The third‐year pediatric clerkship at the University of Utah School of Medicine has a relatively unique inpatient service, the Glasgow Service, which consists of an academic attending, a third‐year pediatric resident, and 4 third‐year medical students, but no interns. (This service was named in honor of Lowell Glasgow, chair of pediatrics, 1972‐82.) This structure was introduced in 1992 by the chair of pediatrics, Michael Simmons, the residency program director, Richard Molteni, and the clerkship director, Karen Hansen. These individuals desired to improve students' inpatient experience by providing greater responsibility for patient care. An additional motive was to increase the total number of patients followed by house staff without increasing the size of the residency program.

This inpatient service is a part of a 6‐week pediatric clerkship. All students perform the 3‐week inpatient portion of their clerkship at Primary Children's Medical Center, a tertiary‐care, freestanding children's hospital. (The students also spend 1 week each in a newborn nursery, an outpatient clinic, and a subspecialty setting). The academic attendings include generalists, hospitalists, and specialists who concurrently have other clinical responsibilities. The students take in‐house call every fourth night, supervised by senior residents who are not necessarily members of their service. All students share the same formal teaching activities, including morning report, a noon conference, and a student conference.

Patients are assigned to the ward services by a senior admitting resident. The admitting resident distributes patients among the services based on the complexity and acuity of the patients' conditions as well as the census on the various services. The senior resident supervising a particular service then assigns patients among the members of that service. Each third‐year medical student is expected to care for 2 or 3 patients at a time.

In addition to the intervention service, students also rotate on 2 similar traditional services. These services are traditional in the sense that they are composed of an academic attending, a community attending, a third‐year pediatric resident, 4 interns, and up to 2 fourth‐year and 2 third‐year medical students. Faculty preferences regarding service assignments were accommodated when possible. Therefore, some faculty attended only on one type of service, intervention or traditional, and others attended on both types. Because they have more members and because interns are capable of caring for more patients than are medical students, the traditional services cared for more patients than the intervention service. Although identical in composition, the 2 traditional services differ with each other in several ways. One service typically admits children 3 years old and younger, whereas the other admits children who are between 3 and 12 years old. The service that admits older children also admits most of the hematology‐oncology patients.

Although other authors have described similar inpatient clerkship structures, to our knowledge, none have evaluated them through a prospective randomized controlled trial.1, 2 The recent literature on ambulatory experiences during third‐year clerkships provided a methodological framework for this study. Collectively, such studies have evaluated outcomes with a variety of measures, including patient logs,35 evaluations,3, 4, 6, 7 examinations,37 surveys,3, 5, 7, 8 and career choices.4, 68 Additional outcomes, such as the effect of educational interventions on patient care, have been emphasized.9

In the light of this research, we conducted a prospective, randomized controlled trial to compare outcomes on the intervention service with those on the traditional services. We hypothesized that, compared with the traditional services, the intervention service would show:

• improved process measures in terms of increased number of patients admitted, number of key diagnoses encountered in the patients cared for, and range of ages of the patients admitted;

• similar or improved student performance, as measured by faculty and resident evaluations and a National Board of Medical Examiners (NBME) subject examination;

• increased student satisfaction, as assessed by an end‐of‐rotation questionnaire;

• increased interest in pediatric and, more broadly, primary care careers, as measured by subinternship and internship selections; and

• comparable or improved resource utilization in terms of length of stay and total charges.

METHODS

All students enrolled in the third‐year pediatric rotation during the 2001‐2003 academic years were individually randomized by the clerkship assistant to the intervention service or 1 of the 2 traditional services without respect to career preference. A 5:3 student randomization ratio was used to fulfill the requirement that 4 students be assigned to the intervention service during every 3‐week block. This permitted the service to have call every fourth night.

To evaluate the adequacy of the randomization process, we obtained baseline student characteristics on age, sex, and United States Medical Licensing Examination (USMLE) Step 1 score from the Dean of Student Affairs. The dean also reported the discipline each student enrolled in for the required fourth‐year subinternship(s) and matched in for internship. These data were reported anonymously and linked to the service to which the student was assigned. In this study, pediatrics, internal medicine, and family practice were all considered primary care, but preliminary or transitional internships were not.

RESULTS

Two hundred and three students enrolled in the third‐year pediatric clerkship during the study period, and all students completed the clerkship on their assigned services. One hundred and twenty‐eight were randomized to the intervention service and 75 to the traditional services. There were no statistically significant differences in median age, percentage of male students, or mean USMLE Step 1 score between the students randomized to the intervention service and those randomized to the traditional services (Table 1).

DISCUSSION

This study's objective was to evaluate a third‐year pediatric clerkship structure that focuses on students, using multiple outcome parameters. Utilizing a robust design, the results of this study have demonstrated that the intervention service is more successful than the traditional services in several outcomes. Students assigned to the intervention service were more satisfied and more likely to select pediatrics as a career. These improvements were accomplished while maintaining similar process measures, student performance, and resource utilization compared with those of the traditional services.

CONCLUSIONS

The intervention service is a structure for the pediatric inpatient rotation of third‐year medical students that, instead of dividing the faculty and supervising resident's attention between interns and students, focuses their attention on the students. Although it has been difficult to demonstrate improvements as a result of the educational interventions, we have shown several improvements in the evaluations of the students. Moreover, the pattern of increased student satisfaction and a tendency toward more student selecting careers in pediatrics are remarkable. This was accomplished with similar resource utilization. Therefore, this program merits being continued at our institution and possibly adopted at other medical schools. Further research is needed to determine which aspects of the intervention are responsible for its effects. Some components, such as focused time with students, may be applicable to traditional services.

The third‐year pediatric clerkship at the University of Utah School of Medicine has a relatively unique inpatient service, the Glasgow Service, which consists of an academic attending, a third‐year pediatric resident, and 4 third‐year medical students, but no interns. (This service was named in honor of Lowell Glasgow, chair of pediatrics, 1972‐82.) This structure was introduced in 1992 by the chair of pediatrics, Michael Simmons, the residency program director, Richard Molteni, and the clerkship director, Karen Hansen. These individuals desired to improve students' inpatient experience by providing greater responsibility for patient care. An additional motive was to increase the total number of patients followed by house staff without increasing the size of the residency program.

This inpatient service is a part of a 6‐week pediatric clerkship. All students perform the 3‐week inpatient portion of their clerkship at Primary Children's Medical Center, a tertiary‐care, freestanding children's hospital. (The students also spend 1 week each in a newborn nursery, an outpatient clinic, and a subspecialty setting). The academic attendings include generalists, hospitalists, and specialists who concurrently have other clinical responsibilities. The students take in‐house call every fourth night, supervised by senior residents who are not necessarily members of their service. All students share the same formal teaching activities, including morning report, a noon conference, and a student conference.

Patients are assigned to the ward services by a senior admitting resident. The admitting resident distributes patients among the services based on the complexity and acuity of the patients' conditions as well as the census on the various services. The senior resident supervising a particular service then assigns patients among the members of that service. Each third‐year medical student is expected to care for 2 or 3 patients at a time.

In addition to the intervention service, students also rotate on 2 similar traditional services. These services are traditional in the sense that they are composed of an academic attending, a community attending, a third‐year pediatric resident, 4 interns, and up to 2 fourth‐year and 2 third‐year medical students. Faculty preferences regarding service assignments were accommodated when possible. Therefore, some faculty attended only on one type of service, intervention or traditional, and others attended on both types. Because they have more members and because interns are capable of caring for more patients than are medical students, the traditional services cared for more patients than the intervention service. Although identical in composition, the 2 traditional services differ with each other in several ways. One service typically admits children 3 years old and younger, whereas the other admits children who are between 3 and 12 years old. The service that admits older children also admits most of the hematology‐oncology patients.

Although other authors have described similar inpatient clerkship structures, to our knowledge, none have evaluated them through a prospective randomized controlled trial.1, 2 The recent literature on ambulatory experiences during third‐year clerkships provided a methodological framework for this study. Collectively, such studies have evaluated outcomes with a variety of measures, including patient logs,35 evaluations,3, 4, 6, 7 examinations,37 surveys,3, 5, 7, 8 and career choices.4, 68 Additional outcomes, such as the effect of educational interventions on patient care, have been emphasized.9

In the light of this research, we conducted a prospective, randomized controlled trial to compare outcomes on the intervention service with those on the traditional services. We hypothesized that, compared with the traditional services, the intervention service would show:

• improved process measures in terms of increased number of patients admitted, number of key diagnoses encountered in the patients cared for, and range of ages of the patients admitted;

• similar or improved student performance, as measured by faculty and resident evaluations and a National Board of Medical Examiners (NBME) subject examination;

• increased student satisfaction, as assessed by an end‐of‐rotation questionnaire;

• increased interest in pediatric and, more broadly, primary care careers, as measured by subinternship and internship selections; and

• comparable or improved resource utilization in terms of length of stay and total charges.

METHODS

All students enrolled in the third‐year pediatric rotation during the 2001‐2003 academic years were individually randomized by the clerkship assistant to the intervention service or 1 of the 2 traditional services without respect to career preference. A 5:3 student randomization ratio was used to fulfill the requirement that 4 students be assigned to the intervention service during every 3‐week block. This permitted the service to have call every fourth night.

