 50 years.³ Despite some progress, improving care transitions remains a national challenge. As the health system evolves from a fee-for-service financial model to payment-for-value,⁴ it is especially important that health care providers improve care for patients by optimizing care transitions.

In this article, we summarize the factors contributing to poor care transitions, highlight programs that improve them, and discuss strategies for successful transitions.

TRANSITION PROBLEMS ARE COMMON

Transitions of care occur when patients move to short-term and long-term acute care hospitals, skilled nursing facilities, primary and specialty care offices, community health centers, rehabilitation facilities, home health agencies, hospice, and their own homes.⁵ Problems can arise at any of these transitions, but the risk is especially high when patients leave the hospital to receive care in another setting or at home.

In the past decade, one in five Medicare patients was rehospitalized within 30 days of discharge from the hospital,² and up to 25% were rehospitalized after being discharged to a skilled nursing facility.⁶ Some diagnoses (eg, sickle cell anemia, gangrene) and procedures (eg, kidney transplantation, ileostomy) are associated with readmission rates of nearly one in three.^7,8

The desire of policymakers to “bend the cost curve” of health care has led to efforts to enhance care coordination by improving transitions between care venues. Through the Patient Protection and Affordable Care Act, a number of federal initiatives are promoting strategies to improve care transitions and prevent readmissions after hospital discharge.

The Hospital Readmission Reduction Program⁹ drives much of this effort. In fiscal year 2013 (beginning October 1, 2012), more than 2,000 hospitals incurred financial penalties of up to 1% of total Medicare diagnosis-related group payments (about $280 million the first year) for excess readmissions.¹⁰ The penalty’s maximum rose to 2% in fiscal year 2014 and could increase to 3% in 2015. The total penalty for 2014 is projected to be $227 million, with 2,225 hospitals affected.¹¹

The Centers for Medicare and Medicaid Innovation has committed hundreds of millions of dollars to Community-based Care Transitions Programs¹² and more than $200 million to Hospital Engagement Networks¹³ to carry out the goals of the Partnership for Patients,¹⁴ aiming to reduce rehospitalizations and other adverse events.

At first, despite these efforts, readmission rates did not appear to change substantially.¹⁵ However, the Centers for Medicare and Medicaid Services reported that hospital readmission rates for Medicare fee-for-service beneficiaries declined in 2012 to 18.4%,¹⁶ although some believe that the reduction is related to an increase in the number of patients admitted for observation in recent years.¹⁷

TRANSITIONS ARE OFTEN POORLY COORDINATED

Although some readmissions are unavoidable—resulting from the inevitable progression of disease or worsening of chronic conditions¹⁸—they may also result from a fumbled transition between care settings. Our current system of care transition has serious deficiencies that endanger patients. Areas that need improvement include communication between providers, patient education about medications and treatments, monitoring of medication adherence and complications, follow-up of pending tests and procedures after discharge, and outpatient follow-up soon after discharge.^19–21

Traditional health care does not have dependable mechanisms for coordinating care across settings; we are all ensconced in “silos” that generally keep the focus within individual venues.²² Lack of coordination blurs the lines of responsibility for patients in the period between discharge from one location and admission to another, leaving them confused about whom to contact for care, especially if symptoms worsen.^23,24

Gaps in coordination are not surprising, given the complexity of the US health care system and the often remarkable number of physicians caring for an individual patient.⁵ Medicare beneficiaries see an average of two primary care physicians and five specialists during a 2-year period; patients with chronic conditions may see up to 16 physicians in 1 year.²⁵ Coordinating care between so many providers in different settings, combined with possible patient factors such as disadvantaged socioeconomic status, lack of caregiver support, and inadequate health literacy, provides many opportunities for failures.

Research has identified several root causes behind most failed care transitions:

Poor provider communication

Multiple studies associate adverse events after discharge with a lack of timely communication between hospital and outpatient providers.²⁶ One study estimated that 80% of serious medical errors involve miscommunication during the hand-off between medical providers.²⁷ Discharge summaries often lack important information such as test results, hospital course, discharge medications, patient counseling, and follow-up plans. Most adverse drug events after hospital discharge result directly from breakdown in communication between hospital staff and patients or primary care physicians.²⁸ Approximately 40% of patients have test results pending at the time of discharge and 10% of these require some action; yet outpatient physicians and patients are often unaware of them.²¹

Ineffective patient and caregiver education

The Institute of Medicine report, Crossing the Quality Chasm: A New Health System for the 21st Century,²⁹ noted that patients leaving one setting for another receive little information on how to care for themselves, when to resume activities, what medication side effects to watch out for, and how to get answers to questions. Of particular concern is that patients and caregivers are sometimes omitted from transition planning and often must suddenly assume new self-care responsibilities upon going home that hospital staff managed before discharge. Too often, patients are discharged with inadequate understanding of their medical condition, self-care plan,^23,24 and who should manage their care.³⁰

Up to 36% of adults in the United States have inadequate health literacy (defined as the inability to understand basic health information needed to make appropriate decisions), hindering patient education efforts.^31–33 Even if they understand, patients and their caregivers must be engaged or “activated” (ie, able and willing to manage one’s health) if we expect them to adhere to appropriate care and behaviors. A review found direct correlations between patient activation and healthy behavior, better health outcomes (eg, achieving normal hemoglobin A_1c and cholesterol levels), and better care experiences.³⁴ This review also noted that multiple studies have documented improved activation scores as a result of specific interventions.

No follow-up with primary care providers

The risk of hospital readmission is significantly lower for patients with chronic obstructive pulmonary disease or heart failure who receive follow-up within 7 days of discharge.^35–38 Of Medicare beneficiaries readmitted to the hospital within 30 days of discharge in 2003–2004, half had no contact with an outpatient physician in the interval between their discharge and their readmission,² and one in three adult patients discharged from a hospital to the community does not see a physician within 30 days of discharge.³⁹ The dearth of primary care providers in many communities can make follow-up care difficult to coordinate.

Failure to address chronic conditions

Analyses of national data sets reveal that patients are commonly rehospitalized for conditions unrelated to their initial hospitalization. According to the Center for Studying Health System Change, more than a quarter of readmissions in the 30 days after discharge are for conditions unrelated to those identified in the index admission, the proportion rising to more than one-third at 1 year.³⁹ Among Medicare beneficiaries readmitted within 30 days of discharge, the proportion readmitted for the same condition was just 35% after hospitalization for heart failure, 10% after hospitalization for acute myocardial infarction, and 22% after hospitalization for pneumonia.⁴⁰

Lack of community support

Multiple social and environmental factors contribute to adverse postdischarge events.^41–43 For socioeconomically disadvantaged patients, care-transition issues are compounded by insufficient access to outpatient care, lack of social support, and lack of transportation. Some studies indicate that between 40% to 50% of readmissions are linked to social problems and inadequate access to community resources.^44–47 Psychosocial issues such as limited health literacy, poor self-management skills, inadequate social support, and living alone are associated with adverse outcomes, including readmission and death.^48,49 Such factors may help explain high levels of “no-shows” to outpatient follow-up visits.

NATIONAL MODELS OF BEST PRACTICES

Efforts to reduce readmissions have traditionally focused on hospitals, but experts now recognize that multiple factors influence readmissions and must be comprehensively addressed. Several evidence-based models seek to improve patient outcomes with interventions aimed at care transitions:

Project BOOST

Project BOOST (Better Outcomes by Optimizing Safe Transitions)⁵⁰ is a national initiative developed by the Society of Hospital Medicine to standardize and optimize the care of patients discharged from hospital to home. The program includes evidence-based clinical interventions that can easily be adopted by any hospital. Interventions are aimed at:

• Identifying patients at high risk on admission
• Targeting risk-specific situations
• Improving information flow between inpatient and outpatient providers
• Improving patient and caregiver education by using the teach-back method
• Achieving timely follow-up after discharge.

The program includes a year of technical support provided by a physician mentor.

Preliminary results from pilot sites showed a 14% reduction in 30-day readmission rates in units using BOOST compared with control units in the same hospital.⁵¹ Mentored implementation was recognized by the Joint Commission and the National Quality Forum with the 2011 John M. Eisenberg Award for Innovation in Patient Safety and Quality.⁵²

Project RED

Project RED (Re-Engineered Discharge)⁵³ evolved from efforts by Dr. Brian Jack and colleagues to re-engineer the hospital workflow process to improve patient safety and reduce rehospitalization rates at Boston Medical Center. The intervention has 12 mutually reinforcing components aimed at improving the discharge process.

In a randomized controlled trial, Project RED led to a 30% decrease in emergency department visits and readmissions within 30 days of discharge from a general medical service of an urban academic medical center.⁵⁴ This study excluded patients admitted from a skilled nursing facility or discharged to one, but a recent study demonstrated that Project RED also led to a lower rate of hospital admission within 30 days of discharge from a skilled nursing facility.⁵⁵

The STAAR initiative

The STAAR initiative (State Action on Avoidable Re-hospitalizations)⁵⁶ was launched in 2009 by the Institute for Healthcare Improvement with the goal of reducing avoidable readmissions in the states of Massachusetts, Michigan, and Washington. Hospital teams focus on improving:

• Assessment of needs after hospital discharge
• Teaching and learning
• Real-time hand-off communication
• Timely follow-up after hospital discharge.

As yet, no published studies other than case reports show a benefit from STAAR.⁵⁷

The Care Transitions Program

The Care Transitions Program,⁵⁸ under the leadership of Dr. Eric Coleman, aims to empower patients and caregivers, who meet with a “transition coach.” The program provides assistance with medication reconciliation and self-management, a patient-centered record owned and maintained by the patient to facilitate cross-site information transfer, timely outpatient follow-up with primary or specialty care, a list of red flags to indicate a worsening condition, and instructions on proper responses.

A randomized controlled trial of the program demonstrated a reduction in hospital readmissions at 30, 90, and 180 days, and lower hospital costs at 90 and 180 days.⁵⁹ This approach also proved effective in a real-world setting.⁶⁰

The Transitional Care Model

Developed by Dr. Mary Naylor and colleagues, the Transitional Care Model⁶¹ also aims at patient and family empowerment, focusing on patients’ stated goals and priorities and ensuring patient engagement. In the program, a transitional care nurse has the job of enhancing patient and caregiver understanding, facilitating patient self-management, and overseeing medication management and transitional care.

A randomized controlled trial demonstrated improved outcomes after hospital discharge for elderly patients with complex medical illnesses, with overall reductions in medical costs through preventing or delaying rehospitalization.⁶² A subsequent real-world study validated this approach.⁶³

The Bridge Model

The Illinois Transitional Care Consortium’s Bridge Model⁶⁴ is for older patients discharged home after hospitalization. It is led by social workers (“bridge care coordinators”) who address barriers to implementing the discharge plan, coordinate resources, and intervene at three points: before discharge, 2 days after discharge, and 30 days after discharge.

An initial study showed no impact on the 30-day rehospitalization rate,⁶⁵ but larger studies are under way with a modified version.

Guided Care

Developed at the Johns Hopkins Bloomberg School of Public Health, Guided Care⁶⁶ involves nurses who work in partnership with physicians and others in primary care to provide patient-centered, cost-effective care to patients with multiple chronic conditions. Nurses conduct in-home assessments, facilitate care planning, promote patient self-management, monitor conditions, coordinate the efforts of all care professionals, and facilitate access to community resources.

A cluster-randomized controlled trial found that this program had mixed results, reducing the use of home health care but having little effect on the use of other health services in the short run. However, in the subgroup of patients covered by Kaiser-Permanente, those who were randomized to the program accrued, on average, 52% fewer skilled nursing facility days, 47% fewer skilled nursing facility admissions, 49% fewer hospital readmissions, and 17% fewer emergency department visits.⁶⁷

The GRACE model

The GRACE model (Geriatric Resources for Assessment and Care of Elders)⁶⁸ was developed to improve the quality of geriatric care, reduce excess health care use, and prevent long-term nursing home placement. Each patient is assigned a support team consisting of a nurse practitioner and a social worker who make home visits, coordinate health care and community services, and develop an individualized care plan.

