The application of Ockham's razor, or the law of parsimony, to clinical reasoning implies selecting the competing hypothesis that makes the fewest new assumptions based on known factors. Thus, the prevailing hypothesis when confronting a patient with a bilateral pleural effusion would be that a single disease likely explains the accumulation of pleural fluid on both sides. Although the principle of diagnostic parsimony has become axiomatic for the differential diagnosis of diseases, it might not hold true for all cases. That is, the counterpart to Ockham's razor, known as Hickam's dictum, states that patients can have as many diseases as they damn well please. An example is Contarini's condition. Francesco Contarini (1556‐1624) died 1 year after he became the 95th Doge of Venice. The postmortem study revealed a right pleural effusion, probably due to heart failure and a contralateral empyema.1 Since then, it became apparent that bilateral pleural effusions might have more than a single explanation. To improve knowledge of this entity, coined as Contarini's condition by Kutty and Varkey in 1978,2 we systematically searched for cases from a large prospectively maintained pleural fluid database at the Arnau de Vilanova University Hospital (Lleida, Spain), a 470‐bed general medical center serving a population of 400,000 inhabitants. An analysis of previously documented cases was also performed.

Information was gathered from all consecutive patients who have undergone pleural fluid aspiration and analysis during the last 16 years at our institution. Medical records were screened of those patients submitted to bilateral thoracentesis, during a single hospital admission, that resulted in pleural fluids with markedly different biochemical characteristics. Written informed consent was obtained from all patients to use their clinical data in future investigations. The local ethics committee approved this study. In addition, the Embase and PubMed databases were searched using the keywords Contarini's condition, Contarini's syndrome, and bilateral pleural effusion to identify all previously reported cases. Pleural effusion etiology and definition of transudate/exudate were established by standard criteria. Specifically, complicated parapneumonic effusion referred to those pneumonia‐associated non‐purulent effusions that needed a tube thoracostomy for resolution.

Of 2605 patients from our database, 546 (21%) had bilateral pleural effusions, mostly due to heart failure (286 patients) and malignancy (102 patients). There were only 5 (0.9%) patients who had bilateral effusions of different etiologies which, added to an additional 7 patients identified via literature review,1‐7 totaled 12 cases. Their characteristics are summarized in Table 1. However, it should be noted that 4 of the 7 previously reported cases were described as the concurrence of chylothorax and malignant effusion.3‐5, 7 This combination may result from a common causative factor (ie, lymphoma or metastatic carcinoma), thus bringing into question their status as valid examples of Contarini's condition. Aside from these cases, bacterial infections (ie, parapneumonics and empyema) represent the most common coexisting disease in Contarini's cases, particularly in association with heart failure (50% of the cases). The reason behind this is that pneumonia may precipitate an acute decompensation of heart failure. In a recent study, 7.4% of 33,130 patients developed heart failure during hospitalization for pneumonia.8

Kalomenidis et al. studied 27 patients with bilateral pleural effusions who underwent bilateral thoracentesis to determine if the findings were the same.9 They found that the main biochemical and cellular features on both sides were generally similar, except for 2 (7.5%) cases which had significantly different pleural fluid lactate dehydrogenase (LDH) levels. Although a plausible explanation for the latter was not given, this circumstance did not change the categorization of the effusions. The authors concluded that bilateral diagnostic thoracenteses were not necessary unless there was a specific clinical indication. The fact that most patients with bilateral pleural effusions are submitted to a unilateral thoracentesis may have resulted in an underestimation of the current prevalence of Contarini's syndrome. In our series, differing lung and pleural computed tomographic (CT) imaging characteristics between both hemithoraces was the primary reason for performing bilateral pleural taps in all 5 cases. After the dual diagnosis, the corresponding patients benefited from an additional therapeutic intervention, mainly treatment for heart failure. Therefore, the rationale to exceptionally consider a bilateral diagnostic thoracentesis is to avoid missing significant pathology by sampling the wrong pleural space (in particular, one caused by heart failure) and thus failing to properly diagnose contralateral exudative effusion with an attendant serious etiology.

In conclusion, Contarini's syndrome is a rare and distinct entity, but probably underdiagnosed. Although a bilateral pleural fluid aspiration is seldom justified in routine clinical practice, it should be considered if any of the following are met: unilateral parenchymal lung involvement, significantly disparate‐sized effusions, markedly different attenuation values (Hounsfield units) or appearance (eg, unilateral pleural loculations or enhancement) on CT, atypical clinical findings (fever or pleuritic chest pain in the context of decompensated heart failure), resolution of pleural effusion only on 1 side, and the diagnosis of pleural diseases usually associated with unilateral effusions (eg, pneumonia). However, it should be stressed that these are expert, rather than evidence‐based, recommendations.

The application of Ockham's razor, or the law of parsimony, to clinical reasoning implies selecting the competing hypothesis that makes the fewest new assumptions based on known factors. Thus, the prevailing hypothesis when confronting a patient with a bilateral pleural effusion would be that a single disease likely explains the accumulation of pleural fluid on both sides. Although the principle of diagnostic parsimony has become axiomatic for the differential diagnosis of diseases, it might not hold true for all cases. That is, the counterpart to Ockham's razor, known as Hickam's dictum, states that patients can have as many diseases as they damn well please. An example is Contarini's condition. Francesco Contarini (1556‐1624) died 1 year after he became the 95th Doge of Venice. The postmortem study revealed a right pleural effusion, probably due to heart failure and a contralateral empyema.1 Since then, it became apparent that bilateral pleural effusions might have more than a single explanation. To improve knowledge of this entity, coined as Contarini's condition by Kutty and Varkey in 1978,2 we systematically searched for cases from a large prospectively maintained pleural fluid database at the Arnau de Vilanova University Hospital (Lleida, Spain), a 470‐bed general medical center serving a population of 400,000 inhabitants. An analysis of previously documented cases was also performed.

Information was gathered from all consecutive patients who have undergone pleural fluid aspiration and analysis during the last 16 years at our institution. Medical records were screened of those patients submitted to bilateral thoracentesis, during a single hospital admission, that resulted in pleural fluids with markedly different biochemical characteristics. Written informed consent was obtained from all patients to use their clinical data in future investigations. The local ethics committee approved this study. In addition, the Embase and PubMed databases were searched using the keywords Contarini's condition, Contarini's syndrome, and bilateral pleural effusion to identify all previously reported cases. Pleural effusion etiology and definition of transudate/exudate were established by standard criteria. Specifically, complicated parapneumonic effusion referred to those pneumonia‐associated non‐purulent effusions that needed a tube thoracostomy for resolution.

Of 2605 patients from our database, 546 (21%) had bilateral pleural effusions, mostly due to heart failure (286 patients) and malignancy (102 patients). There were only 5 (0.9%) patients who had bilateral effusions of different etiologies which, added to an additional 7 patients identified via literature review,1‐7 totaled 12 cases. Their characteristics are summarized in Table 1. However, it should be noted that 4 of the 7 previously reported cases were described as the concurrence of chylothorax and malignant effusion.3‐5, 7 This combination may result from a common causative factor (ie, lymphoma or metastatic carcinoma), thus bringing into question their status as valid examples of Contarini's condition. Aside from these cases, bacterial infections (ie, parapneumonics and empyema) represent the most common coexisting disease in Contarini's cases, particularly in association with heart failure (50% of the cases). The reason behind this is that pneumonia may precipitate an acute decompensation of heart failure. In a recent study, 7.4% of 33,130 patients developed heart failure during hospitalization for pneumonia.8

Kalomenidis et al. studied 27 patients with bilateral pleural effusions who underwent bilateral thoracentesis to determine if the findings were the same.9 They found that the main biochemical and cellular features on both sides were generally similar, except for 2 (7.5%) cases which had significantly different pleural fluid lactate dehydrogenase (LDH) levels. Although a plausible explanation for the latter was not given, this circumstance did not change the categorization of the effusions. The authors concluded that bilateral diagnostic thoracenteses were not necessary unless there was a specific clinical indication. The fact that most patients with bilateral pleural effusions are submitted to a unilateral thoracentesis may have resulted in an underestimation of the current prevalence of Contarini's syndrome. In our series, differing lung and pleural computed tomographic (CT) imaging characteristics between both hemithoraces was the primary reason for performing bilateral pleural taps in all 5 cases. After the dual diagnosis, the corresponding patients benefited from an additional therapeutic intervention, mainly treatment for heart failure. Therefore, the rationale to exceptionally consider a bilateral diagnostic thoracentesis is to avoid missing significant pathology by sampling the wrong pleural space (in particular, one caused by heart failure) and thus failing to properly diagnose contralateral exudative effusion with an attendant serious etiology.

In conclusion, Contarini's syndrome is a rare and distinct entity, but probably underdiagnosed. Although a bilateral pleural fluid aspiration is seldom justified in routine clinical practice, it should be considered if any of the following are met: unilateral parenchymal lung involvement, significantly disparate‐sized effusions, markedly different attenuation values (Hounsfield units) or appearance (eg, unilateral pleural loculations or enhancement) on CT, atypical clinical findings (fever or pleuritic chest pain in the context of decompensated heart failure), resolution of pleural effusion only on 1 side, and the diagnosis of pleural diseases usually associated with unilateral effusions (eg, pneumonia). However, it should be stressed that these are expert, rather than evidence‐based, recommendations.

The application of Ockham's razor, or the law of parsimony, to clinical reasoning implies selecting the competing hypothesis that makes the fewest new assumptions based on known factors. Thus, the prevailing hypothesis when confronting a patient with a bilateral pleural effusion would be that a single disease likely explains the accumulation of pleural fluid on both sides. Although the principle of diagnostic parsimony has become axiomatic for the differential diagnosis of diseases, it might not hold true for all cases. That is, the counterpart to Ockham's razor, known as Hickam's dictum, states that patients can have as many diseases as they damn well please. An example is Contarini's condition. Francesco Contarini (1556‐1624) died 1 year after he became the 95th Doge of Venice. The postmortem study revealed a right pleural effusion, probably due to heart failure and a contralateral empyema.1 Since then, it became apparent that bilateral pleural effusions might have more than a single explanation. To improve knowledge of this entity, coined as Contarini's condition by Kutty and Varkey in 1978,2 we systematically searched for cases from a large prospectively maintained pleural fluid database at the Arnau de Vilanova University Hospital (Lleida, Spain), a 470‐bed general medical center serving a population of 400,000 inhabitants. An analysis of previously documented cases was also performed.

Information was gathered from all consecutive patients who have undergone pleural fluid aspiration and analysis during the last 16 years at our institution. Medical records were screened of those patients submitted to bilateral thoracentesis, during a single hospital admission, that resulted in pleural fluids with markedly different biochemical characteristics. Written informed consent was obtained from all patients to use their clinical data in future investigations. The local ethics committee approved this study. In addition, the Embase and PubMed databases were searched using the keywords Contarini's condition, Contarini's syndrome, and bilateral pleural effusion to identify all previously reported cases. Pleural effusion etiology and definition of transudate/exudate were established by standard criteria. Specifically, complicated parapneumonic effusion referred to those pneumonia‐associated non‐purulent effusions that needed a tube thoracostomy for resolution.

Of 2605 patients from our database, 546 (21%) had bilateral pleural effusions, mostly due to heart failure (286 patients) and malignancy (102 patients). There were only 5 (0.9%) patients who had bilateral effusions of different etiologies which, added to an additional 7 patients identified via literature review,1‐7 totaled 12 cases. Their characteristics are summarized in Table 1. However, it should be noted that 4 of the 7 previously reported cases were described as the concurrence of chylothorax and malignant effusion.3‐5, 7 This combination may result from a common causative factor (ie, lymphoma or metastatic carcinoma), thus bringing into question their status as valid examples of Contarini's condition. Aside from these cases, bacterial infections (ie, parapneumonics and empyema) represent the most common coexisting disease in Contarini's cases, particularly in association with heart failure (50% of the cases). The reason behind this is that pneumonia may precipitate an acute decompensation of heart failure. In a recent study, 7.4% of 33,130 patients developed heart failure during hospitalization for pneumonia.8

Kalomenidis et al. studied 27 patients with bilateral pleural effusions who underwent bilateral thoracentesis to determine if the findings were the same.9 They found that the main biochemical and cellular features on both sides were generally similar, except for 2 (7.5%) cases which had significantly different pleural fluid lactate dehydrogenase (LDH) levels. Although a plausible explanation for the latter was not given, this circumstance did not change the categorization of the effusions. The authors concluded that bilateral diagnostic thoracenteses were not necessary unless there was a specific clinical indication. The fact that most patients with bilateral pleural effusions are submitted to a unilateral thoracentesis may have resulted in an underestimation of the current prevalence of Contarini's syndrome. In our series, differing lung and pleural computed tomographic (CT) imaging characteristics between both hemithoraces was the primary reason for performing bilateral pleural taps in all 5 cases. After the dual diagnosis, the corresponding patients benefited from an additional therapeutic intervention, mainly treatment for heart failure. Therefore, the rationale to exceptionally consider a bilateral diagnostic thoracentesis is to avoid missing significant pathology by sampling the wrong pleural space (in particular, one caused by heart failure) and thus failing to properly diagnose contralateral exudative effusion with an attendant serious etiology.

In conclusion, Contarini's syndrome is a rare and distinct entity, but probably underdiagnosed. Although a bilateral pleural fluid aspiration is seldom justified in routine clinical practice, it should be considered if any of the following are met: unilateral parenchymal lung involvement, significantly disparate‐sized effusions, markedly different attenuation values (Hounsfield units) or appearance (eg, unilateral pleural loculations or enhancement) on CT, atypical clinical findings (fever or pleuritic chest pain in the context of decompensated heart failure), resolution of pleural effusion only on 1 side, and the diagnosis of pleural diseases usually associated with unilateral effusions (eg, pneumonia). However, it should be stressed that these are expert, rather than evidence‐based, recommendations.

Over the past decade, hospital medicine has been the nation's fastest‐growing medical specialty. According to the American Hospital Association's (AHA) 2009 survey, 58% of United States (US) hospitals now have hospital medicine programs, and for hospitals with 200 or more beds, this figure is 89%.1 In 2009, the AHA estimated that the number of US hospitalists would increase to over 34,000 by 2011, over double that of the 16,000 present in 2005.1 Studies demonstrate that, compared to a system where primary care physicians provide inpatient care, the hospitalist model improves efficiency while maintaining at least equal patient outcomes.211 However, scant data exist as to the effects of hospitalists on patient satisfaction.12 Understanding how care models affect patient experience is vital in the current environment of healthcare reform and performance reporting, especially in light of the Centers for Medicare and Medicaid Services' (CMS) efforts to link the patient experience to reimbursement through value‐based purchasing.13 Value‐based purchasing is a strategy to encourage and reward excellence in healthcare delivery through differential reimbursement based on defined performance measures. As one part of value‐based purchasing, hospital reimbursement will be linked to patient‐experience measures, including patient ratings of their doctor's ability to communicate with them and other questions assessing patient satisfaction with their hospital stay.14

In the outpatient setting, trust is the variable most strongly associated with patient satisfaction.1518 In contrast to PCPs, who may develop relationships with patients over years, hospitalists often first meet a patient in the hospital and must engender trust quickly. In addition, hospitalists work in shifts and may not be responsible for the same patients each day. Since continuity is positively related to trust,19, 20 there is reason to believe satisfaction with hospitalist care might be lower than satisfaction with care provided by PCPs. We report on 8295 patients and 6 years experience with hospitalist programs at 3 hospitals. Based on the known link between continuity and patient satisfaction, we hypothesized that patient satisfaction would be lower with hospitalists than with primary care internists.