To evaluate the adequacy of the randomization process, we obtained baseline student characteristics on age, sex, and United States Medical Licensing Examination (USMLE) Step 1 score from the Dean of Student Affairs. The dean also reported the discipline each student enrolled in for the required fourth‐year subinternship(s) and matched in for internship. These data were reported anonymously and linked to the service to which the student was assigned. In this study, pediatrics, internal medicine, and family practice were all considered primary care, but preliminary or transitional internships were not.

RESULTS

Two hundred and three students enrolled in the third‐year pediatric clerkship during the study period, and all students completed the clerkship on their assigned services. One hundred and twenty‐eight were randomized to the intervention service and 75 to the traditional services. There were no statistically significant differences in median age, percentage of male students, or mean USMLE Step 1 score between the students randomized to the intervention service and those randomized to the traditional services (Table 1).

DISCUSSION

This study's objective was to evaluate a third‐year pediatric clerkship structure that focuses on students, using multiple outcome parameters. Utilizing a robust design, the results of this study have demonstrated that the intervention service is more successful than the traditional services in several outcomes. Students assigned to the intervention service were more satisfied and more likely to select pediatrics as a career. These improvements were accomplished while maintaining similar process measures, student performance, and resource utilization compared with those of the traditional services.

CONCLUSIONS

The intervention service is a structure for the pediatric inpatient rotation of third‐year medical students that, instead of dividing the faculty and supervising resident's attention between interns and students, focuses their attention on the students. Although it has been difficult to demonstrate improvements as a result of the educational interventions, we have shown several improvements in the evaluations of the students. Moreover, the pattern of increased student satisfaction and a tendency toward more student selecting careers in pediatrics are remarkable. This was accomplished with similar resource utilization. Therefore, this program merits being continued at our institution and possibly adopted at other medical schools. Further research is needed to determine which aspects of the intervention are responsible for its effects. Some components, such as focused time with students, may be applicable to traditional services.

The third‐year pediatric clerkship at the University of Utah School of Medicine has a relatively unique inpatient service, the Glasgow Service, which consists of an academic attending, a third‐year pediatric resident, and 4 third‐year medical students, but no interns. (This service was named in honor of Lowell Glasgow, chair of pediatrics, 1972‐82.) This structure was introduced in 1992 by the chair of pediatrics, Michael Simmons, the residency program director, Richard Molteni, and the clerkship director, Karen Hansen. These individuals desired to improve students' inpatient experience by providing greater responsibility for patient care. An additional motive was to increase the total number of patients followed by house staff without increasing the size of the residency program.

This inpatient service is a part of a 6‐week pediatric clerkship. All students perform the 3‐week inpatient portion of their clerkship at Primary Children's Medical Center, a tertiary‐care, freestanding children's hospital. (The students also spend 1 week each in a newborn nursery, an outpatient clinic, and a subspecialty setting). The academic attendings include generalists, hospitalists, and specialists who concurrently have other clinical responsibilities. The students take in‐house call every fourth night, supervised by senior residents who are not necessarily members of their service. All students share the same formal teaching activities, including morning report, a noon conference, and a student conference.

Patients are assigned to the ward services by a senior admitting resident. The admitting resident distributes patients among the services based on the complexity and acuity of the patients' conditions as well as the census on the various services. The senior resident supervising a particular service then assigns patients among the members of that service. Each third‐year medical student is expected to care for 2 or 3 patients at a time.

In addition to the intervention service, students also rotate on 2 similar traditional services. These services are traditional in the sense that they are composed of an academic attending, a community attending, a third‐year pediatric resident, 4 interns, and up to 2 fourth‐year and 2 third‐year medical students. Faculty preferences regarding service assignments were accommodated when possible. Therefore, some faculty attended only on one type of service, intervention or traditional, and others attended on both types. Because they have more members and because interns are capable of caring for more patients than are medical students, the traditional services cared for more patients than the intervention service. Although identical in composition, the 2 traditional services differ with each other in several ways. One service typically admits children 3 years old and younger, whereas the other admits children who are between 3 and 12 years old. The service that admits older children also admits most of the hematology‐oncology patients.

Although other authors have described similar inpatient clerkship structures, to our knowledge, none have evaluated them through a prospective randomized controlled trial.1, 2 The recent literature on ambulatory experiences during third‐year clerkships provided a methodological framework for this study. Collectively, such studies have evaluated outcomes with a variety of measures, including patient logs,35 evaluations,3, 4, 6, 7 examinations,37 surveys,3, 5, 7, 8 and career choices.4, 68 Additional outcomes, such as the effect of educational interventions on patient care, have been emphasized.9

In the light of this research, we conducted a prospective, randomized controlled trial to compare outcomes on the intervention service with those on the traditional services. We hypothesized that, compared with the traditional services, the intervention service would show:

• improved process measures in terms of increased number of patients admitted, number of key diagnoses encountered in the patients cared for, and range of ages of the patients admitted;

• similar or improved student performance, as measured by faculty and resident evaluations and a National Board of Medical Examiners (NBME) subject examination;

• increased student satisfaction, as assessed by an end‐of‐rotation questionnaire;

• increased interest in pediatric and, more broadly, primary care careers, as measured by subinternship and internship selections; and

• comparable or improved resource utilization in terms of length of stay and total charges.

METHODS

All students enrolled in the third‐year pediatric rotation during the 2001‐2003 academic years were individually randomized by the clerkship assistant to the intervention service or 1 of the 2 traditional services without respect to career preference. A 5:3 student randomization ratio was used to fulfill the requirement that 4 students be assigned to the intervention service during every 3‐week block. This permitted the service to have call every fourth night.

To evaluate the adequacy of the randomization process, we obtained baseline student characteristics on age, sex, and United States Medical Licensing Examination (USMLE) Step 1 score from the Dean of Student Affairs. The dean also reported the discipline each student enrolled in for the required fourth‐year subinternship(s) and matched in for internship. These data were reported anonymously and linked to the service to which the student was assigned. In this study, pediatrics, internal medicine, and family practice were all considered primary care, but preliminary or transitional internships were not.

RESULTS

Two hundred and three students enrolled in the third‐year pediatric clerkship during the study period, and all students completed the clerkship on their assigned services. One hundred and twenty‐eight were randomized to the intervention service and 75 to the traditional services. There were no statistically significant differences in median age, percentage of male students, or mean USMLE Step 1 score between the students randomized to the intervention service and those randomized to the traditional services (Table 1).

DISCUSSION

This study's objective was to evaluate a third‐year pediatric clerkship structure that focuses on students, using multiple outcome parameters. Utilizing a robust design, the results of this study have demonstrated that the intervention service is more successful than the traditional services in several outcomes. Students assigned to the intervention service were more satisfied and more likely to select pediatrics as a career. These improvements were accomplished while maintaining similar process measures, student performance, and resource utilization compared with those of the traditional services.

CONCLUSIONS

The intervention service is a structure for the pediatric inpatient rotation of third‐year medical students that, instead of dividing the faculty and supervising resident's attention between interns and students, focuses their attention on the students. Although it has been difficult to demonstrate improvements as a result of the educational interventions, we have shown several improvements in the evaluations of the students. Moreover, the pattern of increased student satisfaction and a tendency toward more student selecting careers in pediatrics are remarkable. This was accomplished with similar resource utilization. Therefore, this program merits being continued at our institution and possibly adopted at other medical schools. Further research is needed to determine which aspects of the intervention are responsible for its effects. Some components, such as focused time with students, may be applicable to traditional services.

Alcohol dependence is commonly associated with severe medical disease1 and is common among hospitalized medical patients. A nationally representative hospital sample found current alcohol use disorders to have a prevalence of 7.4%; most of those with these disorders were alcohol dependent.2 However, depending on the communities served by specific hospitals, prevalence can be much higher among medical inpatients,25 with studies finding problem drinking in as many as 28% of such patients. Although heavy drinking and the psychosocial problems that characterize alcohol dependence cause disease and interfere with disease management, remission is often difficult to achieve. As a result, although inpatient care of such patients probably does not differ from the average,6 this population is at high risk for poor health outcomes. as illustrated by factors such as suboptimal chronic disease management,7 preventable hospitalization,8 and increased mortality.9, 10 Remission involves a major behavior change that has been conceived of as a progression of stages, including precontemplation, contemplation, preparation, action, and maintenance.11 Clinically, this process encompasses an initial lack of awareness of the problem, followed by problem recognition and ambivalence about change, an increasing desire to change and concrete attempts at behavior change, and eventually long‐lasting behavioral improvements. The work described in this article was based on the broad hypothesis that acute illness and other alcohol‐related consequences will accelerate the process of change and that it might be possible to utilize this effect of acute illness to improve treatment outcomes for medically ill, alcohol‐dependent patients. If so, then measures of the change process should be correlated with measures of health status and alcohol consequences, correlations that were estimated in this study. If such measures are correlated and future research supports a causal relationship, then the link between illness and desire to change at the time of hospitalization could be exploited to help motivate changes in drinking and involvement in alcohol dependence treatment following hospital discharge.