In one study,⁶⁹ GRACE reduced hospital admission rates for participants at high risk of hospitalization by 12% in the first year of the program and 44% in the second year. GRACE participants also reported higher quality of life compared with the control group.⁶⁹

INTERACT tools

Led by Dr. Joseph Ouslander, INTERACT (Interventions to Reduce Acute Care Transfers)⁷⁰ is a quality-improvement initiative for skilled nursing facilities, designed to facilitate the early identification, evaluation, documentation, and communication of changes in the status of residents. Visitors to its website can download a set of tools and strategies to help them manage conditions before they become serious enough to require a hospital transfer. The tools assist in promoting important communication among providers and enhancing advance-care planning.

A 6-month study in 25 nursing homes showed a 17% reduction in self-reported hospital admissions with this program compared with the same period the previous year.⁷¹

Additional home-based care interventions

Additional innovations are under way in home-based care.

The Home Health Quality Improvement National Campaign is a patient-centered movement to improve the quality of care received by patients residing at home.⁷² Through its Best Practices Intervention Packages, it offers evidence-based educational tools, resources, and interventions for reducing avoidable hospitalizations, improving medication management, and coordinating transitional care.

The Center for Medicare and Medicaid Innovation Independence at Home Demonstration⁷³ is testing whether home-based comprehensive primary care can improve care and reduce hospitalizations for Medicare beneficiaries with multiple chronic conditions.

NO SINGLE INTERVENTION: MULTIPLE STRATEGIES NEEDED

A 2011 review found no single intervention that regularly reduced the 30-day risk of re-hospitalization.⁷⁴ However, other studies have shown that multifaceted interventions can reduce 30-day readmission rates. Randomized controlled trials in short-stay, acute care hospitals indicate that improvement in the following areas can directly reduce hospital readmission rates:

• Comprehensive planning and risk assessment throughout hospitalization
• Quality of care during the initial admission
• Communication with patients, their caregivers, and their clinicians
• Patient education
• Predischarge assessment
• Coordination of care after discharge.

In randomized trials, successful programs reduced the 30-day readmission rates by 20% to 40%,^{54,62,75–79} and a 2011 meta-analysis of randomized clinical trials found evidence that interventions associated with discharge planning helped to reduce readmission rates.⁸⁰

Methods developed by the national care transition models described above can help hospitals optimize patient transitions (Table 1). Although every model has its unique attributes, they have several strategies in common:

Engage a team of key stakeholders that may include patients and caregivers, hospital staff (physicians, nurses, case managers, social workers, and pharmacists), community physicians (primary care, medical homes, and specialists), advance practice providers (physician assistants and nurse practitioners), and postacute care facilities and services (skilled nursing facilities, home health agencies, assisted living residences, hospice, and rehabilitation facilities).

Develop a comprehensive transition plan throughout hospitalization that includes attention to factors that may affect self-care, such as health literacy, chronic conditions, medications, and social support.

Enhance medication reconciliation and management. Obtain the best possible medication history on admission, and ensure that patients understand changes in their medications, how to take each medicine correctly, and important side effects.

Institute daily interdisciplinary communication and care coordination by everyone on the health care team with an emphasis on the care plan, discharge planning, and safety issues.⁸¹

Standardize transition plans, procedures and forms. All discharging physicians should use a standard discharge summary template that includes pertinent diagnoses, active issues, a reconciled medication list with changes highlighted, results from important tests and consultations, pending test results, planned follow-up and required services, warning signs of a worsening condition, and actions needed if a problem arises.

Always send discharge summaries directly to the patient’s primary care physician or next care setting at the time of discharge.

Give the patient a discharge plan that is easy to understand. Enhance patient and family education using health literacy standards⁸² and interactive methods such as teach-back,⁸³ in which patients demonstrate comprehension and skills required for self-care immediately after being taught. Such tools actively teach patients and caregivers to follow a care plan, including managing medications.

Follow up and coordinate support in a timely manner after a patient leaves the care setting. Follow-up visits should be arranged before discharge. Within 1 to 3 days after discharge, the patient should be called or visited by a case manager, social worker, nurse, or other health care provider.

CHALLENGES TO IMPROVING TRANSITIONS

Although several models demonstrated significant reductions of hospital readmissions in trials, challenges remain. Studies do not identify which features of the models are necessary or sufficient, or how applicable they are to different hospital and patient characteristics. A 2012 analysis⁸⁴ of a program designed to reduce readmissions in three states identified key obstacles to successfully improving care transitions:

Collaborative relationships across settings are critical, but very difficult to achieve. It takes time to develop the relationships and trust among providers, and little incentive exists for skilled nursing facilities and physicians outside the hospital to engage in the process.

Infrastructure is lacking, as is experience to implement quality improvements.

We lack proof that models work on a large scale. Confusion exists about which readmissions are preventable and which are not. More evidence is needed to help guide hospitals’ efforts to improve transitions of care and reduce readmissions.

You have spent several days checking on a patient hospitalized for an acute exacerbation of heart failure. You have straightened out her medications and diet and discussed a plan for follow-up with the patient and a family member, and now she is being wheeled out the door. What happens to her next?

Too often, not your desired plan. If she is going home, maybe she understands what she needs to do, maybe not. Maybe she will get your prescriptions filled and take the medications as directed, maybe not. If she is going to a nursing home, maybe the physician covering the nursing home will get your plan, maybe not. There is a good chance she will be back in the emergency room soon, all because of a poor transition of care.

Transitions of care are changes in the level, location, or providers of care as patients move within the health care system. These can be critical junctures in patients’ lives, and if poorly executed can result in many adverse effects—including rehospitalization.¹

Although high rehospitalization rates gained national attention in 2009 after a analysis of Medicare data,² health care providers have known about the lack of coordinated care transitions for more than 50 years.³ Despite some progress, improving care transitions remains a national challenge. As the health system evolves from a fee-for-service financial model to payment-for-value,⁴ it is especially important that health care providers improve care for patients by optimizing care transitions.

In this article, we summarize the factors contributing to poor care transitions, highlight programs that improve them, and discuss strategies for successful transitions.

TRANSITION PROBLEMS ARE COMMON

Transitions of care occur when patients move to short-term and long-term acute care hospitals, skilled nursing facilities, primary and specialty care offices, community health centers, rehabilitation facilities, home health agencies, hospice, and their own homes.⁵ Problems can arise at any of these transitions, but the risk is especially high when patients leave the hospital to receive care in another setting or at home.

In the past decade, one in five Medicare patients was rehospitalized within 30 days of discharge from the hospital,² and up to 25% were rehospitalized after being discharged to a skilled nursing facility.⁶ Some diagnoses (eg, sickle cell anemia, gangrene) and procedures (eg, kidney transplantation, ileostomy) are associated with readmission rates of nearly one in three.^7,8

The desire of policymakers to “bend the cost curve” of health care has led to efforts to enhance care coordination by improving transitions between care venues. Through the Patient Protection and Affordable Care Act, a number of federal initiatives are promoting strategies to improve care transitions and prevent readmissions after hospital discharge.

The Hospital Readmission Reduction Program⁹ drives much of this effort. In fiscal year 2013 (beginning October 1, 2012), more than 2,000 hospitals incurred financial penalties of up to 1% of total Medicare diagnosis-related group payments (about $280 million the first year) for excess readmissions.¹⁰ The penalty’s maximum rose to 2% in fiscal year 2014 and could increase to 3% in 2015. The total penalty for 2014 is projected to be $227 million, with 2,225 hospitals affected.¹¹

The Centers for Medicare and Medicaid Innovation has committed hundreds of millions of dollars to Community-based Care Transitions Programs¹² and more than $200 million to Hospital Engagement Networks¹³ to carry out the goals of the Partnership for Patients,¹⁴ aiming to reduce rehospitalizations and other adverse events.

At first, despite these efforts, readmission rates did not appear to change substantially.¹⁵ However, the Centers for Medicare and Medicaid Services reported that hospital readmission rates for Medicare fee-for-service beneficiaries declined in 2012 to 18.4%,¹⁶ although some believe that the reduction is related to an increase in the number of patients admitted for observation in recent years.¹⁷

TRANSITIONS ARE OFTEN POORLY COORDINATED

Although some readmissions are unavoidable—resulting from the inevitable progression of disease or worsening of chronic conditions¹⁸—they may also result from a fumbled transition between care settings. Our current system of care transition has serious deficiencies that endanger patients. Areas that need improvement include communication between providers, patient education about medications and treatments, monitoring of medication adherence and complications, follow-up of pending tests and procedures after discharge, and outpatient follow-up soon after discharge.^19–21

Traditional health care does not have dependable mechanisms for coordinating care across settings; we are all ensconced in “silos” that generally keep the focus within individual venues.²² Lack of coordination blurs the lines of responsibility for patients in the period between discharge from one location and admission to another, leaving them confused about whom to contact for care, especially if symptoms worsen.^23,24

Gaps in coordination are not surprising, given the complexity of the US health care system and the often remarkable number of physicians caring for an individual patient.⁵ Medicare beneficiaries see an average of two primary care physicians and five specialists during a 2-year period; patients with chronic conditions may see up to 16 physicians in 1 year.²⁵ Coordinating care between so many providers in different settings, combined with possible patient factors such as disadvantaged socioeconomic status, lack of caregiver support, and inadequate health literacy, provides many opportunities for failures.

Research has identified several root causes behind most failed care transitions:

Poor provider communication

Multiple studies associate adverse events after discharge with a lack of timely communication between hospital and outpatient providers.²⁶ One study estimated that 80% of serious medical errors involve miscommunication during the hand-off between medical providers.²⁷ Discharge summaries often lack important information such as test results, hospital course, discharge medications, patient counseling, and follow-up plans. Most adverse drug events after hospital discharge result directly from breakdown in communication between hospital staff and patients or primary care physicians.²⁸ Approximately 40% of patients have test results pending at the time of discharge and 10% of these require some action; yet outpatient physicians and patients are often unaware of them.²¹

Ineffective patient and caregiver education

The Institute of Medicine report, Crossing the Quality Chasm: A New Health System for the 21st Century,²⁹ noted that patients leaving one setting for another receive little information on how to care for themselves, when to resume activities, what medication side effects to watch out for, and how to get answers to questions. Of particular concern is that patients and caregivers are sometimes omitted from transition planning and often must suddenly assume new self-care responsibilities upon going home that hospital staff managed before discharge. Too often, patients are discharged with inadequate understanding of their medical condition, self-care plan,^23,24 and who should manage their care.³⁰

Up to 36% of adults in the United States have inadequate health literacy (defined as the inability to understand basic health information needed to make appropriate decisions), hindering patient education efforts.^31–33 Even if they understand, patients and their caregivers must be engaged or “activated” (ie, able and willing to manage one’s health) if we expect them to adhere to appropriate care and behaviors. A review found direct correlations between patient activation and healthy behavior, better health outcomes (eg, achieving normal hemoglobin A_1c and cholesterol levels), and better care experiences.³⁴ This review also noted that multiple studies have documented improved activation scores as a result of specific interventions.

No follow-up with primary care providers

The risk of hospital readmission is significantly lower for patients with chronic obstructive pulmonary disease or heart failure who receive follow-up within 7 days of discharge.^35–38 Of Medicare beneficiaries readmitted to the hospital within 30 days of discharge in 2003–2004, half had no contact with an outpatient physician in the interval between their discharge and their readmission,² and one in three adult patients discharged from a hospital to the community does not see a physician within 30 days of discharge.³⁹ The dearth of primary care providers in many communities can make follow-up care difficult to coordinate.

