Venous thromboembolism (VTE), which encompasses both deep vein thrombosis (DVT) and pulmonary embolism (PE), is a major cause of the morbidity and mortality of hospitalized medical patients.1 Hospitalization for an acute medical illness has been associated with an 8‐fold increase in the relative risk of VTE and is responsible for approximately a quarter of all VTE cases in the general population.2, 3

Current evidence‐based guidelines, including those from the American College of Chest Physicians (ACCP), recommend prophylaxis with low‐dose unfractionated heparin (UFH) or low‐molecular‐weight heparin (LMWH) for medical patients with risk factors for VTE.4, 5 Mechanical prophylaxis methods including graduated compression stockings and intermittent pneumatic compression are recommended for those patients for whom anticoagulant therapy is contraindicated because of a high risk of bleeding.4, 5 However, several studies have shown that adherence to these guidelines is suboptimal, with many at‐risk patients receiving inadequate prophylaxis (range 32%‐87%).610

Physician‐related factors identified as potential barriers to guideline adherence include not being aware or familiar with the guidelines, not agreeing with the guidelines, or believing the guideline recommendations to be ineffective.11 More specific studies have shown that some physicians may lack basic knowledge regarding the current treatment standards for VTE and may underestimate the significance of VTE.1213 As distinct strategies, education aimed at disseminating VTE prophylaxis guidelines, as well as regular audit‐and‐feedback of physician performance, has been shown to improve rates of VTE prophylaxis in clinical practice.6, 1417 Implementation of educational programs significantly increased the level of appropriate VTE prophylaxis from 59% to 70% of patients in an Australian hospital15 and from 73% to 97% of patients in a Scottish hospital.14 Another strategy, the use of point‐of‐care electronically provided reminders with decision support, has been successful not only in increasing the rates of VTE prophylaxis, but also in decreasing the incidence of clinical VTE events.16 Although highly effective, electronic alerts with computerized decision support do not exist in many hospitals, and other methods of intervention are needed.

In this study, we evaluated adherence to the 2001 ACCP guidelines for VTE prophylaxis among medical patients in our teaching hospital. (The guidelines were updated in 2004, after our study was completed.) After determining that our baseline rates of appropriate VTE prophylaxis were suboptimal, we developed, implemented, and evaluated a multifaceted strategy to improve the rates of appropriate thromboprophylaxis among our medical inpatients.

Six categories of quality improvement strategies have been described: provider education, decision support, audit‐and‐feedback, patient education, organization change, and regulation and policy.18 The intervention we developed was a composite of 3 of these: provider education, decision support, and audit‐and‐feedback.

METHODS

RESULTS

Risk Factors for VTE

Patient risk factors for VTE in each data collection period are summarized in Table 4. Analysis of this data showed that the most prevalent risk factors for VTE in the 3 patient populations were age older than 40 years (262/312, 84.0% of the total patient population) and nonambulatory state (231/271, 85.2% of the total population). Overall, the average number of risk factors for VTE was approximately 3, with more than 60% of patients having 3 or more VTE risk factors (Fig. 1).

Figure 1

Change in Prophylaxis Use

Twelve and 18 months after implementation of the quality improvement program, we observed an increase in the use of any prophylaxis, from 46.9% at baseline to 86.2% and 86.4%, respectively (Table 5; P < .01 in both groups versus baseline). This increase was a result almost entirely of an increase in the proportion of patients receiving pharmacological prophylaxis, which significantly increased, from 44.9% to 81.0% and 80.3%, at the 12‐ and 18‐month time points, respectively (Table 5; P < .01 for both groups versus baseline). Most meaningfully, there was a significant increase in the proportion of patients for whom an appropriate prophylaxis decision was made (from 42.9% to 68.1% and 85.0%, at the 12‐ and 18‐month time points, respectively; Table 6; P < .01 for both groups versus baseline). This represented a trend toward continuing increases in the use of appropriate prophylaxis as the study progressed (Fig. 2). This change was driven mainly by a significant increase in the prescribing of LMWH, almost all of which was prescribed in accordance with the 2001 ACCP guidelines (Table 6).

Figure 2

DISCUSSION

In this study we evaluated the effect of an intervention that combined physician education with a decision support tool and a mechanism for audit‐and‐feedback. We have shown that implementation of such a multifaceted intervention is practical in a teaching hospital and can improve the rates of VTE prophylaxis use in medical patients. In nearly doubling the rate of appropriate prophylaxis, the effect size of our intervention was large, statistically significant, and sustained 18 months after implementation.

More than 60% of our patients had 3 or more risk factors, and more than 80% had at least 2 risk factors. The rate we observed for patients with 3 or more risk factors was 3 times higher than that reported previously.22 Despite the prevalence of high‐risk patients in our study, we observed that the preintervention rate of VTE prophylaxis among medical patients was relatively low at 47%, and only 43% of patients received prophylaxis in accordance with the ACCP guidelines. Our study findings are consistent with those of several other studies that have shown low rates of VTE prophylaxis in medical patients.6, 8, 2324 In a study of 15 hospitals in Massachusetts, only 13%‐19% of medical patients with indications and risk factors for VTE prophylaxis received any prophylaxis prior to an educational intervention.6 Similarly, a study of 368 consecutive medical patients at a Swiss hospital showed that only 22% of those at‐risk received VTE prophylaxis in accordance with the Thromboembolic Risk Factors (THRIFT) I Consensus Group recommendations.8 Results from 2 prospective patient registries also indicated low rates of VTE prophylaxis in medical patients.19, 24 In the IMPROVE registry of acutely ill medical patients, only 39% of patients hospitalized for 3 or more days received VTE prophylaxis19 and in the DVT‐FREE registry only 42% of medical patients with the inpatient diagnosis of DVT had received prophylaxis within 30 days of that diagnosis.24 In a recent retrospective study of 217 medical patients at the University of Utah hospital, just 43% of patients at high risk for VTE received any sort of prophylaxis.23

Physician education was the main intervention in several previous studies aimed at raising rates of VTE prophylaxis. Our study joins those that have also shown significant improvements after implementation of VTE prophylaxis educational initiatives.6, 14, 15, 23 In the study by Anderson et al., a significantly greater increase in the proportion of high‐risk patients receiving effective VTE prophylaxis was seen between 1986 and 1989 in hospitals that participated in a formal continuing medical education program compared with those that did not (increase: 28% versus 11%; P < .001).6 In 3 additional studies, educational interventions were shown to increase the rate of appropriate prophylaxis in at‐risk patients from 59% to 70%, from 55% to 96%, and from 43% to 72%.14, 15, 23

Other studies have cast doubt on the ability of time‐limited educational interventions to achieve a large or sustained effect.27, 28 A recent systematic review of strategies to improve the use of prophylaxis in hospitals concluded that a number of active strategies are likely to achieve optimal outcomes by combining a system for reminding clinicians to assess patients for VTE with assisting the selection of prophylaxis and providing audit‐and‐feedback.29 The large, sustained effect reported in our study might have been a result of the multifaceted and ongoing nature of the intervention, with reintroduction of the material to all incoming house staff each month. An audit from the last quarter of 2005nearly 2 years after the start of our interventionshowed that prophylaxis rates were approaching 100% (data not included in this study).

Another strategy, the use of computerized reminders to physicians, has been shown to increase the rate of VTE prophylaxis in surgical and medical/surgical patients.16, 26 Kucher et al. compared the incidence of DVT or PE in 1255 hospitalized patients whose physicians received an electronic alert of patient risk of DVT with 1251 hospitalized patients whose physicians did not receive such an alert. They found that the computer alert was associated with a significant reduction in the incidence of DVT or PE at 90 days, with a hazard ratio of 0.59 (95% confidence interval: 0.43, 0.81).16 Our study offers one practical alternative for those institutions that, like ours, do not currently have computerized order entry.

We were unable to determine if there was a specific element of the multifaceted VTE prophylaxis intervention program that contributed the most to the improvement in prophylaxis rates. Provider education was ongoing rather than just a single educational campaign. It was further supported by the pocket cards that provided support for decision making on VTE risk factors, risk categories (based on number and type of risk factor), recommended prophylaxis choices, and potential contraindications. In addition, our method of audit‐and‐feedback constructively leveraged the Hawthorne effect: aware that individual behavior was being measured, our physicians likely adjusted their practice accordingly. Taken together, it is likely that the several elements of our intervention were more powerful in combination than they would have been alone.

Although the multifaceted intervention worked well within our urban university teaching hospital, its application and outcome might be different for other types of hospitals. In our audit‐and‐feedback, for instance, review of resident physician performance was conducted by the Division Chief of General Internal Medicine, tapping into a very strong authority gradient. Hierarchical structures are likely to be different in other types of hospitals. It would therefore be valuable to examine whether the audit‐and‐feedback methodology presented in this article can be replicated in other hospital settings.

A potential limitation of this study was the use of retrospective review to determine baseline rates of VTE prophylaxis. This approach relies on medical notes being accurate and complete; such notes may not have been available for each patient. However, random reviews of both patient charts and hospital billing data for comorbidities performed after coding as a quality control step allowed for confirmation of the data or the extraction and addition of missing data. In addition, data collection was limited to a single day in the latter half of the month. It is not clear whether this sampling strategy collects data that are reflective of performance for the entire month. Our study was also limited by the absence of a control group. Without a control group, we cannot exclude the possibility that during the study factors other than the educational intervention might have contributed to the improvement in prophylaxis rates.

In this study we did not address whether an increase in VTE prophylaxis use translates to an improvement in patient outcomes, namely, a reduction in the rate of VTE. Mosen et al. showed that increasing the VTE prophylaxis rate by implementing a computerized reminder system did not decrease the rate of VTE.26 However, the baseline rate of VTE prophylaxis was already very good, and the study was only powered to detect a large difference in VTE rates. Conversely, Kucher et al. recently demonstrated a significant reduction in VTE events 90 days after initiation of a computerized alert program.16 Further studies designed to confirm the inverse relationship between rate of VTE prophylaxis and rate of clinical outcome of VTE would be helpful.

In conclusion, in a setting in which most hospitalized medically ill patients have multiple risk factors for VTE, we have shown that a practical multifaceted intervention can result in a marked increase in the proportion of medical patients receiving VTE prophylaxis, as well as in the proportion of patients receiving prophylaxis commensurate with evidence‐based guidelines.

Venous thromboembolism (VTE), which encompasses both deep vein thrombosis (DVT) and pulmonary embolism (PE), is a major cause of the morbidity and mortality of hospitalized medical patients.1 Hospitalization for an acute medical illness has been associated with an 8‐fold increase in the relative risk of VTE and is responsible for approximately a quarter of all VTE cases in the general population.2, 3

Current evidence‐based guidelines, including those from the American College of Chest Physicians (ACCP), recommend prophylaxis with low‐dose unfractionated heparin (UFH) or low‐molecular‐weight heparin (LMWH) for medical patients with risk factors for VTE.4, 5 Mechanical prophylaxis methods including graduated compression stockings and intermittent pneumatic compression are recommended for those patients for whom anticoagulant therapy is contraindicated because of a high risk of bleeding.4, 5 However, several studies have shown that adherence to these guidelines is suboptimal, with many at‐risk patients receiving inadequate prophylaxis (range 32%‐87%).610

Physician‐related factors identified as potential barriers to guideline adherence include not being aware or familiar with the guidelines, not agreeing with the guidelines, or believing the guideline recommendations to be ineffective.11 More specific studies have shown that some physicians may lack basic knowledge regarding the current treatment standards for VTE and may underestimate the significance of VTE.1213 As distinct strategies, education aimed at disseminating VTE prophylaxis guidelines, as well as regular audit‐and‐feedback of physician performance, has been shown to improve rates of VTE prophylaxis in clinical practice.6, 1417 Implementation of educational programs significantly increased the level of appropriate VTE prophylaxis from 59% to 70% of patients in an Australian hospital15 and from 73% to 97% of patients in a Scottish hospital.14 Another strategy, the use of point‐of‐care electronically provided reminders with decision support, has been successful not only in increasing the rates of VTE prophylaxis, but also in decreasing the incidence of clinical VTE events.16 Although highly effective, electronic alerts with computerized decision support do not exist in many hospitals, and other methods of intervention are needed.

In this study, we evaluated adherence to the 2001 ACCP guidelines for VTE prophylaxis among medical patients in our teaching hospital. (The guidelines were updated in 2004, after our study was completed.) After determining that our baseline rates of appropriate VTE prophylaxis were suboptimal, we developed, implemented, and evaluated a multifaceted strategy to improve the rates of appropriate thromboprophylaxis among our medical inpatients.

Six categories of quality improvement strategies have been described: provider education, decision support, audit‐and‐feedback, patient education, organization change, and regulation and policy.18 The intervention we developed was a composite of 3 of these: provider education, decision support, and audit‐and‐feedback.

METHODS

RESULTS

Risk Factors for VTE

Patient risk factors for VTE in each data collection period are summarized in Table 4. Analysis of this data showed that the most prevalent risk factors for VTE in the 3 patient populations were age older than 40 years (262/312, 84.0% of the total patient population) and nonambulatory state (231/271, 85.2% of the total population). Overall, the average number of risk factors for VTE was approximately 3, with more than 60% of patients having 3 or more VTE risk factors (Fig. 1).

Figure 1

Change in Prophylaxis Use

Twelve and 18 months after implementation of the quality improvement program, we observed an increase in the use of any prophylaxis, from 46.9% at baseline to 86.2% and 86.4%, respectively (Table 5; P < .01 in both groups versus baseline). This increase was a result almost entirely of an increase in the proportion of patients receiving pharmacological prophylaxis, which significantly increased, from 44.9% to 81.0% and 80.3%, at the 12‐ and 18‐month time points, respectively (Table 5; P < .01 for both groups versus baseline). Most meaningfully, there was a significant increase in the proportion of patients for whom an appropriate prophylaxis decision was made (from 42.9% to 68.1% and 85.0%, at the 12‐ and 18‐month time points, respectively; Table 6; P < .01 for both groups versus baseline). This represented a trend toward continuing increases in the use of appropriate prophylaxis as the study progressed (Fig. 2). This change was driven mainly by a significant increase in the prescribing of LMWH, almost all of which was prescribed in accordance with the 2001 ACCP guidelines (Table 6).

Figure 2

DISCUSSION

In this study we evaluated the effect of an intervention that combined physician education with a decision support tool and a mechanism for audit‐and‐feedback. We have shown that implementation of such a multifaceted intervention is practical in a teaching hospital and can improve the rates of VTE prophylaxis use in medical patients. In nearly doubling the rate of appropriate prophylaxis, the effect size of our intervention was large, statistically significant, and sustained 18 months after implementation.