METHODS

RESULTS

We identified 117 potential participants, accounting for 6% of total admissions (n = 1964) during the 7‐month recruitment period (late 2004 through mid‐2005). Of this total, 20 (17%) refused or withdrew prior to completing the study questionnaires, 17 (14.5%) were not eligible because of chronic cognitive impairment, 15 (12.8%) were discharged prior to being interviewed, 14 (12%) did not meet current alcohol dependence criteria (4 of whom met abuse criteria), and 1 (<1%) did not speak English. The remaining 50 subjects were included in this analysis. Characteristics of this group are listed in Table 1. They were primarily male, and socioeconomic status (assessed on the basis of education and health insurance) was low relative to the general population.17 Persons listed as having public insurance only were mainly covered by state Medicaid plans with or without Medicare. The less common diagnoses listed as other in Table 1, pertaining to no more than 2 patients, include arrhythmia, upper gastrointestinal bleeding, gout, electrolyte imbalance, hypoglycemia, diabetic ketoacidosis, diarrhea, stroke, and congestive heart failure. Measures of alcohol consumption were consistent with the clinical diagnosis of current alcohol dependence.

Of the components of the change process measured, the Recognition and Ambivalence subscales tended to have high scores. Thirty percent of subjects had the highest possible score on the Recognition scale, and 16% had the highest possible score on the Ambivalence scale. The scores for taking steps to change were more evenly distributed. A description of the study variables is included in Table 2.

The unadjusted correlations of alcohol problem recognition, ambivalence, and taking steps to change drinking with each perceived health status and alcohol‐consequence variable are shown in Table 3. Problem recognition was modestly and inversely associated with overall perception of physical health as measured by the SF‐12 but was not associated with perceived mental health. All the SIP subscales had strong univariate associations with problem recognition (P < .001 for each subscale). Problem recognition was associated with both percent drinking days (r = 0.39, P = .005) and average drinks per drinking day (r = 0.34, P = .0191). Adjustment for age, sex, and ethnicity did not modify the associations of recognition with perceived health and adverse consequences. Additional adjustment for percent drinking days eliminated the significant association with overall physical health from the SF‐12 (35% reduction in the regression coefficient, adjusted P = .100). A similar reduction in the association between problem recognition and overall physical health was observed for average drinks per drinking day (36% reduction in the regression coefficient, adjusted P = .102). All SIP subscales remained strongly associated with problem recognition despite the additional adjustment for alcohol consumption measures (all adjusted P values 0.001).

Ambivalence was associated with overall perception of physical health (P = .003) but not perceived mental health. All SIP subscales were associated with ambivalence (all P < .010). Alcohol consumption measures were not significantly associated with ambivalence (for percent drinking days, r = 0.25, P = .083; for average drinks per drinking day, r = 0.24, P = .106). Adjustment for age, sex, and ethnicity did not alter these findings.

Taking steps to change drinking behavior was not significantly associated with overall perceptions of physical and mental health. The physical (P = .002), interpersonal (P = .006), and intrapersonal (P = .034) SIP subscales were associated with taking steps to change. Alcohol consumption measures were not significantly associated with taking steps (percent drinking days, r = 0.19, P = .196; average drinks per drinking day, r = 0.24, P = .105). Adjustment for age, sex, and ethnicity, had minimal impact on the associations between taking steps and the physical, interpersonal, and intrapersonal SIP subscales.

DISCUSSION

This study evaluated the association of recognition of problem drinking, ambivalence about change, and taking steps to change with measures of perceived health status and alcohol‐related consequences. The results suggest that most medically hospitalized patients with clinically recognized alcohol dependence are highly cognizant of their drinking problem and wonder about the consequences of their drinking, and many feel they either have taken or will take steps to change their drinking behavior. Overall physical health perceptions during hospitalization were correlated with problem recognition (possibly mediated by heavy drinking) and ambivalence, but not with taking steps to change. Conversely, specific alcohol‐related physical and other consequences were often correlated with each process of change.

The SOCRATES results for this group were similar to those found in a large group of alcohol‐dependent persons participating in a large treatment trial.18 Relative to the distribution of SOCRATES scores in that group, 42% of this hospital sample would be above the median for recognition, 66% for ambivalence, and 44% for taking steps to change.19 This finding, coupled with the correlations of problem recognition and ambivalence (but not taking steps to change) with perceived physical health, suggests that medical hospitalization presents a unique opportunity for fostering change by moving ambivalent patients toward initiating change. However, additional research is needed to establish that these change processes during hospitalization predict participation after hospitalization in available treatment programs or other objective indicators of positive behavioral change.

Several limitations should be considered in interpreting our results. The participants represent clinically recognized and subsequently confirmed alcohol‐dependent patients. Results might differ for those initially detected by systematic screening, for example, by using a heavy‐drinking‐day item, as recommended by the NIH,20 or biomarkers.21 It is possible that such patients would have fewer apparent alcohol‐related consequences and may thus be less contemplative of change. In addition, our sample was recruited from a single hospital, and was primarily male and socioeconomically disadvantaged. Results may not generalize to other groups. Although we did not find any evidence that age, sex, and ethnicity altered the relationships examined in this work, the study was not adequately powered for firm conclusions about this. Finally, although the study demonstrated an association of perceived health status with readiness to change drinking, a causal relationship may not exist. Theoretically, poorer health would increase recognition, ambivalence, and intent to change, but close measurement of these variables over time would be required to establish this.

In conclusion, clinically recognized alcohol‐dependent patients with acute medical illness typically are highly cognizant of their drinking problems and often wish to change at the time of hospitalization. This is linked to perceptions of physical health and a variety of alcohol‐related consequences. The association of recognition, ambivalence, and intention to change at the time of hospitalization with postdischarge treatment participation and outcomes should be evaluated further. If these factors do predict such outcomes, intervention studies aimed at enhancing the process of change during hospitalization will be warranted.

Alcohol dependence is commonly associated with severe medical disease1 and is common among hospitalized medical patients. A nationally representative hospital sample found current alcohol use disorders to have a prevalence of 7.4%; most of those with these disorders were alcohol dependent.2 However, depending on the communities served by specific hospitals, prevalence can be much higher among medical inpatients,25 with studies finding problem drinking in as many as 28% of such patients. Although heavy drinking and the psychosocial problems that characterize alcohol dependence cause disease and interfere with disease management, remission is often difficult to achieve. As a result, although inpatient care of such patients probably does not differ from the average,6 this population is at high risk for poor health outcomes. as illustrated by factors such as suboptimal chronic disease management,7 preventable hospitalization,8 and increased mortality.9, 10 Remission involves a major behavior change that has been conceived of as a progression of stages, including precontemplation, contemplation, preparation, action, and maintenance.11 Clinically, this process encompasses an initial lack of awareness of the problem, followed by problem recognition and ambivalence about change, an increasing desire to change and concrete attempts at behavior change, and eventually long‐lasting behavioral improvements. The work described in this article was based on the broad hypothesis that acute illness and other alcohol‐related consequences will accelerate the process of change and that it might be possible to utilize this effect of acute illness to improve treatment outcomes for medically ill, alcohol‐dependent patients. If so, then measures of the change process should be correlated with measures of health status and alcohol consequences, correlations that were estimated in this study. If such measures are correlated and future research supports a causal relationship, then the link between illness and desire to change at the time of hospitalization could be exploited to help motivate changes in drinking and involvement in alcohol dependence treatment following hospital discharge.

METHODS

RESULTS

We identified 117 potential participants, accounting for 6% of total admissions (n = 1964) during the 7‐month recruitment period (late 2004 through mid‐2005). Of this total, 20 (17%) refused or withdrew prior to completing the study questionnaires, 17 (14.5%) were not eligible because of chronic cognitive impairment, 15 (12.8%) were discharged prior to being interviewed, 14 (12%) did not meet current alcohol dependence criteria (4 of whom met abuse criteria), and 1 (<1%) did not speak English. The remaining 50 subjects were included in this analysis. Characteristics of this group are listed in Table 1. They were primarily male, and socioeconomic status (assessed on the basis of education and health insurance) was low relative to the general population.17 Persons listed as having public insurance only were mainly covered by state Medicaid plans with or without Medicare. The less common diagnoses listed as other in Table 1, pertaining to no more than 2 patients, include arrhythmia, upper gastrointestinal bleeding, gout, electrolyte imbalance, hypoglycemia, diabetic ketoacidosis, diarrhea, stroke, and congestive heart failure. Measures of alcohol consumption were consistent with the clinical diagnosis of current alcohol dependence.

Of the components of the change process measured, the Recognition and Ambivalence subscales tended to have high scores. Thirty percent of subjects had the highest possible score on the Recognition scale, and 16% had the highest possible score on the Ambivalence scale. The scores for taking steps to change were more evenly distributed. A description of the study variables is included in Table 2.

The unadjusted correlations of alcohol problem recognition, ambivalence, and taking steps to change drinking with each perceived health status and alcohol‐consequence variable are shown in Table 3. Problem recognition was modestly and inversely associated with overall perception of physical health as measured by the SF‐12 but was not associated with perceived mental health. All the SIP subscales had strong univariate associations with problem recognition (P < .001 for each subscale). Problem recognition was associated with both percent drinking days (r = 0.39, P = .005) and average drinks per drinking day (r = 0.34, P = .0191). Adjustment for age, sex, and ethnicity did not modify the associations of recognition with perceived health and adverse consequences. Additional adjustment for percent drinking days eliminated the significant association with overall physical health from the SF‐12 (35% reduction in the regression coefficient, adjusted P = .100). A similar reduction in the association between problem recognition and overall physical health was observed for average drinks per drinking day (36% reduction in the regression coefficient, adjusted P = .102). All SIP subscales remained strongly associated with problem recognition despite the additional adjustment for alcohol consumption measures (all adjusted P values 0.001).