METHODS

RESULTS

Of patients who were reached by telephone, 87% agreed to participate in the hospital survey. However, most patients could not be reached by phone; thus our estimated response rate, including those who could not be reached, was 27%. For the subset of patients interviewed using the HCAHPS protocol, the response rate was 40%. Our final sample included 8295 patients (3597 cared for by 59 hospitalists and 4698 by 288 PCPs) interviewed between 2003 and 2009. Three‐quarters of the patients were from the tertiary care center, whereas 17% and 8% were from each of the community hospitals (see Supporting Appendix B in the online version of this article). Patient characteristics appear in Table 1. Patients cared for by hospitalists were similar to those cared for by PCPs in terms of age, sex, marital status, education, and language, but hospitalist patients were more likely to have been admitted through the emergency department (93% vs 84%, P < 0.001) and less likely to be white (83% vs 85%, P = 0.01). Patients cared for by hospitalists also had higher average illness severity score (2.2 0.8 vs 2.0 0.8, P < 0.001), longer average LOS (4.3 4.3 vs 4.0 3.6, P < 0.001), and lower mean perceived health score (2.8 1.2 vs 3.0 1.2, P = 0.01).

Unadjusted patient reported satisfaction with physician care quality was slightly greater for PCPs than hospitalists (4.25 vs 4.19, P = 0.009). After multivariable adjustment, the difference was attenuated but persisted (4.24 vs 4.20, P = 0.04). We found no statistical difference among the hospitals or the specific hospitalist groups in terms of satisfaction with overall physician care quality (Figure 1). There were no statistical differences in patient satisfaction ratings of hospitalist and PCPs for the subdomains of behavior, pain, and communication (Table 2). There were also no differences in the proportion of patients cared for by hospitalists or PCPs who rated their physicians in the highest satisfaction category (79% vs 81%, P = 0.17) or the lowest (5% vs 5%, P = 0.19). Among patients cared for by academic hospitalists, there was no difference in satisfaction rating between those patients who had a designated primary care physician in the outpatient setting and those who did not (4.22 0.94 vs 4.19 0.94, P = 0.97). Finally, satisfaction with both hospitalists and PCPs showed equivalent rates of improvement over time (Figure 2).

Figure 1

Figure 2

DISCUSSION

In this observational study of over 8200 patients cared for over 6 years by 347 physicians at 3 hospitals, we found that patient satisfaction with inpatient care provided by hospitalists and primary care doctors was almost identical. As we hypothesized, overall satisfaction with physician care quality, our primary outcome, was slightly greater with primary care doctors; however, the observed difference, 0.04 on a scale of 1 to 5, cannot be considered clinically significant. All patients were generally satisfied (4.2‐4.3 rating on 5‐point scale) with their inpatient care, and satisfaction scores increased over time. We also found no differences among the specific domains of satisfaction, including communication skills, pain control, and physician behavior. Finally, we found no significant difference in patient satisfaction with physician care quality among the different hospitalist services.

Previous studies of patient satisfaction conducted in the outpatient setting found that continuity of care was an important determinant of trust and, consequently, overall satisfaction.15, 16, 19, 20, 22 Because hospitalist models introduce discontinuity, they might be expected to undermine satisfaction. Surprisingly, few studies have addressed this issue. In a review of the hospitalist studies through 2002, Wachter and Goldman found 19 studies, 5 of which measured patient satisfaction.23 Three of these were conducted on teaching services and compared designated faculty hospitalists to traditional ward attendings, who rotated onto the inpatient services 1 to 2 months per year. Primary care doctors were excluded.2426 A fourth study provided a descriptive narrative of the development of the first hospitalist program in Minneapolis, Minnesota, and anecdotally noted no difference in patient satisfaction between the hospitalist and traditional model, but presented no data because the satisfaction surveys were not designed with publication in mind.27 The only study to actually assess whether patient satisfaction was greater with hospitalists or PCPs was an observational study by Davis et al., conducted in 1 rural hospital during the first year of its hospitalist program. In that study, 2 hospitalists were compared to 17 PCPs, and patient satisfaction surveys were available for approximately 44 patients managed by hospitalists and 168 patients managed by PCPs. Specific data were not reported, but it was noted that there was no statistical difference in satisfaction between those cared for by hospitalists versus PCPs.28 On the basis of these studies, Wachter and Goldman concluded that surveys of patients who were cared for by hospitalists show high levels of satisfaction, no lower than that of similar patients cared for by their own primary physicians.23 Wachter and Goldman's review has been highly cited, and we could find no subsequent studies addressing this issue. Our study provides the first real evidence to support this conclusion, including data from 59 hospitalists practicing in 5 separate hospitalist programs at 3 different hospitals.

Our finding that hospitalists maintain satisfaction despite a lack of continuity suggests that other aspects of care may be more important to patient satisfaction. Larson et al. found that physician ability to meet patient's information needs was positively associated with patient satisfaction.29 Similarly, Tarrant et al. found that patient's trust in a physician improved with increasing communication, interpersonal care, and knowledge of the patient. Interestingly, continuity, ie. the proportion of visits to the usual general practitioner (GP) or duration with the practice, did not correlate with trust.30 Finally, a systematic review of determinants of outpatient satisfaction found that continuity has a variable effect on satisfaction. Subjective continuity measures, such as whether patients saw their regular physician on the day they were surveyed, were consistently associated with patient satisfaction, however, quantitative measures including relationship duration were not.31

It is also possible that patients believe they value continuity more than they actually do. In 1 survey of inpatients with an established PCP yet cared for by a hospitalist, most agreed that patients receive better care and have more trust in physicians with whom they have long‐term relationships. Yet most also had positive opinions of their hospital care.32 Similarly, in a survey of over 2500 outpatients, 92% rated continuity as very important or important, but the majority was unwilling to expend substantial personal time (88%), defined as driving greater than 60 minutes, or money (82%), defined as spending an additional $20 to $40 a month, to maintain continuity with their PCP.33 Our study appears to confirm the lack of connection between continuity and satisfaction. Even those patients who valued continuity, as evidenced by having an established PCP, were as satisfied with hospitalist physician care as patients who had no established PCP.

Our study has several limitations. First, we report on outcomes of 3 institutions within a single healthcare system, within a limited geographic area. Although our sample included a wide range of patient demographics, hundreds of physicians, and multiple hospitalist models, it is possible that some hospitalist models may provide greater or lesser satisfaction than those we observed. Second, our study was observational, and thus subject to selection bias and confounding. Patients cared for by the hospitalists differed in a number of ways from those cared for by PCPs. We controlled for identifiable confounders such as illness severity, self‐perceived health, and admission through the emergency department, but the possibility exists that additional unidentified factors could have affected our results. It is possible other drivers of patient satisfaction, such as amenities, nursing, or food, could have influenced our findings. However, this is unlikely because all patient groups shared these components of hospital experience equally. Third, only a minority of patients could be reached for interview. This is typical for post‐hospitalization surveys, and our response rate of 40% for HCAHPS patients compared favorably to the 2010 HCAHPS national average of 33%.34 Still, the responses of those who could not be reached may have differed from those who were interviewed. Fourth, we identified hospitalists and PCPs by the attending of record, but we were unable to tell who provided care to the patient on any given day. Thus, we could not determine to what extent patients cared for by PCPs were actually seen by their own doctor, as opposed to an associated physician within the practice. Nevertheless, our results are representative of the care model provided by PCPs in the hospital. Similarly, we could not know or compare the number of different attending physicians each patient experienced during their hospitalization. Higher turnover of inpatient physicians may have affected patient satisfaction scores independent of attending physician designation. These are potentially important measures of relationship duration, yet whether duration affects patient satisfaction remains undecided.1618, 20, 28, 30, 32, 33 We assessed satisfaction using HCAHPS questions, in order to provide objective and meaningful comparisons across hospitals. The HCAHPS instrument, however, is intended to assess patient satisfaction with doctors in general, not with subgroups or individuals, and responses in our study were uniformly high. A more sensitive survey instrument may have yielded different results. Finally, it is possible that individual physicians may possess lower satisfaction scores than others, making the results not representative of hospitalist models as much as specific doctors' care quality. We think this is unlikely since surveys reached over 8000 patients, over 6 years, representing the care of 347 individual physicians. However, hospital medicine is a rapidly evolving field with many divergent organizational structures, and patient satisfaction is bound to fluctuate while there exists high variability in how care is provided.

Over the past decade, the hospitalist model has become one of the dominant models for care of medical inpatients. Compared to the traditional model in which PCPs provide inpatient care, the hospitalist model has a number of advantages, including continuous on‐site coverage for increasingly acute patients, specialization, and incentives aligned with the hospital to provide efficient, high‐quality care. One concern that remains, however, is that patients may not trust doctors they first meet in the hospital or may be dissatisfied with the lack of continuity from day to day. Our findings are reassuring in this regard. Although patients cared for by hospitalists were slightly less satisfied, the differences could not be considered clinically meaningful and should be outweighed by gains in quality and efficiency. Furthermore, hospitalists can expect to fare well under value‐based purchasing. Given the rapid ascension of hospital medicine programs, prospective comparisons of hospitalists and PCPs may no longer be feasible. Future research might employ survey instruments designed specifically to measure patient experience under hospitalist care in order to identify methods to maximize patient satisfaction within the hospitalist model.

Over the past decade, hospital medicine has been the nation's fastest‐growing medical specialty. According to the American Hospital Association's (AHA) 2009 survey, 58% of United States (US) hospitals now have hospital medicine programs, and for hospitals with 200 or more beds, this figure is 89%.1 In 2009, the AHA estimated that the number of US hospitalists would increase to over 34,000 by 2011, over double that of the 16,000 present in 2005.1 Studies demonstrate that, compared to a system where primary care physicians provide inpatient care, the hospitalist model improves efficiency while maintaining at least equal patient outcomes.211 However, scant data exist as to the effects of hospitalists on patient satisfaction.12 Understanding how care models affect patient experience is vital in the current environment of healthcare reform and performance reporting, especially in light of the Centers for Medicare and Medicaid Services' (CMS) efforts to link the patient experience to reimbursement through value‐based purchasing.13 Value‐based purchasing is a strategy to encourage and reward excellence in healthcare delivery through differential reimbursement based on defined performance measures. As one part of value‐based purchasing, hospital reimbursement will be linked to patient‐experience measures, including patient ratings of their doctor's ability to communicate with them and other questions assessing patient satisfaction with their hospital stay.14

In the outpatient setting, trust is the variable most strongly associated with patient satisfaction.1518 In contrast to PCPs, who may develop relationships with patients over years, hospitalists often first meet a patient in the hospital and must engender trust quickly. In addition, hospitalists work in shifts and may not be responsible for the same patients each day. Since continuity is positively related to trust,19, 20 there is reason to believe satisfaction with hospitalist care might be lower than satisfaction with care provided by PCPs. We report on 8295 patients and 6 years experience with hospitalist programs at 3 hospitals. Based on the known link between continuity and patient satisfaction, we hypothesized that patient satisfaction would be lower with hospitalists than with primary care internists.

METHODS

RESULTS

Of patients who were reached by telephone, 87% agreed to participate in the hospital survey. However, most patients could not be reached by phone; thus our estimated response rate, including those who could not be reached, was 27%. For the subset of patients interviewed using the HCAHPS protocol, the response rate was 40%. Our final sample included 8295 patients (3597 cared for by 59 hospitalists and 4698 by 288 PCPs) interviewed between 2003 and 2009. Three‐quarters of the patients were from the tertiary care center, whereas 17% and 8% were from each of the community hospitals (see Supporting Appendix B in the online version of this article). Patient characteristics appear in Table 1. Patients cared for by hospitalists were similar to those cared for by PCPs in terms of age, sex, marital status, education, and language, but hospitalist patients were more likely to have been admitted through the emergency department (93% vs 84%, P < 0.001) and less likely to be white (83% vs 85%, P = 0.01). Patients cared for by hospitalists also had higher average illness severity score (2.2 0.8 vs 2.0 0.8, P < 0.001), longer average LOS (4.3 4.3 vs 4.0 3.6, P < 0.001), and lower mean perceived health score (2.8 1.2 vs 3.0 1.2, P = 0.01).

Unadjusted patient reported satisfaction with physician care quality was slightly greater for PCPs than hospitalists (4.25 vs 4.19, P = 0.009). After multivariable adjustment, the difference was attenuated but persisted (4.24 vs 4.20, P = 0.04). We found no statistical difference among the hospitals or the specific hospitalist groups in terms of satisfaction with overall physician care quality (Figure 1). There were no statistical differences in patient satisfaction ratings of hospitalist and PCPs for the subdomains of behavior, pain, and communication (Table 2). There were also no differences in the proportion of patients cared for by hospitalists or PCPs who rated their physicians in the highest satisfaction category (79% vs 81%, P = 0.17) or the lowest (5% vs 5%, P = 0.19). Among patients cared for by academic hospitalists, there was no difference in satisfaction rating between those patients who had a designated primary care physician in the outpatient setting and those who did not (4.22 0.94 vs 4.19 0.94, P = 0.97). Finally, satisfaction with both hospitalists and PCPs showed equivalent rates of improvement over time (Figure 2).

Figure 1

Figure 2

DISCUSSION

In this observational study of over 8200 patients cared for over 6 years by 347 physicians at 3 hospitals, we found that patient satisfaction with inpatient care provided by hospitalists and primary care doctors was almost identical. As we hypothesized, overall satisfaction with physician care quality, our primary outcome, was slightly greater with primary care doctors; however, the observed difference, 0.04 on a scale of 1 to 5, cannot be considered clinically significant. All patients were generally satisfied (4.2‐4.3 rating on 5‐point scale) with their inpatient care, and satisfaction scores increased over time. We also found no differences among the specific domains of satisfaction, including communication skills, pain control, and physician behavior. Finally, we found no significant difference in patient satisfaction with physician care quality among the different hospitalist services.

Previous studies of patient satisfaction conducted in the outpatient setting found that continuity of care was an important determinant of trust and, consequently, overall satisfaction.15, 16, 19, 20, 22 Because hospitalist models introduce discontinuity, they might be expected to undermine satisfaction. Surprisingly, few studies have addressed this issue. In a review of the hospitalist studies through 2002, Wachter and Goldman found 19 studies, 5 of which measured patient satisfaction.23 Three of these were conducted on teaching services and compared designated faculty hospitalists to traditional ward attendings, who rotated onto the inpatient services 1 to 2 months per year. Primary care doctors were excluded.2426 A fourth study provided a descriptive narrative of the development of the first hospitalist program in Minneapolis, Minnesota, and anecdotally noted no difference in patient satisfaction between the hospitalist and traditional model, but presented no data because the satisfaction surveys were not designed with publication in mind.27 The only study to actually assess whether patient satisfaction was greater with hospitalists or PCPs was an observational study by Davis et al., conducted in 1 rural hospital during the first year of its hospitalist program. In that study, 2 hospitalists were compared to 17 PCPs, and patient satisfaction surveys were available for approximately 44 patients managed by hospitalists and 168 patients managed by PCPs. Specific data were not reported, but it was noted that there was no statistical difference in satisfaction between those cared for by hospitalists versus PCPs.28 On the basis of these studies, Wachter and Goldman concluded that surveys of patients who were cared for by hospitalists show high levels of satisfaction, no lower than that of similar patients cared for by their own primary physicians.23 Wachter and Goldman's review has been highly cited, and we could find no subsequent studies addressing this issue. Our study provides the first real evidence to support this conclusion, including data from 59 hospitalists practicing in 5 separate hospitalist programs at 3 different hospitals.