Failure to address chronic conditions

Analyses of national data sets reveal that patients are commonly rehospitalized for conditions unrelated to their initial hospitalization. According to the Center for Studying Health System Change, more than a quarter of readmissions in the 30 days after discharge are for conditions unrelated to those identified in the index admission, the proportion rising to more than one-third at 1 year.³⁹ Among Medicare beneficiaries readmitted within 30 days of discharge, the proportion readmitted for the same condition was just 35% after hospitalization for heart failure, 10% after hospitalization for acute myocardial infarction, and 22% after hospitalization for pneumonia.⁴⁰

Lack of community support

Multiple social and environmental factors contribute to adverse postdischarge events.^41–43 For socioeconomically disadvantaged patients, care-transition issues are compounded by insufficient access to outpatient care, lack of social support, and lack of transportation. Some studies indicate that between 40% to 50% of readmissions are linked to social problems and inadequate access to community resources.^44–47 Psychosocial issues such as limited health literacy, poor self-management skills, inadequate social support, and living alone are associated with adverse outcomes, including readmission and death.^48,49 Such factors may help explain high levels of “no-shows” to outpatient follow-up visits.

NATIONAL MODELS OF BEST PRACTICES

Efforts to reduce readmissions have traditionally focused on hospitals, but experts now recognize that multiple factors influence readmissions and must be comprehensively addressed. Several evidence-based models seek to improve patient outcomes with interventions aimed at care transitions:

Project BOOST

Project BOOST (Better Outcomes by Optimizing Safe Transitions)⁵⁰ is a national initiative developed by the Society of Hospital Medicine to standardize and optimize the care of patients discharged from hospital to home. The program includes evidence-based clinical interventions that can easily be adopted by any hospital. Interventions are aimed at:

• Identifying patients at high risk on admission
• Targeting risk-specific situations
• Improving information flow between inpatient and outpatient providers
• Improving patient and caregiver education by using the teach-back method
• Achieving timely follow-up after discharge.

The program includes a year of technical support provided by a physician mentor.

Preliminary results from pilot sites showed a 14% reduction in 30-day readmission rates in units using BOOST compared with control units in the same hospital.⁵¹ Mentored implementation was recognized by the Joint Commission and the National Quality Forum with the 2011 John M. Eisenberg Award for Innovation in Patient Safety and Quality.⁵²

Project RED

Project RED (Re-Engineered Discharge)⁵³ evolved from efforts by Dr. Brian Jack and colleagues to re-engineer the hospital workflow process to improve patient safety and reduce rehospitalization rates at Boston Medical Center. The intervention has 12 mutually reinforcing components aimed at improving the discharge process.

In a randomized controlled trial, Project RED led to a 30% decrease in emergency department visits and readmissions within 30 days of discharge from a general medical service of an urban academic medical center.⁵⁴ This study excluded patients admitted from a skilled nursing facility or discharged to one, but a recent study demonstrated that Project RED also led to a lower rate of hospital admission within 30 days of discharge from a skilled nursing facility.⁵⁵

The STAAR initiative

The STAAR initiative (State Action on Avoidable Re-hospitalizations)⁵⁶ was launched in 2009 by the Institute for Healthcare Improvement with the goal of reducing avoidable readmissions in the states of Massachusetts, Michigan, and Washington. Hospital teams focus on improving:

• Assessment of needs after hospital discharge
• Teaching and learning
• Real-time hand-off communication
• Timely follow-up after hospital discharge.

As yet, no published studies other than case reports show a benefit from STAAR.⁵⁷

The Care Transitions Program

The Care Transitions Program,⁵⁸ under the leadership of Dr. Eric Coleman, aims to empower patients and caregivers, who meet with a “transition coach.” The program provides assistance with medication reconciliation and self-management, a patient-centered record owned and maintained by the patient to facilitate cross-site information transfer, timely outpatient follow-up with primary or specialty care, a list of red flags to indicate a worsening condition, and instructions on proper responses.

A randomized controlled trial of the program demonstrated a reduction in hospital readmissions at 30, 90, and 180 days, and lower hospital costs at 90 and 180 days.⁵⁹ This approach also proved effective in a real-world setting.⁶⁰

The Transitional Care Model

Developed by Dr. Mary Naylor and colleagues, the Transitional Care Model⁶¹ also aims at patient and family empowerment, focusing on patients’ stated goals and priorities and ensuring patient engagement. In the program, a transitional care nurse has the job of enhancing patient and caregiver understanding, facilitating patient self-management, and overseeing medication management and transitional care.

A randomized controlled trial demonstrated improved outcomes after hospital discharge for elderly patients with complex medical illnesses, with overall reductions in medical costs through preventing or delaying rehospitalization.⁶² A subsequent real-world study validated this approach.⁶³

The Bridge Model

The Illinois Transitional Care Consortium’s Bridge Model⁶⁴ is for older patients discharged home after hospitalization. It is led by social workers (“bridge care coordinators”) who address barriers to implementing the discharge plan, coordinate resources, and intervene at three points: before discharge, 2 days after discharge, and 30 days after discharge.

An initial study showed no impact on the 30-day rehospitalization rate,⁶⁵ but larger studies are under way with a modified version.

Guided Care

Developed at the Johns Hopkins Bloomberg School of Public Health, Guided Care⁶⁶ involves nurses who work in partnership with physicians and others in primary care to provide patient-centered, cost-effective care to patients with multiple chronic conditions. Nurses conduct in-home assessments, facilitate care planning, promote patient self-management, monitor conditions, coordinate the efforts of all care professionals, and facilitate access to community resources.

A cluster-randomized controlled trial found that this program had mixed results, reducing the use of home health care but having little effect on the use of other health services in the short run. However, in the subgroup of patients covered by Kaiser-Permanente, those who were randomized to the program accrued, on average, 52% fewer skilled nursing facility days, 47% fewer skilled nursing facility admissions, 49% fewer hospital readmissions, and 17% fewer emergency department visits.⁶⁷

The GRACE model

The GRACE model (Geriatric Resources for Assessment and Care of Elders)⁶⁸ was developed to improve the quality of geriatric care, reduce excess health care use, and prevent long-term nursing home placement. Each patient is assigned a support team consisting of a nurse practitioner and a social worker who make home visits, coordinate health care and community services, and develop an individualized care plan.

In one study,⁶⁹ GRACE reduced hospital admission rates for participants at high risk of hospitalization by 12% in the first year of the program and 44% in the second year. GRACE participants also reported higher quality of life compared with the control group.⁶⁹

INTERACT tools

Led by Dr. Joseph Ouslander, INTERACT (Interventions to Reduce Acute Care Transfers)⁷⁰ is a quality-improvement initiative for skilled nursing facilities, designed to facilitate the early identification, evaluation, documentation, and communication of changes in the status of residents. Visitors to its website can download a set of tools and strategies to help them manage conditions before they become serious enough to require a hospital transfer. The tools assist in promoting important communication among providers and enhancing advance-care planning.

A 6-month study in 25 nursing homes showed a 17% reduction in self-reported hospital admissions with this program compared with the same period the previous year.⁷¹

Additional home-based care interventions

Additional innovations are under way in home-based care.

The Home Health Quality Improvement National Campaign is a patient-centered movement to improve the quality of care received by patients residing at home.⁷² Through its Best Practices Intervention Packages, it offers evidence-based educational tools, resources, and interventions for reducing avoidable hospitalizations, improving medication management, and coordinating transitional care.

The Center for Medicare and Medicaid Innovation Independence at Home Demonstration⁷³ is testing whether home-based comprehensive primary care can improve care and reduce hospitalizations for Medicare beneficiaries with multiple chronic conditions.

NO SINGLE INTERVENTION: MULTIPLE STRATEGIES NEEDED

A 2011 review found no single intervention that regularly reduced the 30-day risk of re-hospitalization.⁷⁴ However, other studies have shown that multifaceted interventions can reduce 30-day readmission rates. Randomized controlled trials in short-stay, acute care hospitals indicate that improvement in the following areas can directly reduce hospital readmission rates:

• Comprehensive planning and risk assessment throughout hospitalization
• Quality of care during the initial admission
• Communication with patients, their caregivers, and their clinicians
• Patient education
• Predischarge assessment
• Coordination of care after discharge.

In randomized trials, successful programs reduced the 30-day readmission rates by 20% to 40%,^{54,62,75–79} and a 2011 meta-analysis of randomized clinical trials found evidence that interventions associated with discharge planning helped to reduce readmission rates.⁸⁰

Methods developed by the national care transition models described above can help hospitals optimize patient transitions (Table 1). Although every model has its unique attributes, they have several strategies in common:

Engage a team of key stakeholders that may include patients and caregivers, hospital staff (physicians, nurses, case managers, social workers, and pharmacists), community physicians (primary care, medical homes, and specialists), advance practice providers (physician assistants and nurse practitioners), and postacute care facilities and services (skilled nursing facilities, home health agencies, assisted living residences, hospice, and rehabilitation facilities).

Develop a comprehensive transition plan throughout hospitalization that includes attention to factors that may affect self-care, such as health literacy, chronic conditions, medications, and social support.

Enhance medication reconciliation and management. Obtain the best possible medication history on admission, and ensure that patients understand changes in their medications, how to take each medicine correctly, and important side effects.

Institute daily interdisciplinary communication and care coordination by everyone on the health care team with an emphasis on the care plan, discharge planning, and safety issues.⁸¹

Standardize transition plans, procedures and forms. All discharging physicians should use a standard discharge summary template that includes pertinent diagnoses, active issues, a reconciled medication list with changes highlighted, results from important tests and consultations, pending test results, planned follow-up and required services, warning signs of a worsening condition, and actions needed if a problem arises.

Always send discharge summaries directly to the patient’s primary care physician or next care setting at the time of discharge.

Give the patient a discharge plan that is easy to understand. Enhance patient and family education using health literacy standards⁸² and interactive methods such as teach-back,⁸³ in which patients demonstrate comprehension and skills required for self-care immediately after being taught. Such tools actively teach patients and caregivers to follow a care plan, including managing medications.

Follow up and coordinate support in a timely manner after a patient leaves the care setting. Follow-up visits should be arranged before discharge. Within 1 to 3 days after discharge, the patient should be called or visited by a case manager, social worker, nurse, or other health care provider.

CHALLENGES TO IMPROVING TRANSITIONS

Although several models demonstrated significant reductions of hospital readmissions in trials, challenges remain. Studies do not identify which features of the models are necessary or sufficient, or how applicable they are to different hospital and patient characteristics. A 2012 analysis⁸⁴ of a program designed to reduce readmissions in three states identified key obstacles to successfully improving care transitions:

Collaborative relationships across settings are critical, but very difficult to achieve. It takes time to develop the relationships and trust among providers, and little incentive exists for skilled nursing facilities and physicians outside the hospital to engage in the process.

Infrastructure is lacking, as is experience to implement quality improvements.

We lack proof that models work on a large scale. Confusion exists about which readmissions are preventable and which are not. More evidence is needed to help guide hospitals’ efforts to improve transitions of care and reduce readmissions.

You have spent several days checking on a patient hospitalized for an acute exacerbation of heart failure. You have straightened out her medications and diet and discussed a plan for follow-up with the patient and a family member, and now she is being wheeled out the door. What happens to her next?

Too often, not your desired plan. If she is going home, maybe she understands what she needs to do, maybe not. Maybe she will get your prescriptions filled and take the medications as directed, maybe not. If she is going to a nursing home, maybe the physician covering the nursing home will get your plan, maybe not. There is a good chance she will be back in the emergency room soon, all because of a poor transition of care.

Transitions of care are changes in the level, location, or providers of care as patients move within the health care system. These can be critical junctures in patients’ lives, and if poorly executed can result in many adverse effects—including rehospitalization.¹

Although high rehospitalization rates gained national attention in 2009 after a analysis of Medicare data,² health care providers have known about the lack of coordinated care transitions for more than 50 years.³ Despite some progress, improving care transitions remains a national challenge. As the health system evolves from a fee-for-service financial model to payment-for-value,⁴ it is especially important that health care providers improve care for patients by optimizing care transitions.