Ambivalence was associated with overall perception of physical health (P = .003) but not perceived mental health. All SIP subscales were associated with ambivalence (all P < .010). Alcohol consumption measures were not significantly associated with ambivalence (for percent drinking days, r = 0.25, P = .083; for average drinks per drinking day, r = 0.24, P = .106). Adjustment for age, sex, and ethnicity did not alter these findings.

Taking steps to change drinking behavior was not significantly associated with overall perceptions of physical and mental health. The physical (P = .002), interpersonal (P = .006), and intrapersonal (P = .034) SIP subscales were associated with taking steps to change. Alcohol consumption measures were not significantly associated with taking steps (percent drinking days, r = 0.19, P = .196; average drinks per drinking day, r = 0.24, P = .105). Adjustment for age, sex, and ethnicity, had minimal impact on the associations between taking steps and the physical, interpersonal, and intrapersonal SIP subscales.

DISCUSSION

This study evaluated the association of recognition of problem drinking, ambivalence about change, and taking steps to change with measures of perceived health status and alcohol‐related consequences. The results suggest that most medically hospitalized patients with clinically recognized alcohol dependence are highly cognizant of their drinking problem and wonder about the consequences of their drinking, and many feel they either have taken or will take steps to change their drinking behavior. Overall physical health perceptions during hospitalization were correlated with problem recognition (possibly mediated by heavy drinking) and ambivalence, but not with taking steps to change. Conversely, specific alcohol‐related physical and other consequences were often correlated with each process of change.

The SOCRATES results for this group were similar to those found in a large group of alcohol‐dependent persons participating in a large treatment trial.18 Relative to the distribution of SOCRATES scores in that group, 42% of this hospital sample would be above the median for recognition, 66% for ambivalence, and 44% for taking steps to change.19 This finding, coupled with the correlations of problem recognition and ambivalence (but not taking steps to change) with perceived physical health, suggests that medical hospitalization presents a unique opportunity for fostering change by moving ambivalent patients toward initiating change. However, additional research is needed to establish that these change processes during hospitalization predict participation after hospitalization in available treatment programs or other objective indicators of positive behavioral change.

Several limitations should be considered in interpreting our results. The participants represent clinically recognized and subsequently confirmed alcohol‐dependent patients. Results might differ for those initially detected by systematic screening, for example, by using a heavy‐drinking‐day item, as recommended by the NIH,20 or biomarkers.21 It is possible that such patients would have fewer apparent alcohol‐related consequences and may thus be less contemplative of change. In addition, our sample was recruited from a single hospital, and was primarily male and socioeconomically disadvantaged. Results may not generalize to other groups. Although we did not find any evidence that age, sex, and ethnicity altered the relationships examined in this work, the study was not adequately powered for firm conclusions about this. Finally, although the study demonstrated an association of perceived health status with readiness to change drinking, a causal relationship may not exist. Theoretically, poorer health would increase recognition, ambivalence, and intent to change, but close measurement of these variables over time would be required to establish this.

In conclusion, clinically recognized alcohol‐dependent patients with acute medical illness typically are highly cognizant of their drinking problems and often wish to change at the time of hospitalization. This is linked to perceptions of physical health and a variety of alcohol‐related consequences. The association of recognition, ambivalence, and intention to change at the time of hospitalization with postdischarge treatment participation and outcomes should be evaluated further. If these factors do predict such outcomes, intervention studies aimed at enhancing the process of change during hospitalization will be warranted.

Alcohol dependence is commonly associated with severe medical disease1 and is common among hospitalized medical patients. A nationally representative hospital sample found current alcohol use disorders to have a prevalence of 7.4%; most of those with these disorders were alcohol dependent.2 However, depending on the communities served by specific hospitals, prevalence can be much higher among medical inpatients,25 with studies finding problem drinking in as many as 28% of such patients. Although heavy drinking and the psychosocial problems that characterize alcohol dependence cause disease and interfere with disease management, remission is often difficult to achieve. As a result, although inpatient care of such patients probably does not differ from the average,6 this population is at high risk for poor health outcomes. as illustrated by factors such as suboptimal chronic disease management,7 preventable hospitalization,8 and increased mortality.9, 10 Remission involves a major behavior change that has been conceived of as a progression of stages, including precontemplation, contemplation, preparation, action, and maintenance.11 Clinically, this process encompasses an initial lack of awareness of the problem, followed by problem recognition and ambivalence about change, an increasing desire to change and concrete attempts at behavior change, and eventually long‐lasting behavioral improvements. The work described in this article was based on the broad hypothesis that acute illness and other alcohol‐related consequences will accelerate the process of change and that it might be possible to utilize this effect of acute illness to improve treatment outcomes for medically ill, alcohol‐dependent patients. If so, then measures of the change process should be correlated with measures of health status and alcohol consequences, correlations that were estimated in this study. If such measures are correlated and future research supports a causal relationship, then the link between illness and desire to change at the time of hospitalization could be exploited to help motivate changes in drinking and involvement in alcohol dependence treatment following hospital discharge.

METHODS

RESULTS

We identified 117 potential participants, accounting for 6% of total admissions (n = 1964) during the 7‐month recruitment period (late 2004 through mid‐2005). Of this total, 20 (17%) refused or withdrew prior to completing the study questionnaires, 17 (14.5%) were not eligible because of chronic cognitive impairment, 15 (12.8%) were discharged prior to being interviewed, 14 (12%) did not meet current alcohol dependence criteria (4 of whom met abuse criteria), and 1 (<1%) did not speak English. The remaining 50 subjects were included in this analysis. Characteristics of this group are listed in Table 1. They were primarily male, and socioeconomic status (assessed on the basis of education and health insurance) was low relative to the general population.17 Persons listed as having public insurance only were mainly covered by state Medicaid plans with or without Medicare. The less common diagnoses listed as other in Table 1, pertaining to no more than 2 patients, include arrhythmia, upper gastrointestinal bleeding, gout, electrolyte imbalance, hypoglycemia, diabetic ketoacidosis, diarrhea, stroke, and congestive heart failure. Measures of alcohol consumption were consistent with the clinical diagnosis of current alcohol dependence.

Of the components of the change process measured, the Recognition and Ambivalence subscales tended to have high scores. Thirty percent of subjects had the highest possible score on the Recognition scale, and 16% had the highest possible score on the Ambivalence scale. The scores for taking steps to change were more evenly distributed. A description of the study variables is included in Table 2.

The unadjusted correlations of alcohol problem recognition, ambivalence, and taking steps to change drinking with each perceived health status and alcohol‐consequence variable are shown in Table 3. Problem recognition was modestly and inversely associated with overall perception of physical health as measured by the SF‐12 but was not associated with perceived mental health. All the SIP subscales had strong univariate associations with problem recognition (P < .001 for each subscale). Problem recognition was associated with both percent drinking days (r = 0.39, P = .005) and average drinks per drinking day (r = 0.34, P = .0191). Adjustment for age, sex, and ethnicity did not modify the associations of recognition with perceived health and adverse consequences. Additional adjustment for percent drinking days eliminated the significant association with overall physical health from the SF‐12 (35% reduction in the regression coefficient, adjusted P = .100). A similar reduction in the association between problem recognition and overall physical health was observed for average drinks per drinking day (36% reduction in the regression coefficient, adjusted P = .102). All SIP subscales remained strongly associated with problem recognition despite the additional adjustment for alcohol consumption measures (all adjusted P values 0.001).

Ambivalence was associated with overall perception of physical health (P = .003) but not perceived mental health. All SIP subscales were associated with ambivalence (all P < .010). Alcohol consumption measures were not significantly associated with ambivalence (for percent drinking days, r = 0.25, P = .083; for average drinks per drinking day, r = 0.24, P = .106). Adjustment for age, sex, and ethnicity did not alter these findings.

Taking steps to change drinking behavior was not significantly associated with overall perceptions of physical and mental health. The physical (P = .002), interpersonal (P = .006), and intrapersonal (P = .034) SIP subscales were associated with taking steps to change. Alcohol consumption measures were not significantly associated with taking steps (percent drinking days, r = 0.19, P = .196; average drinks per drinking day, r = 0.24, P = .105). Adjustment for age, sex, and ethnicity, had minimal impact on the associations between taking steps and the physical, interpersonal, and intrapersonal SIP subscales.

DISCUSSION

This study evaluated the association of recognition of problem drinking, ambivalence about change, and taking steps to change with measures of perceived health status and alcohol‐related consequences. The results suggest that most medically hospitalized patients with clinically recognized alcohol dependence are highly cognizant of their drinking problem and wonder about the consequences of their drinking, and many feel they either have taken or will take steps to change their drinking behavior. Overall physical health perceptions during hospitalization were correlated with problem recognition (possibly mediated by heavy drinking) and ambivalence, but not with taking steps to change. Conversely, specific alcohol‐related physical and other consequences were often correlated with each process of change.