Our finding that hospitalists maintain satisfaction despite a lack of continuity suggests that other aspects of care may be more important to patient satisfaction. Larson et al. found that physician ability to meet patient's information needs was positively associated with patient satisfaction.29 Similarly, Tarrant et al. found that patient's trust in a physician improved with increasing communication, interpersonal care, and knowledge of the patient. Interestingly, continuity, ie. the proportion of visits to the usual general practitioner (GP) or duration with the practice, did not correlate with trust.30 Finally, a systematic review of determinants of outpatient satisfaction found that continuity has a variable effect on satisfaction. Subjective continuity measures, such as whether patients saw their regular physician on the day they were surveyed, were consistently associated with patient satisfaction, however, quantitative measures including relationship duration were not.31

It is also possible that patients believe they value continuity more than they actually do. In 1 survey of inpatients with an established PCP yet cared for by a hospitalist, most agreed that patients receive better care and have more trust in physicians with whom they have long‐term relationships. Yet most also had positive opinions of their hospital care.32 Similarly, in a survey of over 2500 outpatients, 92% rated continuity as very important or important, but the majority was unwilling to expend substantial personal time (88%), defined as driving greater than 60 minutes, or money (82%), defined as spending an additional $20 to $40 a month, to maintain continuity with their PCP.33 Our study appears to confirm the lack of connection between continuity and satisfaction. Even those patients who valued continuity, as evidenced by having an established PCP, were as satisfied with hospitalist physician care as patients who had no established PCP.

Our study has several limitations. First, we report on outcomes of 3 institutions within a single healthcare system, within a limited geographic area. Although our sample included a wide range of patient demographics, hundreds of physicians, and multiple hospitalist models, it is possible that some hospitalist models may provide greater or lesser satisfaction than those we observed. Second, our study was observational, and thus subject to selection bias and confounding. Patients cared for by the hospitalists differed in a number of ways from those cared for by PCPs. We controlled for identifiable confounders such as illness severity, self‐perceived health, and admission through the emergency department, but the possibility exists that additional unidentified factors could have affected our results. It is possible other drivers of patient satisfaction, such as amenities, nursing, or food, could have influenced our findings. However, this is unlikely because all patient groups shared these components of hospital experience equally. Third, only a minority of patients could be reached for interview. This is typical for post‐hospitalization surveys, and our response rate of 40% for HCAHPS patients compared favorably to the 2010 HCAHPS national average of 33%.34 Still, the responses of those who could not be reached may have differed from those who were interviewed. Fourth, we identified hospitalists and PCPs by the attending of record, but we were unable to tell who provided care to the patient on any given day. Thus, we could not determine to what extent patients cared for by PCPs were actually seen by their own doctor, as opposed to an associated physician within the practice. Nevertheless, our results are representative of the care model provided by PCPs in the hospital. Similarly, we could not know or compare the number of different attending physicians each patient experienced during their hospitalization. Higher turnover of inpatient physicians may have affected patient satisfaction scores independent of attending physician designation. These are potentially important measures of relationship duration, yet whether duration affects patient satisfaction remains undecided.1618, 20, 28, 30, 32, 33 We assessed satisfaction using HCAHPS questions, in order to provide objective and meaningful comparisons across hospitals. The HCAHPS instrument, however, is intended to assess patient satisfaction with doctors in general, not with subgroups or individuals, and responses in our study were uniformly high. A more sensitive survey instrument may have yielded different results. Finally, it is possible that individual physicians may possess lower satisfaction scores than others, making the results not representative of hospitalist models as much as specific doctors' care quality. We think this is unlikely since surveys reached over 8000 patients, over 6 years, representing the care of 347 individual physicians. However, hospital medicine is a rapidly evolving field with many divergent organizational structures, and patient satisfaction is bound to fluctuate while there exists high variability in how care is provided.

Over the past decade, the hospitalist model has become one of the dominant models for care of medical inpatients. Compared to the traditional model in which PCPs provide inpatient care, the hospitalist model has a number of advantages, including continuous on‐site coverage for increasingly acute patients, specialization, and incentives aligned with the hospital to provide efficient, high‐quality care. One concern that remains, however, is that patients may not trust doctors they first meet in the hospital or may be dissatisfied with the lack of continuity from day to day. Our findings are reassuring in this regard. Although patients cared for by hospitalists were slightly less satisfied, the differences could not be considered clinically meaningful and should be outweighed by gains in quality and efficiency. Furthermore, hospitalists can expect to fare well under value‐based purchasing. Given the rapid ascension of hospital medicine programs, prospective comparisons of hospitalists and PCPs may no longer be feasible. Future research might employ survey instruments designed specifically to measure patient experience under hospitalist care in order to identify methods to maximize patient satisfaction within the hospitalist model.

Over the past decade, hospital medicine has been the nation's fastest‐growing medical specialty. According to the American Hospital Association's (AHA) 2009 survey, 58% of United States (US) hospitals now have hospital medicine programs, and for hospitals with 200 or more beds, this figure is 89%.1 In 2009, the AHA estimated that the number of US hospitalists would increase to over 34,000 by 2011, over double that of the 16,000 present in 2005.1 Studies demonstrate that, compared to a system where primary care physicians provide inpatient care, the hospitalist model improves efficiency while maintaining at least equal patient outcomes.211 However, scant data exist as to the effects of hospitalists on patient satisfaction.12 Understanding how care models affect patient experience is vital in the current environment of healthcare reform and performance reporting, especially in light of the Centers for Medicare and Medicaid Services' (CMS) efforts to link the patient experience to reimbursement through value‐based purchasing.13 Value‐based purchasing is a strategy to encourage and reward excellence in healthcare delivery through differential reimbursement based on defined performance measures. As one part of value‐based purchasing, hospital reimbursement will be linked to patient‐experience measures, including patient ratings of their doctor's ability to communicate with them and other questions assessing patient satisfaction with their hospital stay.14

In the outpatient setting, trust is the variable most strongly associated with patient satisfaction.1518 In contrast to PCPs, who may develop relationships with patients over years, hospitalists often first meet a patient in the hospital and must engender trust quickly. In addition, hospitalists work in shifts and may not be responsible for the same patients each day. Since continuity is positively related to trust,19, 20 there is reason to believe satisfaction with hospitalist care might be lower than satisfaction with care provided by PCPs. We report on 8295 patients and 6 years experience with hospitalist programs at 3 hospitals. Based on the known link between continuity and patient satisfaction, we hypothesized that patient satisfaction would be lower with hospitalists than with primary care internists.

METHODS

RESULTS

Of patients who were reached by telephone, 87% agreed to participate in the hospital survey. However, most patients could not be reached by phone; thus our estimated response rate, including those who could not be reached, was 27%. For the subset of patients interviewed using the HCAHPS protocol, the response rate was 40%. Our final sample included 8295 patients (3597 cared for by 59 hospitalists and 4698 by 288 PCPs) interviewed between 2003 and 2009. Three‐quarters of the patients were from the tertiary care center, whereas 17% and 8% were from each of the community hospitals (see Supporting Appendix B in the online version of this article). Patient characteristics appear in Table 1. Patients cared for by hospitalists were similar to those cared for by PCPs in terms of age, sex, marital status, education, and language, but hospitalist patients were more likely to have been admitted through the emergency department (93% vs 84%, P < 0.001) and less likely to be white (83% vs 85%, P = 0.01). Patients cared for by hospitalists also had higher average illness severity score (2.2 0.8 vs 2.0 0.8, P < 0.001), longer average LOS (4.3 4.3 vs 4.0 3.6, P < 0.001), and lower mean perceived health score (2.8 1.2 vs 3.0 1.2, P = 0.01).

Unadjusted patient reported satisfaction with physician care quality was slightly greater for PCPs than hospitalists (4.25 vs 4.19, P = 0.009). After multivariable adjustment, the difference was attenuated but persisted (4.24 vs 4.20, P = 0.04). We found no statistical difference among the hospitals or the specific hospitalist groups in terms of satisfaction with overall physician care quality (Figure 1). There were no statistical differences in patient satisfaction ratings of hospitalist and PCPs for the subdomains of behavior, pain, and communication (Table 2). There were also no differences in the proportion of patients cared for by hospitalists or PCPs who rated their physicians in the highest satisfaction category (79% vs 81%, P = 0.17) or the lowest (5% vs 5%, P = 0.19). Among patients cared for by academic hospitalists, there was no difference in satisfaction rating between those patients who had a designated primary care physician in the outpatient setting and those who did not (4.22 0.94 vs 4.19 0.94, P = 0.97). Finally, satisfaction with both hospitalists and PCPs showed equivalent rates of improvement over time (Figure 2).

Figure 1

Figure 2

DISCUSSION

In this observational study of over 8200 patients cared for over 6 years by 347 physicians at 3 hospitals, we found that patient satisfaction with inpatient care provided by hospitalists and primary care doctors was almost identical. As we hypothesized, overall satisfaction with physician care quality, our primary outcome, was slightly greater with primary care doctors; however, the observed difference, 0.04 on a scale of 1 to 5, cannot be considered clinically significant. All patients were generally satisfied (4.2‐4.3 rating on 5‐point scale) with their inpatient care, and satisfaction scores increased over time. We also found no differences among the specific domains of satisfaction, including communication skills, pain control, and physician behavior. Finally, we found no significant difference in patient satisfaction with physician care quality among the different hospitalist services.

Previous studies of patient satisfaction conducted in the outpatient setting found that continuity of care was an important determinant of trust and, consequently, overall satisfaction.15, 16, 19, 20, 22 Because hospitalist models introduce discontinuity, they might be expected to undermine satisfaction. Surprisingly, few studies have addressed this issue. In a review of the hospitalist studies through 2002, Wachter and Goldman found 19 studies, 5 of which measured patient satisfaction.23 Three of these were conducted on teaching services and compared designated faculty hospitalists to traditional ward attendings, who rotated onto the inpatient services 1 to 2 months per year. Primary care doctors were excluded.2426 A fourth study provided a descriptive narrative of the development of the first hospitalist program in Minneapolis, Minnesota, and anecdotally noted no difference in patient satisfaction between the hospitalist and traditional model, but presented no data because the satisfaction surveys were not designed with publication in mind.27 The only study to actually assess whether patient satisfaction was greater with hospitalists or PCPs was an observational study by Davis et al., conducted in 1 rural hospital during the first year of its hospitalist program. In that study, 2 hospitalists were compared to 17 PCPs, and patient satisfaction surveys were available for approximately 44 patients managed by hospitalists and 168 patients managed by PCPs. Specific data were not reported, but it was noted that there was no statistical difference in satisfaction between those cared for by hospitalists versus PCPs.28 On the basis of these studies, Wachter and Goldman concluded that surveys of patients who were cared for by hospitalists show high levels of satisfaction, no lower than that of similar patients cared for by their own primary physicians.23 Wachter and Goldman's review has been highly cited, and we could find no subsequent studies addressing this issue. Our study provides the first real evidence to support this conclusion, including data from 59 hospitalists practicing in 5 separate hospitalist programs at 3 different hospitals.

Our finding that hospitalists maintain satisfaction despite a lack of continuity suggests that other aspects of care may be more important to patient satisfaction. Larson et al. found that physician ability to meet patient's information needs was positively associated with patient satisfaction.29 Similarly, Tarrant et al. found that patient's trust in a physician improved with increasing communication, interpersonal care, and knowledge of the patient. Interestingly, continuity, ie. the proportion of visits to the usual general practitioner (GP) or duration with the practice, did not correlate with trust.30 Finally, a systematic review of determinants of outpatient satisfaction found that continuity has a variable effect on satisfaction. Subjective continuity measures, such as whether patients saw their regular physician on the day they were surveyed, were consistently associated with patient satisfaction, however, quantitative measures including relationship duration were not.31

It is also possible that patients believe they value continuity more than they actually do. In 1 survey of inpatients with an established PCP yet cared for by a hospitalist, most agreed that patients receive better care and have more trust in physicians with whom they have long‐term relationships. Yet most also had positive opinions of their hospital care.32 Similarly, in a survey of over 2500 outpatients, 92% rated continuity as very important or important, but the majority was unwilling to expend substantial personal time (88%), defined as driving greater than 60 minutes, or money (82%), defined as spending an additional $20 to $40 a month, to maintain continuity with their PCP.33 Our study appears to confirm the lack of connection between continuity and satisfaction. Even those patients who valued continuity, as evidenced by having an established PCP, were as satisfied with hospitalist physician care as patients who had no established PCP.

Our study has several limitations. First, we report on outcomes of 3 institutions within a single healthcare system, within a limited geographic area. Although our sample included a wide range of patient demographics, hundreds of physicians, and multiple hospitalist models, it is possible that some hospitalist models may provide greater or lesser satisfaction than those we observed. Second, our study was observational, and thus subject to selection bias and confounding. Patients cared for by the hospitalists differed in a number of ways from those cared for by PCPs. We controlled for identifiable confounders such as illness severity, self‐perceived health, and admission through the emergency department, but the possibility exists that additional unidentified factors could have affected our results. It is possible other drivers of patient satisfaction, such as amenities, nursing, or food, could have influenced our findings. However, this is unlikely because all patient groups shared these components of hospital experience equally. Third, only a minority of patients could be reached for interview. This is typical for post‐hospitalization surveys, and our response rate of 40% for HCAHPS patients compared favorably to the 2010 HCAHPS national average of 33%.34 Still, the responses of those who could not be reached may have differed from those who were interviewed. Fourth, we identified hospitalists and PCPs by the attending of record, but we were unable to tell who provided care to the patient on any given day. Thus, we could not determine to what extent patients cared for by PCPs were actually seen by their own doctor, as opposed to an associated physician within the practice. Nevertheless, our results are representative of the care model provided by PCPs in the hospital. Similarly, we could not know or compare the number of different attending physicians each patient experienced during their hospitalization. Higher turnover of inpatient physicians may have affected patient satisfaction scores independent of attending physician designation. These are potentially important measures of relationship duration, yet whether duration affects patient satisfaction remains undecided.1618, 20, 28, 30, 32, 33 We assessed satisfaction using HCAHPS questions, in order to provide objective and meaningful comparisons across hospitals. The HCAHPS instrument, however, is intended to assess patient satisfaction with doctors in general, not with subgroups or individuals, and responses in our study were uniformly high. A more sensitive survey instrument may have yielded different results. Finally, it is possible that individual physicians may possess lower satisfaction scores than others, making the results not representative of hospitalist models as much as specific doctors' care quality. We think this is unlikely since surveys reached over 8000 patients, over 6 years, representing the care of 347 individual physicians. However, hospital medicine is a rapidly evolving field with many divergent organizational structures, and patient satisfaction is bound to fluctuate while there exists high variability in how care is provided.

Over the past decade, the hospitalist model has become one of the dominant models for care of medical inpatients. Compared to the traditional model in which PCPs provide inpatient care, the hospitalist model has a number of advantages, including continuous on‐site coverage for increasingly acute patients, specialization, and incentives aligned with the hospital to provide efficient, high‐quality care. One concern that remains, however, is that patients may not trust doctors they first meet in the hospital or may be dissatisfied with the lack of continuity from day to day. Our findings are reassuring in this regard. Although patients cared for by hospitalists were slightly less satisfied, the differences could not be considered clinically meaningful and should be outweighed by gains in quality and efficiency. Furthermore, hospitalists can expect to fare well under value‐based purchasing. Given the rapid ascension of hospital medicine programs, prospective comparisons of hospitalists and PCPs may no longer be feasible. Future research might employ survey instruments designed specifically to measure patient experience under hospitalist care in order to identify methods to maximize patient satisfaction within the hospitalist model.