More than 60% of our patients had 3 or more risk factors, and more than 80% had at least 2 risk factors. The rate we observed for patients with 3 or more risk factors was 3 times higher than that reported previously.22 Despite the prevalence of high‐risk patients in our study, we observed that the preintervention rate of VTE prophylaxis among medical patients was relatively low at 47%, and only 43% of patients received prophylaxis in accordance with the ACCP guidelines. Our study findings are consistent with those of several other studies that have shown low rates of VTE prophylaxis in medical patients.6, 8, 2324 In a study of 15 hospitals in Massachusetts, only 13%‐19% of medical patients with indications and risk factors for VTE prophylaxis received any prophylaxis prior to an educational intervention.6 Similarly, a study of 368 consecutive medical patients at a Swiss hospital showed that only 22% of those at‐risk received VTE prophylaxis in accordance with the Thromboembolic Risk Factors (THRIFT) I Consensus Group recommendations.8 Results from 2 prospective patient registries also indicated low rates of VTE prophylaxis in medical patients.19, 24 In the IMPROVE registry of acutely ill medical patients, only 39% of patients hospitalized for 3 or more days received VTE prophylaxis19 and in the DVT‐FREE registry only 42% of medical patients with the inpatient diagnosis of DVT had received prophylaxis within 30 days of that diagnosis.24 In a recent retrospective study of 217 medical patients at the University of Utah hospital, just 43% of patients at high risk for VTE received any sort of prophylaxis.23

Physician education was the main intervention in several previous studies aimed at raising rates of VTE prophylaxis. Our study joins those that have also shown significant improvements after implementation of VTE prophylaxis educational initiatives.6, 14, 15, 23 In the study by Anderson et al., a significantly greater increase in the proportion of high‐risk patients receiving effective VTE prophylaxis was seen between 1986 and 1989 in hospitals that participated in a formal continuing medical education program compared with those that did not (increase: 28% versus 11%; P < .001).6 In 3 additional studies, educational interventions were shown to increase the rate of appropriate prophylaxis in at‐risk patients from 59% to 70%, from 55% to 96%, and from 43% to 72%.14, 15, 23

Other studies have cast doubt on the ability of time‐limited educational interventions to achieve a large or sustained effect.27, 28 A recent systematic review of strategies to improve the use of prophylaxis in hospitals concluded that a number of active strategies are likely to achieve optimal outcomes by combining a system for reminding clinicians to assess patients for VTE with assisting the selection of prophylaxis and providing audit‐and‐feedback.29 The large, sustained effect reported in our study might have been a result of the multifaceted and ongoing nature of the intervention, with reintroduction of the material to all incoming house staff each month. An audit from the last quarter of 2005nearly 2 years after the start of our interventionshowed that prophylaxis rates were approaching 100% (data not included in this study).

Another strategy, the use of computerized reminders to physicians, has been shown to increase the rate of VTE prophylaxis in surgical and medical/surgical patients.16, 26 Kucher et al. compared the incidence of DVT or PE in 1255 hospitalized patients whose physicians received an electronic alert of patient risk of DVT with 1251 hospitalized patients whose physicians did not receive such an alert. They found that the computer alert was associated with a significant reduction in the incidence of DVT or PE at 90 days, with a hazard ratio of 0.59 (95% confidence interval: 0.43, 0.81).16 Our study offers one practical alternative for those institutions that, like ours, do not currently have computerized order entry.

We were unable to determine if there was a specific element of the multifaceted VTE prophylaxis intervention program that contributed the most to the improvement in prophylaxis rates. Provider education was ongoing rather than just a single educational campaign. It was further supported by the pocket cards that provided support for decision making on VTE risk factors, risk categories (based on number and type of risk factor), recommended prophylaxis choices, and potential contraindications. In addition, our method of audit‐and‐feedback constructively leveraged the Hawthorne effect: aware that individual behavior was being measured, our physicians likely adjusted their practice accordingly. Taken together, it is likely that the several elements of our intervention were more powerful in combination than they would have been alone.

Although the multifaceted intervention worked well within our urban university teaching hospital, its application and outcome might be different for other types of hospitals. In our audit‐and‐feedback, for instance, review of resident physician performance was conducted by the Division Chief of General Internal Medicine, tapping into a very strong authority gradient. Hierarchical structures are likely to be different in other types of hospitals. It would therefore be valuable to examine whether the audit‐and‐feedback methodology presented in this article can be replicated in other hospital settings.

A potential limitation of this study was the use of retrospective review to determine baseline rates of VTE prophylaxis. This approach relies on medical notes being accurate and complete; such notes may not have been available for each patient. However, random reviews of both patient charts and hospital billing data for comorbidities performed after coding as a quality control step allowed for confirmation of the data or the extraction and addition of missing data. In addition, data collection was limited to a single day in the latter half of the month. It is not clear whether this sampling strategy collects data that are reflective of performance for the entire month. Our study was also limited by the absence of a control group. Without a control group, we cannot exclude the possibility that during the study factors other than the educational intervention might have contributed to the improvement in prophylaxis rates.

In this study we did not address whether an increase in VTE prophylaxis use translates to an improvement in patient outcomes, namely, a reduction in the rate of VTE. Mosen et al. showed that increasing the VTE prophylaxis rate by implementing a computerized reminder system did not decrease the rate of VTE.26 However, the baseline rate of VTE prophylaxis was already very good, and the study was only powered to detect a large difference in VTE rates. Conversely, Kucher et al. recently demonstrated a significant reduction in VTE events 90 days after initiation of a computerized alert program.16 Further studies designed to confirm the inverse relationship between rate of VTE prophylaxis and rate of clinical outcome of VTE would be helpful.

In conclusion, in a setting in which most hospitalized medically ill patients have multiple risk factors for VTE, we have shown that a practical multifaceted intervention can result in a marked increase in the proportion of medical patients receiving VTE prophylaxis, as well as in the proportion of patients receiving prophylaxis commensurate with evidence‐based guidelines.

Venous thromboembolism (VTE), which encompasses both deep vein thrombosis (DVT) and pulmonary embolism (PE), is a major cause of the morbidity and mortality of hospitalized medical patients.1 Hospitalization for an acute medical illness has been associated with an 8‐fold increase in the relative risk of VTE and is responsible for approximately a quarter of all VTE cases in the general population.2, 3

Current evidence‐based guidelines, including those from the American College of Chest Physicians (ACCP), recommend prophylaxis with low‐dose unfractionated heparin (UFH) or low‐molecular‐weight heparin (LMWH) for medical patients with risk factors for VTE.4, 5 Mechanical prophylaxis methods including graduated compression stockings and intermittent pneumatic compression are recommended for those patients for whom anticoagulant therapy is contraindicated because of a high risk of bleeding.4, 5 However, several studies have shown that adherence to these guidelines is suboptimal, with many at‐risk patients receiving inadequate prophylaxis (range 32%‐87%).610

Physician‐related factors identified as potential barriers to guideline adherence include not being aware or familiar with the guidelines, not agreeing with the guidelines, or believing the guideline recommendations to be ineffective.11 More specific studies have shown that some physicians may lack basic knowledge regarding the current treatment standards for VTE and may underestimate the significance of VTE.1213 As distinct strategies, education aimed at disseminating VTE prophylaxis guidelines, as well as regular audit‐and‐feedback of physician performance, has been shown to improve rates of VTE prophylaxis in clinical practice.6, 1417 Implementation of educational programs significantly increased the level of appropriate VTE prophylaxis from 59% to 70% of patients in an Australian hospital15 and from 73% to 97% of patients in a Scottish hospital.14 Another strategy, the use of point‐of‐care electronically provided reminders with decision support, has been successful not only in increasing the rates of VTE prophylaxis, but also in decreasing the incidence of clinical VTE events.16 Although highly effective, electronic alerts with computerized decision support do not exist in many hospitals, and other methods of intervention are needed.

In this study, we evaluated adherence to the 2001 ACCP guidelines for VTE prophylaxis among medical patients in our teaching hospital. (The guidelines were updated in 2004, after our study was completed.) After determining that our baseline rates of appropriate VTE prophylaxis were suboptimal, we developed, implemented, and evaluated a multifaceted strategy to improve the rates of appropriate thromboprophylaxis among our medical inpatients.

Six categories of quality improvement strategies have been described: provider education, decision support, audit‐and‐feedback, patient education, organization change, and regulation and policy.18 The intervention we developed was a composite of 3 of these: provider education, decision support, and audit‐and‐feedback.

METHODS

RESULTS

Risk Factors for VTE

Patient risk factors for VTE in each data collection period are summarized in Table 4. Analysis of this data showed that the most prevalent risk factors for VTE in the 3 patient populations were age older than 40 years (262/312, 84.0% of the total patient population) and nonambulatory state (231/271, 85.2% of the total population). Overall, the average number of risk factors for VTE was approximately 3, with more than 60% of patients having 3 or more VTE risk factors (Fig. 1).

Figure 1

Change in Prophylaxis Use

Twelve and 18 months after implementation of the quality improvement program, we observed an increase in the use of any prophylaxis, from 46.9% at baseline to 86.2% and 86.4%, respectively (Table 5; P < .01 in both groups versus baseline). This increase was a result almost entirely of an increase in the proportion of patients receiving pharmacological prophylaxis, which significantly increased, from 44.9% to 81.0% and 80.3%, at the 12‐ and 18‐month time points, respectively (Table 5; P < .01 for both groups versus baseline). Most meaningfully, there was a significant increase in the proportion of patients for whom an appropriate prophylaxis decision was made (from 42.9% to 68.1% and 85.0%, at the 12‐ and 18‐month time points, respectively; Table 6; P < .01 for both groups versus baseline). This represented a trend toward continuing increases in the use of appropriate prophylaxis as the study progressed (Fig. 2). This change was driven mainly by a significant increase in the prescribing of LMWH, almost all of which was prescribed in accordance with the 2001 ACCP guidelines (Table 6).

Figure 2

DISCUSSION

In this study we evaluated the effect of an intervention that combined physician education with a decision support tool and a mechanism for audit‐and‐feedback. We have shown that implementation of such a multifaceted intervention is practical in a teaching hospital and can improve the rates of VTE prophylaxis use in medical patients. In nearly doubling the rate of appropriate prophylaxis, the effect size of our intervention was large, statistically significant, and sustained 18 months after implementation.

More than 60% of our patients had 3 or more risk factors, and more than 80% had at least 2 risk factors. The rate we observed for patients with 3 or more risk factors was 3 times higher than that reported previously.22 Despite the prevalence of high‐risk patients in our study, we observed that the preintervention rate of VTE prophylaxis among medical patients was relatively low at 47%, and only 43% of patients received prophylaxis in accordance with the ACCP guidelines. Our study findings are consistent with those of several other studies that have shown low rates of VTE prophylaxis in medical patients.6, 8, 2324 In a study of 15 hospitals in Massachusetts, only 13%‐19% of medical patients with indications and risk factors for VTE prophylaxis received any prophylaxis prior to an educational intervention.6 Similarly, a study of 368 consecutive medical patients at a Swiss hospital showed that only 22% of those at‐risk received VTE prophylaxis in accordance with the Thromboembolic Risk Factors (THRIFT) I Consensus Group recommendations.8 Results from 2 prospective patient registries also indicated low rates of VTE prophylaxis in medical patients.19, 24 In the IMPROVE registry of acutely ill medical patients, only 39% of patients hospitalized for 3 or more days received VTE prophylaxis19 and in the DVT‐FREE registry only 42% of medical patients with the inpatient diagnosis of DVT had received prophylaxis within 30 days of that diagnosis.24 In a recent retrospective study of 217 medical patients at the University of Utah hospital, just 43% of patients at high risk for VTE received any sort of prophylaxis.23

Physician education was the main intervention in several previous studies aimed at raising rates of VTE prophylaxis. Our study joins those that have also shown significant improvements after implementation of VTE prophylaxis educational initiatives.6, 14, 15, 23 In the study by Anderson et al., a significantly greater increase in the proportion of high‐risk patients receiving effective VTE prophylaxis was seen between 1986 and 1989 in hospitals that participated in a formal continuing medical education program compared with those that did not (increase: 28% versus 11%; P < .001).6 In 3 additional studies, educational interventions were shown to increase the rate of appropriate prophylaxis in at‐risk patients from 59% to 70%, from 55% to 96%, and from 43% to 72%.14, 15, 23

Other studies have cast doubt on the ability of time‐limited educational interventions to achieve a large or sustained effect.27, 28 A recent systematic review of strategies to improve the use of prophylaxis in hospitals concluded that a number of active strategies are likely to achieve optimal outcomes by combining a system for reminding clinicians to assess patients for VTE with assisting the selection of prophylaxis and providing audit‐and‐feedback.29 The large, sustained effect reported in our study might have been a result of the multifaceted and ongoing nature of the intervention, with reintroduction of the material to all incoming house staff each month. An audit from the last quarter of 2005nearly 2 years after the start of our interventionshowed that prophylaxis rates were approaching 100% (data not included in this study).

Another strategy, the use of computerized reminders to physicians, has been shown to increase the rate of VTE prophylaxis in surgical and medical/surgical patients.16, 26 Kucher et al. compared the incidence of DVT or PE in 1255 hospitalized patients whose physicians received an electronic alert of patient risk of DVT with 1251 hospitalized patients whose physicians did not receive such an alert. They found that the computer alert was associated with a significant reduction in the incidence of DVT or PE at 90 days, with a hazard ratio of 0.59 (95% confidence interval: 0.43, 0.81).16 Our study offers one practical alternative for those institutions that, like ours, do not currently have computerized order entry.

We were unable to determine if there was a specific element of the multifaceted VTE prophylaxis intervention program that contributed the most to the improvement in prophylaxis rates. Provider education was ongoing rather than just a single educational campaign. It was further supported by the pocket cards that provided support for decision making on VTE risk factors, risk categories (based on number and type of risk factor), recommended prophylaxis choices, and potential contraindications. In addition, our method of audit‐and‐feedback constructively leveraged the Hawthorne effect: aware that individual behavior was being measured, our physicians likely adjusted their practice accordingly. Taken together, it is likely that the several elements of our intervention were more powerful in combination than they would have been alone.

Although the multifaceted intervention worked well within our urban university teaching hospital, its application and outcome might be different for other types of hospitals. In our audit‐and‐feedback, for instance, review of resident physician performance was conducted by the Division Chief of General Internal Medicine, tapping into a very strong authority gradient. Hierarchical structures are likely to be different in other types of hospitals. It would therefore be valuable to examine whether the audit‐and‐feedback methodology presented in this article can be replicated in other hospital settings.

A potential limitation of this study was the use of retrospective review to determine baseline rates of VTE prophylaxis. This approach relies on medical notes being accurate and complete; such notes may not have been available for each patient. However, random reviews of both patient charts and hospital billing data for comorbidities performed after coding as a quality control step allowed for confirmation of the data or the extraction and addition of missing data. In addition, data collection was limited to a single day in the latter half of the month. It is not clear whether this sampling strategy collects data that are reflective of performance for the entire month. Our study was also limited by the absence of a control group. Without a control group, we cannot exclude the possibility that during the study factors other than the educational intervention might have contributed to the improvement in prophylaxis rates.

In this study we did not address whether an increase in VTE prophylaxis use translates to an improvement in patient outcomes, namely, a reduction in the rate of VTE. Mosen et al. showed that increasing the VTE prophylaxis rate by implementing a computerized reminder system did not decrease the rate of VTE.26 However, the baseline rate of VTE prophylaxis was already very good, and the study was only powered to detect a large difference in VTE rates. Conversely, Kucher et al. recently demonstrated a significant reduction in VTE events 90 days after initiation of a computerized alert program.16 Further studies designed to confirm the inverse relationship between rate of VTE prophylaxis and rate of clinical outcome of VTE would be helpful.

In conclusion, in a setting in which most hospitalized medically ill patients have multiple risk factors for VTE, we have shown that a practical multifaceted intervention can result in a marked increase in the proportion of medical patients receiving VTE prophylaxis, as well as in the proportion of patients receiving prophylaxis commensurate with evidence‐based guidelines.

A 53‐year‐old Korean woman was admitted to the hospital with a diagnosis of cellulitis (thin arrow) and rule out vasculitis. Further history obtained with the assistance of a Korean translator revealed that the patient, though untrained in Chinese medicine, had attempted scarring direct moxibustion for intermittent headaches. She was treated with intravenous antibiotics for 24 hours for her cellulitis and discharged in good condition on oral antibiotics.

Moxibustion is a traditional Chinese medical technique that involves burning the herb mugwort (Artemesia vulgaris) to relieve cold or stagnant conditions by stimulating circulation. Moxibustion can be performed indirectly or directly. Indirect moxibustion involves application of the burning moxa to the end of an acupuncture needle or by holding the moxa close to the skin. In direct moxibustion, a cone‐shaped moxa is held over an acupuncture point. Direct moxibustion can be divided into scarring and nonscarring types. With nonscarring direct moxibustion, moxa is placed on top of an acupuncture point, lit, and then removed before it burns the skin. With scarring moxibustion, the burning moxa is left on the skin until it burns out, leading to burns and scarring.