The SOCRATES results for this group were similar to those found in a large group of alcohol‐dependent persons participating in a large treatment trial.18 Relative to the distribution of SOCRATES scores in that group, 42% of this hospital sample would be above the median for recognition, 66% for ambivalence, and 44% for taking steps to change.19 This finding, coupled with the correlations of problem recognition and ambivalence (but not taking steps to change) with perceived physical health, suggests that medical hospitalization presents a unique opportunity for fostering change by moving ambivalent patients toward initiating change. However, additional research is needed to establish that these change processes during hospitalization predict participation after hospitalization in available treatment programs or other objective indicators of positive behavioral change.

Several limitations should be considered in interpreting our results. The participants represent clinically recognized and subsequently confirmed alcohol‐dependent patients. Results might differ for those initially detected by systematic screening, for example, by using a heavy‐drinking‐day item, as recommended by the NIH,20 or biomarkers.21 It is possible that such patients would have fewer apparent alcohol‐related consequences and may thus be less contemplative of change. In addition, our sample was recruited from a single hospital, and was primarily male and socioeconomically disadvantaged. Results may not generalize to other groups. Although we did not find any evidence that age, sex, and ethnicity altered the relationships examined in this work, the study was not adequately powered for firm conclusions about this. Finally, although the study demonstrated an association of perceived health status with readiness to change drinking, a causal relationship may not exist. Theoretically, poorer health would increase recognition, ambivalence, and intent to change, but close measurement of these variables over time would be required to establish this.

In conclusion, clinically recognized alcohol‐dependent patients with acute medical illness typically are highly cognizant of their drinking problems and often wish to change at the time of hospitalization. This is linked to perceptions of physical health and a variety of alcohol‐related consequences. The association of recognition, ambivalence, and intention to change at the time of hospitalization with postdischarge treatment participation and outcomes should be evaluated further. If these factors do predict such outcomes, intervention studies aimed at enhancing the process of change during hospitalization will be warranted.

As modern medicine developed the technological capacity to deliver aggressive life‐sustaining interventionsthrough methods such as cardiopulmonary resuscitation (CPR), intensive care units, and mechanical ventilationthe concept of do‐not‐resuscitate (DNR) orders emerged to allow individual patients to choose to forego selected treatments. To encourage patients to articulate these preferences, Congress passed the Patient Self‐Determination Act in 1991, a measure that required health care facilities to discuss advance directives with patients as they enter their system.1 Although the act has had less of an impact on the quality of DNR discussions than originally hoped for,25 its passage was evidence of the importance our society places on patientclinician discussions regarding goals of care. In addition to this legislative push, many organizations and advocacy groups use a variety of marketing campaigns, accreditation standards,6 and standard instruments and tools79 to promote the use of advance directives

Despite all these efforts, fewer than 30% of Americans (54% older than age 65) have completed advance directives.10 Nevertheless, many patientsparticularly those at highest risk for requiring end‐of‐life caredo express preferences regarding resuscitation at the time of hospital admission. In an ideal world, these preferences would be available for all providers to view, respect, and act on.

Unfortunately, research on patient safety and quality has demonstrated wide gaps between ideal and actual practice.1112 In the context of DNR wishes, despite strong efforts to collect patients' preferences, no current regulation provides or mandates a best practice on making these preferences operational. There are also few data that indicate whether patients' preferences are in fact transmitted to providers at the point of care and in an accurate and reliable manner.

Past research on proper identification of DNR orders is limited, with much of the focus on prehospital protocols.1315 Anecdotally, hospitals seem to employ varying strategies to highlight DNR orders using a combination of paper or electronic documentation and color‐coded patient wristbands. There have been several reports of errors involving this issue, including patients receiving CPR despite stated DNR preferences and a patient having CPR withheld because the wrong chart (of another patient with a DNR order) was mistakenly pulled.1617

The patient safety field emphasizes standardization as a key strategy to prevent errors. Because of problems articulating DNR orders (and other important patient‐related information), several hospitals promote the use of color‐coded wristbands to denote preferences for resuscitation. However, without national regulations or standards, the possibility remains that one safety hazard (advance directives on a paper chart distant from a patient's room) may be traded for another hazard (front‐line providers interpreting a color‐coded wristband incorrectly). In addition to the ethical problems inherent in failing to adhere to patients' resuscitation preferences, errors in following advance directives may also create legal liability.18 With all this in mind, we conducted a national survey to determine practice variations in the identification of DNR orders and the use of color‐coded patient wristbands. We hypothesized that there is considerable variation both in identification practices and in the use of color‐coded wristbands across academic medical centers.

METHODS

The project was approved by the University of California, San Francisco Committee on Human Research. We anonymously surveyed nursing executives who are members of the University HealthSystem Consortium (UHC), an alliance of 97 academic medical centers and their affiliated hospitals representing 90% of the nation's nonprofit academic medical centers.19 The nursing executives are senior nursing leaders at participating UHC institutions and members of a dedicated UHC Chief Nursing Officer Council E‐mail listserv. We designed a brief survey and distributed it via their E‐mail listserv using an online commercial survey administration tool.20 Respondents were asked to complete the survey or have one of their colleagues familiar with local DNR identification practices complete it on their behalf. The online tool also provided summary reports and descriptive findings to meet the study objectives. We provided a 1‐month window (during summer 2006) with 1 interval E‐mail reminder to complete the surveys.

RESULTS

Survey announcements were E‐mailed to 127 nursing executives, 69 of whom completed it (response rate 54%). The respondents represented mostly academic medical centers (87%; another 13% represented affiliated community teaching hospitals), public institutions (89%), and large facilities (60% with more than 400 beds; 40% with 201‐400 beds). More than half the respondents (56%) reported their hospitals use paper chart documentation as the only method of identifying patients with a DNR order, whereas 16% reported their hospitals use only electronic health record (EHR) documentation (Fig. 1). Twenty‐five percent of hospitals (n = 17) use a color‐coded patient wristband in addition to either paper or electronic documentation. Of these 17 hospitals, a total of 8 colors or color schemes were employed to designate DNR status (Table 1).

Figure 1

The use of color‐coded wristbands was not limited to identification of DNR status. Fifty‐five percent of hospitals (n = 31) use color‐coded wristbands to indicate another piece of patient‐related data such as an allergy, fall risk, or same last name alert (Table 2). In fact, 12 indications were depicted by various colors, with variations in both the color choice for a given indication (eg, allergy wristbands red at one hospital and yellow at another) and across indications (eg, red for allergy at one hospital and red for bleeding risk at another). Nearly 3 of 4 respondents (n = 48) reported being aware of a case at your institution in which confusion about a DNR order led to problems or confusion in patient care. A few respondents shared a brief anecdote of the event, illustrating the spectrum of clinical scenarios that lead to potential confusion (Table 3). Respondents reporting a case of confusion were not more likely to be from an institution that used color‐coded wristbands.

When asked whether most (greater than 75%) physicians and nurses could properly identify the color associated with a DNR patient wristband, responses differed by discipline. Eight‐five percent of respondents believed that most nurses at their institutions could correctly report the color for DNR, whereas only 15% believed physicians could do the same. Only 22% of respondents anticipated a change in the current system within the next 2 years; all these changes were a transition from paper to electronic documentation systems.

DISCUSSION

Regardless of whether the DNR documentation occurs in paper or electronic form (and our study demonstrates significant practice variation in the documentation method), the risk that a hospitalized patient may suddenly stop breathing or become pulseless is ever present. When such a patient is discovered, providers race to the bedside and initiate care, but immediately ask, Is the patient a full code? In these often‐chaotic moments, accurate and timely information about DNR status is critical to respecting a patient's preferences and avoiding a potentially devastating error. A number of the anecdotes shared by survey respondents and highlighted in Table 3 reinforce this concern. Many of these scenarios occur in the middle of the night or off a patient's primary unit (ie, at a test or procedure area), increasing the need for quick and easy identification of DNR status.

Our study demonstrates that a logical point‐of‐care solutiona color‐coded DNR patient wristbandmay create its own safety hazards, particularly if the color designations are not known by all providers (including floating and traveling nurses or trainees who rotate at different hospitals) and if the colors being employed represent different indications at a given hospital (see accompanying Images Dx, page 445). We found that approximately 1 in 4 surveyed hospitals depict DNR status by a color‐coded wristband. We also discovered remarkable variation in the colors chosen and the degree to which institutions use color‐coded wristbands to signal a panoply of other patient‐related issues. Human factors research demonstrates that even well‐meaning patient safety solutions may cause harm in new ways if they are poorly implemented or if the interface between the technology and human work patterns is not well appreciated. For example, recent studies illustrate unintended consequences from safety‐driven solutions, such as the implementation of computerized order entry,2122 quality measurement,23 adoption of EHRs,24 and bar code medication administration systems.25 Because standardization is a key mechanism for decreasing the opportunities for error, our findings raise serious concerns about current wristband use.

Interestingly, the lack of standardization and its related risk of failing to recall the conditions associated with color‐coded wristbands are complicated by societal trends. In December 2004 the issue of patient wristbands made headlines in Florida, when hospitals using yellow DNR wristbands (as was the case in 3 hospitals in our sample) reported several near‐misses among patients wearing yellow Lance Armstrong Livestrong bracelets.2627 Given recent estimates that nearly 1 in 5 Americans wears these bracelets to support people living with cancer,28 even safety‐minded journals and national newspapers have highlighted the issue.2930 Most hospitals that continue to use yellow DNR wristbands now either remove or cover Livestrong bracelets at the time of hospital admission. Furthermore, many other self‐help organizations now issue wristbands in a variety of colors as well, creating a potential hazard for any person wearing one in the hospital. Although patients do not mind wearing color‐coded wristbands,31 they might feel differently if they knew the potential for confusion.