In this article, we summarize the factors contributing to poor care transitions, highlight programs that improve them, and discuss strategies for successful transitions.

TRANSITION PROBLEMS ARE COMMON

Transitions of care occur when patients move to short-term and long-term acute care hospitals, skilled nursing facilities, primary and specialty care offices, community health centers, rehabilitation facilities, home health agencies, hospice, and their own homes.⁵ Problems can arise at any of these transitions, but the risk is especially high when patients leave the hospital to receive care in another setting or at home.

In the past decade, one in five Medicare patients was rehospitalized within 30 days of discharge from the hospital,² and up to 25% were rehospitalized after being discharged to a skilled nursing facility.⁶ Some diagnoses (eg, sickle cell anemia, gangrene) and procedures (eg, kidney transplantation, ileostomy) are associated with readmission rates of nearly one in three.^7,8

The desire of policymakers to “bend the cost curve” of health care has led to efforts to enhance care coordination by improving transitions between care venues. Through the Patient Protection and Affordable Care Act, a number of federal initiatives are promoting strategies to improve care transitions and prevent readmissions after hospital discharge.

The Hospital Readmission Reduction Program⁹ drives much of this effort. In fiscal year 2013 (beginning October 1, 2012), more than 2,000 hospitals incurred financial penalties of up to 1% of total Medicare diagnosis-related group payments (about $280 million the first year) for excess readmissions.¹⁰ The penalty’s maximum rose to 2% in fiscal year 2014 and could increase to 3% in 2015. The total penalty for 2014 is projected to be $227 million, with 2,225 hospitals affected.¹¹

The Centers for Medicare and Medicaid Innovation has committed hundreds of millions of dollars to Community-based Care Transitions Programs¹² and more than $200 million to Hospital Engagement Networks¹³ to carry out the goals of the Partnership for Patients,¹⁴ aiming to reduce rehospitalizations and other adverse events.

At first, despite these efforts, readmission rates did not appear to change substantially.¹⁵ However, the Centers for Medicare and Medicaid Services reported that hospital readmission rates for Medicare fee-for-service beneficiaries declined in 2012 to 18.4%,¹⁶ although some believe that the reduction is related to an increase in the number of patients admitted for observation in recent years.¹⁷

TRANSITIONS ARE OFTEN POORLY COORDINATED

Although some readmissions are unavoidable—resulting from the inevitable progression of disease or worsening of chronic conditions¹⁸—they may also result from a fumbled transition between care settings. Our current system of care transition has serious deficiencies that endanger patients. Areas that need improvement include communication between providers, patient education about medications and treatments, monitoring of medication adherence and complications, follow-up of pending tests and procedures after discharge, and outpatient follow-up soon after discharge.^19–21

Traditional health care does not have dependable mechanisms for coordinating care across settings; we are all ensconced in “silos” that generally keep the focus within individual venues.²² Lack of coordination blurs the lines of responsibility for patients in the period between discharge from one location and admission to another, leaving them confused about whom to contact for care, especially if symptoms worsen.^23,24

Gaps in coordination are not surprising, given the complexity of the US health care system and the often remarkable number of physicians caring for an individual patient.⁵ Medicare beneficiaries see an average of two primary care physicians and five specialists during a 2-year period; patients with chronic conditions may see up to 16 physicians in 1 year.²⁵ Coordinating care between so many providers in different settings, combined with possible patient factors such as disadvantaged socioeconomic status, lack of caregiver support, and inadequate health literacy, provides many opportunities for failures.

Research has identified several root causes behind most failed care transitions:

Poor provider communication

Multiple studies associate adverse events after discharge with a lack of timely communication between hospital and outpatient providers.²⁶ One study estimated that 80% of serious medical errors involve miscommunication during the hand-off between medical providers.²⁷ Discharge summaries often lack important information such as test results, hospital course, discharge medications, patient counseling, and follow-up plans. Most adverse drug events after hospital discharge result directly from breakdown in communication between hospital staff and patients or primary care physicians.²⁸ Approximately 40% of patients have test results pending at the time of discharge and 10% of these require some action; yet outpatient physicians and patients are often unaware of them.²¹

Ineffective patient and caregiver education

The Institute of Medicine report, Crossing the Quality Chasm: A New Health System for the 21st Century,²⁹ noted that patients leaving one setting for another receive little information on how to care for themselves, when to resume activities, what medication side effects to watch out for, and how to get answers to questions. Of particular concern is that patients and caregivers are sometimes omitted from transition planning and often must suddenly assume new self-care responsibilities upon going home that hospital staff managed before discharge. Too often, patients are discharged with inadequate understanding of their medical condition, self-care plan,^23,24 and who should manage their care.³⁰

Up to 36% of adults in the United States have inadequate health literacy (defined as the inability to understand basic health information needed to make appropriate decisions), hindering patient education efforts.^31–33 Even if they understand, patients and their caregivers must be engaged or “activated” (ie, able and willing to manage one’s health) if we expect them to adhere to appropriate care and behaviors. A review found direct correlations between patient activation and healthy behavior, better health outcomes (eg, achieving normal hemoglobin A_1c and cholesterol levels), and better care experiences.³⁴ This review also noted that multiple studies have documented improved activation scores as a result of specific interventions.

No follow-up with primary care providers

The risk of hospital readmission is significantly lower for patients with chronic obstructive pulmonary disease or heart failure who receive follow-up within 7 days of discharge.^35–38 Of Medicare beneficiaries readmitted to the hospital within 30 days of discharge in 2003–2004, half had no contact with an outpatient physician in the interval between their discharge and their readmission,² and one in three adult patients discharged from a hospital to the community does not see a physician within 30 days of discharge.³⁹ The dearth of primary care providers in many communities can make follow-up care difficult to coordinate.

Failure to address chronic conditions

Analyses of national data sets reveal that patients are commonly rehospitalized for conditions unrelated to their initial hospitalization. According to the Center for Studying Health System Change, more than a quarter of readmissions in the 30 days after discharge are for conditions unrelated to those identified in the index admission, the proportion rising to more than one-third at 1 year.³⁹ Among Medicare beneficiaries readmitted within 30 days of discharge, the proportion readmitted for the same condition was just 35% after hospitalization for heart failure, 10% after hospitalization for acute myocardial infarction, and 22% after hospitalization for pneumonia.⁴⁰

Lack of community support

Multiple social and environmental factors contribute to adverse postdischarge events.^41–43 For socioeconomically disadvantaged patients, care-transition issues are compounded by insufficient access to outpatient care, lack of social support, and lack of transportation. Some studies indicate that between 40% to 50% of readmissions are linked to social problems and inadequate access to community resources.^44–47 Psychosocial issues such as limited health literacy, poor self-management skills, inadequate social support, and living alone are associated with adverse outcomes, including readmission and death.^48,49 Such factors may help explain high levels of “no-shows” to outpatient follow-up visits.

NATIONAL MODELS OF BEST PRACTICES

Efforts to reduce readmissions have traditionally focused on hospitals, but experts now recognize that multiple factors influence readmissions and must be comprehensively addressed. Several evidence-based models seek to improve patient outcomes with interventions aimed at care transitions:

Project BOOST

Project BOOST (Better Outcomes by Optimizing Safe Transitions)⁵⁰ is a national initiative developed by the Society of Hospital Medicine to standardize and optimize the care of patients discharged from hospital to home. The program includes evidence-based clinical interventions that can easily be adopted by any hospital. Interventions are aimed at:

• Identifying patients at high risk on admission
• Targeting risk-specific situations
• Improving information flow between inpatient and outpatient providers
• Improving patient and caregiver education by using the teach-back method
• Achieving timely follow-up after discharge.

The program includes a year of technical support provided by a physician mentor.

Preliminary results from pilot sites showed a 14% reduction in 30-day readmission rates in units using BOOST compared with control units in the same hospital.⁵¹ Mentored implementation was recognized by the Joint Commission and the National Quality Forum with the 2011 John M. Eisenberg Award for Innovation in Patient Safety and Quality.⁵²

Project RED

Project RED (Re-Engineered Discharge)⁵³ evolved from efforts by Dr. Brian Jack and colleagues to re-engineer the hospital workflow process to improve patient safety and reduce rehospitalization rates at Boston Medical Center. The intervention has 12 mutually reinforcing components aimed at improving the discharge process.

In a randomized controlled trial, Project RED led to a 30% decrease in emergency department visits and readmissions within 30 days of discharge from a general medical service of an urban academic medical center.⁵⁴ This study excluded patients admitted from a skilled nursing facility or discharged to one, but a recent study demonstrated that Project RED also led to a lower rate of hospital admission within 30 days of discharge from a skilled nursing facility.⁵⁵

The STAAR initiative

The STAAR initiative (State Action on Avoidable Re-hospitalizations)⁵⁶ was launched in 2009 by the Institute for Healthcare Improvement with the goal of reducing avoidable readmissions in the states of Massachusetts, Michigan, and Washington. Hospital teams focus on improving:

• Assessment of needs after hospital discharge
• Teaching and learning
• Real-time hand-off communication
• Timely follow-up after hospital discharge.

As yet, no published studies other than case reports show a benefit from STAAR.⁵⁷

The Care Transitions Program

The Care Transitions Program,⁵⁸ under the leadership of Dr. Eric Coleman, aims to empower patients and caregivers, who meet with a “transition coach.” The program provides assistance with medication reconciliation and self-management, a patient-centered record owned and maintained by the patient to facilitate cross-site information transfer, timely outpatient follow-up with primary or specialty care, a list of red flags to indicate a worsening condition, and instructions on proper responses.

A randomized controlled trial of the program demonstrated a reduction in hospital readmissions at 30, 90, and 180 days, and lower hospital costs at 90 and 180 days.⁵⁹ This approach also proved effective in a real-world setting.⁶⁰

The Transitional Care Model

Developed by Dr. Mary Naylor and colleagues, the Transitional Care Model⁶¹ also aims at patient and family empowerment, focusing on patients’ stated goals and priorities and ensuring patient engagement. In the program, a transitional care nurse has the job of enhancing patient and caregiver understanding, facilitating patient self-management, and overseeing medication management and transitional care.

A randomized controlled trial demonstrated improved outcomes after hospital discharge for elderly patients with complex medical illnesses, with overall reductions in medical costs through preventing or delaying rehospitalization.⁶² A subsequent real-world study validated this approach.⁶³

The Bridge Model

The Illinois Transitional Care Consortium’s Bridge Model⁶⁴ is for older patients discharged home after hospitalization. It is led by social workers (“bridge care coordinators”) who address barriers to implementing the discharge plan, coordinate resources, and intervene at three points: before discharge, 2 days after discharge, and 30 days after discharge.

An initial study showed no impact on the 30-day rehospitalization rate,⁶⁵ but larger studies are under way with a modified version.

Guided Care

Developed at the Johns Hopkins Bloomberg School of Public Health, Guided Care⁶⁶ involves nurses who work in partnership with physicians and others in primary care to provide patient-centered, cost-effective care to patients with multiple chronic conditions. Nurses conduct in-home assessments, facilitate care planning, promote patient self-management, monitor conditions, coordinate the efforts of all care professionals, and facilitate access to community resources.

A cluster-randomized controlled trial found that this program had mixed results, reducing the use of home health care but having little effect on the use of other health services in the short run. However, in the subgroup of patients covered by Kaiser-Permanente, those who were randomized to the program accrued, on average, 52% fewer skilled nursing facility days, 47% fewer skilled nursing facility admissions, 49% fewer hospital readmissions, and 17% fewer emergency department visits.⁶⁷

The GRACE model

The GRACE model (Geriatric Resources for Assessment and Care of Elders)⁶⁸ was developed to improve the quality of geriatric care, reduce excess health care use, and prevent long-term nursing home placement. Each patient is assigned a support team consisting of a nurse practitioner and a social worker who make home visits, coordinate health care and community services, and develop an individualized care plan.