This case demonstrates the importance of obtaining an accurate history when making a clinical diagnosis and, in patients who are not fluent in English, the critical role that translators serve in the management of patients. The differential diagnosis of skin ulcers encompasses many other conditions in addition to infection, including iatrogenic causes of traditional as well as alternative medical therapies. 0

Figure 1

0

Figure 2

A 53‐year‐old Korean woman was admitted to the hospital with a diagnosis of cellulitis (thin arrow) and rule out vasculitis. Further history obtained with the assistance of a Korean translator revealed that the patient, though untrained in Chinese medicine, had attempted scarring direct moxibustion for intermittent headaches. She was treated with intravenous antibiotics for 24 hours for her cellulitis and discharged in good condition on oral antibiotics.

Moxibustion is a traditional Chinese medical technique that involves burning the herb mugwort (Artemesia vulgaris) to relieve cold or stagnant conditions by stimulating circulation. Moxibustion can be performed indirectly or directly. Indirect moxibustion involves application of the burning moxa to the end of an acupuncture needle or by holding the moxa close to the skin. In direct moxibustion, a cone‐shaped moxa is held over an acupuncture point. Direct moxibustion can be divided into scarring and nonscarring types. With nonscarring direct moxibustion, moxa is placed on top of an acupuncture point, lit, and then removed before it burns the skin. With scarring moxibustion, the burning moxa is left on the skin until it burns out, leading to burns and scarring.

This case demonstrates the importance of obtaining an accurate history when making a clinical diagnosis and, in patients who are not fluent in English, the critical role that translators serve in the management of patients. The differential diagnosis of skin ulcers encompasses many other conditions in addition to infection, including iatrogenic causes of traditional as well as alternative medical therapies. 0

Figure 1

0

Figure 2

A 53‐year‐old Korean woman was admitted to the hospital with a diagnosis of cellulitis (thin arrow) and rule out vasculitis. Further history obtained with the assistance of a Korean translator revealed that the patient, though untrained in Chinese medicine, had attempted scarring direct moxibustion for intermittent headaches. She was treated with intravenous antibiotics for 24 hours for her cellulitis and discharged in good condition on oral antibiotics.

Moxibustion is a traditional Chinese medical technique that involves burning the herb mugwort (Artemesia vulgaris) to relieve cold or stagnant conditions by stimulating circulation. Moxibustion can be performed indirectly or directly. Indirect moxibustion involves application of the burning moxa to the end of an acupuncture needle or by holding the moxa close to the skin. In direct moxibustion, a cone‐shaped moxa is held over an acupuncture point. Direct moxibustion can be divided into scarring and nonscarring types. With nonscarring direct moxibustion, moxa is placed on top of an acupuncture point, lit, and then removed before it burns the skin. With scarring moxibustion, the burning moxa is left on the skin until it burns out, leading to burns and scarring.

This case demonstrates the importance of obtaining an accurate history when making a clinical diagnosis and, in patients who are not fluent in English, the critical role that translators serve in the management of patients. The differential diagnosis of skin ulcers encompasses many other conditions in addition to infection, including iatrogenic causes of traditional as well as alternative medical therapies. 0

Figure 1

0

Figure 2

A 20‐year‐old woman presented to the emergency department after 2 days of epistaxis and vaginal bleeding.

A young woman is more likely to present with infection, toxic exposure, or rheumatologic disease than with a degenerative disease or malignancy. Her bleeding may relate to a platelet abnormality, either quantitative or qualitative. I would pursue her bleeding and menstrual history further.

The patient was healthy until 2 months previously, when she noted arthralgia of her shoulders, wrists, elbows, knees, and ankles. She was examined by a rheumatologist who detected mild arthritis in her left wrist and proximal interphalangeal joints. The rest of her joints were normal. Rheumatoid factor and ANA were positive, and the erythrocyte sedimentation rate was 122 mm/hour. She was diagnosed with possible systemic lupus erythematosus and was placed on a nonsteroidal anti‐inflammatory agent. At a follow‐up visit 1 month prior to admission, her arthralgia had markedly improved. Two weeks prior to admission, the patient began to feel fatigued. Two days prior to admission, she developed epistaxis and what she thought was her menses, though bleeding was heavier than usual and associated with the passage of red clots. On the day of admission the vaginal bleeding worsened, and emergency personnel transported the patient to the hospital.

The diagnosis of systemic lupus erythematosus (SLE) is not engraved in stone. One must be vigilant for other diseases masquerading as SLE while continuing to build a case for it. As more criteria are fulfilled, the probability of lupus increases, yet no findings, alone or in combination, are pathognomonic of this protean disease. This patient's age, sex, and serology are compatible with SLE; otherwise, her presentation is nonspecific. I would request a complete blood count, coagulation tests, and additional serological tests.

The quantity of the bleeding is described, but this does not help decipher its etiology. Excess bleeding may be a result of one or more of 3 broad etiologies: problems with platelets (quantitative or qualitative), with clotting factors (quantitative or qualitative), or with blood vessels (trauma, vasculitis, or diseases affecting collagen). Because quantitative and qualitative factor disorders generally do not present with mucosal bleeding, I am thinking more about platelet problems and about processes that damage the microvasculature. If this woman has lupus, immunologic thrombocytopenia may be the cause of mucosal bleeding.

The patient had no previous medical problems and had never been pregnant. Her only medication was sulindac twice daily for the past month. She was born in Hong Kong, graduated from high school in San Francisco, and attended junior college. She lived with her parents and brother and denied alcohol, tobacco, or recreational drug use but had recently obtained a tattoo on her lower back. There was no family history of autoimmune or bleeding disorders, and a review of systems was notable for dyspnea with minimal exertion and fatigue which worsened in the past 2 days. She had no prior episodes of abnormal bleeding or clotting.

Tattoos may be surrogates for other high‐risk behaviors and suggest an increased risk of hepatitis and sexually transmitted diseases. I want to know her sexual history and other risk factors for human immunodeficiency virus infection. The dyspnea and fatigue are likely the result of anemia, but I am also considering cardiac disease. Though SLE remains a possibility, I cannot assume the presence of a lupus anticoagulant with antiphospholipid syndrome without a history of infertility or recurrent miscarriages.

On arrival at the emergency department, the patient had a blood pressure of 78/46 mm Hg, a pulse of 120 beats/min, a temperature of 34C, 14 respirations per minute, and oxygen saturation of 99% while breathing supplemental oxygen through a nonrebreather mask. Systolic blood pressure improved to 90 mmHg after 4 L of normal saline was administered. The patient was pale but alert. There was crusted blood in her mouth and nostrils without active bleeding or petechiae. Her tongue was pierced with a ring, and sclerae were anicteric. Bleeding was noted from both nipples. There was no heart murmur or gallop, and jugular venous pressure was not elevated. Pulmonary exam revealed bibasilar crackles. Abdomen was soft, not tender, and without hepatosplenomegaly, and her umbilicus was pierced by a ring. Genitourinary exam revealed scant vaginal discharge and clotted blood in the vagina. Skin demonstrated no petechiae, ecchymoses, or stigmata of liver disease. Neurological and joint exams were normal.

It is hard to conceive of vaginal bleeding producing this profound a degree of hypotension. The patient may have additional occult sites of bleeding, or she may have a distributive cause of hypotension such as sepsis or adrenal hemorrhage with resultant adrenal insufficiency. Breast bleeding is unusual, even with profound thrombocytopenia, and I wonder about a concomitant factor deficiency. Furthermore, if thrombocytopenia was the sole reason for the bleeding, I would have expected petechiae. Diffuse vascular injury, such as from lupus or vasculitis, would be an unusual cause of profound bleeding unless there was also disseminated intravascular coagulation.

Laboratory studies revealed a white count of 2000/mm³, of which 42% were neutrophils, 40% bands, 8% lymphocytes, and 10% monocytes. Hematocrit was 17.6%, platelets 35,000/mm³. Sodium was 124 mmol/L, potassium 6 mmol/L, chloride 92 mmol/L, bicarbonate 10 mmol/L, blood urea nitrogen 122 mg/dL (43.5 mmol/L), and creatinine 3.4 mg/dL (300 mol/L). Blood glucose was 44 mg/dL (2.44 mmol/L). Total bilirubin was 3.0 mg/dL (51.3 mol/L; normal range, 0.1‐1.5), alkaline phosphatase 105 U/L (normal range, 39‐117), aspartate aminotransferase 849 U/L (normal range, 8‐31), alanine aminotransferase 261 U/L (normal range, 7‐31), international normalized unit (INR) 2.9, and partial thromboplastin time (PTT) 34.2 seconds.

The combination of profound hypotension, electrolyte abnormalities, hypoglycemia, and hypothermia makes adrenal insufficiency a consideration. I would perform a cortrosyn stimulation test and start glucocorticoid and perhaps mineralocorticoid replacement. In addition, there is renal failure and metabolic acidosis, with a calculated anion gap of 22. The anion gap may be from lactic acidosis secondary to hypotension and hypoperfusion. The abnormal transaminases and bilirubin could relate to infectious hepatitis or systemic infection. Although ischemia could explain these findings, it is rare for a 20‐year‐old to develop ischemic hepatopathy. Thrombocytopenia this moderate may augment the volume of blood loss, but spontaneous bleeding because of thrombocytopenia is unusual until the platelet count falls below 20,000/mm³. Furthermore, the elevated INR points to a mixed coagulopathy. Interpretation of the INR is complicated by the fact she has liver disease, and I am most concerned about acute disseminated intravascular coagulation (DIC) or impending fulminant hepatic failure. This is not the pattern seen with antiphospholipid antibody syndrome, in which the INR tends to be preserved and the PTT prolonged.

Urine dipstick testing demonstrated a specific gravity of 1.015, trace leukocyte esterase, 2+ protein, and 3+ blood, and microscopy revealed 2 white blood cells and 38 red blood cells per high‐power field, many bacteria, and no casts. Creatine kinase was 20,599 U/L, with a myocardial fraction of 1.4%. Lipase was normal, lactate was 7.3 mmol/L, and serum pregnancy test was negative.

Although there is proteinuria and hematuria, we do not have solid evidence of glomerulonephritis. Although the red cells could be a contaminant from her vaginal bleeding, I would examine her sediment carefully for dysmorphic red cells, recognizing that only a quarter of people with glomerulonephritis have red‐cell casts. A urine protein‐to‐creatinine ratio would be useful for estimating the degree of proteinuria. The elevated creatine kinase indicates rhabdomyolysis. In a previously healthy young woman without evidence of cardiogenic shock, it would be unusual for hypotension to result in rhabdomyolysis. Infection and metabolic derangements are possible etiologies of rhabdomyolysis. Alternatively, coagulopathy might have produced intramuscular bleeding. The constellation of thrombocytopenia, anemia, and renal failure raises my suspicion that there is a thrombotic microangiopathy, such as thrombotic thrombocytopenic purpura (TTP) or hemolytic uremic syndrome (HUS). I would inspect a peripheral‐blood smear for schistocytes and evidence of microangiopathy.

The chest radiograph demonstrated low lung volumes, patchy areas of consolidation, and pulmonary edema. Heart size was normal, and there were no pleural effusions. On the first hospital day the patient required mechanical ventilation because of respiratory failure. She received 5 units of packed red blood cells, 2 units of fresh frozen plasma, and 1 unit of platelets. Vasopressor infusion was started, and a vascular catheter was placed for hemodialysis. Blood, respiratory, and urine cultures were sent, and methylprednisolone, piperacillin/tazobactam, and vancomycin were administered. _D‐dimer was greater than 10,000 ng/mL, fibrinogen was 178 mg/dL, and lactate dehydrogenase was 1671 U/L (27 kat/L). The peripheral‐blood smear demonstrated 1+ schistocytes and no spherocytes. There were fewer white blood cells with bands and myelocytes, but no blasts.

The presence of schistocytes and the elevated lactate dehydrogenase point to a microangiopathic hemolytic process. Causes of microangiopathic hemolytic anemia include TTP, HUS, DIC, paraneoplastic conditions, and endothelial damage from malignant hypertension or scleroderma renal crisis. The INR and PTT will usually be normal in TTP and HUS. The depressed fibrinogen and elevated _D‐dimer suggest that in response to severe bleeding, she is also clotting. DIC, possibly from a severe infection, would explain these findings. Alternatively, the multisystem organ failure may represent progression of SLE.

Additional serology studies detected antinuclear antibodies at 1:320 with a speckled pattern. Rheumatoid factor was not present, but antidouble‐stranded DNA and antiSmith antibodies were elevated. C3 was 30 mg/dL (normal range, 90‐180), C4 was 24 mg/dL (normal range, 16‐47), and the erythrocyte sedimentation rate was 53 mm/h.

The results of the additional lab tests support a diagnosis of lupus and thus a lupus flare, but I agree that antibiotics should be empirically administered while searching for an underlying infection that might mimic lupus. Apart from infection, severe lupus may be complicated by widespread vasculitis or catastrophic antiphospholipid antibody syndrome, which would necessitate high‐dose immunosuppressive therapy and anticoagulation, respectively.

Tests for antiphospholipid antibodies including lupus anticoagulant and for anticardiolipin antibodies were negative. The patient continued to require vasopressors, hemodialysis, and mechanical ventilation. On the fourth hospital day she developed a morbilliform rash over her trunk, face, and extremities. Skin over her right buttock became indurated and tender. On the sixth day of hospitalization the skin on her face, extremities, and palms began to desquamate (Fig. 1).

Figure 1

Regarding the rash, it is hard to differentiate the chicken from the egg. The rash may be a reaction to medication, or it may be a clue to a multiorgan disease. I am considering severe skin reactions like Stevens‐Johnson as well as bacterial toxin‐mediated diseases such as toxic shock syndrome. The criteria for toxic shock syndrome with multisystem involvement are very similar to those for lupus. In this case, a desquamating rash occurring on the heel of a multiorgan illness definitely points to toxic shock syndrome. In staphylococcal toxic shock cases, blood cultures are frequently negative, and the origin may elude detection, but of the sources identified, most have been wounds and soft‐tissue infections.

On hospital day 4, blood cultures from admission grew oxacillin‐sensitive Staphylococcus aureus in 4 of the 4 bottles. Magnetic resonance imaging of the thigh demonstrated extensive necrosis of multiple muscles (Fig. 2). The patient underwent muscle debridement in the operating room, and Gram's stain of the debrided muscle revealed Gram‐positive cocci. Following surgery, she rapidly improved. She no longer required dialysis and was eventually discharged home after completing a prolonged course of intravenous anti‐Staphylococcal antibiotics at a rehabilitation facility. Follow‐up urine testing on 2 occasions revealed 1.6 and 1.4 g of protein in 24‐hour collections, but serum creatinine remained normal, and microscopy demonstrated no dysmorphic red cells or red‐cell casts. Performance of a kidney biopsy was deferred. Other than transient arthralgia and malar rash, her lupus has been quiescent, and her prednisone dose was tapered to 5 mg daily. Six months after discharge she returned to school.