Diabetes care in the inpatient setting requires coordination between multiple service providers. Breakdowns in this process occur at all levels leading to potential serious harm.1 Error rates focusing on multiple areas related to diabetes care, including the inpatient provision of insulin, have been described as high as 19.5% in 14,000 patients surveyed in the United Kingdom.2 Missteps are important, as insulin prescribing errors are more commonly associated with patient harm.3 In the United States, medication errors related to provision of care to critically ill patients has been documented, but, to our knowledge, no such reports regarding general medical or surgical wards exist.4

Insulin errors can result from a wide range of possible reasons including: incorrect medication reconciliation, prescribing errors, dispensing errors, administration errors, suboptimal meal timing, or errors in communication for discharge plans regarding diabetes care. Examining each of these areas as a whole could be a daunting task. As such, we sought to examine 1 portion of insulin provision as an initial focus for performance improvement at our institution. Our purpose was to describe the rates of errors associated with insulin administration at our single academic medical center on general medicine and surgical wards.

Methods

Study patients for this observational, prospective snapshot were identified by electronic medical records in 4 consecutive weeks in April 2009 at Barnes‐Jewish Hospital (St Louis, MO), a 1200 bed academic medical center. This study was approved by the Washington University in St Louis School of Medicine Human Studies Committee, and the requirement for informed consent was waived.

On day 1 of each snapshot period, all patients on the identified wards were examined to determine if insulin was currently active as part of the inpatient medication orders. If active, this patient was enrolled into the evaluation data set. No patients were excluded if insulin was currently ordered. Four inpatient areas were selected to provide a representation of the non‐critically ill patient population at our institution. The 4 areas selected were: a cardiac care ward (typical census = 24), a general medicine ward (typical census = 24), an abdominal transplant ward (typical census = 18), and a general surgery ward (typical census = 22). Taken together, these areas represent about 20% of the total non‐critically ill patient population at our hospital. The transplant area was chosen because it represents a high‐risk population with medication (corticosteroid)‐induced diabetes. Nursing and physician care are typically exclusive in these areas, and very little crossover among these healthcare providers would have occurred among the units surveyed during the study period.

Each patient included on day 1 of each audit period was followed for a total of up to 7 days. Patients were only enrolled on day 1 of each audit period. Four survey periods were conducted, providing an evaluation of 28 days of insulin therapy in the studied units. Four periods were selected to pick up more patients on day 1 of each audit period. Electronic records of medication administration and evaluation of paper chart orders provided the information for insulin administration error rates. Additionally, physician notes regarding patients' histories and home insulin use were reviewed for background information for our patient population. Prospective daily assessments of insulin orders, doses charted, nursing notes, and blood glucose values were conducted for potential errors in insulin administration.

All definitions of insulin administration errors were defined prior to data collection. The investigators reviewed available literature involving insulin errors, and found no standardized definitions or previously published assessments at the time of inception of our study. As such, we examined our own clinical practice for areas of potential concern related to insulin administration. The following error categories were identified: transcription errors (eg, insulin glargine 10 units qpm written, but order transcribed and carried out as 20 units qpm); greater than 1 hour between obtained point‐of‐care blood glucose value and provision of correctional (sliding‐scale) insulin; insulin held without a physician order present in the medical records; missing documentation of insulin doses (glucose value of 150 mg/dL present, but no documented correctional dose corresponding to this value present in medical record); premeal and correctional insulin given at separate times; and no documentation of physician notification for hypoglycemia. Other reasons for potential insulin administration errors were collected if deemed pertinent by the individual auditors.

At the time of our survey, a standardized subcutaneous insulin administration order set was utilized in all of the surveyed units. As computerized physician order entry was not yet available at our institution, all orders were transcribed electronically from paper orders. This insulin order set has been in place for 5 years. Once initiated, all portions of the order set are initiated, including communication to nurses regarding glucose measuring times, requirement for documentation of hypoglycemia, and proposed glucose goals. A survey of insulin orders during the audit time revealed that >97% of all insulin orders were initiated from this standardized order set. These order sets encouraged the provision of physiological insulin (basal‐bolus) using insulin glargine and insulin aspart in eligible individuals. Although no systematic, standardized goal for glucose attainment was promoted, a fasting blood glucose of 90‐130 mg/dL and post‐prandial value of <180 mg/dL was encouraged. The order sets had a stated requirement of physician contact for all blood glucose values <70 mg/dL. Although lack of documentation of hypoglycemia may not be directly considered an error associated with administration of insulin, the research group decided to include this provision in the definition of administration errors, given the ability of this parameter to provide a sense of overall completeness of insulin orders and as a marker of collaborative practice in the management of inpatient hyperglycemia.

Nurses documented glucose values and responses in electronic medical administration records as a matter of routine. Point‐of‐care glucose values were obtained by either patient care technicians or nurses on each individual ward. As an academic medical institution, physicians were frequently paged by other members of the healthcare team.

Each auditor (E.N.D., A.L., L.L.W., K.A.H.) reviewed 1 consistent unit during the audit period. All data for insulin administration errors were tabulated, and descriptive rates of errors were used on a per‐patient or per‐stay basis

Results

A total of 116 patient‐audit periods were identified during the 28‐day study period (Table 1). Sixty‐five patients were on surgical services, and 51 were on medicine services, representing 378 inpatient days. Median length of stay was 3.5 days. Home insulin use was evident in 49% of the surveyed population. Patients' mean A1C (data available within 3 months prior to admission) was 8.1% (n = 41). Inpatient insulin regimens on day 1 included correctional insulin only (51.7% of cases). Regimens containing neutral protamine Hagedorn (NPH) or glargine also included correctional insulin in 95% of cases, and premeal insulin in 35%. Regimens including both premeal insulin and correctional insulin occurred in 25% of the patients. Diet status indicated that 83% of the population was taking an oral diet on day 1, and 13% were nil per os (nothing by mouth [NPO]).

A total of 199 administration errors occurred at a rate of 1.72 errors/patient‐period and 0.53 errors/patient day (Table 2). Missing documentation of doses (15.5% of all patients) and insulin being held without an order (25% of patients) were the most frequently occurring events. Errors classified as other were found in 13.1% of the defined events. These other errors consisted of not carrying out correctional dose insulin orders appropriately (eg, blood sugar value of 149 mg/dL should have resulted in a correctional dose of 2 units, but 3 units were documented as given instead), timing errors related to provision of mealtime insulin apart from documented provision of a meal, or not following the required documentation for insulin pumps.

Forty‐two patients (36%) experienced no errors in insulin administration, 18 patients experienced 1 error, 21 patients had 2 errors, and 11 patients had 3 errors. The remainder of the patients (n = 23; 19.9%) had 4 or more errors during their observation period. Were similar across the units surveyed. Frequency of errors remained consistent regardless of reason for admission, history of diabetes or insulin use at home, or length of stay. Most errors occurred on days 1 and 2 of the hospital stay. Error rates and types were consistent across all units surveyed.

Discussion/Conclusion

We found that insulin administration errors were common in our inpatient snapshot of non‐critically ill patients. In our observational evaluation, 64% of patients had at least 1 error related to insulin administration. Errors related to missing documentation of scheduled doses, or doses held without a prescriber order, were the most common. Implications of missed or held doses could range from unclear approaches for dose adjustment due to missing information, incorrect titration due to incomplete information, or hypoglycemia and hyperglycemia.

This observed rate of error is much higher than the described error rate of 19.5% reported in the United Kingdom.2 This difference in error rates most likely reflects a difference in focus, as investigators in that national effort focused on prescriber error, aberrations in blood glucose values, and readmission rates. Our evaluation in assessing error rates regarding insulin administration supports the use of personnel keenly aware of the processes related to insulin administration, and provides insight into the importance of evaluating small portions of insulin provision (administration vs prescribing, etc) in assessing grounds for improvement in care. It is important to note that our findings may be exaggerated and are not entirely comparable to a study with a different scope and size.

Our snapshot tool and baseline evaluation is a simple method that could be undertaken at many institutions. As such, this methodology and error estimate serves as a gauge for future comparisons and areas for intervention. Limitations of our assessment include the small portion of patients audited during our evaluation versus using a snapshot of our entire hospital, utilizing nonstandardized criteria for determination of insulin errors, and the lack of correlation of clinical significance (aberrations in glucose values) with errors observed. Also, this single‐institution review may not be generalizable to all institutions. Additionally, we only examined errors related to administration of insulin. Other areas that would complete the picture, related to diabetic therapies and outcomes, would need to include prescribing errors or dispensing errors and relate these to glycemic outcomes. Assessment of these additional errors may have revealed more clinically important events that were not revealed in this small snapshot. Lastly, clinical endpoints such as intensive care unit (ICU) transfers, mortality, or readmissions were not assessed in this small study.

We are fortunate that many of these errors were apparently clinically silent, but in a subset of patients, the risk is real and life‐threatening. Risk occurs at both ends of the glucose spectrum, with the low end receiving the greatest attention. Severe hypoglycemia with harm and inpatient diabetic ketoacidosis have been qualified as newer events by Medicare. Hypoglycemia in the ICU population (<40 mg/dL) is an independent marker of mortality.5 Hypoglycemia (<50 mg/dL) has been associated with heart attacks, strokes, and death in the outpatient setting.6

The ability to safely control blood sugar in the hospital requires that medications are administered on time, and that communication occurs between the prescribing provider and the nursing staff providing care. Along with the case‐by‐case implications regarding the need for accurate administration of insulin for subsequent titration and determination of discharge prescriptions for patients with diabetes, there are many implications regarding the assessment of inpatient provision of insulin on determining institutional practices based on previous performance. If insulin administration is not accurately provided or documented, institutions will find it difficult to correctly make changes to insulin protocols for targeting future improvements. Our evaluation indicates an obvious need for quality improvement with 18.1% of the errors reflecting holding insulin without an order, and 12.6% of the errors showing no documentation for the physician being notified of hypoglycemia requiring treatment. The need to foster structured nurse‐physician communication will play a critical role in any process improvement. Communication is key for the optimal provision of insulin in the inpatient setting. Computerized order entry and bar‐code guided administration of doses of insulin may fix some types of the errors (transcription and missed documentation, respectively). That said, one of the largest impacts of this survey may reveal that these errors may not be fixed by technology, but may require more targeted and difficult interventions, such as continuing education and holding clinicians accountable. This study provides insight into the complicated issues regarding inpatient insulin administration and, due to its systematic approach, has given direction for process and system improvements.

Diabetes care in the inpatient setting requires coordination between multiple service providers. Breakdowns in this process occur at all levels leading to potential serious harm.1 Error rates focusing on multiple areas related to diabetes care, including the inpatient provision of insulin, have been described as high as 19.5% in 14,000 patients surveyed in the United Kingdom.2 Missteps are important, as insulin prescribing errors are more commonly associated with patient harm.3 In the United States, medication errors related to provision of care to critically ill patients has been documented, but, to our knowledge, no such reports regarding general medical or surgical wards exist.4

Insulin errors can result from a wide range of possible reasons including: incorrect medication reconciliation, prescribing errors, dispensing errors, administration errors, suboptimal meal timing, or errors in communication for discharge plans regarding diabetes care. Examining each of these areas as a whole could be a daunting task. As such, we sought to examine 1 portion of insulin provision as an initial focus for performance improvement at our institution. Our purpose was to describe the rates of errors associated with insulin administration at our single academic medical center on general medicine and surgical wards.

Methods

Study patients for this observational, prospective snapshot were identified by electronic medical records in 4 consecutive weeks in April 2009 at Barnes‐Jewish Hospital (St Louis, MO), a 1200 bed academic medical center. This study was approved by the Washington University in St Louis School of Medicine Human Studies Committee, and the requirement for informed consent was waived.

On day 1 of each snapshot period, all patients on the identified wards were examined to determine if insulin was currently active as part of the inpatient medication orders. If active, this patient was enrolled into the evaluation data set. No patients were excluded if insulin was currently ordered. Four inpatient areas were selected to provide a representation of the non‐critically ill patient population at our institution. The 4 areas selected were: a cardiac care ward (typical census = 24), a general medicine ward (typical census = 24), an abdominal transplant ward (typical census = 18), and a general surgery ward (typical census = 22). Taken together, these areas represent about 20% of the total non‐critically ill patient population at our hospital. The transplant area was chosen because it represents a high‐risk population with medication (corticosteroid)‐induced diabetes. Nursing and physician care are typically exclusive in these areas, and very little crossover among these healthcare providers would have occurred among the units surveyed during the study period.

Each patient included on day 1 of each audit period was followed for a total of up to 7 days. Patients were only enrolled on day 1 of each audit period. Four survey periods were conducted, providing an evaluation of 28 days of insulin therapy in the studied units. Four periods were selected to pick up more patients on day 1 of each audit period. Electronic records of medication administration and evaluation of paper chart orders provided the information for insulin administration error rates. Additionally, physician notes regarding patients' histories and home insulin use were reviewed for background information for our patient population. Prospective daily assessments of insulin orders, doses charted, nursing notes, and blood glucose values were conducted for potential errors in insulin administration.

All definitions of insulin administration errors were defined prior to data collection. The investigators reviewed available literature involving insulin errors, and found no standardized definitions or previously published assessments at the time of inception of our study. As such, we examined our own clinical practice for areas of potential concern related to insulin administration. The following error categories were identified: transcription errors (eg, insulin glargine 10 units qpm written, but order transcribed and carried out as 20 units qpm); greater than 1 hour between obtained point‐of‐care blood glucose value and provision of correctional (sliding‐scale) insulin; insulin held without a physician order present in the medical records; missing documentation of insulin doses (glucose value of 150 mg/dL present, but no documented correctional dose corresponding to this value present in medical record); premeal and correctional insulin given at separate times; and no documentation of physician notification for hypoglycemia. Other reasons for potential insulin administration errors were collected if deemed pertinent by the individual auditors.

At the time of our survey, a standardized subcutaneous insulin administration order set was utilized in all of the surveyed units. As computerized physician order entry was not yet available at our institution, all orders were transcribed electronically from paper orders. This insulin order set has been in place for 5 years. Once initiated, all portions of the order set are initiated, including communication to nurses regarding glucose measuring times, requirement for documentation of hypoglycemia, and proposed glucose goals. A survey of insulin orders during the audit time revealed that >97% of all insulin orders were initiated from this standardized order set. These order sets encouraged the provision of physiological insulin (basal‐bolus) using insulin glargine and insulin aspart in eligible individuals. Although no systematic, standardized goal for glucose attainment was promoted, a fasting blood glucose of 90‐130 mg/dL and post‐prandial value of <180 mg/dL was encouraged. The order sets had a stated requirement of physician contact for all blood glucose values <70 mg/dL. Although lack of documentation of hypoglycemia may not be directly considered an error associated with administration of insulin, the research group decided to include this provision in the definition of administration errors, given the ability of this parameter to provide a sense of overall completeness of insulin orders and as a marker of collaborative practice in the management of inpatient hyperglycemia.

Nurses documented glucose values and responses in electronic medical administration records as a matter of routine. Point‐of‐care glucose values were obtained by either patient care technicians or nurses on each individual ward. As an academic medical institution, physicians were frequently paged by other members of the healthcare team.

Each auditor (E.N.D., A.L., L.L.W., K.A.H.) reviewed 1 consistent unit during the audit period. All data for insulin administration errors were tabulated, and descriptive rates of errors were used on a per‐patient or per‐stay basis

Results

A total of 116 patient‐audit periods were identified during the 28‐day study period (Table 1). Sixty‐five patients were on surgical services, and 51 were on medicine services, representing 378 inpatient days. Median length of stay was 3.5 days. Home insulin use was evident in 49% of the surveyed population. Patients' mean A1C (data available within 3 months prior to admission) was 8.1% (n = 41). Inpatient insulin regimens on day 1 included correctional insulin only (51.7% of cases). Regimens containing neutral protamine Hagedorn (NPH) or glargine also included correctional insulin in 95% of cases, and premeal insulin in 35%. Regimens including both premeal insulin and correctional insulin occurred in 25% of the patients. Diet status indicated that 83% of the population was taking an oral diet on day 1, and 13% were nil per os (nothing by mouth [NPO]).