In one study,⁶⁹ GRACE reduced hospital admission rates for participants at high risk of hospitalization by 12% in the first year of the program and 44% in the second year. GRACE participants also reported higher quality of life compared with the control group.⁶⁹

INTERACT tools

Led by Dr. Joseph Ouslander, INTERACT (Interventions to Reduce Acute Care Transfers)⁷⁰ is a quality-improvement initiative for skilled nursing facilities, designed to facilitate the early identification, evaluation, documentation, and communication of changes in the status of residents. Visitors to its website can download a set of tools and strategies to help them manage conditions before they become serious enough to require a hospital transfer. The tools assist in promoting important communication among providers and enhancing advance-care planning.

A 6-month study in 25 nursing homes showed a 17% reduction in self-reported hospital admissions with this program compared with the same period the previous year.⁷¹

Additional home-based care interventions

Additional innovations are under way in home-based care.

The Home Health Quality Improvement National Campaign is a patient-centered movement to improve the quality of care received by patients residing at home.⁷² Through its Best Practices Intervention Packages, it offers evidence-based educational tools, resources, and interventions for reducing avoidable hospitalizations, improving medication management, and coordinating transitional care.

The Center for Medicare and Medicaid Innovation Independence at Home Demonstration⁷³ is testing whether home-based comprehensive primary care can improve care and reduce hospitalizations for Medicare beneficiaries with multiple chronic conditions.

NO SINGLE INTERVENTION: MULTIPLE STRATEGIES NEEDED

A 2011 review found no single intervention that regularly reduced the 30-day risk of re-hospitalization.⁷⁴ However, other studies have shown that multifaceted interventions can reduce 30-day readmission rates. Randomized controlled trials in short-stay, acute care hospitals indicate that improvement in the following areas can directly reduce hospital readmission rates:

• Comprehensive planning and risk assessment throughout hospitalization
• Quality of care during the initial admission
• Communication with patients, their caregivers, and their clinicians
• Patient education
• Predischarge assessment
• Coordination of care after discharge.

In randomized trials, successful programs reduced the 30-day readmission rates by 20% to 40%,^{54,62,75–79} and a 2011 meta-analysis of randomized clinical trials found evidence that interventions associated with discharge planning helped to reduce readmission rates.⁸⁰

Methods developed by the national care transition models described above can help hospitals optimize patient transitions (Table 1). Although every model has its unique attributes, they have several strategies in common:

Engage a team of key stakeholders that may include patients and caregivers, hospital staff (physicians, nurses, case managers, social workers, and pharmacists), community physicians (primary care, medical homes, and specialists), advance practice providers (physician assistants and nurse practitioners), and postacute care facilities and services (skilled nursing facilities, home health agencies, assisted living residences, hospice, and rehabilitation facilities).

Develop a comprehensive transition plan throughout hospitalization that includes attention to factors that may affect self-care, such as health literacy, chronic conditions, medications, and social support.

Enhance medication reconciliation and management. Obtain the best possible medication history on admission, and ensure that patients understand changes in their medications, how to take each medicine correctly, and important side effects.

Institute daily interdisciplinary communication and care coordination by everyone on the health care team with an emphasis on the care plan, discharge planning, and safety issues.⁸¹

Standardize transition plans, procedures and forms. All discharging physicians should use a standard discharge summary template that includes pertinent diagnoses, active issues, a reconciled medication list with changes highlighted, results from important tests and consultations, pending test results, planned follow-up and required services, warning signs of a worsening condition, and actions needed if a problem arises.

Always send discharge summaries directly to the patient’s primary care physician or next care setting at the time of discharge.

Give the patient a discharge plan that is easy to understand. Enhance patient and family education using health literacy standards⁸² and interactive methods such as teach-back,⁸³ in which patients demonstrate comprehension and skills required for self-care immediately after being taught. Such tools actively teach patients and caregivers to follow a care plan, including managing medications.

Follow up and coordinate support in a timely manner after a patient leaves the care setting. Follow-up visits should be arranged before discharge. Within 1 to 3 days after discharge, the patient should be called or visited by a case manager, social worker, nurse, or other health care provider.

CHALLENGES TO IMPROVING TRANSITIONS

Although several models demonstrated significant reductions of hospital readmissions in trials, challenges remain. Studies do not identify which features of the models are necessary or sufficient, or how applicable they are to different hospital and patient characteristics. A 2012 analysis⁸⁴ of a program designed to reduce readmissions in three states identified key obstacles to successfully improving care transitions:

Collaborative relationships across settings are critical, but very difficult to achieve. It takes time to develop the relationships and trust among providers, and little incentive exists for skilled nursing facilities and physicians outside the hospital to engage in the process.

Infrastructure is lacking, as is experience to implement quality improvements.

We lack proof that models work on a large scale. Confusion exists about which readmissions are preventable and which are not. More evidence is needed to help guide hospitals’ efforts to improve transitions of care and reduce readmissions.

No. Based on current evidence and guidelines, routine acid-suppressive therapy to prevent stress ulcers has no benefit in hospitalized patients outside the critical-care setting. Only critically ill patients who meet specific criteria, as described in the guidelines of the American Society of Health System Pharmacists, should receive acid-suppressive therapy.

Unfortunately, routine stress ulcer prophylaxis is common in US hospitals, unnecessarily putting patients at risk of complications and adding costs.

STRESS ULCER AND CRITICAL ILLNESS

Stress ulcers—ulcerations of the upper part of the gastrointestinal (GI) mucosa in the setting of acute disease—usually involve the fundus and body of the stomach. The stomach is lined with a glycoprotein mucous layer rich in bicarbonates, forming a physiologic barrier to protect the gastric wall from acid insult by neutralizing hydrogen ions. Disruption of this protective layer can occur in critically ill patients (eg, those with shock or sepsis) through overproduction of uremic toxins, increased reflux of bile salts, compromised blood flow, and increased stomach acidity through gastrin stimulation of parietal cells.

More than 75% of patients with major burns or cranial trauma develop endoscopic mucosal abnormalities within 72 hours of injury.¹ In critically ill patients, the risk of ulcer-related overt bleeding is estimated to be 5% to 25%. Furthermore, 1% to 5% of stress ulcers can be deep enough to erode into the submucosa, causing clinically significant GI bleeding, defined as bleeding complicated by hemodynamic compromise or a drop in hemoglobin that requires a blood transfusion.² In contrast, in inpatients who are not critically ill, the risk of overt bleeding from stress ulcers is less than 1%.³

ADDRESSING RISK

A multicenter prospective cohort study of 2,252 intensive care patients² reported two main risk factors for significant bleeding caused by stress ulcers: mechanical ventilation for more than 48 hours and coagulopathy, defined as a platelet count below 50 × 10⁹/L, an international normalized ratio greater than 1.5, or a partial thromboplastin time more than twice the control value.⁴ In hemodynamically stable patients receiving anticoagulation in a general medical or surgical ward, the risk of GI bleeding was low, and acid suppression failed to lower the rate of stress ulcer occurrence.³

Other risk factors include severe sepsis, shock, liver failure, kidney failure, burns over 35% of the total body surface, organ transplantation, cranial trauma, spinal cord trauma, history of peptic ulcer disease, and history of upper GI bleeding.^3,5,6 Steroid therapy is not considered a risk factor for stress ulcers unless it is used in the presence of another risk factor such as use of aspirin or nonsteroidal antiinflammatory drugs (NSAIDs).²

INDICATIONS FOR PROPHYLAXIS

Prophylaxis with a proton pump inhibitor (PPI) is indicated in specific conditions—ie, peptic ulcer disease, gastroesophageal reflux disease, chronic NSAID therapy, and Zollinger-Ellison syndrome—and to eradicate Helicobacter pylori infection.⁷ But in the United States, stress ulcer prophylaxis is overused in general-care floors despite the lack of supporting evidence.

The American Society of Health System Pharmacists guidelines recommend it in the intensive care unit for patients with any of the following: coagulopathy, prolonged mechanical ventilation (more than 48 hours), GI ulcer or bleeding within the past year, sepsis, a stay longer than 1 week in the intensive care unit, occult GI bleeding for 6 or more days, and steroid therapy with more than 250 mg of hydrocortisone daily.⁸ Hemodynamically stable patients admitted to general-care floors should not receive stress ulcer prophylaxis, as it only negligibly decreases the rate of GI bleeding, from 0.33% to 0.22%.⁹

WHY ROUTINE ULCER PROPHYLAXIS IS NOT FOR ALL HOSPITALIZED PATIENTS

Although stress ulcer prophylaxis is often considered benign, its lack of proven benefit, additional cost, and risk of adverse effects, including interactions with foods and other drugs, preclude using it routinely for all hospitalized patients.^10,11 Chronic use of PPIs has been associated with complications, as discussed below.

Infection

Acid suppression may impair the destruction of ingested microorganisms, resulting in overgrowth of bacteria.¹² Overuse of PPIs may increase the risk of several infections:

• Diarrhea due to Clostridium difficile¹²
• Community-acquired pneumonia, from increased microaspiration of overgrown microorganisms into the lung.¹²
• Spontaneous bacterial peritonitis in patients with cirrhosis,¹³ although the mechanism is not clear. (Small-bowel bacterial overgrowth is the hypothesized cause.)

Bone fracture

PPIs lower gastric acidity, and this can inhibit intestinal calcium absorption. Furthermore, PPIs may directly inhibit bone resorption by osteoclasts.¹⁴

Reduction in clopidogrel efficacy

PPIs may reduce the efficacy of clopidogrel as a result of competitive inhibition of cytochrome CYP2C19, which is necessary to metabolize clopidogrel to its active forms. Therefore, concomitant use of clopidogrel with omeprazole, esomeprazole, or other CYP2C19 inhibitors is not recommended.¹⁵

Nutritional deficiencies

The overgrown microorganisms consume cobalamin in the stomach, resulting in vitamin B₁₂ deficiency. Acid-suppressive therapy can also reduce the absorption of magnesium and iron.¹²

Unnecessary cost

Heidelbaugh and Inadomi¹⁶ reviewed the non-evidence-based use of stress ulcer prophylaxis in patients admitted to a large university hospital and estimated that it entailed a cost to the hospital of $111,791 over the course of a year.

WHICH ULCER PROPHYLAXIS SHOULD BE USED IN CRITICALLY ILL PATIENTS?

Studies have shown histamine-2 blockers to be superior to antacids and sucralfate in preventing stress ulcer and GI bleeding,^8,15 but no study has compared PPIs with sucralfate and antacids.

When indicated, an oral PPI is preferred over an oral histamine-2 blocker for GI prophylaxis.¹⁷ This practice is considered cost-effective and is associated with lower rates of stress ulcer and GI bleeding. In intubated patients, however, an intravenous histamine-2 blocker is preferable to an intravenous PPI.^3,8,11 Interestingly, no difference was reported between PPIs and histamine-2 blockers in terms of mortality rate or reduction in the incidence of nosocomial pneumonia.¹⁷

OUR RECOMMENDATION

Only critically ill patients who meet the specific criteria described here should receive stress ulcer prophylaxis. More effort is needed to educate residents, medical staff, and pharmacists about current guidelines. Computerized ordering templates and reminders to discontinue prophylaxis at discharge or step-down may decrease overall use, reduce costs, and limit potential side effects.¹⁸

No. Based on current evidence and guidelines, routine acid-suppressive therapy to prevent stress ulcers has no benefit in hospitalized patients outside the critical-care setting. Only critically ill patients who meet specific criteria, as described in the guidelines of the American Society of Health System Pharmacists, should receive acid-suppressive therapy.

Unfortunately, routine stress ulcer prophylaxis is common in US hospitals, unnecessarily putting patients at risk of complications and adding costs.