Figure 2

COMMENTARY

Using the American College of Rheumatology (ACR) definition, systemic lupus erythematosus (SLE) is diagnosed when at least 4 criteria are met with a sensitivity and specificity above 95%. These criteria were developed for study purposes to differentiate SLE from other rheumatic diseases. At disease onset a patient may not meet the ACR threshold, but delaying treatment may be harmful. Data conflict on the probability of such patients eventually being classified as having SLE, with estimates ranging from less than 10% to more than 60%.1, 2 With SLE prominent in the differential diagnosis of a critically ill patient, hospitalists must consider the 3 most common causes of death in lupus patients: lupus crisis, severe infection, and thrombosis.3

Most exacerbations of SLE occur in one system, most commonly the musculoskeletal system, and are mild. However, 10% of patients a year will require high‐dose corticosteroids or cytotoxic agents for severe flares that can occur in any system affected by lupus and in 15% of cases may involve multiple sites simultaneously.4, 5 Diagnosing lupus flares remains challenging. Although pulmonary hemorrhage and red blood cell casts may strongly implicate active lupus in the lungs or kidneys, specific clinical and laboratory markers of lupus crisis are lacking. Several global indices reliably measure current disease status but are cumbersome, cannot be relied on solely for treatment decisions and have not been well studied in hospitalized patients.68 Fever, once a dependable harbinger of active lupus,9 cannot reliably discriminate lupus flares from infection. In 2 studies, Rovin et al. found that infection accounted for fever in all but one SLE outpatient taking prednisone and that in hospitalized SLE patients, failure of fevers to resolve within 48 hours of administering 20‐40 mg of prednisone daily strongly suggested infection.10 The laboratory findings provided general support for there being an SLE flare or an infection, but, as the discussant pointed out, these cannot be relied on exclusively to discriminate between the two. Results that suggest infection in an SLE patient include leukocytosis, increased band forms or metamyelocytes, and possibly elevated C‐reactive protein. Findings favoring SLE flare include leukopenia, low C3 or C4 (particularly for nephritis or hematologic flares) and elevated anti‐double‐stranded DNA antibodies for nephritis.1113 Without a clear gold standard for definitively determining a lupus crisis, it is diagnosed when clinical manifestations fit a pattern seen in SLE (nephritis, cerebritis, serositis, vasculitis, pneumonitis), the results of serology studies support this conclusion, and other plausible diagnoses are excluded.

Infection and active disease account for most ICU admissions of lupus patients. SLE and infection intertwine in 3 ways. First, SLE patients are predisposed to infection, possibly because of a variety of identified genetic abnormalities of immune function.14 Although community‐acquired bacteria and viruses account for most infections, lupus patients are vulnerable to a wide array of atypical and opportunistic pathogens. Clinical factors that augment this intrinsic risk include severity of the underlying SLE, flares of the central nervous system or kidneys, and use of immunosuppressive agents.14 The latter deserves particular attention, as a recent study found more than 90% of SLE patients admitted to an ICU with severe infection were taking corticosteroids prior to hospitalization.15 Second, infection may trigger a lupus flare. Third, features of severe lupus flares and infection may overlap. Differentiating between the 2 may be difficult, and the stakes are high, as SLE patients admitted to ICUs have a risk of death that is substantially higher (47%) than that of those without SLE (29%) and much greater than the overall risk of death for those with SLE, for whom 10‐year survival exceeds 90%.15

In addition to lupus crisis and infection, the differential diagnosis of acute multisystem disease in a patient with SLE includes catastrophic antiphospholipid syndrome (APS) and thrombotic thrombocytopenic purpura, 2 thrombotic microangiopathies to which SLE patients are predisposed. Thrombocytopenia and hemolytic anemia with schistocytes should raise suspicion of these diagnoses. Additional findings for TTP include fevers, altered mental status, acute renal failure, and elevated serum lactate dehydrogenase; however, prothrombin time should not be prolonged. Lupus anticoagulant or anticardiolipin antibodies are found in up to 30% of lupus patients, of whom 50%‐70% develop APS within 20 years, characterized by thrombosis or spontaneous abortions in the presence of antiphospholipid antibodies.16 Catastrophic APS is a rare subset of APS involving thromboses of multiple organs simultaneously and has a mortality rate of 50%.

In the present patient, an elevated INR, bleeding, hypotension, and the absence of antiphospholipid antibodies argued against TTP and APS, leading the discussant to focus on SLE and sepsis. Arthralgia, cytopenia, and the results of serology studies suggested a lupus crisis, but hypothermia, hypotension, and DIC pointed to severe infection. Empiric treatment of both conditions with corticosteroids and broad‐spectrum antibiotics was indicated, and ultimately the patient's condition was found to meet criteria for toxic shock syndrome (TSS) and SLE. TSS has rarely been reported in SLE1718 and poses a particularly difficult diagnostic challenge because a severe lupus flare can meet the diagnostic criteria for TSS (Table 1), especially early on, before the characteristic desquamating rash appears. Acuity of the illness increased the ante in this challenging case. Afraid not to treat a potentially life‐threatening condition, empiric treatment of severe lupus and sepsis was initiated. Attention then shifted to fraying, or unraveling, the knot linking infection and lupus. Ultimately, diagnoses of both TSS and SLE were established.

A 20‐year‐old woman presented to the emergency department after 2 days of epistaxis and vaginal bleeding.

A young woman is more likely to present with infection, toxic exposure, or rheumatologic disease than with a degenerative disease or malignancy. Her bleeding may relate to a platelet abnormality, either quantitative or qualitative. I would pursue her bleeding and menstrual history further.

The patient was healthy until 2 months previously, when she noted arthralgia of her shoulders, wrists, elbows, knees, and ankles. She was examined by a rheumatologist who detected mild arthritis in her left wrist and proximal interphalangeal joints. The rest of her joints were normal. Rheumatoid factor and ANA were positive, and the erythrocyte sedimentation rate was 122 mm/hour. She was diagnosed with possible systemic lupus erythematosus and was placed on a nonsteroidal anti‐inflammatory agent. At a follow‐up visit 1 month prior to admission, her arthralgia had markedly improved. Two weeks prior to admission, the patient began to feel fatigued. Two days prior to admission, she developed epistaxis and what she thought was her menses, though bleeding was heavier than usual and associated with the passage of red clots. On the day of admission the vaginal bleeding worsened, and emergency personnel transported the patient to the hospital.

The diagnosis of systemic lupus erythematosus (SLE) is not engraved in stone. One must be vigilant for other diseases masquerading as SLE while continuing to build a case for it. As more criteria are fulfilled, the probability of lupus increases, yet no findings, alone or in combination, are pathognomonic of this protean disease. This patient's age, sex, and serology are compatible with SLE; otherwise, her presentation is nonspecific. I would request a complete blood count, coagulation tests, and additional serological tests.

The quantity of the bleeding is described, but this does not help decipher its etiology. Excess bleeding may be a result of one or more of 3 broad etiologies: problems with platelets (quantitative or qualitative), with clotting factors (quantitative or qualitative), or with blood vessels (trauma, vasculitis, or diseases affecting collagen). Because quantitative and qualitative factor disorders generally do not present with mucosal bleeding, I am thinking more about platelet problems and about processes that damage the microvasculature. If this woman has lupus, immunologic thrombocytopenia may be the cause of mucosal bleeding.

The patient had no previous medical problems and had never been pregnant. Her only medication was sulindac twice daily for the past month. She was born in Hong Kong, graduated from high school in San Francisco, and attended junior college. She lived with her parents and brother and denied alcohol, tobacco, or recreational drug use but had recently obtained a tattoo on her lower back. There was no family history of autoimmune or bleeding disorders, and a review of systems was notable for dyspnea with minimal exertion and fatigue which worsened in the past 2 days. She had no prior episodes of abnormal bleeding or clotting.

Tattoos may be surrogates for other high‐risk behaviors and suggest an increased risk of hepatitis and sexually transmitted diseases. I want to know her sexual history and other risk factors for human immunodeficiency virus infection. The dyspnea and fatigue are likely the result of anemia, but I am also considering cardiac disease. Though SLE remains a possibility, I cannot assume the presence of a lupus anticoagulant with antiphospholipid syndrome without a history of infertility or recurrent miscarriages.

On arrival at the emergency department, the patient had a blood pressure of 78/46 mm Hg, a pulse of 120 beats/min, a temperature of 34C, 14 respirations per minute, and oxygen saturation of 99% while breathing supplemental oxygen through a nonrebreather mask. Systolic blood pressure improved to 90 mmHg after 4 L of normal saline was administered. The patient was pale but alert. There was crusted blood in her mouth and nostrils without active bleeding or petechiae. Her tongue was pierced with a ring, and sclerae were anicteric. Bleeding was noted from both nipples. There was no heart murmur or gallop, and jugular venous pressure was not elevated. Pulmonary exam revealed bibasilar crackles. Abdomen was soft, not tender, and without hepatosplenomegaly, and her umbilicus was pierced by a ring. Genitourinary exam revealed scant vaginal discharge and clotted blood in the vagina. Skin demonstrated no petechiae, ecchymoses, or stigmata of liver disease. Neurological and joint exams were normal.

It is hard to conceive of vaginal bleeding producing this profound a degree of hypotension. The patient may have additional occult sites of bleeding, or she may have a distributive cause of hypotension such as sepsis or adrenal hemorrhage with resultant adrenal insufficiency. Breast bleeding is unusual, even with profound thrombocytopenia, and I wonder about a concomitant factor deficiency. Furthermore, if thrombocytopenia was the sole reason for the bleeding, I would have expected petechiae. Diffuse vascular injury, such as from lupus or vasculitis, would be an unusual cause of profound bleeding unless there was also disseminated intravascular coagulation.

Laboratory studies revealed a white count of 2000/mm³, of which 42% were neutrophils, 40% bands, 8% lymphocytes, and 10% monocytes. Hematocrit was 17.6%, platelets 35,000/mm³. Sodium was 124 mmol/L, potassium 6 mmol/L, chloride 92 mmol/L, bicarbonate 10 mmol/L, blood urea nitrogen 122 mg/dL (43.5 mmol/L), and creatinine 3.4 mg/dL (300 mol/L). Blood glucose was 44 mg/dL (2.44 mmol/L). Total bilirubin was 3.0 mg/dL (51.3 mol/L; normal range, 0.1‐1.5), alkaline phosphatase 105 U/L (normal range, 39‐117), aspartate aminotransferase 849 U/L (normal range, 8‐31), alanine aminotransferase 261 U/L (normal range, 7‐31), international normalized unit (INR) 2.9, and partial thromboplastin time (PTT) 34.2 seconds.

The combination of profound hypotension, electrolyte abnormalities, hypoglycemia, and hypothermia makes adrenal insufficiency a consideration. I would perform a cortrosyn stimulation test and start glucocorticoid and perhaps mineralocorticoid replacement. In addition, there is renal failure and metabolic acidosis, with a calculated anion gap of 22. The anion gap may be from lactic acidosis secondary to hypotension and hypoperfusion. The abnormal transaminases and bilirubin could relate to infectious hepatitis or systemic infection. Although ischemia could explain these findings, it is rare for a 20‐year‐old to develop ischemic hepatopathy. Thrombocytopenia this moderate may augment the volume of blood loss, but spontaneous bleeding because of thrombocytopenia is unusual until the platelet count falls below 20,000/mm³. Furthermore, the elevated INR points to a mixed coagulopathy. Interpretation of the INR is complicated by the fact she has liver disease, and I am most concerned about acute disseminated intravascular coagulation (DIC) or impending fulminant hepatic failure. This is not the pattern seen with antiphospholipid antibody syndrome, in which the INR tends to be preserved and the PTT prolonged.

Urine dipstick testing demonstrated a specific gravity of 1.015, trace leukocyte esterase, 2+ protein, and 3+ blood, and microscopy revealed 2 white blood cells and 38 red blood cells per high‐power field, many bacteria, and no casts. Creatine kinase was 20,599 U/L, with a myocardial fraction of 1.4%. Lipase was normal, lactate was 7.3 mmol/L, and serum pregnancy test was negative.

Although there is proteinuria and hematuria, we do not have solid evidence of glomerulonephritis. Although the red cells could be a contaminant from her vaginal bleeding, I would examine her sediment carefully for dysmorphic red cells, recognizing that only a quarter of people with glomerulonephritis have red‐cell casts. A urine protein‐to‐creatinine ratio would be useful for estimating the degree of proteinuria. The elevated creatine kinase indicates rhabdomyolysis. In a previously healthy young woman without evidence of cardiogenic shock, it would be unusual for hypotension to result in rhabdomyolysis. Infection and metabolic derangements are possible etiologies of rhabdomyolysis. Alternatively, coagulopathy might have produced intramuscular bleeding. The constellation of thrombocytopenia, anemia, and renal failure raises my suspicion that there is a thrombotic microangiopathy, such as thrombotic thrombocytopenic purpura (TTP) or hemolytic uremic syndrome (HUS). I would inspect a peripheral‐blood smear for schistocytes and evidence of microangiopathy.

The chest radiograph demonstrated low lung volumes, patchy areas of consolidation, and pulmonary edema. Heart size was normal, and there were no pleural effusions. On the first hospital day the patient required mechanical ventilation because of respiratory failure. She received 5 units of packed red blood cells, 2 units of fresh frozen plasma, and 1 unit of platelets. Vasopressor infusion was started, and a vascular catheter was placed for hemodialysis. Blood, respiratory, and urine cultures were sent, and methylprednisolone, piperacillin/tazobactam, and vancomycin were administered. _D‐dimer was greater than 10,000 ng/mL, fibrinogen was 178 mg/dL, and lactate dehydrogenase was 1671 U/L (27 kat/L). The peripheral‐blood smear demonstrated 1+ schistocytes and no spherocytes. There were fewer white blood cells with bands and myelocytes, but no blasts.

The presence of schistocytes and the elevated lactate dehydrogenase point to a microangiopathic hemolytic process. Causes of microangiopathic hemolytic anemia include TTP, HUS, DIC, paraneoplastic conditions, and endothelial damage from malignant hypertension or scleroderma renal crisis. The INR and PTT will usually be normal in TTP and HUS. The depressed fibrinogen and elevated _D‐dimer suggest that in response to severe bleeding, she is also clotting. DIC, possibly from a severe infection, would explain these findings. Alternatively, the multisystem organ failure may represent progression of SLE.

Additional serology studies detected antinuclear antibodies at 1:320 with a speckled pattern. Rheumatoid factor was not present, but antidouble‐stranded DNA and antiSmith antibodies were elevated. C3 was 30 mg/dL (normal range, 90‐180), C4 was 24 mg/dL (normal range, 16‐47), and the erythrocyte sedimentation rate was 53 mm/h.

The results of the additional lab tests support a diagnosis of lupus and thus a lupus flare, but I agree that antibiotics should be empirically administered while searching for an underlying infection that might mimic lupus. Apart from infection, severe lupus may be complicated by widespread vasculitis or catastrophic antiphospholipid antibody syndrome, which would necessitate high‐dose immunosuppressive therapy and anticoagulation, respectively.

Tests for antiphospholipid antibodies including lupus anticoagulant and for anticardiolipin antibodies were negative. The patient continued to require vasopressors, hemodialysis, and mechanical ventilation. On the fourth hospital day she developed a morbilliform rash over her trunk, face, and extremities. Skin over her right buttock became indurated and tender. On the sixth day of hospitalization the skin on her face, extremities, and palms began to desquamate (Fig. 1).

Figure 1

Regarding the rash, it is hard to differentiate the chicken from the egg. The rash may be a reaction to medication, or it may be a clue to a multiorgan disease. I am considering severe skin reactions like Stevens‐Johnson as well as bacterial toxin‐mediated diseases such as toxic shock syndrome. The criteria for toxic shock syndrome with multisystem involvement are very similar to those for lupus. In this case, a desquamating rash occurring on the heel of a multiorgan illness definitely points to toxic shock syndrome. In staphylococcal toxic shock cases, blood cultures are frequently negative, and the origin may elude detection, but of the sources identified, most have been wounds and soft‐tissue infections.