A total of 199 administration errors occurred at a rate of 1.72 errors/patient‐period and 0.53 errors/patient day (Table 2). Missing documentation of doses (15.5% of all patients) and insulin being held without an order (25% of patients) were the most frequently occurring events. Errors classified as other were found in 13.1% of the defined events. These other errors consisted of not carrying out correctional dose insulin orders appropriately (eg, blood sugar value of 149 mg/dL should have resulted in a correctional dose of 2 units, but 3 units were documented as given instead), timing errors related to provision of mealtime insulin apart from documented provision of a meal, or not following the required documentation for insulin pumps.

Forty‐two patients (36%) experienced no errors in insulin administration, 18 patients experienced 1 error, 21 patients had 2 errors, and 11 patients had 3 errors. The remainder of the patients (n = 23; 19.9%) had 4 or more errors during their observation period. Were similar across the units surveyed. Frequency of errors remained consistent regardless of reason for admission, history of diabetes or insulin use at home, or length of stay. Most errors occurred on days 1 and 2 of the hospital stay. Error rates and types were consistent across all units surveyed.

Discussion/Conclusion

We found that insulin administration errors were common in our inpatient snapshot of non‐critically ill patients. In our observational evaluation, 64% of patients had at least 1 error related to insulin administration. Errors related to missing documentation of scheduled doses, or doses held without a prescriber order, were the most common. Implications of missed or held doses could range from unclear approaches for dose adjustment due to missing information, incorrect titration due to incomplete information, or hypoglycemia and hyperglycemia.

This observed rate of error is much higher than the described error rate of 19.5% reported in the United Kingdom.2 This difference in error rates most likely reflects a difference in focus, as investigators in that national effort focused on prescriber error, aberrations in blood glucose values, and readmission rates. Our evaluation in assessing error rates regarding insulin administration supports the use of personnel keenly aware of the processes related to insulin administration, and provides insight into the importance of evaluating small portions of insulin provision (administration vs prescribing, etc) in assessing grounds for improvement in care. It is important to note that our findings may be exaggerated and are not entirely comparable to a study with a different scope and size.

Our snapshot tool and baseline evaluation is a simple method that could be undertaken at many institutions. As such, this methodology and error estimate serves as a gauge for future comparisons and areas for intervention. Limitations of our assessment include the small portion of patients audited during our evaluation versus using a snapshot of our entire hospital, utilizing nonstandardized criteria for determination of insulin errors, and the lack of correlation of clinical significance (aberrations in glucose values) with errors observed. Also, this single‐institution review may not be generalizable to all institutions. Additionally, we only examined errors related to administration of insulin. Other areas that would complete the picture, related to diabetic therapies and outcomes, would need to include prescribing errors or dispensing errors and relate these to glycemic outcomes. Assessment of these additional errors may have revealed more clinically important events that were not revealed in this small snapshot. Lastly, clinical endpoints such as intensive care unit (ICU) transfers, mortality, or readmissions were not assessed in this small study.

We are fortunate that many of these errors were apparently clinically silent, but in a subset of patients, the risk is real and life‐threatening. Risk occurs at both ends of the glucose spectrum, with the low end receiving the greatest attention. Severe hypoglycemia with harm and inpatient diabetic ketoacidosis have been qualified as newer events by Medicare. Hypoglycemia in the ICU population (<40 mg/dL) is an independent marker of mortality.5 Hypoglycemia (<50 mg/dL) has been associated with heart attacks, strokes, and death in the outpatient setting.6

The ability to safely control blood sugar in the hospital requires that medications are administered on time, and that communication occurs between the prescribing provider and the nursing staff providing care. Along with the case‐by‐case implications regarding the need for accurate administration of insulin for subsequent titration and determination of discharge prescriptions for patients with diabetes, there are many implications regarding the assessment of inpatient provision of insulin on determining institutional practices based on previous performance. If insulin administration is not accurately provided or documented, institutions will find it difficult to correctly make changes to insulin protocols for targeting future improvements. Our evaluation indicates an obvious need for quality improvement with 18.1% of the errors reflecting holding insulin without an order, and 12.6% of the errors showing no documentation for the physician being notified of hypoglycemia requiring treatment. The need to foster structured nurse‐physician communication will play a critical role in any process improvement. Communication is key for the optimal provision of insulin in the inpatient setting. Computerized order entry and bar‐code guided administration of doses of insulin may fix some types of the errors (transcription and missed documentation, respectively). That said, one of the largest impacts of this survey may reveal that these errors may not be fixed by technology, but may require more targeted and difficult interventions, such as continuing education and holding clinicians accountable. This study provides insight into the complicated issues regarding inpatient insulin administration and, due to its systematic approach, has given direction for process and system improvements.

Diabetes care in the inpatient setting requires coordination between multiple service providers. Breakdowns in this process occur at all levels leading to potential serious harm.1 Error rates focusing on multiple areas related to diabetes care, including the inpatient provision of insulin, have been described as high as 19.5% in 14,000 patients surveyed in the United Kingdom.2 Missteps are important, as insulin prescribing errors are more commonly associated with patient harm.3 In the United States, medication errors related to provision of care to critically ill patients has been documented, but, to our knowledge, no such reports regarding general medical or surgical wards exist.4

Insulin errors can result from a wide range of possible reasons including: incorrect medication reconciliation, prescribing errors, dispensing errors, administration errors, suboptimal meal timing, or errors in communication for discharge plans regarding diabetes care. Examining each of these areas as a whole could be a daunting task. As such, we sought to examine 1 portion of insulin provision as an initial focus for performance improvement at our institution. Our purpose was to describe the rates of errors associated with insulin administration at our single academic medical center on general medicine and surgical wards.

Methods

Study patients for this observational, prospective snapshot were identified by electronic medical records in 4 consecutive weeks in April 2009 at Barnes‐Jewish Hospital (St Louis, MO), a 1200 bed academic medical center. This study was approved by the Washington University in St Louis School of Medicine Human Studies Committee, and the requirement for informed consent was waived.

On day 1 of each snapshot period, all patients on the identified wards were examined to determine if insulin was currently active as part of the inpatient medication orders. If active, this patient was enrolled into the evaluation data set. No patients were excluded if insulin was currently ordered. Four inpatient areas were selected to provide a representation of the non‐critically ill patient population at our institution. The 4 areas selected were: a cardiac care ward (typical census = 24), a general medicine ward (typical census = 24), an abdominal transplant ward (typical census = 18), and a general surgery ward (typical census = 22). Taken together, these areas represent about 20% of the total non‐critically ill patient population at our hospital. The transplant area was chosen because it represents a high‐risk population with medication (corticosteroid)‐induced diabetes. Nursing and physician care are typically exclusive in these areas, and very little crossover among these healthcare providers would have occurred among the units surveyed during the study period.

Each patient included on day 1 of each audit period was followed for a total of up to 7 days. Patients were only enrolled on day 1 of each audit period. Four survey periods were conducted, providing an evaluation of 28 days of insulin therapy in the studied units. Four periods were selected to pick up more patients on day 1 of each audit period. Electronic records of medication administration and evaluation of paper chart orders provided the information for insulin administration error rates. Additionally, physician notes regarding patients' histories and home insulin use were reviewed for background information for our patient population. Prospective daily assessments of insulin orders, doses charted, nursing notes, and blood glucose values were conducted for potential errors in insulin administration.

All definitions of insulin administration errors were defined prior to data collection. The investigators reviewed available literature involving insulin errors, and found no standardized definitions or previously published assessments at the time of inception of our study. As such, we examined our own clinical practice for areas of potential concern related to insulin administration. The following error categories were identified: transcription errors (eg, insulin glargine 10 units qpm written, but order transcribed and carried out as 20 units qpm); greater than 1 hour between obtained point‐of‐care blood glucose value and provision of correctional (sliding‐scale) insulin; insulin held without a physician order present in the medical records; missing documentation of insulin doses (glucose value of 150 mg/dL present, but no documented correctional dose corresponding to this value present in medical record); premeal and correctional insulin given at separate times; and no documentation of physician notification for hypoglycemia. Other reasons for potential insulin administration errors were collected if deemed pertinent by the individual auditors.

At the time of our survey, a standardized subcutaneous insulin administration order set was utilized in all of the surveyed units. As computerized physician order entry was not yet available at our institution, all orders were transcribed electronically from paper orders. This insulin order set has been in place for 5 years. Once initiated, all portions of the order set are initiated, including communication to nurses regarding glucose measuring times, requirement for documentation of hypoglycemia, and proposed glucose goals. A survey of insulin orders during the audit time revealed that >97% of all insulin orders were initiated from this standardized order set. These order sets encouraged the provision of physiological insulin (basal‐bolus) using insulin glargine and insulin aspart in eligible individuals. Although no systematic, standardized goal for glucose attainment was promoted, a fasting blood glucose of 90‐130 mg/dL and post‐prandial value of <180 mg/dL was encouraged. The order sets had a stated requirement of physician contact for all blood glucose values <70 mg/dL. Although lack of documentation of hypoglycemia may not be directly considered an error associated with administration of insulin, the research group decided to include this provision in the definition of administration errors, given the ability of this parameter to provide a sense of overall completeness of insulin orders and as a marker of collaborative practice in the management of inpatient hyperglycemia.

Nurses documented glucose values and responses in electronic medical administration records as a matter of routine. Point‐of‐care glucose values were obtained by either patient care technicians or nurses on each individual ward. As an academic medical institution, physicians were frequently paged by other members of the healthcare team.

Each auditor (E.N.D., A.L., L.L.W., K.A.H.) reviewed 1 consistent unit during the audit period. All data for insulin administration errors were tabulated, and descriptive rates of errors were used on a per‐patient or per‐stay basis

Results

A total of 116 patient‐audit periods were identified during the 28‐day study period (Table 1). Sixty‐five patients were on surgical services, and 51 were on medicine services, representing 378 inpatient days. Median length of stay was 3.5 days. Home insulin use was evident in 49% of the surveyed population. Patients' mean A1C (data available within 3 months prior to admission) was 8.1% (n = 41). Inpatient insulin regimens on day 1 included correctional insulin only (51.7% of cases). Regimens containing neutral protamine Hagedorn (NPH) or glargine also included correctional insulin in 95% of cases, and premeal insulin in 35%. Regimens including both premeal insulin and correctional insulin occurred in 25% of the patients. Diet status indicated that 83% of the population was taking an oral diet on day 1, and 13% were nil per os (nothing by mouth [NPO]).

A total of 199 administration errors occurred at a rate of 1.72 errors/patient‐period and 0.53 errors/patient day (Table 2). Missing documentation of doses (15.5% of all patients) and insulin being held without an order (25% of patients) were the most frequently occurring events. Errors classified as other were found in 13.1% of the defined events. These other errors consisted of not carrying out correctional dose insulin orders appropriately (eg, blood sugar value of 149 mg/dL should have resulted in a correctional dose of 2 units, but 3 units were documented as given instead), timing errors related to provision of mealtime insulin apart from documented provision of a meal, or not following the required documentation for insulin pumps.

Forty‐two patients (36%) experienced no errors in insulin administration, 18 patients experienced 1 error, 21 patients had 2 errors, and 11 patients had 3 errors. The remainder of the patients (n = 23; 19.9%) had 4 or more errors during their observation period. Were similar across the units surveyed. Frequency of errors remained consistent regardless of reason for admission, history of diabetes or insulin use at home, or length of stay. Most errors occurred on days 1 and 2 of the hospital stay. Error rates and types were consistent across all units surveyed.

Discussion/Conclusion

We found that insulin administration errors were common in our inpatient snapshot of non‐critically ill patients. In our observational evaluation, 64% of patients had at least 1 error related to insulin administration. Errors related to missing documentation of scheduled doses, or doses held without a prescriber order, were the most common. Implications of missed or held doses could range from unclear approaches for dose adjustment due to missing information, incorrect titration due to incomplete information, or hypoglycemia and hyperglycemia.

This observed rate of error is much higher than the described error rate of 19.5% reported in the United Kingdom.2 This difference in error rates most likely reflects a difference in focus, as investigators in that national effort focused on prescriber error, aberrations in blood glucose values, and readmission rates. Our evaluation in assessing error rates regarding insulin administration supports the use of personnel keenly aware of the processes related to insulin administration, and provides insight into the importance of evaluating small portions of insulin provision (administration vs prescribing, etc) in assessing grounds for improvement in care. It is important to note that our findings may be exaggerated and are not entirely comparable to a study with a different scope and size.

Our snapshot tool and baseline evaluation is a simple method that could be undertaken at many institutions. As such, this methodology and error estimate serves as a gauge for future comparisons and areas for intervention. Limitations of our assessment include the small portion of patients audited during our evaluation versus using a snapshot of our entire hospital, utilizing nonstandardized criteria for determination of insulin errors, and the lack of correlation of clinical significance (aberrations in glucose values) with errors observed. Also, this single‐institution review may not be generalizable to all institutions. Additionally, we only examined errors related to administration of insulin. Other areas that would complete the picture, related to diabetic therapies and outcomes, would need to include prescribing errors or dispensing errors and relate these to glycemic outcomes. Assessment of these additional errors may have revealed more clinically important events that were not revealed in this small snapshot. Lastly, clinical endpoints such as intensive care unit (ICU) transfers, mortality, or readmissions were not assessed in this small study.

We are fortunate that many of these errors were apparently clinically silent, but in a subset of patients, the risk is real and life‐threatening. Risk occurs at both ends of the glucose spectrum, with the low end receiving the greatest attention. Severe hypoglycemia with harm and inpatient diabetic ketoacidosis have been qualified as newer events by Medicare. Hypoglycemia in the ICU population (<40 mg/dL) is an independent marker of mortality.5 Hypoglycemia (<50 mg/dL) has been associated with heart attacks, strokes, and death in the outpatient setting.6

The ability to safely control blood sugar in the hospital requires that medications are administered on time, and that communication occurs between the prescribing provider and the nursing staff providing care. Along with the case‐by‐case implications regarding the need for accurate administration of insulin for subsequent titration and determination of discharge prescriptions for patients with diabetes, there are many implications regarding the assessment of inpatient provision of insulin on determining institutional practices based on previous performance. If insulin administration is not accurately provided or documented, institutions will find it difficult to correctly make changes to insulin protocols for targeting future improvements. Our evaluation indicates an obvious need for quality improvement with 18.1% of the errors reflecting holding insulin without an order, and 12.6% of the errors showing no documentation for the physician being notified of hypoglycemia requiring treatment. The need to foster structured nurse‐physician communication will play a critical role in any process improvement. Communication is key for the optimal provision of insulin in the inpatient setting. Computerized order entry and bar‐code guided administration of doses of insulin may fix some types of the errors (transcription and missed documentation, respectively). That said, one of the largest impacts of this survey may reveal that these errors may not be fixed by technology, but may require more targeted and difficult interventions, such as continuing education and holding clinicians accountable. This study provides insight into the complicated issues regarding inpatient insulin administration and, due to its systematic approach, has given direction for process and system improvements.