STRESS ULCER AND CRITICAL ILLNESS

Stress ulcers—ulcerations of the upper part of the gastrointestinal (GI) mucosa in the setting of acute disease—usually involve the fundus and body of the stomach. The stomach is lined with a glycoprotein mucous layer rich in bicarbonates, forming a physiologic barrier to protect the gastric wall from acid insult by neutralizing hydrogen ions. Disruption of this protective layer can occur in critically ill patients (eg, those with shock or sepsis) through overproduction of uremic toxins, increased reflux of bile salts, compromised blood flow, and increased stomach acidity through gastrin stimulation of parietal cells.

More than 75% of patients with major burns or cranial trauma develop endoscopic mucosal abnormalities within 72 hours of injury.¹ In critically ill patients, the risk of ulcer-related overt bleeding is estimated to be 5% to 25%. Furthermore, 1% to 5% of stress ulcers can be deep enough to erode into the submucosa, causing clinically significant GI bleeding, defined as bleeding complicated by hemodynamic compromise or a drop in hemoglobin that requires a blood transfusion.² In contrast, in inpatients who are not critically ill, the risk of overt bleeding from stress ulcers is less than 1%.³

ADDRESSING RISK

A multicenter prospective cohort study of 2,252 intensive care patients² reported two main risk factors for significant bleeding caused by stress ulcers: mechanical ventilation for more than 48 hours and coagulopathy, defined as a platelet count below 50 × 10⁹/L, an international normalized ratio greater than 1.5, or a partial thromboplastin time more than twice the control value.⁴ In hemodynamically stable patients receiving anticoagulation in a general medical or surgical ward, the risk of GI bleeding was low, and acid suppression failed to lower the rate of stress ulcer occurrence.³

Other risk factors include severe sepsis, shock, liver failure, kidney failure, burns over 35% of the total body surface, organ transplantation, cranial trauma, spinal cord trauma, history of peptic ulcer disease, and history of upper GI bleeding.^3,5,6 Steroid therapy is not considered a risk factor for stress ulcers unless it is used in the presence of another risk factor such as use of aspirin or nonsteroidal antiinflammatory drugs (NSAIDs).²

INDICATIONS FOR PROPHYLAXIS

Prophylaxis with a proton pump inhibitor (PPI) is indicated in specific conditions—ie, peptic ulcer disease, gastroesophageal reflux disease, chronic NSAID therapy, and Zollinger-Ellison syndrome—and to eradicate Helicobacter pylori infection.⁷ But in the United States, stress ulcer prophylaxis is overused in general-care floors despite the lack of supporting evidence.

The American Society of Health System Pharmacists guidelines recommend it in the intensive care unit for patients with any of the following: coagulopathy, prolonged mechanical ventilation (more than 48 hours), GI ulcer or bleeding within the past year, sepsis, a stay longer than 1 week in the intensive care unit, occult GI bleeding for 6 or more days, and steroid therapy with more than 250 mg of hydrocortisone daily.⁸ Hemodynamically stable patients admitted to general-care floors should not receive stress ulcer prophylaxis, as it only negligibly decreases the rate of GI bleeding, from 0.33% to 0.22%.⁹

WHY ROUTINE ULCER PROPHYLAXIS IS NOT FOR ALL HOSPITALIZED PATIENTS

Although stress ulcer prophylaxis is often considered benign, its lack of proven benefit, additional cost, and risk of adverse effects, including interactions with foods and other drugs, preclude using it routinely for all hospitalized patients.^10,11 Chronic use of PPIs has been associated with complications, as discussed below.

Infection

Acid suppression may impair the destruction of ingested microorganisms, resulting in overgrowth of bacteria.¹² Overuse of PPIs may increase the risk of several infections:

• Diarrhea due to Clostridium difficile¹²
• Community-acquired pneumonia, from increased microaspiration of overgrown microorganisms into the lung.¹²
• Spontaneous bacterial peritonitis in patients with cirrhosis,¹³ although the mechanism is not clear. (Small-bowel bacterial overgrowth is the hypothesized cause.)

Bone fracture

PPIs lower gastric acidity, and this can inhibit intestinal calcium absorption. Furthermore, PPIs may directly inhibit bone resorption by osteoclasts.¹⁴

Reduction in clopidogrel efficacy

PPIs may reduce the efficacy of clopidogrel as a result of competitive inhibition of cytochrome CYP2C19, which is necessary to metabolize clopidogrel to its active forms. Therefore, concomitant use of clopidogrel with omeprazole, esomeprazole, or other CYP2C19 inhibitors is not recommended.¹⁵

Nutritional deficiencies

The overgrown microorganisms consume cobalamin in the stomach, resulting in vitamin B₁₂ deficiency. Acid-suppressive therapy can also reduce the absorption of magnesium and iron.¹²

Unnecessary cost

Heidelbaugh and Inadomi¹⁶ reviewed the non-evidence-based use of stress ulcer prophylaxis in patients admitted to a large university hospital and estimated that it entailed a cost to the hospital of $111,791 over the course of a year.

WHICH ULCER PROPHYLAXIS SHOULD BE USED IN CRITICALLY ILL PATIENTS?

Studies have shown histamine-2 blockers to be superior to antacids and sucralfate in preventing stress ulcer and GI bleeding,^8,15 but no study has compared PPIs with sucralfate and antacids.

When indicated, an oral PPI is preferred over an oral histamine-2 blocker for GI prophylaxis.¹⁷ This practice is considered cost-effective and is associated with lower rates of stress ulcer and GI bleeding. In intubated patients, however, an intravenous histamine-2 blocker is preferable to an intravenous PPI.^3,8,11 Interestingly, no difference was reported between PPIs and histamine-2 blockers in terms of mortality rate or reduction in the incidence of nosocomial pneumonia.¹⁷

OUR RECOMMENDATION

Only critically ill patients who meet the specific criteria described here should receive stress ulcer prophylaxis. More effort is needed to educate residents, medical staff, and pharmacists about current guidelines. Computerized ordering templates and reminders to discontinue prophylaxis at discharge or step-down may decrease overall use, reduce costs, and limit potential side effects.¹⁸

No. Based on current evidence and guidelines, routine acid-suppressive therapy to prevent stress ulcers has no benefit in hospitalized patients outside the critical-care setting. Only critically ill patients who meet specific criteria, as described in the guidelines of the American Society of Health System Pharmacists, should receive acid-suppressive therapy.

Unfortunately, routine stress ulcer prophylaxis is common in US hospitals, unnecessarily putting patients at risk of complications and adding costs.

STRESS ULCER AND CRITICAL ILLNESS

Stress ulcers—ulcerations of the upper part of the gastrointestinal (GI) mucosa in the setting of acute disease—usually involve the fundus and body of the stomach. The stomach is lined with a glycoprotein mucous layer rich in bicarbonates, forming a physiologic barrier to protect the gastric wall from acid insult by neutralizing hydrogen ions. Disruption of this protective layer can occur in critically ill patients (eg, those with shock or sepsis) through overproduction of uremic toxins, increased reflux of bile salts, compromised blood flow, and increased stomach acidity through gastrin stimulation of parietal cells.

More than 75% of patients with major burns or cranial trauma develop endoscopic mucosal abnormalities within 72 hours of injury.¹ In critically ill patients, the risk of ulcer-related overt bleeding is estimated to be 5% to 25%. Furthermore, 1% to 5% of stress ulcers can be deep enough to erode into the submucosa, causing clinically significant GI bleeding, defined as bleeding complicated by hemodynamic compromise or a drop in hemoglobin that requires a blood transfusion.² In contrast, in inpatients who are not critically ill, the risk of overt bleeding from stress ulcers is less than 1%.³

ADDRESSING RISK

A multicenter prospective cohort study of 2,252 intensive care patients² reported two main risk factors for significant bleeding caused by stress ulcers: mechanical ventilation for more than 48 hours and coagulopathy, defined as a platelet count below 50 × 10⁹/L, an international normalized ratio greater than 1.5, or a partial thromboplastin time more than twice the control value.⁴ In hemodynamically stable patients receiving anticoagulation in a general medical or surgical ward, the risk of GI bleeding was low, and acid suppression failed to lower the rate of stress ulcer occurrence.³

Other risk factors include severe sepsis, shock, liver failure, kidney failure, burns over 35% of the total body surface, organ transplantation, cranial trauma, spinal cord trauma, history of peptic ulcer disease, and history of upper GI bleeding.^3,5,6 Steroid therapy is not considered a risk factor for stress ulcers unless it is used in the presence of another risk factor such as use of aspirin or nonsteroidal antiinflammatory drugs (NSAIDs).²

INDICATIONS FOR PROPHYLAXIS

Prophylaxis with a proton pump inhibitor (PPI) is indicated in specific conditions—ie, peptic ulcer disease, gastroesophageal reflux disease, chronic NSAID therapy, and Zollinger-Ellison syndrome—and to eradicate Helicobacter pylori infection.⁷ But in the United States, stress ulcer prophylaxis is overused in general-care floors despite the lack of supporting evidence.

The American Society of Health System Pharmacists guidelines recommend it in the intensive care unit for patients with any of the following: coagulopathy, prolonged mechanical ventilation (more than 48 hours), GI ulcer or bleeding within the past year, sepsis, a stay longer than 1 week in the intensive care unit, occult GI bleeding for 6 or more days, and steroid therapy with more than 250 mg of hydrocortisone daily.⁸ Hemodynamically stable patients admitted to general-care floors should not receive stress ulcer prophylaxis, as it only negligibly decreases the rate of GI bleeding, from 0.33% to 0.22%.⁹

WHY ROUTINE ULCER PROPHYLAXIS IS NOT FOR ALL HOSPITALIZED PATIENTS

Although stress ulcer prophylaxis is often considered benign, its lack of proven benefit, additional cost, and risk of adverse effects, including interactions with foods and other drugs, preclude using it routinely for all hospitalized patients.^10,11 Chronic use of PPIs has been associated with complications, as discussed below.

Infection

Acid suppression may impair the destruction of ingested microorganisms, resulting in overgrowth of bacteria.¹² Overuse of PPIs may increase the risk of several infections:

• Diarrhea due to Clostridium difficile¹²
• Community-acquired pneumonia, from increased microaspiration of overgrown microorganisms into the lung.¹²
• Spontaneous bacterial peritonitis in patients with cirrhosis,¹³ although the mechanism is not clear. (Small-bowel bacterial overgrowth is the hypothesized cause.)

Bone fracture

PPIs lower gastric acidity, and this can inhibit intestinal calcium absorption. Furthermore, PPIs may directly inhibit bone resorption by osteoclasts.¹⁴

Reduction in clopidogrel efficacy

PPIs may reduce the efficacy of clopidogrel as a result of competitive inhibition of cytochrome CYP2C19, which is necessary to metabolize clopidogrel to its active forms. Therefore, concomitant use of clopidogrel with omeprazole, esomeprazole, or other CYP2C19 inhibitors is not recommended.¹⁵

Nutritional deficiencies

The overgrown microorganisms consume cobalamin in the stomach, resulting in vitamin B₁₂ deficiency. Acid-suppressive therapy can also reduce the absorption of magnesium and iron.¹²

Unnecessary cost

Heidelbaugh and Inadomi¹⁶ reviewed the non-evidence-based use of stress ulcer prophylaxis in patients admitted to a large university hospital and estimated that it entailed a cost to the hospital of $111,791 over the course of a year.

WHICH ULCER PROPHYLAXIS SHOULD BE USED IN CRITICALLY ILL PATIENTS?

Studies have shown histamine-2 blockers to be superior to antacids and sucralfate in preventing stress ulcer and GI bleeding,^8,15 but no study has compared PPIs with sucralfate and antacids.