On hospital day 4, blood cultures from admission grew oxacillin‐sensitive Staphylococcus aureus in 4 of the 4 bottles. Magnetic resonance imaging of the thigh demonstrated extensive necrosis of multiple muscles (Fig. 2). The patient underwent muscle debridement in the operating room, and Gram's stain of the debrided muscle revealed Gram‐positive cocci. Following surgery, she rapidly improved. She no longer required dialysis and was eventually discharged home after completing a prolonged course of intravenous anti‐Staphylococcal antibiotics at a rehabilitation facility. Follow‐up urine testing on 2 occasions revealed 1.6 and 1.4 g of protein in 24‐hour collections, but serum creatinine remained normal, and microscopy demonstrated no dysmorphic red cells or red‐cell casts. Performance of a kidney biopsy was deferred. Other than transient arthralgia and malar rash, her lupus has been quiescent, and her prednisone dose was tapered to 5 mg daily. Six months after discharge she returned to school.

Figure 2

COMMENTARY

Using the American College of Rheumatology (ACR) definition, systemic lupus erythematosus (SLE) is diagnosed when at least 4 criteria are met with a sensitivity and specificity above 95%. These criteria were developed for study purposes to differentiate SLE from other rheumatic diseases. At disease onset a patient may not meet the ACR threshold, but delaying treatment may be harmful. Data conflict on the probability of such patients eventually being classified as having SLE, with estimates ranging from less than 10% to more than 60%.1, 2 With SLE prominent in the differential diagnosis of a critically ill patient, hospitalists must consider the 3 most common causes of death in lupus patients: lupus crisis, severe infection, and thrombosis.3

Most exacerbations of SLE occur in one system, most commonly the musculoskeletal system, and are mild. However, 10% of patients a year will require high‐dose corticosteroids or cytotoxic agents for severe flares that can occur in any system affected by lupus and in 15% of cases may involve multiple sites simultaneously.4, 5 Diagnosing lupus flares remains challenging. Although pulmonary hemorrhage and red blood cell casts may strongly implicate active lupus in the lungs or kidneys, specific clinical and laboratory markers of lupus crisis are lacking. Several global indices reliably measure current disease status but are cumbersome, cannot be relied on solely for treatment decisions and have not been well studied in hospitalized patients.68 Fever, once a dependable harbinger of active lupus,9 cannot reliably discriminate lupus flares from infection. In 2 studies, Rovin et al. found that infection accounted for fever in all but one SLE outpatient taking prednisone and that in hospitalized SLE patients, failure of fevers to resolve within 48 hours of administering 20‐40 mg of prednisone daily strongly suggested infection.10 The laboratory findings provided general support for there being an SLE flare or an infection, but, as the discussant pointed out, these cannot be relied on exclusively to discriminate between the two. Results that suggest infection in an SLE patient include leukocytosis, increased band forms or metamyelocytes, and possibly elevated C‐reactive protein. Findings favoring SLE flare include leukopenia, low C3 or C4 (particularly for nephritis or hematologic flares) and elevated anti‐double‐stranded DNA antibodies for nephritis.1113 Without a clear gold standard for definitively determining a lupus crisis, it is diagnosed when clinical manifestations fit a pattern seen in SLE (nephritis, cerebritis, serositis, vasculitis, pneumonitis), the results of serology studies support this conclusion, and other plausible diagnoses are excluded.

Infection and active disease account for most ICU admissions of lupus patients. SLE and infection intertwine in 3 ways. First, SLE patients are predisposed to infection, possibly because of a variety of identified genetic abnormalities of immune function.14 Although community‐acquired bacteria and viruses account for most infections, lupus patients are vulnerable to a wide array of atypical and opportunistic pathogens. Clinical factors that augment this intrinsic risk include severity of the underlying SLE, flares of the central nervous system or kidneys, and use of immunosuppressive agents.14 The latter deserves particular attention, as a recent study found more than 90% of SLE patients admitted to an ICU with severe infection were taking corticosteroids prior to hospitalization.15 Second, infection may trigger a lupus flare. Third, features of severe lupus flares and infection may overlap. Differentiating between the 2 may be difficult, and the stakes are high, as SLE patients admitted to ICUs have a risk of death that is substantially higher (47%) than that of those without SLE (29%) and much greater than the overall risk of death for those with SLE, for whom 10‐year survival exceeds 90%.15

In addition to lupus crisis and infection, the differential diagnosis of acute multisystem disease in a patient with SLE includes catastrophic antiphospholipid syndrome (APS) and thrombotic thrombocytopenic purpura, 2 thrombotic microangiopathies to which SLE patients are predisposed. Thrombocytopenia and hemolytic anemia with schistocytes should raise suspicion of these diagnoses. Additional findings for TTP include fevers, altered mental status, acute renal failure, and elevated serum lactate dehydrogenase; however, prothrombin time should not be prolonged. Lupus anticoagulant or anticardiolipin antibodies are found in up to 30% of lupus patients, of whom 50%‐70% develop APS within 20 years, characterized by thrombosis or spontaneous abortions in the presence of antiphospholipid antibodies.16 Catastrophic APS is a rare subset of APS involving thromboses of multiple organs simultaneously and has a mortality rate of 50%.

In the present patient, an elevated INR, bleeding, hypotension, and the absence of antiphospholipid antibodies argued against TTP and APS, leading the discussant to focus on SLE and sepsis. Arthralgia, cytopenia, and the results of serology studies suggested a lupus crisis, but hypothermia, hypotension, and DIC pointed to severe infection. Empiric treatment of both conditions with corticosteroids and broad‐spectrum antibiotics was indicated, and ultimately the patient's condition was found to meet criteria for toxic shock syndrome (TSS) and SLE. TSS has rarely been reported in SLE1718 and poses a particularly difficult diagnostic challenge because a severe lupus flare can meet the diagnostic criteria for TSS (Table 1), especially early on, before the characteristic desquamating rash appears. Acuity of the illness increased the ante in this challenging case. Afraid not to treat a potentially life‐threatening condition, empiric treatment of severe lupus and sepsis was initiated. Attention then shifted to fraying, or unraveling, the knot linking infection and lupus. Ultimately, diagnoses of both TSS and SLE were established.

A 20‐year‐old woman presented to the emergency department after 2 days of epistaxis and vaginal bleeding.

A young woman is more likely to present with infection, toxic exposure, or rheumatologic disease than with a degenerative disease or malignancy. Her bleeding may relate to a platelet abnormality, either quantitative or qualitative. I would pursue her bleeding and menstrual history further.

The patient was healthy until 2 months previously, when she noted arthralgia of her shoulders, wrists, elbows, knees, and ankles. She was examined by a rheumatologist who detected mild arthritis in her left wrist and proximal interphalangeal joints. The rest of her joints were normal. Rheumatoid factor and ANA were positive, and the erythrocyte sedimentation rate was 122 mm/hour. She was diagnosed with possible systemic lupus erythematosus and was placed on a nonsteroidal anti‐inflammatory agent. At a follow‐up visit 1 month prior to admission, her arthralgia had markedly improved. Two weeks prior to admission, the patient began to feel fatigued. Two days prior to admission, she developed epistaxis and what she thought was her menses, though bleeding was heavier than usual and associated with the passage of red clots. On the day of admission the vaginal bleeding worsened, and emergency personnel transported the patient to the hospital.

The diagnosis of systemic lupus erythematosus (SLE) is not engraved in stone. One must be vigilant for other diseases masquerading as SLE while continuing to build a case for it. As more criteria are fulfilled, the probability of lupus increases, yet no findings, alone or in combination, are pathognomonic of this protean disease. This patient's age, sex, and serology are compatible with SLE; otherwise, her presentation is nonspecific. I would request a complete blood count, coagulation tests, and additional serological tests.

The quantity of the bleeding is described, but this does not help decipher its etiology. Excess bleeding may be a result of one or more of 3 broad etiologies: problems with platelets (quantitative or qualitative), with clotting factors (quantitative or qualitative), or with blood vessels (trauma, vasculitis, or diseases affecting collagen). Because quantitative and qualitative factor disorders generally do not present with mucosal bleeding, I am thinking more about platelet problems and about processes that damage the microvasculature. If this woman has lupus, immunologic thrombocytopenia may be the cause of mucosal bleeding.

The patient had no previous medical problems and had never been pregnant. Her only medication was sulindac twice daily for the past month. She was born in Hong Kong, graduated from high school in San Francisco, and attended junior college. She lived with her parents and brother and denied alcohol, tobacco, or recreational drug use but had recently obtained a tattoo on her lower back. There was no family history of autoimmune or bleeding disorders, and a review of systems was notable for dyspnea with minimal exertion and fatigue which worsened in the past 2 days. She had no prior episodes of abnormal bleeding or clotting.

Tattoos may be surrogates for other high‐risk behaviors and suggest an increased risk of hepatitis and sexually transmitted diseases. I want to know her sexual history and other risk factors for human immunodeficiency virus infection. The dyspnea and fatigue are likely the result of anemia, but I am also considering cardiac disease. Though SLE remains a possibility, I cannot assume the presence of a lupus anticoagulant with antiphospholipid syndrome without a history of infertility or recurrent miscarriages.

On arrival at the emergency department, the patient had a blood pressure of 78/46 mm Hg, a pulse of 120 beats/min, a temperature of 34C, 14 respirations per minute, and oxygen saturation of 99% while breathing supplemental oxygen through a nonrebreather mask. Systolic blood pressure improved to 90 mmHg after 4 L of normal saline was administered. The patient was pale but alert. There was crusted blood in her mouth and nostrils without active bleeding or petechiae. Her tongue was pierced with a ring, and sclerae were anicteric. Bleeding was noted from both nipples. There was no heart murmur or gallop, and jugular venous pressure was not elevated. Pulmonary exam revealed bibasilar crackles. Abdomen was soft, not tender, and without hepatosplenomegaly, and her umbilicus was pierced by a ring. Genitourinary exam revealed scant vaginal discharge and clotted blood in the vagina. Skin demonstrated no petechiae, ecchymoses, or stigmata of liver disease. Neurological and joint exams were normal.

It is hard to conceive of vaginal bleeding producing this profound a degree of hypotension. The patient may have additional occult sites of bleeding, or she may have a distributive cause of hypotension such as sepsis or adrenal hemorrhage with resultant adrenal insufficiency. Breast bleeding is unusual, even with profound thrombocytopenia, and I wonder about a concomitant factor deficiency. Furthermore, if thrombocytopenia was the sole reason for the bleeding, I would have expected petechiae. Diffuse vascular injury, such as from lupus or vasculitis, would be an unusual cause of profound bleeding unless there was also disseminated intravascular coagulation.

Laboratory studies revealed a white count of 2000/mm³, of which 42% were neutrophils, 40% bands, 8% lymphocytes, and 10% monocytes. Hematocrit was 17.6%, platelets 35,000/mm³. Sodium was 124 mmol/L, potassium 6 mmol/L, chloride 92 mmol/L, bicarbonate 10 mmol/L, blood urea nitrogen 122 mg/dL (43.5 mmol/L), and creatinine 3.4 mg/dL (300 mol/L). Blood glucose was 44 mg/dL (2.44 mmol/L). Total bilirubin was 3.0 mg/dL (51.3 mol/L; normal range, 0.1‐1.5), alkaline phosphatase 105 U/L (normal range, 39‐117), aspartate aminotransferase 849 U/L (normal range, 8‐31), alanine aminotransferase 261 U/L (normal range, 7‐31), international normalized unit (INR) 2.9, and partial thromboplastin time (PTT) 34.2 seconds.

The combination of profound hypotension, electrolyte abnormalities, hypoglycemia, and hypothermia makes adrenal insufficiency a consideration. I would perform a cortrosyn stimulation test and start glucocorticoid and perhaps mineralocorticoid replacement. In addition, there is renal failure and metabolic acidosis, with a calculated anion gap of 22. The anion gap may be from lactic acidosis secondary to hypotension and hypoperfusion. The abnormal transaminases and bilirubin could relate to infectious hepatitis or systemic infection. Although ischemia could explain these findings, it is rare for a 20‐year‐old to develop ischemic hepatopathy. Thrombocytopenia this moderate may augment the volume of blood loss, but spontaneous bleeding because of thrombocytopenia is unusual until the platelet count falls below 20,000/mm³. Furthermore, the elevated INR points to a mixed coagulopathy. Interpretation of the INR is complicated by the fact she has liver disease, and I am most concerned about acute disseminated intravascular coagulation (DIC) or impending fulminant hepatic failure. This is not the pattern seen with antiphospholipid antibody syndrome, in which the INR tends to be preserved and the PTT prolonged.

Urine dipstick testing demonstrated a specific gravity of 1.015, trace leukocyte esterase, 2+ protein, and 3+ blood, and microscopy revealed 2 white blood cells and 38 red blood cells per high‐power field, many bacteria, and no casts. Creatine kinase was 20,599 U/L, with a myocardial fraction of 1.4%. Lipase was normal, lactate was 7.3 mmol/L, and serum pregnancy test was negative.

Although there is proteinuria and hematuria, we do not have solid evidence of glomerulonephritis. Although the red cells could be a contaminant from her vaginal bleeding, I would examine her sediment carefully for dysmorphic red cells, recognizing that only a quarter of people with glomerulonephritis have red‐cell casts. A urine protein‐to‐creatinine ratio would be useful for estimating the degree of proteinuria. The elevated creatine kinase indicates rhabdomyolysis. In a previously healthy young woman without evidence of cardiogenic shock, it would be unusual for hypotension to result in rhabdomyolysis. Infection and metabolic derangements are possible etiologies of rhabdomyolysis. Alternatively, coagulopathy might have produced intramuscular bleeding. The constellation of thrombocytopenia, anemia, and renal failure raises my suspicion that there is a thrombotic microangiopathy, such as thrombotic thrombocytopenic purpura (TTP) or hemolytic uremic syndrome (HUS). I would inspect a peripheral‐blood smear for schistocytes and evidence of microangiopathy.

The chest radiograph demonstrated low lung volumes, patchy areas of consolidation, and pulmonary edema. Heart size was normal, and there were no pleural effusions. On the first hospital day the patient required mechanical ventilation because of respiratory failure. She received 5 units of packed red blood cells, 2 units of fresh frozen plasma, and 1 unit of platelets. Vasopressor infusion was started, and a vascular catheter was placed for hemodialysis. Blood, respiratory, and urine cultures were sent, and methylprednisolone, piperacillin/tazobactam, and vancomycin were administered. _D‐dimer was greater than 10,000 ng/mL, fibrinogen was 178 mg/dL, and lactate dehydrogenase was 1671 U/L (27 kat/L). The peripheral‐blood smear demonstrated 1+ schistocytes and no spherocytes. There were fewer white blood cells with bands and myelocytes, but no blasts.

The presence of schistocytes and the elevated lactate dehydrogenase point to a microangiopathic hemolytic process. Causes of microangiopathic hemolytic anemia include TTP, HUS, DIC, paraneoplastic conditions, and endothelial damage from malignant hypertension or scleroderma renal crisis. The INR and PTT will usually be normal in TTP and HUS. The depressed fibrinogen and elevated _D‐dimer suggest that in response to severe bleeding, she is also clotting. DIC, possibly from a severe infection, would explain these findings. Alternatively, the multisystem organ failure may represent progression of SLE.

Additional serology studies detected antinuclear antibodies at 1:320 with a speckled pattern. Rheumatoid factor was not present, but antidouble‐stranded DNA and antiSmith antibodies were elevated. C3 was 30 mg/dL (normal range, 90‐180), C4 was 24 mg/dL (normal range, 16‐47), and the erythrocyte sedimentation rate was 53 mm/h.

The results of the additional lab tests support a diagnosis of lupus and thus a lupus flare, but I agree that antibiotics should be empirically administered while searching for an underlying infection that might mimic lupus. Apart from infection, severe lupus may be complicated by widespread vasculitis or catastrophic antiphospholipid antibody syndrome, which would necessitate high‐dose immunosuppressive therapy and anticoagulation, respectively.

Tests for antiphospholipid antibodies including lupus anticoagulant and for anticardiolipin antibodies were negative. The patient continued to require vasopressors, hemodialysis, and mechanical ventilation. On the fourth hospital day she developed a morbilliform rash over her trunk, face, and extremities. Skin over her right buttock became indurated and tender. On the sixth day of hospitalization the skin on her face, extremities, and palms began to desquamate (Fig. 1).