After these anecdotal reports of identification mistakes surfaced, several states, most notably Arizona and Pennsylvania, launched initiatives to address the problem.3233 Arizona, after discovering 8 colors being used in the state, developed plans for a purple DNR color‐coded wristband. The choice of purple, and the careful decision to avoid blue, occurred because many hospitals call their resuscitative efforts a code blue, creating yet another potential source of confusion if a blue wristband is associated with a DNR order. The Pennsylvania Patient Safety Authority also found tremendous color variations in patient wristbands used in a statewide survey. Both states ultimately promoted standardized colors and indications and provided tool kits and implementation manuals.3233

Although statewide initiatives represent a step forward, we believe that a national standard for color‐coded wristbands would improve patient safety. Precedents for this call to action exist. For many years, anecdotal information circulated about the errors caused by ambiguous use of abbreviations, such as qd instead of daily or U instead of units. Individual hospitals often banned or limited the use of such abbreviations, but no standard list of high‐risk abbreviations guided practice or required adherence, and cross‐hospital variation undoubtedly led to confusion. In 2004 the Joint Commission created a uniform list of high‐risk abbreviations as part of their National Patient Safety Goals, which instantly ended the debate about which abbreviations to ban and mandated compliance with the safety practice.34 A national group of stakeholders should similarly be convened to develop a list of colors and associated conditions that should be widely disseminated and enforced by the Joint Commission or a similar body. The statewide efforts by Arizona and Pennsylvania are instructive in this regard. Despite being guided by the goal of standardization, these 2 states chose different colors for DNR identification (interestingly, Pennsylvania chose blue for DNR, perhaps for the same reason that Arizona avoided itcode blue), further supporting the need for national guidelines (Table 4).

Our study represents the first national sample of DNR identification practices. Although it targeted academic health centers and affiliated institutions, we believe that these practice variations likely exist in all health care settings. Our study limitations included reliance on self‐reported institutional practices rather than direct review of existing policies and limited information about the surveyed population, making it impossible to compare respondents and nonrespondents. However, we have no reason to believe that these groups differed sufficiently to influence the study's main findings.

In the future, better technology may ultimately replace color‐coded wristbands. For instance, the time may come when wireless technologies seamlessly linked to the electronic health record will alert providers to a patient's DNR status when entering the patient's room. However, for today, point‐of‐care solutions using color‐coded wristbands remain a reasonable solution. Creating a nationally enforced standardized methodology, understandable and memorable to providers and free of stigma to patients (eg, a black wristband for DNR or writing DNR on a wristband) should be a patient safety priority. Because simplification is another key characteristic of safe systems, it seems prudent to aim for a national system that involves a maximum of 3‐4 colors.

CONCLUSIONS

Patients and families dedicate tremendous energy to making decisions about their advance directives, and discussions of these issues often create considerable angst and sadness. Health care providers are trained to elicit and advocate for such directives so they can act with patients' wishes in mind. Despite the high stakes, all these efforts can be undermined when the system for making providers aware of a patient's DNR status is flawed. Our data confirm the tremendous variability in the systems used to indicate DNR status (and other types of indications), variability that may place patients at risk from catastrophic errors. Following the lead of a few states, we call for a national mandate to standardize the identification of DNR orders and to make the colors of wristbands for a small set of indications uniform in every hospital across the country.

As modern medicine developed the technological capacity to deliver aggressive life‐sustaining interventionsthrough methods such as cardiopulmonary resuscitation (CPR), intensive care units, and mechanical ventilationthe concept of do‐not‐resuscitate (DNR) orders emerged to allow individual patients to choose to forego selected treatments. To encourage patients to articulate these preferences, Congress passed the Patient Self‐Determination Act in 1991, a measure that required health care facilities to discuss advance directives with patients as they enter their system.1 Although the act has had less of an impact on the quality of DNR discussions than originally hoped for,25 its passage was evidence of the importance our society places on patientclinician discussions regarding goals of care. In addition to this legislative push, many organizations and advocacy groups use a variety of marketing campaigns, accreditation standards,6 and standard instruments and tools79 to promote the use of advance directives

Despite all these efforts, fewer than 30% of Americans (54% older than age 65) have completed advance directives.10 Nevertheless, many patientsparticularly those at highest risk for requiring end‐of‐life caredo express preferences regarding resuscitation at the time of hospital admission. In an ideal world, these preferences would be available for all providers to view, respect, and act on.

Unfortunately, research on patient safety and quality has demonstrated wide gaps between ideal and actual practice.1112 In the context of DNR wishes, despite strong efforts to collect patients' preferences, no current regulation provides or mandates a best practice on making these preferences operational. There are also few data that indicate whether patients' preferences are in fact transmitted to providers at the point of care and in an accurate and reliable manner.

Past research on proper identification of DNR orders is limited, with much of the focus on prehospital protocols.1315 Anecdotally, hospitals seem to employ varying strategies to highlight DNR orders using a combination of paper or electronic documentation and color‐coded patient wristbands. There have been several reports of errors involving this issue, including patients receiving CPR despite stated DNR preferences and a patient having CPR withheld because the wrong chart (of another patient with a DNR order) was mistakenly pulled.1617

The patient safety field emphasizes standardization as a key strategy to prevent errors. Because of problems articulating DNR orders (and other important patient‐related information), several hospitals promote the use of color‐coded wristbands to denote preferences for resuscitation. However, without national regulations or standards, the possibility remains that one safety hazard (advance directives on a paper chart distant from a patient's room) may be traded for another hazard (front‐line providers interpreting a color‐coded wristband incorrectly). In addition to the ethical problems inherent in failing to adhere to patients' resuscitation preferences, errors in following advance directives may also create legal liability.18 With all this in mind, we conducted a national survey to determine practice variations in the identification of DNR orders and the use of color‐coded patient wristbands. We hypothesized that there is considerable variation both in identification practices and in the use of color‐coded wristbands across academic medical centers.

METHODS

The project was approved by the University of California, San Francisco Committee on Human Research. We anonymously surveyed nursing executives who are members of the University HealthSystem Consortium (UHC), an alliance of 97 academic medical centers and their affiliated hospitals representing 90% of the nation's nonprofit academic medical centers.19 The nursing executives are senior nursing leaders at participating UHC institutions and members of a dedicated UHC Chief Nursing Officer Council E‐mail listserv. We designed a brief survey and distributed it via their E‐mail listserv using an online commercial survey administration tool.20 Respondents were asked to complete the survey or have one of their colleagues familiar with local DNR identification practices complete it on their behalf. The online tool also provided summary reports and descriptive findings to meet the study objectives. We provided a 1‐month window (during summer 2006) with 1 interval E‐mail reminder to complete the surveys.

RESULTS

Survey announcements were E‐mailed to 127 nursing executives, 69 of whom completed it (response rate 54%). The respondents represented mostly academic medical centers (87%; another 13% represented affiliated community teaching hospitals), public institutions (89%), and large facilities (60% with more than 400 beds; 40% with 201‐400 beds). More than half the respondents (56%) reported their hospitals use paper chart documentation as the only method of identifying patients with a DNR order, whereas 16% reported their hospitals use only electronic health record (EHR) documentation (Fig. 1). Twenty‐five percent of hospitals (n = 17) use a color‐coded patient wristband in addition to either paper or electronic documentation. Of these 17 hospitals, a total of 8 colors or color schemes were employed to designate DNR status (Table 1).

Figure 1

The use of color‐coded wristbands was not limited to identification of DNR status. Fifty‐five percent of hospitals (n = 31) use color‐coded wristbands to indicate another piece of patient‐related data such as an allergy, fall risk, or same last name alert (Table 2). In fact, 12 indications were depicted by various colors, with variations in both the color choice for a given indication (eg, allergy wristbands red at one hospital and yellow at another) and across indications (eg, red for allergy at one hospital and red for bleeding risk at another). Nearly 3 of 4 respondents (n = 48) reported being aware of a case at your institution in which confusion about a DNR order led to problems or confusion in patient care. A few respondents shared a brief anecdote of the event, illustrating the spectrum of clinical scenarios that lead to potential confusion (Table 3). Respondents reporting a case of confusion were not more likely to be from an institution that used color‐coded wristbands.

When asked whether most (greater than 75%) physicians and nurses could properly identify the color associated with a DNR patient wristband, responses differed by discipline. Eight‐five percent of respondents believed that most nurses at their institutions could correctly report the color for DNR, whereas only 15% believed physicians could do the same. Only 22% of respondents anticipated a change in the current system within the next 2 years; all these changes were a transition from paper to electronic documentation systems.