When indicated, an oral PPI is preferred over an oral histamine-2 blocker for GI prophylaxis.¹⁷ This practice is considered cost-effective and is associated with lower rates of stress ulcer and GI bleeding. In intubated patients, however, an intravenous histamine-2 blocker is preferable to an intravenous PPI.^3,8,11 Interestingly, no difference was reported between PPIs and histamine-2 blockers in terms of mortality rate or reduction in the incidence of nosocomial pneumonia.¹⁷

OUR RECOMMENDATION

Only critically ill patients who meet the specific criteria described here should receive stress ulcer prophylaxis. More effort is needed to educate residents, medical staff, and pharmacists about current guidelines. Computerized ordering templates and reminders to discontinue prophylaxis at discharge or step-down may decrease overall use, reduce costs, and limit potential side effects.¹⁸

In 2012, a task force of the European Society of Cardiology, the American College of Cardiology Foundation, the American Heart Association, and the World Heart Federation released its “third universal definition” of myocardial infarction (MI),¹ replacing the previous (2007) definition. The new consensus definition reflects the increasing sensitivity of available troponin assays, which are commonly elevated in other conditions and after uncomplicated percutaneous coronary intervention or cardiac surgery. With a more appropriate definition of the troponin threshold after these procedures, benign myocardial injury can be differentiated from pathologic MI.

TROPONINS: THE PREFERRED MARKERS

Symptoms of MI such as nausea, chest pain, epigastric discomfort, syncope, and diaphoresis may be nonspecific, and findings on electrocardiography or imaging studies may be nondiagnostic. We thus rely on biomarker elevations to identify patients who need treatment.

Cardiac troponin I and cardiac troponin T have become the preferred markers for detecting MI, as they are more sensitive and tissue-specific than their main competitor, the MB fraction of creatine kinase (CK-MB).² But the newer troponin assays, which are even more sensitive than earlier ones, have raised concerns about their ability to differentiate patients who truly have acute coronary syndromes from those with other causes of troponin elevation. This can have major effects on treatment, patient psyche, and hospital costs.

Troponin elevations can occur in patients with heart failure, end-stage renal disease, sepsis, acute pulmonary embolism, myopericarditis, arrhythmias, and many other conditions. As noted by the task force, these cases of elevated troponin in the absence of clinical supportive evidence should not be labeled as an MI but rather as myocardial injury.

Troponins bind actin and myosin filaments in a trimeric complex composed of troponins I, C, and T. Troponins are present in all muscle cells, but the cardiac isoforms are specific to myocardial tissue.

As a result, both cardiac troponin I and cardiac troponin T, as measured by fourth-generation assays, are highly sensitive (75.2%, 95% confidence interval [CI] 66.8%–83.4%) and specific (94.6%, 95% CI 93.4%–96.3%) for detecting pathologic processes involving the heart.^3,4 Nonetheless, increases in cardiac troponin T (but not I) have been documented in patients with disease of skeletal muscles, likely secondary to re-expressed isoforms of the troponin C gene present in both cardiac and skeletal myocytes.³ There has been no evidence to suggest that either cardiac troponin I nor cardiac troponin T is superior to the other as a marker of MI.

Serum troponin levels detectably rise by 2 to 3 hours after myocardial injury. This temporal pattern is similar to that of CK-MB, which rises at about 2 hours and reaches a peak in 4 to 6 hours. However, troponins are more sensitive than CK-MB during this early time period, since a greater proportion is released from the heart during times of cardiac injury.

The definition of an abnormal troponin value is set by the precision of each individual assay. The task force has designated the optimal precision for troponin assays to be at a coefficient of variation of less than 10% when describing a value exceeding the 99th percentile in a reference population. The 99th percentile, which is the upper reference limit, corresponds to a value near 0.035 μg/L for fourth-generation troponin I and troponin T assays.⁵ Most assays have been adapted to ensure that they meet such criteria.

High-sensitivity assays

Over the past few years, “high-sensitivity” assays have been developed that can detect nanogram levels of troponin.

In one study, an algorithm that incorporated high-sensitivity cardiac troponin T levels was able to rule in or rule out acute MI in 77% of patients with chest pain within 1 hour.⁶ The algorithm had a sensitivity and negative predictive value of 100%.

Other studies have shown a sensitivity of 100.0%, a specificity of 34.0%, and a negative predictive value of 100.0% when using a cardiac troponin T cutoff of 3 ng/L, while a cutoff of 14 ng/L yielded a sensitivity of 85.4%, a specificity of 82.4%, and a negative predictive value of 96.1%.⁴ With cutoffs as low as 3 ng/L, some assays detect elevated troponin in up to 90% of people in normal reference populations without MI.⁷

Physicians thus need to be aware that high-sensitivity troponin assays should mainly be used to rule out acute coronary syndrome, as their high sensitivity substantially compromises their specificity. The appropriate thresholds for various patient populations, the appropriate testing procedures with high-sensitivity assays as compared with the fourth-generation troponin assays (ie, frequency of testing, change in level, and rise), and the cost and clinical outcomes of care based on algorithms that use these values remain unclear and will require further study.^8,9

TYPES OF MYOCARDIAL INFARCTION

The task force defines the following categories of MI (Table 1):

Type 1: Spontaneous myocardial infarction

Type 1, or “spontaneous” MI, is an acute coronary syndrome, colloquially called a “heart attack.” It is primarily the result of rupture, fissuring, erosion, or dissection of atherosclerotic plaque. Most are the result of underlying atherosclerotic coronary artery disease, although some (ie, those caused by coronary dissection) are not.

To diagnose type 1 MI, a blood sample must detect a rise or fall (or both) of cardiac biomarker values (preferably cardiac troponin), with at least one value above the 99th percentile. However, an elevated troponin level is not sufficient. At least one of the following criteria must also be met:

• Symptoms of ischemia
• New ST-segment or T-wave changes or new left bundle branch block
• Development of pathologic Q waves
• Imaging evidence of new loss of viable myocardium or new wall-motion abnormality
• Finding of an intracoronary thrombus by angiography or autopsy.

Type 1 MI therapy requires antithrombotic drugs and, with the additional findings, revascularization.

Type 2 MI is caused by a supply-demand imbalance in myocardial perfusion, resulting in ischemic damage. This specifically excludes acute coronary thrombosis, but can result from marked changes in demand or supply (eg, sepsis) or from a combination of acute changes and chronic conditions (eg, tachycardia with baseline coronary artery disease). Baseline stable coronary artery disease, left ventricular hypertrophy, endothelial dysfunction, coronary artery spasm, coronary embolism, arrhythmias, anemia, respiratory failure, hypotension, and hypertension can all contribute to a supply-demand mismatch sufficient to cause permanent myocardial damage.

The criteria for diagnosing type 2 MI are the same as for type 1: both elevated troponin levels and one of the clinical criteria (symptoms of ischemia, electrocardiographic changes, new wall-motion abnormality, or intracoronary thrombus) must be present.

Of importance, unlike those with type 1 MI, most patients with type 2 MI are unlikely to immediately benefit from antithrombotic therapy, as they typically have no acute thrombosis (except in cases of coronary embolism). Therapy should instead be directed at the underlying supply-demand imbalance and may include volume resuscitation, blood pressure support or control, or control of tachyarrhythmias.

In the long term, treatment to resolve or prevent supply-demand imbalances may also include revascularization or antithrombotic drugs, but these may be contraindicated in the acute setting.

Type 3: Sudden cardiac death from MI

The third type of MI occurs when myocardial ischemia results in sudden cardiac death before blood samples can be obtained. Before dying, the patient should have had symptoms suggesting myocardial ischemia and should have had presumed new ischemic electrocardiographic changes or new left bundle branch block.

This definition of MI is not very useful clinically but is important for population-based research studies.

Type 4a: Due to percutaneous coronary intervention

A rise in CK-MB levels after percutaneous coronary intervention has been associated with a higher rate of death or recurrent MI.¹⁰ Previously, type 4 MI was defined as an elevation of cardiac biomarker values (> 3 times the 99th percentile) after percutaneous coronary intervention in a patient who had a normal baseline value (< 99th percentile).¹¹

Unfortunately, using troponin at this threshold, the number of cases is five times higher than when CK-MB is used, without a consistent correlation with the outcomes of death or complications.¹² Currently, the increase in cardiac troponin after percutaneous coronary intervention is best interpreted as a marker of the patient’s atherothrombotic burden more than as a predictor of adverse outcomes.¹³

The updated definition of MI associated with percutaneous coronary intervention now requires an elevation of cardiac troponin values greater than 5 times the 99th percentile in a patient who had normal baseline values or an increase of more than 20% from baseline within 48 hours of the procedure. As this value has been arbitrarily assigned rather than based on an established threshold with clinical outcomes, a true MI must further meet one of the following criteria:

• Symptoms suggesting myocardial ischemia
• New ischemic electrocardiographic changes or new left bundle branch block
• Angiographic loss of patency of a major coronary artery or a side branch or persistent slow-flow or no-flow or embolization
• Imaging evidence of a new loss of viable myocardium or a new wall-motion abnormality.

Given that troponin levels may be elevated in up to 65% of patients after uncomplicated percutaneous coronary intervention and this elevation may be unavoidable,¹⁴ a higher troponin threshold to diagnose MI and the clear requirement of clinical correlates may resonate with physicians as a more appropriate definition. In turn, such guidelines may better identify those with an adverse event, while partly reducing unnecessary hospitalization and observation time in those for whom it is not necessary.

Type 4b: Due to stent thrombosis

Type 4b MI is MI caused by stent thrombosis. The thrombosis must be detected by coronary angiography or autopsy in the setting of myocardial ischemia and a rise or fall of cardiac biomarker values, with at least one value above the 99th percentile.

Type 4c: Due to restenosis

Proposed is the addition of type 4c MI, ie, MI resulting from restenosis of more than 50%, because restenosis after percutaneous coronary intervention can lead to MI without thrombosis.¹⁵

Type 5: After coronary artery bypass grafting

Similar to the situation after percutaneous coronary intervention, increased CK-MB levels after coronary artery bypass graft surgery are associated with poor outcomes.¹⁶ Although some studies have indicated that increased troponin levels within 24 hours of this surgery are associated with higher death rates, no study has established a troponin threshold that correlates with outcomes.¹⁷

The task force acknowledged this lack of prognostic value but arbitrarily defined type 5 MI as requiring biomarker elevations greater than 10 times the 99th percentile during the first 48 hours after surgery, with a normal baseline value. One of the following additional criteria must also be met:

• New pathologic Q waves or new left bundle branch block
• Angiographically documented new occlusion in the graft or native coronary artery
• Imaging evidence of new loss of viable myocardium or new wall-motion abnormality.

CHANGES FROM THE 2007 DEFINITIONS

Updates to the definitions of the MI types since the 2007 task force definition can be found in Table 1.

In type 1 and 2 MI, the finding of an intracoronary thrombus by angiography or autopsy was added as one of the possible criteria for evidence of myocardial ischemia.

In type 3 MI, the definition was simplified by deleting the former criterion of finding a fresh thrombus by angiography or autopsy.

In type 4a MI, by requiring clinical correlates, the updated definition in particular moves away from relying solely on troponin levels to diagnose an infarction after percutaneous coronary intervention, as was the case in 2007. Other changes from the 2007 definition: the troponin MI threshold was previously 3 times the 99th percentile, now it is 5 times. Also, if the patient had an elevated baseline value, he or she can now still qualify as having an MI if the level increases by more than 20%.

In type 5 MI, changes to the definition similarly reflect the need to address overly sensitive troponin values when diagnosing an MI after coronary artery bypass grafting. To address such concerns, the required cardiac biomarker values were increased from more than 5 to more than 10 times the 99th percentile.

The task force raised the troponin thresholds for type 4 and type 5 MI in response to evidence showing that troponins are excessively sensitive to minimal myocardial damage during revascularization, and the lack of a troponin threshold that correlates with clinical outcomes.¹² Although higher, these values remain arbitrary, so physicians will need to exercise clinical judgment when deciding whether patients are experiencing benign myocardial injury or rather a true MI after revascularization procedures.