Figure 1

Regarding the rash, it is hard to differentiate the chicken from the egg. The rash may be a reaction to medication, or it may be a clue to a multiorgan disease. I am considering severe skin reactions like Stevens‐Johnson as well as bacterial toxin‐mediated diseases such as toxic shock syndrome. The criteria for toxic shock syndrome with multisystem involvement are very similar to those for lupus. In this case, a desquamating rash occurring on the heel of a multiorgan illness definitely points to toxic shock syndrome. In staphylococcal toxic shock cases, blood cultures are frequently negative, and the origin may elude detection, but of the sources identified, most have been wounds and soft‐tissue infections.

On hospital day 4, blood cultures from admission grew oxacillin‐sensitive Staphylococcus aureus in 4 of the 4 bottles. Magnetic resonance imaging of the thigh demonstrated extensive necrosis of multiple muscles (Fig. 2). The patient underwent muscle debridement in the operating room, and Gram's stain of the debrided muscle revealed Gram‐positive cocci. Following surgery, she rapidly improved. She no longer required dialysis and was eventually discharged home after completing a prolonged course of intravenous anti‐Staphylococcal antibiotics at a rehabilitation facility. Follow‐up urine testing on 2 occasions revealed 1.6 and 1.4 g of protein in 24‐hour collections, but serum creatinine remained normal, and microscopy demonstrated no dysmorphic red cells or red‐cell casts. Performance of a kidney biopsy was deferred. Other than transient arthralgia and malar rash, her lupus has been quiescent, and her prednisone dose was tapered to 5 mg daily. Six months after discharge she returned to school.

Figure 2

COMMENTARY

Using the American College of Rheumatology (ACR) definition, systemic lupus erythematosus (SLE) is diagnosed when at least 4 criteria are met with a sensitivity and specificity above 95%. These criteria were developed for study purposes to differentiate SLE from other rheumatic diseases. At disease onset a patient may not meet the ACR threshold, but delaying treatment may be harmful. Data conflict on the probability of such patients eventually being classified as having SLE, with estimates ranging from less than 10% to more than 60%.1, 2 With SLE prominent in the differential diagnosis of a critically ill patient, hospitalists must consider the 3 most common causes of death in lupus patients: lupus crisis, severe infection, and thrombosis.3

Most exacerbations of SLE occur in one system, most commonly the musculoskeletal system, and are mild. However, 10% of patients a year will require high‐dose corticosteroids or cytotoxic agents for severe flares that can occur in any system affected by lupus and in 15% of cases may involve multiple sites simultaneously.4, 5 Diagnosing lupus flares remains challenging. Although pulmonary hemorrhage and red blood cell casts may strongly implicate active lupus in the lungs or kidneys, specific clinical and laboratory markers of lupus crisis are lacking. Several global indices reliably measure current disease status but are cumbersome, cannot be relied on solely for treatment decisions and have not been well studied in hospitalized patients.68 Fever, once a dependable harbinger of active lupus,9 cannot reliably discriminate lupus flares from infection. In 2 studies, Rovin et al. found that infection accounted for fever in all but one SLE outpatient taking prednisone and that in hospitalized SLE patients, failure of fevers to resolve within 48 hours of administering 20‐40 mg of prednisone daily strongly suggested infection.10 The laboratory findings provided general support for there being an SLE flare or an infection, but, as the discussant pointed out, these cannot be relied on exclusively to discriminate between the two. Results that suggest infection in an SLE patient include leukocytosis, increased band forms or metamyelocytes, and possibly elevated C‐reactive protein. Findings favoring SLE flare include leukopenia, low C3 or C4 (particularly for nephritis or hematologic flares) and elevated anti‐double‐stranded DNA antibodies for nephritis.1113 Without a clear gold standard for definitively determining a lupus crisis, it is diagnosed when clinical manifestations fit a pattern seen in SLE (nephritis, cerebritis, serositis, vasculitis, pneumonitis), the results of serology studies support this conclusion, and other plausible diagnoses are excluded.

Infection and active disease account for most ICU admissions of lupus patients. SLE and infection intertwine in 3 ways. First, SLE patients are predisposed to infection, possibly because of a variety of identified genetic abnormalities of immune function.14 Although community‐acquired bacteria and viruses account for most infections, lupus patients are vulnerable to a wide array of atypical and opportunistic pathogens. Clinical factors that augment this intrinsic risk include severity of the underlying SLE, flares of the central nervous system or kidneys, and use of immunosuppressive agents.14 The latter deserves particular attention, as a recent study found more than 90% of SLE patients admitted to an ICU with severe infection were taking corticosteroids prior to hospitalization.15 Second, infection may trigger a lupus flare. Third, features of severe lupus flares and infection may overlap. Differentiating between the 2 may be difficult, and the stakes are high, as SLE patients admitted to ICUs have a risk of death that is substantially higher (47%) than that of those without SLE (29%) and much greater than the overall risk of death for those with SLE, for whom 10‐year survival exceeds 90%.15

In addition to lupus crisis and infection, the differential diagnosis of acute multisystem disease in a patient with SLE includes catastrophic antiphospholipid syndrome (APS) and thrombotic thrombocytopenic purpura, 2 thrombotic microangiopathies to which SLE patients are predisposed. Thrombocytopenia and hemolytic anemia with schistocytes should raise suspicion of these diagnoses. Additional findings for TTP include fevers, altered mental status, acute renal failure, and elevated serum lactate dehydrogenase; however, prothrombin time should not be prolonged. Lupus anticoagulant or anticardiolipin antibodies are found in up to 30% of lupus patients, of whom 50%‐70% develop APS within 20 years, characterized by thrombosis or spontaneous abortions in the presence of antiphospholipid antibodies.16 Catastrophic APS is a rare subset of APS involving thromboses of multiple organs simultaneously and has a mortality rate of 50%.

In the present patient, an elevated INR, bleeding, hypotension, and the absence of antiphospholipid antibodies argued against TTP and APS, leading the discussant to focus on SLE and sepsis. Arthralgia, cytopenia, and the results of serology studies suggested a lupus crisis, but hypothermia, hypotension, and DIC pointed to severe infection. Empiric treatment of both conditions with corticosteroids and broad‐spectrum antibiotics was indicated, and ultimately the patient's condition was found to meet criteria for toxic shock syndrome (TSS) and SLE. TSS has rarely been reported in SLE1718 and poses a particularly difficult diagnostic challenge because a severe lupus flare can meet the diagnostic criteria for TSS (Table 1), especially early on, before the characteristic desquamating rash appears. Acuity of the illness increased the ante in this challenging case. Afraid not to treat a potentially life‐threatening condition, empiric treatment of severe lupus and sepsis was initiated. Attention then shifted to fraying, or unraveling, the knot linking infection and lupus. Ultimately, diagnoses of both TSS and SLE were established.

It was the second week after finishing my internal medicine residency. What a daunting experience it was to be a newly appointed attending physician. My hospital rounds were painfully slow because I would consult the Tarsacon pharmacopoeia and Uptodate prior to writing any orders or making clinical decisions. During this keystone phase of my career I had the privilege of taking care of Ms. S. It has been almost 2 years since, and I still think of her and what I learned taking care of her.

Ms. S was a woman in her eighties with end‐stage chronic obstructive pulmonary disease (COPD). She was a frequent flyer, as evidenced by the multiple discharge summaries appended to her chart. Her hospital course was predictably punctuated by frequent inpatient exacerbations of COPD, and every time I told her that she'd be discharged the next day I had to eat my own words. After much effort and pharmaceutical gymnastics she finally seemed to be improving. She went nearly 3 days without a significant exacerbation of her condition. I believed that with my medical prowess, I would be mankind's next savior. As I told her yet again that she'd be discharged the next day, she thanked me and said she hoped not to be readmitted any time soon. Making small talk I learned that she had been a nurse at the very same hospital, where she had spent close to 40 years discharging her responsibilities. We didn't know smoking was bad back theneverybody did it, she said. I commiserated and assured her that she was well on her way to recovery.

The next day as I zealously sauntered into the hospital, I thought of the fantastic job I had done managing Ms. S's COPD. I mentally patted myself on the back; after all I was a smart guy. As I went into her room, I saw to my utter horror that she was in the midst of a severe COPD exacerbation. I swung into action, barked orders to start nebulizers, gave her a huge dose of steroids, and put her on a monitor. Through the clutter of nurses starting their IVs and the beeping of monitors, she said her time had come and she was ready to die. I gently chided her, assuring her this was just another episode, no different from her previous ones. She looked frail and tired, and her eyes appeared sad and forlorn. I departed from her room to call the resident in the ICU, where I thought surely she would be better served.

The unit unsurprisingly had no beds immediately available. She would be moved as soon as the unit had a bed; meanwhile, the pulmonary physician was on his way to see her. I dashed back to her room to check on her. To my dismay she was in respiratory distress, unable to talk and using all she had just to breathe. On the table next to the bed she had scribbled on a piece of scrap paper: Let me go please, it is OK. Knowing how exhausted she was, it must have taken superhuman effort to write this. She already had advance directives and was a DNR/DNI, but now she was precipitously declining in front of my eyes. Visibly trembling, I went to the nursing station and called her sister to apprise her of the waning of Ms. S's condition. I had never been faced with a scenario like this before. Taking time to compose myself, I wrote an order to put Ms. S on a morphine drip. All the while I couldn't shake off a sense of being ineffectual. Was modern medicine powerless to help people like her? I hoped I was doing the right thing. The pulmonary physician concurred with what I was doing, giving me the validation I was seeking.

Her sister arrived expeditiously, and I filled her in on what had happened. She nodded in understanding and stated her sister had always said when her time was up, she wanted to be let go. Her next question was the one I dreaded: How long do you think she has? I was evasive, reflecting my discomfiture at being totally unprepared in such situations. It is hard to tell, maybe 24 to 48 hours, but these things are hard to predict, I answered her. Writing orders for comfort measures, I couldn't help feeling unqualified to be a doctor. This wasn't something I thought I'd have to grapple with.

Soon all of Ms. S's relatives near and far came in to see her. They spent time at her bedside, but the morphine had taken effect. She was sleeping and looked comfortable but was unable to participate in conversation.

The next day as I arrived at the hospital, there was a cloud of dread in my mind. Ms. S probably had passed away some time during the night. At the nursing station, I was informed that the predictable hadn't happened. I went into her room to find it full of her loved ones. Her sister looked haggard but calm, and Ms. S was sleeping. I asked how things were, and Ms. S's sister's eyes lit up. She woke up at 1 a.m. and spoke to us for 2 hours. We were all able to tell her how much we loved her. She asked about all the children in the family and told us how much she loved them. Thereafter she went to sleep. She woke up at 6 a.m., looked around, and said, Why the hell am I still alive? She went to sleep soon and has been sleeping ever since. Laughter broke out in the room. I laughed with them. This was all a new experience. So Ms. S had closure before she died. Her family seemed content in this sad hour.

Ms. S passed away shortly after that. Her family thanked me for making her comfortable. But initially I felt I had done nothing much. Then the realization of my role as a doctor finally hit home. Sometimes it is apposite to hold a patient's hand and let the inevitable happen rather than fret on the shortcomings of medicine. It is all right not to overintellectualize every clinical event. After all it is all about treating people, not a constellation of bodily organs. Sometimes less is more and all that is needed. I always look back on this experience every time I feel smug. Since then I have incorporated discussion about goals of care into my repertoire. Patients with multiple admissions for incurable diseases need special attention. I hope my personal growth involves approaching such patients with the empathy and compassion they need. Having such discussions also helps to prevent the frenzied panicking that usually accompanies the inevitable decline of patients with terminal illness. The increasing emphasis of medical school curricula on end‐of‐life care issues reflects these sentiments. A growing geriatric cohort with multiple and often chronic medical diagnoses makes these skills indispensable.

My personal credo has been profoundly influenced by Sir Robert Hutchison's Physician's Prayer. I first read it in medical school but revisited it shortly after my experience with Ms S. To my mind, it is still relevant and serves to keep me on an even keel when making difficult decisions.

From inability to let well alone, from too much zeal for the new and contempt for what is old, from putting knowledge before wisdom, science before art and cleverness before common sense, from treating patients as cases and from making the cure of the disease more grievous than the endurance of the same, good Lord deliver us.

Sir Robert Hutchison (1871‐1960)

Print and Online Resources

1. Oxford Textbook of Palliative Medicine.

2. Center to Advance Palliative Care: www.capc.org.

3. American Academy of Hospice and Palliative Medicine: www.abhpm.org.

4. International Association for Hospice and Palliative Care: www.hospicecare.com.

It was the second week after finishing my internal medicine residency. What a daunting experience it was to be a newly appointed attending physician. My hospital rounds were painfully slow because I would consult the Tarsacon pharmacopoeia and Uptodate prior to writing any orders or making clinical decisions. During this keystone phase of my career I had the privilege of taking care of Ms. S. It has been almost 2 years since, and I still think of her and what I learned taking care of her.

Ms. S was a woman in her eighties with end‐stage chronic obstructive pulmonary disease (COPD). She was a frequent flyer, as evidenced by the multiple discharge summaries appended to her chart. Her hospital course was predictably punctuated by frequent inpatient exacerbations of COPD, and every time I told her that she'd be discharged the next day I had to eat my own words. After much effort and pharmaceutical gymnastics she finally seemed to be improving. She went nearly 3 days without a significant exacerbation of her condition. I believed that with my medical prowess, I would be mankind's next savior. As I told her yet again that she'd be discharged the next day, she thanked me and said she hoped not to be readmitted any time soon. Making small talk I learned that she had been a nurse at the very same hospital, where she had spent close to 40 years discharging her responsibilities. We didn't know smoking was bad back theneverybody did it, she said. I commiserated and assured her that she was well on her way to recovery.

The next day as I zealously sauntered into the hospital, I thought of the fantastic job I had done managing Ms. S's COPD. I mentally patted myself on the back; after all I was a smart guy. As I went into her room, I saw to my utter horror that she was in the midst of a severe COPD exacerbation. I swung into action, barked orders to start nebulizers, gave her a huge dose of steroids, and put her on a monitor. Through the clutter of nurses starting their IVs and the beeping of monitors, she said her time had come and she was ready to die. I gently chided her, assuring her this was just another episode, no different from her previous ones. She looked frail and tired, and her eyes appeared sad and forlorn. I departed from her room to call the resident in the ICU, where I thought surely she would be better served.

The unit unsurprisingly had no beds immediately available. She would be moved as soon as the unit had a bed; meanwhile, the pulmonary physician was on his way to see her. I dashed back to her room to check on her. To my dismay she was in respiratory distress, unable to talk and using all she had just to breathe. On the table next to the bed she had scribbled on a piece of scrap paper: Let me go please, it is OK. Knowing how exhausted she was, it must have taken superhuman effort to write this. She already had advance directives and was a DNR/DNI, but now she was precipitously declining in front of my eyes. Visibly trembling, I went to the nursing station and called her sister to apprise her of the waning of Ms. S's condition. I had never been faced with a scenario like this before. Taking time to compose myself, I wrote an order to put Ms. S on a morphine drip. All the while I couldn't shake off a sense of being ineffectual. Was modern medicine powerless to help people like her? I hoped I was doing the right thing. The pulmonary physician concurred with what I was doing, giving me the validation I was seeking.

Her sister arrived expeditiously, and I filled her in on what had happened. She nodded in understanding and stated her sister had always said when her time was up, she wanted to be let go. Her next question was the one I dreaded: How long do you think she has? I was evasive, reflecting my discomfiture at being totally unprepared in such situations. It is hard to tell, maybe 24 to 48 hours, but these things are hard to predict, I answered her. Writing orders for comfort measures, I couldn't help feeling unqualified to be a doctor. This wasn't something I thought I'd have to grapple with.