DISCUSSION

Regardless of whether the DNR documentation occurs in paper or electronic form (and our study demonstrates significant practice variation in the documentation method), the risk that a hospitalized patient may suddenly stop breathing or become pulseless is ever present. When such a patient is discovered, providers race to the bedside and initiate care, but immediately ask, Is the patient a full code? In these often‐chaotic moments, accurate and timely information about DNR status is critical to respecting a patient's preferences and avoiding a potentially devastating error. A number of the anecdotes shared by survey respondents and highlighted in Table 3 reinforce this concern. Many of these scenarios occur in the middle of the night or off a patient's primary unit (ie, at a test or procedure area), increasing the need for quick and easy identification of DNR status.

Our study demonstrates that a logical point‐of‐care solutiona color‐coded DNR patient wristbandmay create its own safety hazards, particularly if the color designations are not known by all providers (including floating and traveling nurses or trainees who rotate at different hospitals) and if the colors being employed represent different indications at a given hospital (see accompanying Images Dx, page 445). We found that approximately 1 in 4 surveyed hospitals depict DNR status by a color‐coded wristband. We also discovered remarkable variation in the colors chosen and the degree to which institutions use color‐coded wristbands to signal a panoply of other patient‐related issues. Human factors research demonstrates that even well‐meaning patient safety solutions may cause harm in new ways if they are poorly implemented or if the interface between the technology and human work patterns is not well appreciated. For example, recent studies illustrate unintended consequences from safety‐driven solutions, such as the implementation of computerized order entry,2122 quality measurement,23 adoption of EHRs,24 and bar code medication administration systems.25 Because standardization is a key mechanism for decreasing the opportunities for error, our findings raise serious concerns about current wristband use.

Interestingly, the lack of standardization and its related risk of failing to recall the conditions associated with color‐coded wristbands are complicated by societal trends. In December 2004 the issue of patient wristbands made headlines in Florida, when hospitals using yellow DNR wristbands (as was the case in 3 hospitals in our sample) reported several near‐misses among patients wearing yellow Lance Armstrong Livestrong bracelets.2627 Given recent estimates that nearly 1 in 5 Americans wears these bracelets to support people living with cancer,28 even safety‐minded journals and national newspapers have highlighted the issue.2930 Most hospitals that continue to use yellow DNR wristbands now either remove or cover Livestrong bracelets at the time of hospital admission. Furthermore, many other self‐help organizations now issue wristbands in a variety of colors as well, creating a potential hazard for any person wearing one in the hospital. Although patients do not mind wearing color‐coded wristbands,31 they might feel differently if they knew the potential for confusion.

After these anecdotal reports of identification mistakes surfaced, several states, most notably Arizona and Pennsylvania, launched initiatives to address the problem.3233 Arizona, after discovering 8 colors being used in the state, developed plans for a purple DNR color‐coded wristband. The choice of purple, and the careful decision to avoid blue, occurred because many hospitals call their resuscitative efforts a code blue, creating yet another potential source of confusion if a blue wristband is associated with a DNR order. The Pennsylvania Patient Safety Authority also found tremendous color variations in patient wristbands used in a statewide survey. Both states ultimately promoted standardized colors and indications and provided tool kits and implementation manuals.3233

Although statewide initiatives represent a step forward, we believe that a national standard for color‐coded wristbands would improve patient safety. Precedents for this call to action exist. For many years, anecdotal information circulated about the errors caused by ambiguous use of abbreviations, such as qd instead of daily or U instead of units. Individual hospitals often banned or limited the use of such abbreviations, but no standard list of high‐risk abbreviations guided practice or required adherence, and cross‐hospital variation undoubtedly led to confusion. In 2004 the Joint Commission created a uniform list of high‐risk abbreviations as part of their National Patient Safety Goals, which instantly ended the debate about which abbreviations to ban and mandated compliance with the safety practice.34 A national group of stakeholders should similarly be convened to develop a list of colors and associated conditions that should be widely disseminated and enforced by the Joint Commission or a similar body. The statewide efforts by Arizona and Pennsylvania are instructive in this regard. Despite being guided by the goal of standardization, these 2 states chose different colors for DNR identification (interestingly, Pennsylvania chose blue for DNR, perhaps for the same reason that Arizona avoided itcode blue), further supporting the need for national guidelines (Table 4).

Our study represents the first national sample of DNR identification practices. Although it targeted academic health centers and affiliated institutions, we believe that these practice variations likely exist in all health care settings. Our study limitations included reliance on self‐reported institutional practices rather than direct review of existing policies and limited information about the surveyed population, making it impossible to compare respondents and nonrespondents. However, we have no reason to believe that these groups differed sufficiently to influence the study's main findings.

In the future, better technology may ultimately replace color‐coded wristbands. For instance, the time may come when wireless technologies seamlessly linked to the electronic health record will alert providers to a patient's DNR status when entering the patient's room. However, for today, point‐of‐care solutions using color‐coded wristbands remain a reasonable solution. Creating a nationally enforced standardized methodology, understandable and memorable to providers and free of stigma to patients (eg, a black wristband for DNR or writing DNR on a wristband) should be a patient safety priority. Because simplification is another key characteristic of safe systems, it seems prudent to aim for a national system that involves a maximum of 3‐4 colors.

CONCLUSIONS

Patients and families dedicate tremendous energy to making decisions about their advance directives, and discussions of these issues often create considerable angst and sadness. Health care providers are trained to elicit and advocate for such directives so they can act with patients' wishes in mind. Despite the high stakes, all these efforts can be undermined when the system for making providers aware of a patient's DNR status is flawed. Our data confirm the tremendous variability in the systems used to indicate DNR status (and other types of indications), variability that may place patients at risk from catastrophic errors. Following the lead of a few states, we call for a national mandate to standardize the identification of DNR orders and to make the colors of wristbands for a small set of indications uniform in every hospital across the country.

As modern medicine developed the technological capacity to deliver aggressive life‐sustaining interventionsthrough methods such as cardiopulmonary resuscitation (CPR), intensive care units, and mechanical ventilationthe concept of do‐not‐resuscitate (DNR) orders emerged to allow individual patients to choose to forego selected treatments. To encourage patients to articulate these preferences, Congress passed the Patient Self‐Determination Act in 1991, a measure that required health care facilities to discuss advance directives with patients as they enter their system.1 Although the act has had less of an impact on the quality of DNR discussions than originally hoped for,25 its passage was evidence of the importance our society places on patientclinician discussions regarding goals of care. In addition to this legislative push, many organizations and advocacy groups use a variety of marketing campaigns, accreditation standards,6 and standard instruments and tools79 to promote the use of advance directives

Despite all these efforts, fewer than 30% of Americans (54% older than age 65) have completed advance directives.10 Nevertheless, many patientsparticularly those at highest risk for requiring end‐of‐life caredo express preferences regarding resuscitation at the time of hospital admission. In an ideal world, these preferences would be available for all providers to view, respect, and act on.

Unfortunately, research on patient safety and quality has demonstrated wide gaps between ideal and actual practice.1112 In the context of DNR wishes, despite strong efforts to collect patients' preferences, no current regulation provides or mandates a best practice on making these preferences operational. There are also few data that indicate whether patients' preferences are in fact transmitted to providers at the point of care and in an accurate and reliable manner.

Past research on proper identification of DNR orders is limited, with much of the focus on prehospital protocols.1315 Anecdotally, hospitals seem to employ varying strategies to highlight DNR orders using a combination of paper or electronic documentation and color‐coded patient wristbands. There have been several reports of errors involving this issue, including patients receiving CPR despite stated DNR preferences and a patient having CPR withheld because the wrong chart (of another patient with a DNR order) was mistakenly pulled.1617

The patient safety field emphasizes standardization as a key strategy to prevent errors. Because of problems articulating DNR orders (and other important patient‐related information), several hospitals promote the use of color‐coded wristbands to denote preferences for resuscitation. However, without national regulations or standards, the possibility remains that one safety hazard (advance directives on a paper chart distant from a patient's room) may be traded for another hazard (front‐line providers interpreting a color‐coded wristband incorrectly). In addition to the ethical problems inherent in failing to adhere to patients' resuscitation preferences, errors in following advance directives may also create legal liability.18 With all this in mind, we conducted a national survey to determine practice variations in the identification of DNR orders and the use of color‐coded patient wristbands. We hypothesized that there is considerable variation both in identification practices and in the use of color‐coded wristbands across academic medical centers.

METHODS

The project was approved by the University of California, San Francisco Committee on Human Research. We anonymously surveyed nursing executives who are members of the University HealthSystem Consortium (UHC), an alliance of 97 academic medical centers and their affiliated hospitals representing 90% of the nation's nonprofit academic medical centers.19 The nursing executives are senior nursing leaders at participating UHC institutions and members of a dedicated UHC Chief Nursing Officer Council E‐mail listserv. We designed a brief survey and distributed it via their E‐mail listserv using an online commercial survey administration tool.20 Respondents were asked to complete the survey or have one of their colleagues familiar with local DNR identification practices complete it on their behalf. The online tool also provided summary reports and descriptive findings to meet the study objectives. We provided a 1‐month window (during summer 2006) with 1 interval E‐mail reminder to complete the surveys.

RESULTS

Survey announcements were E‐mailed to 127 nursing executives, 69 of whom completed it (response rate 54%). The respondents represented mostly academic medical centers (87%; another 13% represented affiliated community teaching hospitals), public institutions (89%), and large facilities (60% with more than 400 beds; 40% with 201‐400 beds). More than half the respondents (56%) reported their hospitals use paper chart documentation as the only method of identifying patients with a DNR order, whereas 16% reported their hospitals use only electronic health record (EHR) documentation (Fig. 1). Twenty‐five percent of hospitals (n = 17) use a color‐coded patient wristband in addition to either paper or electronic documentation. Of these 17 hospitals, a total of 8 colors or color schemes were employed to designate DNR status (Table 1).