OTHER CONDITIONS THAT RAISE TROPONIN LEVELS

As troponin is a marker not only for MI but also for any form of cardiac injury, its levels are elevated in numerous conditions, such as heart failure, renal failure, and left ventricular hypertrophy. The task force identifies distinct troponin elevations above basal levels as the best indication of new pathology, yet several conditions other than acute coronary syndromes can also cause dynamic changes in troponin levels.

Troponin is a sensitive marker for ruling out MI and has tissue specificity for cardiac injury, but it is not specific for acute coronary syndrome as the cause of such injury. Troponin assays were tested and validated in patients in whom there was a high clinical suspicion of acute coronary syndrome, but when ordered indiscriminately, they have a poor positive predictive value (53%) for this disorder.¹⁸

Physicians must distinguish between acute coronary syndrome and other causes when deciding to give antithrombotics. Table 2 lists common causes of increased troponin other than acute coronary syndrome.

Heart failure

Some patients with acute congestive heart failure have elevated troponin levels. In one study, 6.2% of such patients had troponin I levels of 1 μg/L or higher or troponin T levels of 0.1 μg/L or higher, and these patients had poorer outcomes and more severe symptoms.¹⁹ Levels can also be elevated in patients with chronic heart failure, in whom they correlate with impaired hemodynamics, progressive ventricular dysfunction, and death.²⁰ In an overview of two large trials of patients with chronic congestive heart failure, 86% and 98% tested positive for cardiac troponin using high-sensitivity assays.²¹

Troponin levels can rise from baseline and subsequently fall in congestive heart failure due to small amounts of myocardial injury, which may be very difficult to distinguish from MI based on the similar presenting symptoms of dyspnea and chest pressure.^1,22 The increased troponin levels in chronic congestive heart failure may reflect apoptosis secondary to wall stretch or direct cell toxicity by neurohormones, alcohol, chemotherapy agents, or infiltrative disorders.^23–26

End-stage renal disease

Troponin levels are increased in end-stage renal disease, with 25% to 75% of patients having elevated levels using currently available assays.^27–29 With the advent of high-sensitivity assays, however, cardiac troponin T levels higher than the 99th percentile are found in 100% of patients who have end-stage renal disease without cardiac symptoms.³⁰

Troponin values above the 99th percentile are therefore not diagnostic of MI in this population. Rather, a diagnosis of MI in patients with end-stage renal disease requires clinical signs and symptoms and serial changes in troponin levels from baseline levels. The task force and the National Academy of Clinical Biochemistry recommend requiring an elevation of more than 20% from baseline, representing a change in troponin of more than 3 standard deviations.³¹

Increases in troponin in renal failure are thought to be the result of chronic cardiac structural changes such as coronary artery disease, left ventricular hypertrophy, and elevated left ventricular end-diastolic pressure, rather than decreased clearance.^32,33

In stable patients with end-stage renal disease, those who have high levels of cardiac troponin T have a higher mortality rate.³⁴ Although the mechanism is not completely clear, decreased clearance of uremic toxins may contribute to myocardial damage beyond that of the cardiac structural changes.³⁴

Sepsis

Approximately 50% of patients admitted to an intensive care unit with sepsis without acute coronary syndrome have elevated troponin levels.³⁵

Elevated troponin in sepsis patients has been associated with left ventricular dysfunction, most likely from hemodynamic stress, direct cytotoxicity of bacterial endotoxins, and reperfusion injury.^35,36 Critical illness places high demands on the myocardium, while oxygen supply may be diminished by hypotension, pulmonary edema, and intravascular volume depletion. This supply-demand mismatch is similar to the physiology of type 2 MI, with clinical signs and symptoms of MI potentially being the only differentiating factor.

Elevated troponin levels may represent either reversible or irreversible myocardial injury in patients with sepsis and are a predictor of severe illness and death.³⁷ However, what to do about elevated troponin in patients with sepsis is not clear. When patients are in the intensive care unit with single-organ or multi-organ failure, the diagnosis and treatment of troponin elevations may not take priority.¹ Diagnosing MI is further complicated by the inability of critically ill patients to communicate signs and symptoms. Physicians should also remember that diagnostic testing (electrocardiography, echocardiography) is often necessary to meet the clinical criteria for a type 1 or 2 MI in critically ill patients, and that treatment options may be limited.

Pulmonary embolism

Pulmonary embolism is a leading noncardiac cause of troponin elevation in patients in whom the clinical suspicion of acute coronary syndrome is initially high.³⁸ It is thought that increased troponin levels in patients with pulmonary embolism are caused by increased right ventricular strain secondary to increased pulmonary artery resistance.

The signs and symptoms of MI and of pulmonary embolism overlap, and troponin can be elevated in both conditions, making the initial diagnosis difficult. Electrocardiography and early bedside echocardiography can identify the predominant right-sided dilatation and strain in the heart secondary to pulmonary embolism. Computed tomography should be performed if there is even a moderate clinical suspicion of pulmonary embolism.

The appropriate use of thrombolytics in a normotensive patient with pulmonary embolism remains controversial. The significant risks of hemorrhage need to be balanced with the risk of hemodynamic deterioration. For these patients, the combination of cardiac troponin I measurement and echocardiography provides more prognostic information than each does individually.³⁹ Troponin elevation may therefore be a marker for poor outcomes without aggressive treatment with thrombolytics.

However, single troponin measurements in patients hospitalized early with pulmonary embolism can lead to substantial risk of misdiagnosing them with MI. Although the intensity of the peak is not particularly useful in the setting of pulmonary embolism, two consecutive troponin values 8 hours apart will allow for more appropriate risk stratification for pulmonary embolism patients, who may have a delay between right heart injury and troponin release.⁴⁰

It is reasonable to expect that myocarditis—inflammation of the myocardium—would cause release of troponin from myocytes.⁴¹ Interestingly, however, troponin levels can also be elevated in pericarditis.⁴² The reasons are not clear but have been hypothesized as being caused by nonspecific inflammation during pericarditis that also includes the superficial myocardium—hence, “myopericarditis.”

We have only limited data on the outcomes of patients who have pericarditis with troponin elevation, but troponin levels did correlate with an adverse prognosis in one study.⁴³

Arrhythmias

A number of arrhythmias have been associated with elevated troponin levels. Some studies have shown arrhythmias to be the most common cause of high troponin levels in patients who are not experiencing an acute coronary syndrome.^44,45

The reasons proposed for increased troponins in tachyarrhythmia are similar to those in other conditions of oxygen supply-demand mismatch.⁴⁶ Tachycardia alone may lead to troponin release in the absence of myodepressive factors, inflammatory mediators, or coronary artery disease.⁴⁶

Studies have provided only mixed data as to whether troponin levels predict newonset arrhythmias or recurrence of arrhythmias.^47,48 Nonetheless, elevated troponin (≥ 0.040 μg/L) in patients with atrial fibrillation has independently correlated with increased risk of stroke or systemic embolism, death, and other cardiovascular events. This is clinically important, as troponin elevations higher than these levels adds prognostic information to that given by the CHADS2 stroke score (congestive heart failure, hypertension, age ≥ 75 years diabetes mellitus, and prior stroke or transient ischemic attack) and thus can inform appropriate anticoagulation therapy.⁴⁹

USE OF TROPONIN VALUES

Troponins are highly sensitive assays with high tissue specificity for myocardial injury, but levels can be elevated in non-MI conditions and in MIs other than type 1. As with any diagnostic test applied to a population with a low prevalence of the disease, troponin elevation has a low positive predictive value—53% for acute coronary syndrome.¹⁸

Unfortunately, in clinical practice, troponins are measured in up to 50% of admitted patients, a small proportion of whom have clinical signs or symptoms of MI.⁵⁰ Often, clinicians are left with a positive troponin of unknown significance, potentially leading to unnecessary diagnostic testing that detracts from the primary diagnosis.

Dynamic changes in troponin values (eg, a change of more than 20% in a patient with end-stage renal disease) are helpful in distinguishing acute from chronic causes of troponin elevation. However, such changes can also occur with acute or chronic congestive heart failure, tachycardia, hypotension, or other conditions other than acute coronary syndrome.

The absolute numerical value of troponin can help assess the significance of troponin elevation. In most non-MI and non-acute coronary syndrome causes of troponin elevation, the troponin level tends to be lower than 1 μg/mL (Figure 1). Occasional exceptions occur, especially when multiple conditions coexist (end-stage renal disease and congestive heart failure, for example). In contrast, most patients with acute coronary syndromes have either clear symptoms or electrocardiographic changes consistent with MI and a troponin that rises above 0.5 μg/mL.

The task force discourages the use of secondary thresholds for MI, as there is no level of troponin that is considered benign. While any troponin elevation carries a negative prognosis, such prognostic knowledge may not be particularly helpful in deciding whether to anticoagulate patients or attempt revascularization procedures.

We thus recommend using a threshold higher than the 99th percentile to distinguish acute coronary syndromes from other causes of troponin elevations. The particular threshold for decision-making should vary, depending on how strongly one clinically suspects an acute coronary syndrome. For instance, a cardiac troponin I level of 0.2 μg/mL in an otherwise healthy patient with chest pain and ST-segment depression is more than sufficient to diagnose acute coronary syndrome. In contrast, an end-stage renal disease patient with hypertensive cardiomyopathy who presents only with nausea should have a level markedly higher than his or her baseline value (and likely > 0.8 μg/mL) before acute coronary syndrome should be diagnosed.

CK-MB’S ROLE IN THE TROPONIN ERA

Some proponents of troponin assays, including those on the task force, have suggested that CK-MB may no longer be necessary in the evaluation of acute MI.⁵¹ In the past, CK-MB had more research supporting its use in quantifying myocardial damage and in diagnosing reinfarction, but some data suggest that troponin may be equally useful for these applications.^52,53

These comments aside, CK-MB measurements are still widely ordered with troponin, a probable response to the clinical difficulty of determining the cause and significance of troponin elevations. Although likely less common with recent assays, a small subgroup of patients with acute coronary syndrome will be CK-MB–positive and troponin-negative and at higher risk of morbidity and death than those who are troponin- and CK-MB–negative.^54,55

Troponin levels are elevated in many chronic conditions, whereas CK-MB levels may be unaffected or less affected. In some cases, such as congestive heart failure or renal failure, troponins may be both chronically elevated and more than 20% higher than at baseline. In a clinical context in which a false-positive troponin assay is likely, the addition of a CK-MB assay may help determine if a rise (and possibly a subsequent fall) in the troponin level represents true MI. More importantly, deciding on antithrombotic therapy or revascularization is often based on whether a patient has acute coronary syndrome, rather than a small MI from demand ischemia. CK-MB may thus serve as a less sensitive but more specific marker for the larger amount of myocardial damage that one might expect from an acute coronary syndrome.

CK-MB testing also may help determine the acuity of an acute coronary syndrome for patients with known causes of increased troponin. A negative CK-MB value in the presence of a troponin value elevated above baseline could indicate an event a few days prior.

Finally, the approach of ordering both troponin and CK-MB may be particularly helpful in diagnosing type 4 and 5 MIs, as current guidelines suggest that more research is needed to determine whether current troponin thresholds lead to clinical outcomes.

CLINICAL JUDGMENT IS NECESSARY

The updated definition raises the biomarker threshold required to diagnose MI after revascularization procedures and reemphasizes the need to look for other signs of infarction. This change reflects the sometimes excessive sensitivity of troponin assays for minimal and often unavoidable myocardial damage that occurs in numerous conditions.

With sensitive troponin assays, clinical judgment is essential for separating true MI from myocardial injury, and acute coronary syndrome from demand ischemia. Clinicians will now be forced to be cognizant of their suspicion for acute coronary syndrome in the presence of multiple noncoronary causes of increased troponin with little practical guideline guidance. In settings in which troponin elevation is expected (eg, congestive heart failure, end-stage renal failure, shock), a higher cardiac troponin threshold or CK-MB may be useful as a less sensitive but more specific marker of significant myocardial damage requiring aggressive treatment.