Interest in healthcare teams has surged in recent years. A majority of the interest has been devoted to teamwork in the interdisciplinary clinical teams that staff operating rooms,1 emergency departments,2 and other inpatient settings.3 Interventions that enhance elements of teamwork like communication, mutual support among team members, and leadership have demonstrated effectiveness.4

Less attention has been paid to improving the success of hospital quality improvement (QI) teams, which gather individuals from different disciplines to improve a defined aspect of care. Studies suggest that QI teams can enable transformational change in healthcare systems,57 and that interdisciplinary representation,8, 9 physician involvement,10, 11 and clear goals12, 13 are associated with successful QI efforts. However, few studies have examined the behaviors of the QI teams that planned and implemented these efforts. Understanding how QI teams work to achieve their goals will allow hospitals to encourage these behaviors, and allow researchers to design interventions to augment these behaviors.

Accordingly, we sought to characterize the behaviors of successful interdisciplinary hospital QI teams. We previously reported on the strategies used by hospitals to reduce door‐to‐balloon times for patients with ST‐elevation myocardial infarction (STEMI)14, 15 to the evidence‐based guideline of 90 minutes.16 Our objective is to examine how QI teams designed and implemented these strategies. We believe that studying high‐performing QI teams is a first step to developing testable hypotheses about the effectiveness of QI team behaviors and mechanisms by which these behaviors might produce positive team outcomes.

METHODS

We designed a qualitative study using in‐depth interviews. We selected a qualitative methodology, since behaviors, social norms, and interpersonal interactions can be most appropriately examined using qualitative methods.17, 18 In addition, we used a positive deviance approach,19 where we focused on hospitals with top performance and the most improvement in door‐to‐balloon times. We sampled from hospitals in the National Registry of Myocardial Infarction (NRMI) who perform percutaneous coronary intervention (PCI, n = 151). We selected hospitals whose median door‐to‐balloon times were 90 minutes (n = 35). Then, we ranked hospitals in descending order according to their improvement during the previous 3 years (19992002). We sampled hospitals in descending order until we reached theoretical saturation where, as recommended for qualitative inquiry,2022 additional site visits did not uncover new concepts or patterns regarding our study questions. All sampled hospitals agreed to participate.

The first contact at each hospital was typically the director of QI. We asked to interview anyone with substantial involvement in the effort to reduce door‐to‐balloon times, and suggested that a wide variety of disciplines and roles be represented. We also used the snowball technique,22 where we asked participants to provide the names of individuals with substantial involvement in the reducing door‐to‐balloon times. Participants had varied levels of participation in QI teams. We purposely asked for minority and dissenting views from all participants.

At least 2 members of the research team conducted in‐depth interviews during hospital site visits. Interviews were conducted individually or in small groups, and lasted 1 to 1.5 hours. All data were audiotaped after verbal consent. Our interviews began with the grand tour question: What, if anything, has this hospital done to reduce its door‐to‐balloon times for patients with STEMI? The research team used standardized probes20, 23 to guide the discussion and achieve a complete understanding of the phenomena under study, including leadership and activities of the QI teams, and recommendations to other hospitals that wished to reduce door‐to‐balloon times. As recommended by experts,23 our interview guide was purposefully open‐ended to capture the range of experiences with QI teams. We did not specifically probe for facilitating or challenging behaviors. Audiotapes were transcribed by an independent, professional transcriptionist.

For this analysis, we defined QI teams as groups of administrators, providers, and staff who designed, implemented, and monitored processes to reduce door‐to‐balloon times. Each analysis team member independently cataloged quotes about team behaviors using a list of concepts (or codes). We then analyzed the quotes to identify recurrent themes relevant to the behaviors of interdisciplinary QI teams. We used the constant comparative method of analysis,20, 24, 25 which stipulates that the initial list of codes is refined as new transcripts are analyzed, and the final list is applied to all the transcripts. The analysis team included experts in QI, medicine, qualitative and health services research, as well as organizational psychology, and one of the interviewers. The presence of diverse perspectives in the analysis team,21 and a detailed audit trail20 to document the emergence of codes and themes, helped enhance researcher neutrality, data accuracy, and validity. We used Atlas.ti version 5.2 (Scientific Software Development GMbH, Berlin, Germany) to assist in the analysis.

RESULTS

Our sample (n = 11) included hospitals that varied on several characteristics (eg, geographic location), and median door‐to‐balloon times ranged from 55.5 to 89.5 minutes (Table 1). Hospitals in our sample had higher mean improvements in door‐to‐balloon times compared with non‐sampled NRMI hospitals (n = 140, 24 minutes vs 3 minutes over 3 years). Our interview participants (n = 122) included physicians, nurses, QI personnel, and administrative staff (Table 2). Five behaviors emerged from the data analysis. We found that interdisciplinary QI teams in successful hospitals focused on: (1) motivating involved hospital staff towards a shared goal, (2) creating opportunities for learning and problem‐solving, (3) addressing the impact of changes in care processes on staff, (4) protecting the integrity of the newly developed care processes, and (5) representing each involved clinical discipline effectively. These behaviors were recurrent across our diverse set of hospitals.

Discussion

We identified 5 behaviors of successful interdisciplinary QI teams based on our analysis of hospitals that reduced door‐to‐balloon times for patients with STEMI. These QI teams: (1) motivated involved hospital staff to consider lowering door‐to‐balloon times, a shared goal, (2) created opportunities for learning and problem‐solving, (3) addressed the impact of changes to care processes for patients with STEMI on staff, (4) protected the integrity of new care processes, and (5) represented each clinical discipline effectively by having members with in‐depth knowledge and authority.

Experts suggest that the key elements of effective teamwork in healthcare include prioritizing team over individual goals, mutual understanding, leadership, adaptability, and anticipation of the needs of others.26 These elements are supported by mutual trust and closed‐loop communication. The behaviors of QI teams in our study represent adaptive responses to the unique demands of QI in a complex organization. These teams went beyond an improvement model of identifying and analyzing a problem, and then developing and testing solutions by: (1) motivating and gathering information from each discipline, regardless of interdisciplinary conflicts; (2) responding to the concerns of front‐line staff, while maintaining control over the improvement process; and (3) sharing information across the hospital hierarchy. Table 3 illustrates potential relationships between the team behaviors in our data, the demands on hospital QI teams, and known elements of effective teamwork.

The behaviors in our study suggest effective teamwork strategies for QI. For example, our data suggest that successful interdisciplinary QI teams need effective representation from each involved discipline. This representation is necessary for motivation of front‐line staff, gathering of detailed information about processes, and the effective implementation of changes. Although this level of representation might challenge the cohesiveness of some teams,27 the teams in our sample managed conflict among disciplines without sacrificing the shared goal. By allocating attention and resources to the concerns of each discipline, the teams we studied prioritized team over individual goals and promoted mutual understanding.

Similarly, deciding when to modify the new protocols required leadership, adaptability, and anticipation of the needs of others. Successful QI teams in our sample modified protocols based on data and feedback, and created the mutual trust environment that is known to facilitate learning among disciplines.2830 Their willingness to learn, however, did not deter teams from protecting the integrity of new protocols. Lastly, participants stressed the importance of managing information across hierarchical boundaries. Managing reliable, timely, and accurate information across all levels is crucial to teamwork, and to the power and influence of a team.31

Our conclusions should be interpreted in light of several limitations. First, our study did not include a comparison group of low‐performing hospitals. We followed the recommendations of qualitative research experts23 who recommend sampling those with the most information on, and experience with, the phenomena under study (QI teams in high‐performing hospitals). The hypotheses we present here require further testing in quantitative studies of hospitals with diversity in QI team outcomes. Second, it is possible that sampled participants favored responses that they considered more desirable. To minimize this bias, we interviewed multiple participants per hospital, assured their confidentiality, and asked them to elaborate their responses. We sampled participants with a wide range of clinical and operational roles in each hospital, and also used the snowball sampling method to augment our sample. The range of responses collected, including frank discussions about setbacks, argues against the existence of contrasting behaviors to those captured. Third, although our sample included hospitals of various size and location, our findings might not reflect those of a larger sample of US hospitals. Last, the behaviors of QI teams may differ for other clinical processes.

Translating these findings into practice will require future studies of the impact of QI team behaviors on sustainability of quality gains. Since QI teams are not typically permanent, additional research is needed to identify behaviors associated with sustainable improvements. In addition, we must test whether the relationship between behaviors and team outcomes depends on whether the QI team strives to reach an evidence‐based goal or to improve a process as much as possible. Our sample demonstrated a combined approach, where the evidence‐based goal was followed by a desire to continue to further reduce door‐to‐balloon times. Similarly, the relationship between behaviors and team outcomes might depend on the catalyst for improvement (eg, regulatory pressure, an adverse event). The confluence of strong evidence and regulatory pressure that fueled these teams might not be true for other measures. Lastly, studies of teamwork in QI teams will require objective measures of team behaviors. A combination of surveys and direct team observation will likely be required to measure these behaviors, especially effective representation.

Our study highlights behaviors common to successful interdisciplinary QI teams in high‐performing hospitals. Previous studies have identified elements of teamwork and the importance of teams to QI, but have not examined team behaviors. In the era of an ever‐growing list of quality measures and of movement toward performance‐based reimbursement models,3234 hospitals have embraced the use of interdisciplinary teams as a key component of QI efforts. Our findings suggest that hospitals could enhance QI team effectiveness by promoting behaviors associated with successful interdisciplinary teams. When applied to QI teams, teamwork training could be supplemented with knowledge, attitudes, and skills regarding information‐gathering, problem‐solving, and communication across disciplines and levels of the hospital hierarchy.

Interest in healthcare teams has surged in recent years. A majority of the interest has been devoted to teamwork in the interdisciplinary clinical teams that staff operating rooms,1 emergency departments,2 and other inpatient settings.3 Interventions that enhance elements of teamwork like communication, mutual support among team members, and leadership have demonstrated effectiveness.4

Less attention has been paid to improving the success of hospital quality improvement (QI) teams, which gather individuals from different disciplines to improve a defined aspect of care. Studies suggest that QI teams can enable transformational change in healthcare systems,57 and that interdisciplinary representation,8, 9 physician involvement,10, 11 and clear goals12, 13 are associated with successful QI efforts. However, few studies have examined the behaviors of the QI teams that planned and implemented these efforts. Understanding how QI teams work to achieve their goals will allow hospitals to encourage these behaviors, and allow researchers to design interventions to augment these behaviors.

Accordingly, we sought to characterize the behaviors of successful interdisciplinary hospital QI teams. We previously reported on the strategies used by hospitals to reduce door‐to‐balloon times for patients with ST‐elevation myocardial infarction (STEMI)14, 15 to the evidence‐based guideline of 90 minutes.16 Our objective is to examine how QI teams designed and implemented these strategies. We believe that studying high‐performing QI teams is a first step to developing testable hypotheses about the effectiveness of QI team behaviors and mechanisms by which these behaviors might produce positive team outcomes.

METHODS

We designed a qualitative study using in‐depth interviews. We selected a qualitative methodology, since behaviors, social norms, and interpersonal interactions can be most appropriately examined using qualitative methods.17, 18 In addition, we used a positive deviance approach,19 where we focused on hospitals with top performance and the most improvement in door‐to‐balloon times. We sampled from hospitals in the National Registry of Myocardial Infarction (NRMI) who perform percutaneous coronary intervention (PCI, n = 151). We selected hospitals whose median door‐to‐balloon times were 90 minutes (n = 35). Then, we ranked hospitals in descending order according to their improvement during the previous 3 years (19992002). We sampled hospitals in descending order until we reached theoretical saturation where, as recommended for qualitative inquiry,2022 additional site visits did not uncover new concepts or patterns regarding our study questions. All sampled hospitals agreed to participate.

The first contact at each hospital was typically the director of QI. We asked to interview anyone with substantial involvement in the effort to reduce door‐to‐balloon times, and suggested that a wide variety of disciplines and roles be represented. We also used the snowball technique,22 where we asked participants to provide the names of individuals with substantial involvement in the reducing door‐to‐balloon times. Participants had varied levels of participation in QI teams. We purposely asked for minority and dissenting views from all participants.

At least 2 members of the research team conducted in‐depth interviews during hospital site visits. Interviews were conducted individually or in small groups, and lasted 1 to 1.5 hours. All data were audiotaped after verbal consent. Our interviews began with the grand tour question: What, if anything, has this hospital done to reduce its door‐to‐balloon times for patients with STEMI? The research team used standardized probes20, 23 to guide the discussion and achieve a complete understanding of the phenomena under study, including leadership and activities of the QI teams, and recommendations to other hospitals that wished to reduce door‐to‐balloon times. As recommended by experts,23 our interview guide was purposefully open‐ended to capture the range of experiences with QI teams. We did not specifically probe for facilitating or challenging behaviors. Audiotapes were transcribed by an independent, professional transcriptionist.

For this analysis, we defined QI teams as groups of administrators, providers, and staff who designed, implemented, and monitored processes to reduce door‐to‐balloon times. Each analysis team member independently cataloged quotes about team behaviors using a list of concepts (or codes). We then analyzed the quotes to identify recurrent themes relevant to the behaviors of interdisciplinary QI teams. We used the constant comparative method of analysis,20, 24, 25 which stipulates that the initial list of codes is refined as new transcripts are analyzed, and the final list is applied to all the transcripts. The analysis team included experts in QI, medicine, qualitative and health services research, as well as organizational psychology, and one of the interviewers. The presence of diverse perspectives in the analysis team,21 and a detailed audit trail20 to document the emergence of codes and themes, helped enhance researcher neutrality, data accuracy, and validity. We used Atlas.ti version 5.2 (Scientific Software Development GMbH, Berlin, Germany) to assist in the analysis.

RESULTS

Our sample (n = 11) included hospitals that varied on several characteristics (eg, geographic location), and median door‐to‐balloon times ranged from 55.5 to 89.5 minutes (Table 1). Hospitals in our sample had higher mean improvements in door‐to‐balloon times compared with non‐sampled NRMI hospitals (n = 140, 24 minutes vs 3 minutes over 3 years). Our interview participants (n = 122) included physicians, nurses, QI personnel, and administrative staff (Table 2). Five behaviors emerged from the data analysis. We found that interdisciplinary QI teams in successful hospitals focused on: (1) motivating involved hospital staff towards a shared goal, (2) creating opportunities for learning and problem‐solving, (3) addressing the impact of changes in care processes on staff, (4) protecting the integrity of the newly developed care processes, and (5) representing each involved clinical discipline effectively. These behaviors were recurrent across our diverse set of hospitals.

Discussion

We identified 5 behaviors of successful interdisciplinary QI teams based on our analysis of hospitals that reduced door‐to‐balloon times for patients with STEMI. These QI teams: (1) motivated involved hospital staff to consider lowering door‐to‐balloon times, a shared goal, (2) created opportunities for learning and problem‐solving, (3) addressed the impact of changes to care processes for patients with STEMI on staff, (4) protected the integrity of new care processes, and (5) represented each clinical discipline effectively by having members with in‐depth knowledge and authority.

Experts suggest that the key elements of effective teamwork in healthcare include prioritizing team over individual goals, mutual understanding, leadership, adaptability, and anticipation of the needs of others.26 These elements are supported by mutual trust and closed‐loop communication. The behaviors of QI teams in our study represent adaptive responses to the unique demands of QI in a complex organization. These teams went beyond an improvement model of identifying and analyzing a problem, and then developing and testing solutions by: (1) motivating and gathering information from each discipline, regardless of interdisciplinary conflicts; (2) responding to the concerns of front‐line staff, while maintaining control over the improvement process; and (3) sharing information across the hospital hierarchy. Table 3 illustrates potential relationships between the team behaviors in our data, the demands on hospital QI teams, and known elements of effective teamwork.