Soon all of Ms. S's relatives near and far came in to see her. They spent time at her bedside, but the morphine had taken effect. She was sleeping and looked comfortable but was unable to participate in conversation.

The next day as I arrived at the hospital, there was a cloud of dread in my mind. Ms. S probably had passed away some time during the night. At the nursing station, I was informed that the predictable hadn't happened. I went into her room to find it full of her loved ones. Her sister looked haggard but calm, and Ms. S was sleeping. I asked how things were, and Ms. S's sister's eyes lit up. She woke up at 1 a.m. and spoke to us for 2 hours. We were all able to tell her how much we loved her. She asked about all the children in the family and told us how much she loved them. Thereafter she went to sleep. She woke up at 6 a.m., looked around, and said, Why the hell am I still alive? She went to sleep soon and has been sleeping ever since. Laughter broke out in the room. I laughed with them. This was all a new experience. So Ms. S had closure before she died. Her family seemed content in this sad hour.

Ms. S passed away shortly after that. Her family thanked me for making her comfortable. But initially I felt I had done nothing much. Then the realization of my role as a doctor finally hit home. Sometimes it is apposite to hold a patient's hand and let the inevitable happen rather than fret on the shortcomings of medicine. It is all right not to overintellectualize every clinical event. After all it is all about treating people, not a constellation of bodily organs. Sometimes less is more and all that is needed. I always look back on this experience every time I feel smug. Since then I have incorporated discussion about goals of care into my repertoire. Patients with multiple admissions for incurable diseases need special attention. I hope my personal growth involves approaching such patients with the empathy and compassion they need. Having such discussions also helps to prevent the frenzied panicking that usually accompanies the inevitable decline of patients with terminal illness. The increasing emphasis of medical school curricula on end‐of‐life care issues reflects these sentiments. A growing geriatric cohort with multiple and often chronic medical diagnoses makes these skills indispensable.

My personal credo has been profoundly influenced by Sir Robert Hutchison's Physician's Prayer. I first read it in medical school but revisited it shortly after my experience with Ms S. To my mind, it is still relevant and serves to keep me on an even keel when making difficult decisions.

From inability to let well alone, from too much zeal for the new and contempt for what is old, from putting knowledge before wisdom, science before art and cleverness before common sense, from treating patients as cases and from making the cure of the disease more grievous than the endurance of the same, good Lord deliver us.

Sir Robert Hutchison (1871‐1960)

Print and Online Resources

1. Oxford Textbook of Palliative Medicine.

2. Center to Advance Palliative Care: www.capc.org.

3. American Academy of Hospice and Palliative Medicine: www.abhpm.org.

4. International Association for Hospice and Palliative Care: www.hospicecare.com.

It was the second week after finishing my internal medicine residency. What a daunting experience it was to be a newly appointed attending physician. My hospital rounds were painfully slow because I would consult the Tarsacon pharmacopoeia and Uptodate prior to writing any orders or making clinical decisions. During this keystone phase of my career I had the privilege of taking care of Ms. S. It has been almost 2 years since, and I still think of her and what I learned taking care of her.

Ms. S was a woman in her eighties with end‐stage chronic obstructive pulmonary disease (COPD). She was a frequent flyer, as evidenced by the multiple discharge summaries appended to her chart. Her hospital course was predictably punctuated by frequent inpatient exacerbations of COPD, and every time I told her that she'd be discharged the next day I had to eat my own words. After much effort and pharmaceutical gymnastics she finally seemed to be improving. She went nearly 3 days without a significant exacerbation of her condition. I believed that with my medical prowess, I would be mankind's next savior. As I told her yet again that she'd be discharged the next day, she thanked me and said she hoped not to be readmitted any time soon. Making small talk I learned that she had been a nurse at the very same hospital, where she had spent close to 40 years discharging her responsibilities. We didn't know smoking was bad back theneverybody did it, she said. I commiserated and assured her that she was well on her way to recovery.

The next day as I zealously sauntered into the hospital, I thought of the fantastic job I had done managing Ms. S's COPD. I mentally patted myself on the back; after all I was a smart guy. As I went into her room, I saw to my utter horror that she was in the midst of a severe COPD exacerbation. I swung into action, barked orders to start nebulizers, gave her a huge dose of steroids, and put her on a monitor. Through the clutter of nurses starting their IVs and the beeping of monitors, she said her time had come and she was ready to die. I gently chided her, assuring her this was just another episode, no different from her previous ones. She looked frail and tired, and her eyes appeared sad and forlorn. I departed from her room to call the resident in the ICU, where I thought surely she would be better served.

The unit unsurprisingly had no beds immediately available. She would be moved as soon as the unit had a bed; meanwhile, the pulmonary physician was on his way to see her. I dashed back to her room to check on her. To my dismay she was in respiratory distress, unable to talk and using all she had just to breathe. On the table next to the bed she had scribbled on a piece of scrap paper: Let me go please, it is OK. Knowing how exhausted she was, it must have taken superhuman effort to write this. She already had advance directives and was a DNR/DNI, but now she was precipitously declining in front of my eyes. Visibly trembling, I went to the nursing station and called her sister to apprise her of the waning of Ms. S's condition. I had never been faced with a scenario like this before. Taking time to compose myself, I wrote an order to put Ms. S on a morphine drip. All the while I couldn't shake off a sense of being ineffectual. Was modern medicine powerless to help people like her? I hoped I was doing the right thing. The pulmonary physician concurred with what I was doing, giving me the validation I was seeking.

Her sister arrived expeditiously, and I filled her in on what had happened. She nodded in understanding and stated her sister had always said when her time was up, she wanted to be let go. Her next question was the one I dreaded: How long do you think she has? I was evasive, reflecting my discomfiture at being totally unprepared in such situations. It is hard to tell, maybe 24 to 48 hours, but these things are hard to predict, I answered her. Writing orders for comfort measures, I couldn't help feeling unqualified to be a doctor. This wasn't something I thought I'd have to grapple with.

Soon all of Ms. S's relatives near and far came in to see her. They spent time at her bedside, but the morphine had taken effect. She was sleeping and looked comfortable but was unable to participate in conversation.

The next day as I arrived at the hospital, there was a cloud of dread in my mind. Ms. S probably had passed away some time during the night. At the nursing station, I was informed that the predictable hadn't happened. I went into her room to find it full of her loved ones. Her sister looked haggard but calm, and Ms. S was sleeping. I asked how things were, and Ms. S's sister's eyes lit up. She woke up at 1 a.m. and spoke to us for 2 hours. We were all able to tell her how much we loved her. She asked about all the children in the family and told us how much she loved them. Thereafter she went to sleep. She woke up at 6 a.m., looked around, and said, Why the hell am I still alive? She went to sleep soon and has been sleeping ever since. Laughter broke out in the room. I laughed with them. This was all a new experience. So Ms. S had closure before she died. Her family seemed content in this sad hour.

Ms. S passed away shortly after that. Her family thanked me for making her comfortable. But initially I felt I had done nothing much. Then the realization of my role as a doctor finally hit home. Sometimes it is apposite to hold a patient's hand and let the inevitable happen rather than fret on the shortcomings of medicine. It is all right not to overintellectualize every clinical event. After all it is all about treating people, not a constellation of bodily organs. Sometimes less is more and all that is needed. I always look back on this experience every time I feel smug. Since then I have incorporated discussion about goals of care into my repertoire. Patients with multiple admissions for incurable diseases need special attention. I hope my personal growth involves approaching such patients with the empathy and compassion they need. Having such discussions also helps to prevent the frenzied panicking that usually accompanies the inevitable decline of patients with terminal illness. The increasing emphasis of medical school curricula on end‐of‐life care issues reflects these sentiments. A growing geriatric cohort with multiple and often chronic medical diagnoses makes these skills indispensable.

My personal credo has been profoundly influenced by Sir Robert Hutchison's Physician's Prayer. I first read it in medical school but revisited it shortly after my experience with Ms S. To my mind, it is still relevant and serves to keep me on an even keel when making difficult decisions.

From inability to let well alone, from too much zeal for the new and contempt for what is old, from putting knowledge before wisdom, science before art and cleverness before common sense, from treating patients as cases and from making the cure of the disease more grievous than the endurance of the same, good Lord deliver us.

Sir Robert Hutchison (1871‐1960)

Print and Online Resources

1. Oxford Textbook of Palliative Medicine.

2. Center to Advance Palliative Care: www.capc.org.

3. American Academy of Hospice and Palliative Medicine: www.abhpm.org.

4. International Association for Hospice and Palliative Care: www.hospicecare.com.

As we approach the 6th year since the Institute of Medicine's Crossing the Quality Chasm offered a new vision for the American health care system, we still have a marked mismatch between the demand for health care quality and the supply of know‐how to deliver it. What the field of quality improvement (QI) still needs is merely this: QI practitioners in every care setting, a working vocabulary, a predictive framework for the mechanisms of reliable care, and rational therapies rigorously studied.13 Fortunately, the field of QI has attracted enough empiricistsworking in the lab of the hospital and other care settingsto lurch forward. But few would argue that we still have far less insight into the delivery of quality care than into the delivery of myocardial blood flow.

For ischemic heart disease we have classes of therapies, each of which is grounded in basic and clinical science: antiplatelets, beta‐blockers, vasodilators, lipid‐lowering agents. For care delivery we have the makings of analogous therapy classes, derived and introduced rather recently in a large review, Closing the Quality Gap: A Critical Analysis of the Quality Improvement Literature.4, 5 To facilitate their review of the evidence, the authors, including 2 prominent hospitalists, developed a new taxonomy of QI strategies (see Table 1). Though their effect size, relative efficacy, and interactions are not yet clear, many of these strategies can be applied to the inpatient setting, perhaps no less rationally than a well‐constructed antianginal regimen.

Where in the pathophysiology of a hospital do these QI therapies act? A plurality target the level of the provider: provider education, provider reminders, audit‐and‐feedback of provider performance, and facilitated relay of clinical data to providers. Remaining strategies target the patient (patient education, promotion of self‐management, and patient reminders), the immediate system within which care is delivered (organizational change), and the methodology of problem solving (continuous quality improvement). Only one strategy (financial incentives, regulation, and policy) fails to act directly at the level of the patient or provider, arguably the only level at which care actually can be improved.6

The value of the Quality Gap taxonomy is still largely untapped. If QI researchers and practitioners were to adopt its language as a standard, we could ramp up the power with which we communicate, interpret, and ultimately conduct improvement initiatives. In this issue of the Journal of Hospital Medicine, Cohn and colleagues profile a quality improvement initiative that achieved an impressive new level of performance. For an inpatient metric with a baseline institutional performance of 47%and an international benchmark of 39%the investigators executed a QI initiative that appears to have raised the rate of VTE prophylaxis to 85%.7 Despite a study design that weakens validity (beforeafter without controls) and a setting that diminishes applicability (medical patients in a single academic center), the authors have made a solid contribution to the QI literature simply by using the Quality Gap taxonomy. The authors specifically name and profile at least 3 distinct classes of QI strategies: provider education, a provider reminder element (ie, decision support), and an audit‐and‐feedback layer.

Even though provider education is unlikely to be sufficient as a lone QI strategya large review showed consistent but only modest benefitsit is often necessary.8 The provider education executed by Cohn and colleagues was frequent and regular. In the beginning of each month the chief resident oriented incoming house staff about venous thromboembolism (VTE) risk factors and the need for prophylaxis. They were given decision support pocket cards. Posters on display in nurse and physician work areas highlighted VTE risk factors. The provider education element also included discussions with the division chief about the topic. As robust as it was, however, the provider education was just a single component of the larger QI effort.

The second element, decision support, included VTE risk factor pocket cards with prophylaxis options listed. Introduced initially with the provider education, the pocket cards were handed out monthly by the chief resident. It is critical to recognize this decision support layer as a distinct core QI strategyand that it may even be fundamental to the success of the other strategies. Placed into the clinical workflow as a durable item, the decision support pocket card has the power to overcome provider uncertainty at moments of medical decision making. Generally speaking, a decision support layer, whether a pocket card, computer alert, or algorithm on a preprinted order form, can function as a shared baseline. Shared baselines or protocols reduce unnecessary variation in practice, a common source of poor quality care. Any mechanism that encourages groups of providers to deliver the same recommended care to groups of similarly at‐risk patients, while allowing customization of the protocol to meet the special needs of any individual patient, will have the net effect of raising overall quality of care.

This QI initiative may have achieved its greatest performance gainsas well as its greatest loss in terms of applicability to other settingsfrom its third facet, the audit‐and‐feedback layer. As a QI strategy audit‐and‐feedback has been defined as a summary of clinical performance for health care providers or institutions, performed for a specific period of time and reported either publicly or confidentially.1 It has demonstrated small to moderate benefits, with variations in effect most likely related to the format.9 As profiled in this study, it is hard to imagine a more powerful audit‐and‐feedback arrangement. The division chief of General Internal Medicine not only performed the audits, but also directly delivered the feedback to the house staff. In a deliberate, systematic, and successful way the investigators constructively used an existing authority gradient to leverage the Hawthorne effect, a change in worker behavior triggered by knowledge of being observed. Although it contributed to the impressive new VTE prophylaxis rates, this component did diminish generalizability and sustainability. Nonacademic centers may struggle to replicate these results, a point the authors dutifully point out. But even other academic centers might struggle in the absence of an authority figure with comparable influence and dedication to VTE prophylaxis. At the study hospital itself, similar rates of improvement would not be expected in patient populations outside the purview of the division chief.

Several alternatives to the before‐after study design could have produced richer information. Simultaneous data on VTE prophylaxis rates in a nonintervention population in the same or a similar hospital could have controlled for background or secular effects. An interrupted time series design may even have been feasible and could have provided more confidence in causality and more information on effect size. For example, what would be the effect on performance, if any, with removal of the decision support pocket card at 10 or 15 months? How much would performance rebound after its reintroduction? What could we have learned had the authors chosen instead to measure performance after sequentially introducing each component?

Using the language of the Quality Gap taxonomy, what conclusions can we draw from this improvement initiative? The introduction of a portable provider reminder (the decision support pocket card), when preceded by a program of provider education and followed by high‐intensity audit‐and‐feedback within an existing provider hierarchy, may have the power to raise VTE prophylaxis rates to 85% over an 18‐month period. With the large effect size somewhat mitigating the design flaws that weaken causality, we might risk an inference that these 3 classes of QI strategies can be reasonably successful in combination. But would we introduce them in our own medical centers? Using the clarity afforded by the taxonomy, we can identify several potential limitations, all attributable to the specifics of the audit‐and‐feedback arrangement: stringent preconditions of the practice setting, guaranteed inability to spread the initiative to other patient populations within the same medical center, limited scalability to include other QI projects, and reliance on the role of a single individual. Although clopidogrel 300 mg daily in the last week of every month is one way to pursue antiplatelet activity, other schedules or alternative agents may be preferable for the vast majority of patients.

The taxonomy can be used to compare, contrast, and more fully understand other QI studies. For example, among acutely ill medical inpatients not receiving VTE prophylaxis, Kucher and colleagues found that an electronic alert nearly doubled prophylaxis rates compared to those in a control group.10 Before trying to emulate their experience, a similarly equipped hospital would do well to recognize that the electronic alert was deployed as a composite of provider education, provider reminder, and facilitated relay of clinical information strategies, increasing prophylaxis rates for high risk patients from a baseline of 85% to 88%.