The behaviors in our study suggest effective teamwork strategies for QI. For example, our data suggest that successful interdisciplinary QI teams need effective representation from each involved discipline. This representation is necessary for motivation of front‐line staff, gathering of detailed information about processes, and the effective implementation of changes. Although this level of representation might challenge the cohesiveness of some teams,27 the teams in our sample managed conflict among disciplines without sacrificing the shared goal. By allocating attention and resources to the concerns of each discipline, the teams we studied prioritized team over individual goals and promoted mutual understanding.

Similarly, deciding when to modify the new protocols required leadership, adaptability, and anticipation of the needs of others. Successful QI teams in our sample modified protocols based on data and feedback, and created the mutual trust environment that is known to facilitate learning among disciplines.2830 Their willingness to learn, however, did not deter teams from protecting the integrity of new protocols. Lastly, participants stressed the importance of managing information across hierarchical boundaries. Managing reliable, timely, and accurate information across all levels is crucial to teamwork, and to the power and influence of a team.31

Our conclusions should be interpreted in light of several limitations. First, our study did not include a comparison group of low‐performing hospitals. We followed the recommendations of qualitative research experts23 who recommend sampling those with the most information on, and experience with, the phenomena under study (QI teams in high‐performing hospitals). The hypotheses we present here require further testing in quantitative studies of hospitals with diversity in QI team outcomes. Second, it is possible that sampled participants favored responses that they considered more desirable. To minimize this bias, we interviewed multiple participants per hospital, assured their confidentiality, and asked them to elaborate their responses. We sampled participants with a wide range of clinical and operational roles in each hospital, and also used the snowball sampling method to augment our sample. The range of responses collected, including frank discussions about setbacks, argues against the existence of contrasting behaviors to those captured. Third, although our sample included hospitals of various size and location, our findings might not reflect those of a larger sample of US hospitals. Last, the behaviors of QI teams may differ for other clinical processes.

Translating these findings into practice will require future studies of the impact of QI team behaviors on sustainability of quality gains. Since QI teams are not typically permanent, additional research is needed to identify behaviors associated with sustainable improvements. In addition, we must test whether the relationship between behaviors and team outcomes depends on whether the QI team strives to reach an evidence‐based goal or to improve a process as much as possible. Our sample demonstrated a combined approach, where the evidence‐based goal was followed by a desire to continue to further reduce door‐to‐balloon times. Similarly, the relationship between behaviors and team outcomes might depend on the catalyst for improvement (eg, regulatory pressure, an adverse event). The confluence of strong evidence and regulatory pressure that fueled these teams might not be true for other measures. Lastly, studies of teamwork in QI teams will require objective measures of team behaviors. A combination of surveys and direct team observation will likely be required to measure these behaviors, especially effective representation.

Our study highlights behaviors common to successful interdisciplinary QI teams in high‐performing hospitals. Previous studies have identified elements of teamwork and the importance of teams to QI, but have not examined team behaviors. In the era of an ever‐growing list of quality measures and of movement toward performance‐based reimbursement models,3234 hospitals have embraced the use of interdisciplinary teams as a key component of QI efforts. Our findings suggest that hospitals could enhance QI team effectiveness by promoting behaviors associated with successful interdisciplinary teams. When applied to QI teams, teamwork training could be supplemented with knowledge, attitudes, and skills regarding information‐gathering, problem‐solving, and communication across disciplines and levels of the hospital hierarchy.

Interest in healthcare teams has surged in recent years. A majority of the interest has been devoted to teamwork in the interdisciplinary clinical teams that staff operating rooms,1 emergency departments,2 and other inpatient settings.3 Interventions that enhance elements of teamwork like communication, mutual support among team members, and leadership have demonstrated effectiveness.4

Less attention has been paid to improving the success of hospital quality improvement (QI) teams, which gather individuals from different disciplines to improve a defined aspect of care. Studies suggest that QI teams can enable transformational change in healthcare systems,57 and that interdisciplinary representation,8, 9 physician involvement,10, 11 and clear goals12, 13 are associated with successful QI efforts. However, few studies have examined the behaviors of the QI teams that planned and implemented these efforts. Understanding how QI teams work to achieve their goals will allow hospitals to encourage these behaviors, and allow researchers to design interventions to augment these behaviors.

Accordingly, we sought to characterize the behaviors of successful interdisciplinary hospital QI teams. We previously reported on the strategies used by hospitals to reduce door‐to‐balloon times for patients with ST‐elevation myocardial infarction (STEMI)14, 15 to the evidence‐based guideline of 90 minutes.16 Our objective is to examine how QI teams designed and implemented these strategies. We believe that studying high‐performing QI teams is a first step to developing testable hypotheses about the effectiveness of QI team behaviors and mechanisms by which these behaviors might produce positive team outcomes.

METHODS

We designed a qualitative study using in‐depth interviews. We selected a qualitative methodology, since behaviors, social norms, and interpersonal interactions can be most appropriately examined using qualitative methods.17, 18 In addition, we used a positive deviance approach,19 where we focused on hospitals with top performance and the most improvement in door‐to‐balloon times. We sampled from hospitals in the National Registry of Myocardial Infarction (NRMI) who perform percutaneous coronary intervention (PCI, n = 151). We selected hospitals whose median door‐to‐balloon times were 90 minutes (n = 35). Then, we ranked hospitals in descending order according to their improvement during the previous 3 years (19992002). We sampled hospitals in descending order until we reached theoretical saturation where, as recommended for qualitative inquiry,2022 additional site visits did not uncover new concepts or patterns regarding our study questions. All sampled hospitals agreed to participate.

The first contact at each hospital was typically the director of QI. We asked to interview anyone with substantial involvement in the effort to reduce door‐to‐balloon times, and suggested that a wide variety of disciplines and roles be represented. We also used the snowball technique,22 where we asked participants to provide the names of individuals with substantial involvement in the reducing door‐to‐balloon times. Participants had varied levels of participation in QI teams. We purposely asked for minority and dissenting views from all participants.

At least 2 members of the research team conducted in‐depth interviews during hospital site visits. Interviews were conducted individually or in small groups, and lasted 1 to 1.5 hours. All data were audiotaped after verbal consent. Our interviews began with the grand tour question: What, if anything, has this hospital done to reduce its door‐to‐balloon times for patients with STEMI? The research team used standardized probes20, 23 to guide the discussion and achieve a complete understanding of the phenomena under study, including leadership and activities of the QI teams, and recommendations to other hospitals that wished to reduce door‐to‐balloon times. As recommended by experts,23 our interview guide was purposefully open‐ended to capture the range of experiences with QI teams. We did not specifically probe for facilitating or challenging behaviors. Audiotapes were transcribed by an independent, professional transcriptionist.

For this analysis, we defined QI teams as groups of administrators, providers, and staff who designed, implemented, and monitored processes to reduce door‐to‐balloon times. Each analysis team member independently cataloged quotes about team behaviors using a list of concepts (or codes). We then analyzed the quotes to identify recurrent themes relevant to the behaviors of interdisciplinary QI teams. We used the constant comparative method of analysis,20, 24, 25 which stipulates that the initial list of codes is refined as new transcripts are analyzed, and the final list is applied to all the transcripts. The analysis team included experts in QI, medicine, qualitative and health services research, as well as organizational psychology, and one of the interviewers. The presence of diverse perspectives in the analysis team,21 and a detailed audit trail20 to document the emergence of codes and themes, helped enhance researcher neutrality, data accuracy, and validity. We used Atlas.ti version 5.2 (Scientific Software Development GMbH, Berlin, Germany) to assist in the analysis.

RESULTS

Our sample (n = 11) included hospitals that varied on several characteristics (eg, geographic location), and median door‐to‐balloon times ranged from 55.5 to 89.5 minutes (Table 1). Hospitals in our sample had higher mean improvements in door‐to‐balloon times compared with non‐sampled NRMI hospitals (n = 140, 24 minutes vs 3 minutes over 3 years). Our interview participants (n = 122) included physicians, nurses, QI personnel, and administrative staff (Table 2). Five behaviors emerged from the data analysis. We found that interdisciplinary QI teams in successful hospitals focused on: (1) motivating involved hospital staff towards a shared goal, (2) creating opportunities for learning and problem‐solving, (3) addressing the impact of changes in care processes on staff, (4) protecting the integrity of the newly developed care processes, and (5) representing each involved clinical discipline effectively. These behaviors were recurrent across our diverse set of hospitals.

Discussion

We identified 5 behaviors of successful interdisciplinary QI teams based on our analysis of hospitals that reduced door‐to‐balloon times for patients with STEMI. These QI teams: (1) motivated involved hospital staff to consider lowering door‐to‐balloon times, a shared goal, (2) created opportunities for learning and problem‐solving, (3) addressed the impact of changes to care processes for patients with STEMI on staff, (4) protected the integrity of new care processes, and (5) represented each clinical discipline effectively by having members with in‐depth knowledge and authority.

Experts suggest that the key elements of effective teamwork in healthcare include prioritizing team over individual goals, mutual understanding, leadership, adaptability, and anticipation of the needs of others.26 These elements are supported by mutual trust and closed‐loop communication. The behaviors of QI teams in our study represent adaptive responses to the unique demands of QI in a complex organization. These teams went beyond an improvement model of identifying and analyzing a problem, and then developing and testing solutions by: (1) motivating and gathering information from each discipline, regardless of interdisciplinary conflicts; (2) responding to the concerns of front‐line staff, while maintaining control over the improvement process; and (3) sharing information across the hospital hierarchy. Table 3 illustrates potential relationships between the team behaviors in our data, the demands on hospital QI teams, and known elements of effective teamwork.

The behaviors in our study suggest effective teamwork strategies for QI. For example, our data suggest that successful interdisciplinary QI teams need effective representation from each involved discipline. This representation is necessary for motivation of front‐line staff, gathering of detailed information about processes, and the effective implementation of changes. Although this level of representation might challenge the cohesiveness of some teams,27 the teams in our sample managed conflict among disciplines without sacrificing the shared goal. By allocating attention and resources to the concerns of each discipline, the teams we studied prioritized team over individual goals and promoted mutual understanding.

Similarly, deciding when to modify the new protocols required leadership, adaptability, and anticipation of the needs of others. Successful QI teams in our sample modified protocols based on data and feedback, and created the mutual trust environment that is known to facilitate learning among disciplines.2830 Their willingness to learn, however, did not deter teams from protecting the integrity of new protocols. Lastly, participants stressed the importance of managing information across hierarchical boundaries. Managing reliable, timely, and accurate information across all levels is crucial to teamwork, and to the power and influence of a team.31

Our conclusions should be interpreted in light of several limitations. First, our study did not include a comparison group of low‐performing hospitals. We followed the recommendations of qualitative research experts23 who recommend sampling those with the most information on, and experience with, the phenomena under study (QI teams in high‐performing hospitals). The hypotheses we present here require further testing in quantitative studies of hospitals with diversity in QI team outcomes. Second, it is possible that sampled participants favored responses that they considered more desirable. To minimize this bias, we interviewed multiple participants per hospital, assured their confidentiality, and asked them to elaborate their responses. We sampled participants with a wide range of clinical and operational roles in each hospital, and also used the snowball sampling method to augment our sample. The range of responses collected, including frank discussions about setbacks, argues against the existence of contrasting behaviors to those captured. Third, although our sample included hospitals of various size and location, our findings might not reflect those of a larger sample of US hospitals. Last, the behaviors of QI teams may differ for other clinical processes.

Translating these findings into practice will require future studies of the impact of QI team behaviors on sustainability of quality gains. Since QI teams are not typically permanent, additional research is needed to identify behaviors associated with sustainable improvements. In addition, we must test whether the relationship between behaviors and team outcomes depends on whether the QI team strives to reach an evidence‐based goal or to improve a process as much as possible. Our sample demonstrated a combined approach, where the evidence‐based goal was followed by a desire to continue to further reduce door‐to‐balloon times. Similarly, the relationship between behaviors and team outcomes might depend on the catalyst for improvement (eg, regulatory pressure, an adverse event). The confluence of strong evidence and regulatory pressure that fueled these teams might not be true for other measures. Lastly, studies of teamwork in QI teams will require objective measures of team behaviors. A combination of surveys and direct team observation will likely be required to measure these behaviors, especially effective representation.

Our study highlights behaviors common to successful interdisciplinary QI teams in high‐performing hospitals. Previous studies have identified elements of teamwork and the importance of teams to QI, but have not examined team behaviors. In the era of an ever‐growing list of quality measures and of movement toward performance‐based reimbursement models,3234 hospitals have embraced the use of interdisciplinary teams as a key component of QI efforts. Our findings suggest that hospitals could enhance QI team effectiveness by promoting behaviors associated with successful interdisciplinary teams. When applied to QI teams, teamwork training could be supplemented with knowledge, attitudes, and skills regarding information‐gathering, problem‐solving, and communication across disciplines and levels of the hospital hierarchy.

Healthcare quality is defined as the degree to which health services for individuals and populations increase the likelihood of desired health outcomes and are consistent with current professional knowledge.1 Delivering high quality care to patients in the hospital setting is especially challenging, given the rapid pace of clinical care, the severity and multitude of patient conditions, and the interdependence of complex processes within the hospital system. Research has shown that hospitalized patients do not consistently receive recommended care2 and are at risk for experiencing preventable harm.3 In an effort to stimulate improvement, stakeholders have called for increased accountability, including enhanced transparency and differential payment based on performance. A growing number of hospital process and outcome measures are readily available to the public via the Internet.46 The Joint Commission, which accredits US hospitals, requires the collection of core quality measure data7 and sets the expectation that National Patient Safety Goals be met to maintain accreditation.8 Moreover, the Center for Medicare and Medicaid Services (CMS) has developed a Value‐Based Purchasing (VBP) plan intended to adjust hospital payment based on quality measures and the occurrence of certain hospital‐acquired conditions.9, 10

Because of their clinical expertise, understanding of hospital clinical operations, leadership of multidisciplinary inpatient teams, and vested interest to improve the systems in which they work, hospitalists are perfectly positioned to collaborate with their institutions to improve the quality of care delivered to inpatients. However, many hospitalists are inadequately prepared to engage in efforts to improve quality, because medical schools and residency programs have not traditionally included or emphasized healthcare quality and patient safety in their curricula.1113 In a survey of 389 internal medicine‐trained hospitalists, significant educational deficiencies were identified in the area of systems‐based practice.14 Specifically, the topics of quality improvement, team management, practice guideline development, health information systems management, and coordination of care between healthcare settings were listed as essential skills for hospitalist practice but underemphasized in residency training. Recognizing the gap between the needs of practicing physicians and current medical education provided in healthcare quality, professional societies have recently published position papers calling for increased training in quality, safety, and systems, both in medical school11 and residency training.15, 16

The Society of Hospital Medicine (SHM) convened a Quality Summit in December 2008 to develop strategic plans related to healthcare quality. Summit attendees felt that most hospitalists lack the formal training necessary to evaluate, implement, and sustain system changes within the hospital. In response, the SHM Hospital Quality and Patient Safety (HQPS) Committee formed a Quality Improvement Education (QIE) subcommittee in 2009 to assess the needs of hospitalists with respect to hospital quality and patient safety, and to evaluate and expand upon existing educational programs in this area. Membership of the QIE subcommittee consisted of hospitalists with extensive experience in healthcare quality and medical education. The QIE subcommittee refined and expanded upon the healthcare quality and patient safety‐related competencies initially described in the Core Competencies in Hospital Medicine.17 The purpose of this report is to describe the development, provide definitions, and make recommendations on the use of the Hospital Quality and Patient Safety (HQPS) Competencies.