On the 21st‐century side of the quality chasm there is still something to be learned from QI research that falls short of recently proposed standards.3 This may be true as long as the key question remains: what are the mechanisms of reliable and sustainable performance improvement? We have not yet reached the day where a predictive framework, the clarity of our inquiry, the rigor of our study design, and the strength of our evidence churn out coherent answers. But we do have insights from a wealth of ongoing QI activity triggered by such forces as the Institute for Healthcare Improvement's 100,000 Lives Campaign and the advent of mandatory public reporting of hospital performance measures. By adding to this primordial mix the taxonomy offered by Closing the Quality Gap and its uptake into our vernacular by reports such as the one by Cohn in this issue of the Journal of Hospital Medicine, we are acquiring the language and experience to conduct intelligent and intelligible QI research.

As we approach the 6th year since the Institute of Medicine's Crossing the Quality Chasm offered a new vision for the American health care system, we still have a marked mismatch between the demand for health care quality and the supply of know‐how to deliver it. What the field of quality improvement (QI) still needs is merely this: QI practitioners in every care setting, a working vocabulary, a predictive framework for the mechanisms of reliable care, and rational therapies rigorously studied.13 Fortunately, the field of QI has attracted enough empiricistsworking in the lab of the hospital and other care settingsto lurch forward. But few would argue that we still have far less insight into the delivery of quality care than into the delivery of myocardial blood flow.

For ischemic heart disease we have classes of therapies, each of which is grounded in basic and clinical science: antiplatelets, beta‐blockers, vasodilators, lipid‐lowering agents. For care delivery we have the makings of analogous therapy classes, derived and introduced rather recently in a large review, Closing the Quality Gap: A Critical Analysis of the Quality Improvement Literature.4, 5 To facilitate their review of the evidence, the authors, including 2 prominent hospitalists, developed a new taxonomy of QI strategies (see Table 1). Though their effect size, relative efficacy, and interactions are not yet clear, many of these strategies can be applied to the inpatient setting, perhaps no less rationally than a well‐constructed antianginal regimen.

Where in the pathophysiology of a hospital do these QI therapies act? A plurality target the level of the provider: provider education, provider reminders, audit‐and‐feedback of provider performance, and facilitated relay of clinical data to providers. Remaining strategies target the patient (patient education, promotion of self‐management, and patient reminders), the immediate system within which care is delivered (organizational change), and the methodology of problem solving (continuous quality improvement). Only one strategy (financial incentives, regulation, and policy) fails to act directly at the level of the patient or provider, arguably the only level at which care actually can be improved.6

The value of the Quality Gap taxonomy is still largely untapped. If QI researchers and practitioners were to adopt its language as a standard, we could ramp up the power with which we communicate, interpret, and ultimately conduct improvement initiatives. In this issue of the Journal of Hospital Medicine, Cohn and colleagues profile a quality improvement initiative that achieved an impressive new level of performance. For an inpatient metric with a baseline institutional performance of 47%and an international benchmark of 39%the investigators executed a QI initiative that appears to have raised the rate of VTE prophylaxis to 85%.7 Despite a study design that weakens validity (beforeafter without controls) and a setting that diminishes applicability (medical patients in a single academic center), the authors have made a solid contribution to the QI literature simply by using the Quality Gap taxonomy. The authors specifically name and profile at least 3 distinct classes of QI strategies: provider education, a provider reminder element (ie, decision support), and an audit‐and‐feedback layer.

Even though provider education is unlikely to be sufficient as a lone QI strategya large review showed consistent but only modest benefitsit is often necessary.8 The provider education executed by Cohn and colleagues was frequent and regular. In the beginning of each month the chief resident oriented incoming house staff about venous thromboembolism (VTE) risk factors and the need for prophylaxis. They were given decision support pocket cards. Posters on display in nurse and physician work areas highlighted VTE risk factors. The provider education element also included discussions with the division chief about the topic. As robust as it was, however, the provider education was just a single component of the larger QI effort.

The second element, decision support, included VTE risk factor pocket cards with prophylaxis options listed. Introduced initially with the provider education, the pocket cards were handed out monthly by the chief resident. It is critical to recognize this decision support layer as a distinct core QI strategyand that it may even be fundamental to the success of the other strategies. Placed into the clinical workflow as a durable item, the decision support pocket card has the power to overcome provider uncertainty at moments of medical decision making. Generally speaking, a decision support layer, whether a pocket card, computer alert, or algorithm on a preprinted order form, can function as a shared baseline. Shared baselines or protocols reduce unnecessary variation in practice, a common source of poor quality care. Any mechanism that encourages groups of providers to deliver the same recommended care to groups of similarly at‐risk patients, while allowing customization of the protocol to meet the special needs of any individual patient, will have the net effect of raising overall quality of care.

This QI initiative may have achieved its greatest performance gainsas well as its greatest loss in terms of applicability to other settingsfrom its third facet, the audit‐and‐feedback layer. As a QI strategy audit‐and‐feedback has been defined as a summary of clinical performance for health care providers or institutions, performed for a specific period of time and reported either publicly or confidentially.1 It has demonstrated small to moderate benefits, with variations in effect most likely related to the format.9 As profiled in this study, it is hard to imagine a more powerful audit‐and‐feedback arrangement. The division chief of General Internal Medicine not only performed the audits, but also directly delivered the feedback to the house staff. In a deliberate, systematic, and successful way the investigators constructively used an existing authority gradient to leverage the Hawthorne effect, a change in worker behavior triggered by knowledge of being observed. Although it contributed to the impressive new VTE prophylaxis rates, this component did diminish generalizability and sustainability. Nonacademic centers may struggle to replicate these results, a point the authors dutifully point out. But even other academic centers might struggle in the absence of an authority figure with comparable influence and dedication to VTE prophylaxis. At the study hospital itself, similar rates of improvement would not be expected in patient populations outside the purview of the division chief.

Several alternatives to the before‐after study design could have produced richer information. Simultaneous data on VTE prophylaxis rates in a nonintervention population in the same or a similar hospital could have controlled for background or secular effects. An interrupted time series design may even have been feasible and could have provided more confidence in causality and more information on effect size. For example, what would be the effect on performance, if any, with removal of the decision support pocket card at 10 or 15 months? How much would performance rebound after its reintroduction? What could we have learned had the authors chosen instead to measure performance after sequentially introducing each component?

Using the language of the Quality Gap taxonomy, what conclusions can we draw from this improvement initiative? The introduction of a portable provider reminder (the decision support pocket card), when preceded by a program of provider education and followed by high‐intensity audit‐and‐feedback within an existing provider hierarchy, may have the power to raise VTE prophylaxis rates to 85% over an 18‐month period. With the large effect size somewhat mitigating the design flaws that weaken causality, we might risk an inference that these 3 classes of QI strategies can be reasonably successful in combination. But would we introduce them in our own medical centers? Using the clarity afforded by the taxonomy, we can identify several potential limitations, all attributable to the specifics of the audit‐and‐feedback arrangement: stringent preconditions of the practice setting, guaranteed inability to spread the initiative to other patient populations within the same medical center, limited scalability to include other QI projects, and reliance on the role of a single individual. Although clopidogrel 300 mg daily in the last week of every month is one way to pursue antiplatelet activity, other schedules or alternative agents may be preferable for the vast majority of patients.

The taxonomy can be used to compare, contrast, and more fully understand other QI studies. For example, among acutely ill medical inpatients not receiving VTE prophylaxis, Kucher and colleagues found that an electronic alert nearly doubled prophylaxis rates compared to those in a control group.10 Before trying to emulate their experience, a similarly equipped hospital would do well to recognize that the electronic alert was deployed as a composite of provider education, provider reminder, and facilitated relay of clinical information strategies, increasing prophylaxis rates for high risk patients from a baseline of 85% to 88%.

On the 21st‐century side of the quality chasm there is still something to be learned from QI research that falls short of recently proposed standards.3 This may be true as long as the key question remains: what are the mechanisms of reliable and sustainable performance improvement? We have not yet reached the day where a predictive framework, the clarity of our inquiry, the rigor of our study design, and the strength of our evidence churn out coherent answers. But we do have insights from a wealth of ongoing QI activity triggered by such forces as the Institute for Healthcare Improvement's 100,000 Lives Campaign and the advent of mandatory public reporting of hospital performance measures. By adding to this primordial mix the taxonomy offered by Closing the Quality Gap and its uptake into our vernacular by reports such as the one by Cohn in this issue of the Journal of Hospital Medicine, we are acquiring the language and experience to conduct intelligent and intelligible QI research.

As we approach the 6th year since the Institute of Medicine's Crossing the Quality Chasm offered a new vision for the American health care system, we still have a marked mismatch between the demand for health care quality and the supply of know‐how to deliver it. What the field of quality improvement (QI) still needs is merely this: QI practitioners in every care setting, a working vocabulary, a predictive framework for the mechanisms of reliable care, and rational therapies rigorously studied.13 Fortunately, the field of QI has attracted enough empiricistsworking in the lab of the hospital and other care settingsto lurch forward. But few would argue that we still have far less insight into the delivery of quality care than into the delivery of myocardial blood flow.

For ischemic heart disease we have classes of therapies, each of which is grounded in basic and clinical science: antiplatelets, beta‐blockers, vasodilators, lipid‐lowering agents. For care delivery we have the makings of analogous therapy classes, derived and introduced rather recently in a large review, Closing the Quality Gap: A Critical Analysis of the Quality Improvement Literature.4, 5 To facilitate their review of the evidence, the authors, including 2 prominent hospitalists, developed a new taxonomy of QI strategies (see Table 1). Though their effect size, relative efficacy, and interactions are not yet clear, many of these strategies can be applied to the inpatient setting, perhaps no less rationally than a well‐constructed antianginal regimen.

Where in the pathophysiology of a hospital do these QI therapies act? A plurality target the level of the provider: provider education, provider reminders, audit‐and‐feedback of provider performance, and facilitated relay of clinical data to providers. Remaining strategies target the patient (patient education, promotion of self‐management, and patient reminders), the immediate system within which care is delivered (organizational change), and the methodology of problem solving (continuous quality improvement). Only one strategy (financial incentives, regulation, and policy) fails to act directly at the level of the patient or provider, arguably the only level at which care actually can be improved.6

The value of the Quality Gap taxonomy is still largely untapped. If QI researchers and practitioners were to adopt its language as a standard, we could ramp up the power with which we communicate, interpret, and ultimately conduct improvement initiatives. In this issue of the Journal of Hospital Medicine, Cohn and colleagues profile a quality improvement initiative that achieved an impressive new level of performance. For an inpatient metric with a baseline institutional performance of 47%and an international benchmark of 39%the investigators executed a QI initiative that appears to have raised the rate of VTE prophylaxis to 85%.7 Despite a study design that weakens validity (beforeafter without controls) and a setting that diminishes applicability (medical patients in a single academic center), the authors have made a solid contribution to the QI literature simply by using the Quality Gap taxonomy. The authors specifically name and profile at least 3 distinct classes of QI strategies: provider education, a provider reminder element (ie, decision support), and an audit‐and‐feedback layer.

Even though provider education is unlikely to be sufficient as a lone QI strategya large review showed consistent but only modest benefitsit is often necessary.8 The provider education executed by Cohn and colleagues was frequent and regular. In the beginning of each month the chief resident oriented incoming house staff about venous thromboembolism (VTE) risk factors and the need for prophylaxis. They were given decision support pocket cards. Posters on display in nurse and physician work areas highlighted VTE risk factors. The provider education element also included discussions with the division chief about the topic. As robust as it was, however, the provider education was just a single component of the larger QI effort.

The second element, decision support, included VTE risk factor pocket cards with prophylaxis options listed. Introduced initially with the provider education, the pocket cards were handed out monthly by the chief resident. It is critical to recognize this decision support layer as a distinct core QI strategyand that it may even be fundamental to the success of the other strategies. Placed into the clinical workflow as a durable item, the decision support pocket card has the power to overcome provider uncertainty at moments of medical decision making. Generally speaking, a decision support layer, whether a pocket card, computer alert, or algorithm on a preprinted order form, can function as a shared baseline. Shared baselines or protocols reduce unnecessary variation in practice, a common source of poor quality care. Any mechanism that encourages groups of providers to deliver the same recommended care to groups of similarly at‐risk patients, while allowing customization of the protocol to meet the special needs of any individual patient, will have the net effect of raising overall quality of care.

This QI initiative may have achieved its greatest performance gainsas well as its greatest loss in terms of applicability to other settingsfrom its third facet, the audit‐and‐feedback layer. As a QI strategy audit‐and‐feedback has been defined as a summary of clinical performance for health care providers or institutions, performed for a specific period of time and reported either publicly or confidentially.1 It has demonstrated small to moderate benefits, with variations in effect most likely related to the format.9 As profiled in this study, it is hard to imagine a more powerful audit‐and‐feedback arrangement. The division chief of General Internal Medicine not only performed the audits, but also directly delivered the feedback to the house staff. In a deliberate, systematic, and successful way the investigators constructively used an existing authority gradient to leverage the Hawthorne effect, a change in worker behavior triggered by knowledge of being observed. Although it contributed to the impressive new VTE prophylaxis rates, this component did diminish generalizability and sustainability. Nonacademic centers may struggle to replicate these results, a point the authors dutifully point out. But even other academic centers might struggle in the absence of an authority figure with comparable influence and dedication to VTE prophylaxis. At the study hospital itself, similar rates of improvement would not be expected in patient populations outside the purview of the division chief.

Several alternatives to the before‐after study design could have produced richer information. Simultaneous data on VTE prophylaxis rates in a nonintervention population in the same or a similar hospital could have controlled for background or secular effects. An interrupted time series design may even have been feasible and could have provided more confidence in causality and more information on effect size. For example, what would be the effect on performance, if any, with removal of the decision support pocket card at 10 or 15 months? How much would performance rebound after its reintroduction? What could we have learned had the authors chosen instead to measure performance after sequentially introducing each component?

Using the language of the Quality Gap taxonomy, what conclusions can we draw from this improvement initiative? The introduction of a portable provider reminder (the decision support pocket card), when preceded by a program of provider education and followed by high‐intensity audit‐and‐feedback within an existing provider hierarchy, may have the power to raise VTE prophylaxis rates to 85% over an 18‐month period. With the large effect size somewhat mitigating the design flaws that weaken causality, we might risk an inference that these 3 classes of QI strategies can be reasonably successful in combination. But would we introduce them in our own medical centers? Using the clarity afforded by the taxonomy, we can identify several potential limitations, all attributable to the specifics of the audit‐and‐feedback arrangement: stringent preconditions of the practice setting, guaranteed inability to spread the initiative to other patient populations within the same medical center, limited scalability to include other QI projects, and reliance on the role of a single individual. Although clopidogrel 300 mg daily in the last week of every month is one way to pursue antiplatelet activity, other schedules or alternative agents may be preferable for the vast majority of patients.

The taxonomy can be used to compare, contrast, and more fully understand other QI studies. For example, among acutely ill medical inpatients not receiving VTE prophylaxis, Kucher and colleagues found that an electronic alert nearly doubled prophylaxis rates compared to those in a control group.10 Before trying to emulate their experience, a similarly equipped hospital would do well to recognize that the electronic alert was deployed as a composite of provider education, provider reminder, and facilitated relay of clinical information strategies, increasing prophylaxis rates for high risk patients from a baseline of 85% to 88%.

On the 21st‐century side of the quality chasm there is still something to be learned from QI research that falls short of recently proposed standards.3 This may be true as long as the key question remains: what are the mechanisms of reliable and sustainable performance improvement? We have not yet reached the day where a predictive framework, the clarity of our inquiry, the rigor of our study design, and the strength of our evidence churn out coherent answers. But we do have insights from a wealth of ongoing QI activity triggered by such forces as the Institute for Healthcare Improvement's 100,000 Lives Campaign and the advent of mandatory public reporting of hospital performance measures. By adding to this primordial mix the taxonomy offered by Closing the Quality Gap and its uptake into our vernacular by reports such as the one by Cohn in this issue of the Journal of Hospital Medicine, we are acquiring the language and experience to conduct intelligent and intelligible QI research.