Development of The Hospital Quality and Patient Safety Competencies

The multistep process used by the SHM QIE subcommittee to develop the HQPS Competencies is summarized in Figure 1. We performed an in‐depth evaluation of current educational materials and offerings, including a review of the Core Competencies in Hospital Medicine, past annual SHM Quality Improvement Pre‐Course objectives, and the content of training courses offered by other organizations.1722 Throughout our analysis, we emphasized the identification of gaps in content relevant to hospitalists. We then used the Institute of Medicine's (IOM) 6 aims for healthcare quality as a foundation for developing the HQPS Competencies.1 Specifically, the IOM states that healthcare should be safe, effective, patient‐centered, timely, efficient, and equitable. Additionally, we reviewed and integrated elements of the Practice‐Based Learning and Improvement (PBLI) and Systems‐Based Practice (SBP) competencies as defined by the Accreditation Council for Graduate Medical Education (ACGME).23 We defined general areas of competence and specific standards for knowledge, skills, and attitudes within each area. Subcommittee members reflected on their own experience, as clinicians, educators, and leaders in healthcare quality and patient safety, to inform and refine the competency definitions and standards. Acknowledging that some hospitalists may serve as collaborators or clinical content experts, while others may serve as leaders of hospital quality initiatives, 3 levels of expertise were established: basic, intermediate, and advanced.

Figure 1

The QIE subcommittee presented a draft version of the HQPS Competencies to the HQPS Committee in the fall of 2009 and incorporated suggested revisions. The revised set of competencies was then reviewed by members of the Leadership and Education Committees during the winter of 2009‐2010, and additional recommendations were included in the final version now described.

Description of The Competencies

The 8 areas of competence include: Quality Measurement and Stakeholder Interests, Data Acquisition and Interpretation, Organizational Knowledge and Leadership Skills, Patient Safety Principles, Teamwork and Communication, Quality and Safety Improvement Methods, Health Information Systems, and Patient Centeredness. Three levels of competence and standards within each level and area are defined in Table 1. Standards use carefully selected action verbs to reflect educational goals for hospitalists at each level.24 The basic level represents a minimum level of competency for all practicing hospitalists. The intermediate level represents a hospitalist who is prepared to meaningfully engage and collaborate with his or her institution in quality improvement efforts. A hospitalist at this level may also lead uncomplicated improvement projects for his or her medical center and/or hospital medicine group. The advanced level represents a hospitalist prepared to lead quality improvement efforts for his or her institution and/or hospital medicine group. Many hospitalists at this level will have, or will be prepared to have, leadership positions in quality and patient safety at their institutions. Advanced level hospitalists will also have the expertise to teach and mentor other individuals in their quality improvement efforts.

Recommended Use of The Competencies

The HQPS Competencies provide a framework for curricula and other professional development experiences in healthcare quality and patient safety. We recommend a step‐wise approach to curriculum development which includes conducting a targeted needs assessment, defining goals and specific learning objectives, and evaluation of the curriculum.25 The HQPS Competencies can be used at each step and provide educational targets for learners across a range of interest and experience.

Conclusion

Hospitalists are poised to have a tremendous impact on improving the quality of care for hospitalized patients. The lack of training in quality improvement in traditional medical education programs, in which most current hospitalists were trained, can be overcome through appropriate use of the HQPS Competencies. Formal incorporation of the HQPS Competencies into professional development programs, and innovative educational initiatives and curricula, will help provide current hospitalists and the next generations of hospitalists with the needed skills to be successful.

Healthcare quality is defined as the degree to which health services for individuals and populations increase the likelihood of desired health outcomes and are consistent with current professional knowledge.1 Delivering high quality care to patients in the hospital setting is especially challenging, given the rapid pace of clinical care, the severity and multitude of patient conditions, and the interdependence of complex processes within the hospital system. Research has shown that hospitalized patients do not consistently receive recommended care2 and are at risk for experiencing preventable harm.3 In an effort to stimulate improvement, stakeholders have called for increased accountability, including enhanced transparency and differential payment based on performance. A growing number of hospital process and outcome measures are readily available to the public via the Internet.46 The Joint Commission, which accredits US hospitals, requires the collection of core quality measure data7 and sets the expectation that National Patient Safety Goals be met to maintain accreditation.8 Moreover, the Center for Medicare and Medicaid Services (CMS) has developed a Value‐Based Purchasing (VBP) plan intended to adjust hospital payment based on quality measures and the occurrence of certain hospital‐acquired conditions.9, 10

Because of their clinical expertise, understanding of hospital clinical operations, leadership of multidisciplinary inpatient teams, and vested interest to improve the systems in which they work, hospitalists are perfectly positioned to collaborate with their institutions to improve the quality of care delivered to inpatients. However, many hospitalists are inadequately prepared to engage in efforts to improve quality, because medical schools and residency programs have not traditionally included or emphasized healthcare quality and patient safety in their curricula.1113 In a survey of 389 internal medicine‐trained hospitalists, significant educational deficiencies were identified in the area of systems‐based practice.14 Specifically, the topics of quality improvement, team management, practice guideline development, health information systems management, and coordination of care between healthcare settings were listed as essential skills for hospitalist practice but underemphasized in residency training. Recognizing the gap between the needs of practicing physicians and current medical education provided in healthcare quality, professional societies have recently published position papers calling for increased training in quality, safety, and systems, both in medical school11 and residency training.15, 16

The Society of Hospital Medicine (SHM) convened a Quality Summit in December 2008 to develop strategic plans related to healthcare quality. Summit attendees felt that most hospitalists lack the formal training necessary to evaluate, implement, and sustain system changes within the hospital. In response, the SHM Hospital Quality and Patient Safety (HQPS) Committee formed a Quality Improvement Education (QIE) subcommittee in 2009 to assess the needs of hospitalists with respect to hospital quality and patient safety, and to evaluate and expand upon existing educational programs in this area. Membership of the QIE subcommittee consisted of hospitalists with extensive experience in healthcare quality and medical education. The QIE subcommittee refined and expanded upon the healthcare quality and patient safety‐related competencies initially described in the Core Competencies in Hospital Medicine.17 The purpose of this report is to describe the development, provide definitions, and make recommendations on the use of the Hospital Quality and Patient Safety (HQPS) Competencies.

Development of The Hospital Quality and Patient Safety Competencies

The multistep process used by the SHM QIE subcommittee to develop the HQPS Competencies is summarized in Figure 1. We performed an in‐depth evaluation of current educational materials and offerings, including a review of the Core Competencies in Hospital Medicine, past annual SHM Quality Improvement Pre‐Course objectives, and the content of training courses offered by other organizations.1722 Throughout our analysis, we emphasized the identification of gaps in content relevant to hospitalists. We then used the Institute of Medicine's (IOM) 6 aims for healthcare quality as a foundation for developing the HQPS Competencies.1 Specifically, the IOM states that healthcare should be safe, effective, patient‐centered, timely, efficient, and equitable. Additionally, we reviewed and integrated elements of the Practice‐Based Learning and Improvement (PBLI) and Systems‐Based Practice (SBP) competencies as defined by the Accreditation Council for Graduate Medical Education (ACGME).23 We defined general areas of competence and specific standards for knowledge, skills, and attitudes within each area. Subcommittee members reflected on their own experience, as clinicians, educators, and leaders in healthcare quality and patient safety, to inform and refine the competency definitions and standards. Acknowledging that some hospitalists may serve as collaborators or clinical content experts, while others may serve as leaders of hospital quality initiatives, 3 levels of expertise were established: basic, intermediate, and advanced.

Figure 1

The QIE subcommittee presented a draft version of the HQPS Competencies to the HQPS Committee in the fall of 2009 and incorporated suggested revisions. The revised set of competencies was then reviewed by members of the Leadership and Education Committees during the winter of 2009‐2010, and additional recommendations were included in the final version now described.

Description of The Competencies

The 8 areas of competence include: Quality Measurement and Stakeholder Interests, Data Acquisition and Interpretation, Organizational Knowledge and Leadership Skills, Patient Safety Principles, Teamwork and Communication, Quality and Safety Improvement Methods, Health Information Systems, and Patient Centeredness. Three levels of competence and standards within each level and area are defined in Table 1. Standards use carefully selected action verbs to reflect educational goals for hospitalists at each level.24 The basic level represents a minimum level of competency for all practicing hospitalists. The intermediate level represents a hospitalist who is prepared to meaningfully engage and collaborate with his or her institution in quality improvement efforts. A hospitalist at this level may also lead uncomplicated improvement projects for his or her medical center and/or hospital medicine group. The advanced level represents a hospitalist prepared to lead quality improvement efforts for his or her institution and/or hospital medicine group. Many hospitalists at this level will have, or will be prepared to have, leadership positions in quality and patient safety at their institutions. Advanced level hospitalists will also have the expertise to teach and mentor other individuals in their quality improvement efforts.

Recommended Use of The Competencies

The HQPS Competencies provide a framework for curricula and other professional development experiences in healthcare quality and patient safety. We recommend a step‐wise approach to curriculum development which includes conducting a targeted needs assessment, defining goals and specific learning objectives, and evaluation of the curriculum.25 The HQPS Competencies can be used at each step and provide educational targets for learners across a range of interest and experience.

Conclusion

Hospitalists are poised to have a tremendous impact on improving the quality of care for hospitalized patients. The lack of training in quality improvement in traditional medical education programs, in which most current hospitalists were trained, can be overcome through appropriate use of the HQPS Competencies. Formal incorporation of the HQPS Competencies into professional development programs, and innovative educational initiatives and curricula, will help provide current hospitalists and the next generations of hospitalists with the needed skills to be successful.

Healthcare quality is defined as the degree to which health services for individuals and populations increase the likelihood of desired health outcomes and are consistent with current professional knowledge.1 Delivering high quality care to patients in the hospital setting is especially challenging, given the rapid pace of clinical care, the severity and multitude of patient conditions, and the interdependence of complex processes within the hospital system. Research has shown that hospitalized patients do not consistently receive recommended care2 and are at risk for experiencing preventable harm.3 In an effort to stimulate improvement, stakeholders have called for increased accountability, including enhanced transparency and differential payment based on performance. A growing number of hospital process and outcome measures are readily available to the public via the Internet.46 The Joint Commission, which accredits US hospitals, requires the collection of core quality measure data7 and sets the expectation that National Patient Safety Goals be met to maintain accreditation.8 Moreover, the Center for Medicare and Medicaid Services (CMS) has developed a Value‐Based Purchasing (VBP) plan intended to adjust hospital payment based on quality measures and the occurrence of certain hospital‐acquired conditions.9, 10

Because of their clinical expertise, understanding of hospital clinical operations, leadership of multidisciplinary inpatient teams, and vested interest to improve the systems in which they work, hospitalists are perfectly positioned to collaborate with their institutions to improve the quality of care delivered to inpatients. However, many hospitalists are inadequately prepared to engage in efforts to improve quality, because medical schools and residency programs have not traditionally included or emphasized healthcare quality and patient safety in their curricula.1113 In a survey of 389 internal medicine‐trained hospitalists, significant educational deficiencies were identified in the area of systems‐based practice.14 Specifically, the topics of quality improvement, team management, practice guideline development, health information systems management, and coordination of care between healthcare settings were listed as essential skills for hospitalist practice but underemphasized in residency training. Recognizing the gap between the needs of practicing physicians and current medical education provided in healthcare quality, professional societies have recently published position papers calling for increased training in quality, safety, and systems, both in medical school11 and residency training.15, 16

The Society of Hospital Medicine (SHM) convened a Quality Summit in December 2008 to develop strategic plans related to healthcare quality. Summit attendees felt that most hospitalists lack the formal training necessary to evaluate, implement, and sustain system changes within the hospital. In response, the SHM Hospital Quality and Patient Safety (HQPS) Committee formed a Quality Improvement Education (QIE) subcommittee in 2009 to assess the needs of hospitalists with respect to hospital quality and patient safety, and to evaluate and expand upon existing educational programs in this area. Membership of the QIE subcommittee consisted of hospitalists with extensive experience in healthcare quality and medical education. The QIE subcommittee refined and expanded upon the healthcare quality and patient safety‐related competencies initially described in the Core Competencies in Hospital Medicine.17 The purpose of this report is to describe the development, provide definitions, and make recommendations on the use of the Hospital Quality and Patient Safety (HQPS) Competencies.

Development of The Hospital Quality and Patient Safety Competencies

The multistep process used by the SHM QIE subcommittee to develop the HQPS Competencies is summarized in Figure 1. We performed an in‐depth evaluation of current educational materials and offerings, including a review of the Core Competencies in Hospital Medicine, past annual SHM Quality Improvement Pre‐Course objectives, and the content of training courses offered by other organizations.1722 Throughout our analysis, we emphasized the identification of gaps in content relevant to hospitalists. We then used the Institute of Medicine's (IOM) 6 aims for healthcare quality as a foundation for developing the HQPS Competencies.1 Specifically, the IOM states that healthcare should be safe, effective, patient‐centered, timely, efficient, and equitable. Additionally, we reviewed and integrated elements of the Practice‐Based Learning and Improvement (PBLI) and Systems‐Based Practice (SBP) competencies as defined by the Accreditation Council for Graduate Medical Education (ACGME).23 We defined general areas of competence and specific standards for knowledge, skills, and attitudes within each area. Subcommittee members reflected on their own experience, as clinicians, educators, and leaders in healthcare quality and patient safety, to inform and refine the competency definitions and standards. Acknowledging that some hospitalists may serve as collaborators or clinical content experts, while others may serve as leaders of hospital quality initiatives, 3 levels of expertise were established: basic, intermediate, and advanced.

Figure 1

The QIE subcommittee presented a draft version of the HQPS Competencies to the HQPS Committee in the fall of 2009 and incorporated suggested revisions. The revised set of competencies was then reviewed by members of the Leadership and Education Committees during the winter of 2009‐2010, and additional recommendations were included in the final version now described.

Description of The Competencies

The 8 areas of competence include: Quality Measurement and Stakeholder Interests, Data Acquisition and Interpretation, Organizational Knowledge and Leadership Skills, Patient Safety Principles, Teamwork and Communication, Quality and Safety Improvement Methods, Health Information Systems, and Patient Centeredness. Three levels of competence and standards within each level and area are defined in Table 1. Standards use carefully selected action verbs to reflect educational goals for hospitalists at each level.24 The basic level represents a minimum level of competency for all practicing hospitalists. The intermediate level represents a hospitalist who is prepared to meaningfully engage and collaborate with his or her institution in quality improvement efforts. A hospitalist at this level may also lead uncomplicated improvement projects for his or her medical center and/or hospital medicine group. The advanced level represents a hospitalist prepared to lead quality improvement efforts for his or her institution and/or hospital medicine group. Many hospitalists at this level will have, or will be prepared to have, leadership positions in quality and patient safety at their institutions. Advanced level hospitalists will also have the expertise to teach and mentor other individuals in their quality improvement efforts.

Recommended Use of The Competencies

The HQPS Competencies provide a framework for curricula and other professional development experiences in healthcare quality and patient safety. We recommend a step‐wise approach to curriculum development which includes conducting a targeted needs assessment, defining goals and specific learning objectives, and evaluation of the curriculum.25 The HQPS Competencies can be used at each step and provide educational targets for learners across a range of interest and experience.

Conclusion

Hospitalists are poised to have a tremendous impact on improving the quality of care for hospitalized patients. The lack of training in quality improvement in traditional medical education programs, in which most current hospitalists were trained, can be overcome through appropriate use of the HQPS Competencies. Formal incorporation of the HQPS Competencies into professional development programs, and innovative educational initiatives and curricula, will help provide current hospitalists and the next generations of hospitalists with the needed skills to be successful.