Figure 1

The use of color‐coded wristbands was not limited to identification of DNR status. Fifty‐five percent of hospitals (n = 31) use color‐coded wristbands to indicate another piece of patient‐related data such as an allergy, fall risk, or same last name alert (Table 2). In fact, 12 indications were depicted by various colors, with variations in both the color choice for a given indication (eg, allergy wristbands red at one hospital and yellow at another) and across indications (eg, red for allergy at one hospital and red for bleeding risk at another). Nearly 3 of 4 respondents (n = 48) reported being aware of a case at your institution in which confusion about a DNR order led to problems or confusion in patient care. A few respondents shared a brief anecdote of the event, illustrating the spectrum of clinical scenarios that lead to potential confusion (Table 3). Respondents reporting a case of confusion were not more likely to be from an institution that used color‐coded wristbands.

When asked whether most (greater than 75%) physicians and nurses could properly identify the color associated with a DNR patient wristband, responses differed by discipline. Eight‐five percent of respondents believed that most nurses at their institutions could correctly report the color for DNR, whereas only 15% believed physicians could do the same. Only 22% of respondents anticipated a change in the current system within the next 2 years; all these changes were a transition from paper to electronic documentation systems.

DISCUSSION

Regardless of whether the DNR documentation occurs in paper or electronic form (and our study demonstrates significant practice variation in the documentation method), the risk that a hospitalized patient may suddenly stop breathing or become pulseless is ever present. When such a patient is discovered, providers race to the bedside and initiate care, but immediately ask, Is the patient a full code? In these often‐chaotic moments, accurate and timely information about DNR status is critical to respecting a patient's preferences and avoiding a potentially devastating error. A number of the anecdotes shared by survey respondents and highlighted in Table 3 reinforce this concern. Many of these scenarios occur in the middle of the night or off a patient's primary unit (ie, at a test or procedure area), increasing the need for quick and easy identification of DNR status.

Our study demonstrates that a logical point‐of‐care solutiona color‐coded DNR patient wristbandmay create its own safety hazards, particularly if the color designations are not known by all providers (including floating and traveling nurses or trainees who rotate at different hospitals) and if the colors being employed represent different indications at a given hospital (see accompanying Images Dx, page 445). We found that approximately 1 in 4 surveyed hospitals depict DNR status by a color‐coded wristband. We also discovered remarkable variation in the colors chosen and the degree to which institutions use color‐coded wristbands to signal a panoply of other patient‐related issues. Human factors research demonstrates that even well‐meaning patient safety solutions may cause harm in new ways if they are poorly implemented or if the interface between the technology and human work patterns is not well appreciated. For example, recent studies illustrate unintended consequences from safety‐driven solutions, such as the implementation of computerized order entry,2122 quality measurement,23 adoption of EHRs,24 and bar code medication administration systems.25 Because standardization is a key mechanism for decreasing the opportunities for error, our findings raise serious concerns about current wristband use.

Interestingly, the lack of standardization and its related risk of failing to recall the conditions associated with color‐coded wristbands are complicated by societal trends. In December 2004 the issue of patient wristbands made headlines in Florida, when hospitals using yellow DNR wristbands (as was the case in 3 hospitals in our sample) reported several near‐misses among patients wearing yellow Lance Armstrong Livestrong bracelets.2627 Given recent estimates that nearly 1 in 5 Americans wears these bracelets to support people living with cancer,28 even safety‐minded journals and national newspapers have highlighted the issue.2930 Most hospitals that continue to use yellow DNR wristbands now either remove or cover Livestrong bracelets at the time of hospital admission. Furthermore, many other self‐help organizations now issue wristbands in a variety of colors as well, creating a potential hazard for any person wearing one in the hospital. Although patients do not mind wearing color‐coded wristbands,31 they might feel differently if they knew the potential for confusion.

After these anecdotal reports of identification mistakes surfaced, several states, most notably Arizona and Pennsylvania, launched initiatives to address the problem.3233 Arizona, after discovering 8 colors being used in the state, developed plans for a purple DNR color‐coded wristband. The choice of purple, and the careful decision to avoid blue, occurred because many hospitals call their resuscitative efforts a code blue, creating yet another potential source of confusion if a blue wristband is associated with a DNR order. The Pennsylvania Patient Safety Authority also found tremendous color variations in patient wristbands used in a statewide survey. Both states ultimately promoted standardized colors and indications and provided tool kits and implementation manuals.3233

Although statewide initiatives represent a step forward, we believe that a national standard for color‐coded wristbands would improve patient safety. Precedents for this call to action exist. For many years, anecdotal information circulated about the errors caused by ambiguous use of abbreviations, such as qd instead of daily or U instead of units. Individual hospitals often banned or limited the use of such abbreviations, but no standard list of high‐risk abbreviations guided practice or required adherence, and cross‐hospital variation undoubtedly led to confusion. In 2004 the Joint Commission created a uniform list of high‐risk abbreviations as part of their National Patient Safety Goals, which instantly ended the debate about which abbreviations to ban and mandated compliance with the safety practice.34 A national group of stakeholders should similarly be convened to develop a list of colors and associated conditions that should be widely disseminated and enforced by the Joint Commission or a similar body. The statewide efforts by Arizona and Pennsylvania are instructive in this regard. Despite being guided by the goal of standardization, these 2 states chose different colors for DNR identification (interestingly, Pennsylvania chose blue for DNR, perhaps for the same reason that Arizona avoided itcode blue), further supporting the need for national guidelines (Table 4).

Our study represents the first national sample of DNR identification practices. Although it targeted academic health centers and affiliated institutions, we believe that these practice variations likely exist in all health care settings. Our study limitations included reliance on self‐reported institutional practices rather than direct review of existing policies and limited information about the surveyed population, making it impossible to compare respondents and nonrespondents. However, we have no reason to believe that these groups differed sufficiently to influence the study's main findings.

In the future, better technology may ultimately replace color‐coded wristbands. For instance, the time may come when wireless technologies seamlessly linked to the electronic health record will alert providers to a patient's DNR status when entering the patient's room. However, for today, point‐of‐care solutions using color‐coded wristbands remain a reasonable solution. Creating a nationally enforced standardized methodology, understandable and memorable to providers and free of stigma to patients (eg, a black wristband for DNR or writing DNR on a wristband) should be a patient safety priority. Because simplification is another key characteristic of safe systems, it seems prudent to aim for a national system that involves a maximum of 3‐4 colors.

CONCLUSIONS

Patients and families dedicate tremendous energy to making decisions about their advance directives, and discussions of these issues often create considerable angst and sadness. Health care providers are trained to elicit and advocate for such directives so they can act with patients' wishes in mind. Despite the high stakes, all these efforts can be undermined when the system for making providers aware of a patient's DNR status is flawed. Our data confirm the tremendous variability in the systems used to indicate DNR status (and other types of indications), variability that may place patients at risk from catastrophic errors. Following the lead of a few states, we call for a national mandate to standardize the identification of DNR orders and to make the colors of wristbands for a small set of indications uniform in every hospital across the country.

Shortcomings in the quality of care of hospitalized patients at the end of life, especially in the final days, are well documented.1, 2 Recent studies have highlighted inadequate pain and symptom control for hospitalized terminally ill patients,24 poor communication about treatment preferences,57 and limited or delayed referral for hospice care.810 Efforts to improve the quality of end‐of‐life care have been diverse, including increased educational programs,1113 development of palliative care units in hospitals,14, 15 and greater exposure to palliative care for physicians during residency training.16 Despite these efforts, studies assessing the attitudes and knowledge of physicians about hospice and palliative care continue to show deficits in knowledge about managing pain17, 18 as well as hospice policies and services.9

Among the interventions aimed at improving hospital care, the hospitalist movement has emerged as a model of care for improving the quality and cost efficiency of hospital care.1922 Because hospitalists spend substantial time on inpatient services,23 they are often involved in the care of patients with terminal illness, with potential to improve the quality of care that these patients receive while hospitalized. However, little is known about what specific knowledge and perspectives hospitalists and residents have about the care of patients with terminal illness. Although many studies have been conducted among physicians in private practice,9, 10, 2426 they have not focused on the knowledge, reported practices, and attitudes of hospitalists and residents concerning key aspects of end‐of‐life care and hospice. Such information can help to identify potential areas for improving knowledge and addressing common barriers highlighted in linking hospital and posthospital hospice care.

METHODS

RESULTS

DISCUSSION

This study demonstrated that, among hospitalists and residents, there are several misconceptions about fundamental aspects of caring for terminally ill patients. Given the potential importance of the role hospitalists play in improving the quality of inpatient care,1922 it is critical to identify and address these misconceptions. Additionally, physicians in this study indicated that more and earlier communication with patients and families about prognosis and about the option of hospice would be beneficial, but they themselves did not feel knowledgeable enough to discuss hospice and palliative care with patients and their families.

The nature of the misconceptions identified in this study shed light on the well‐documented phenomena of inadequate pain control24, 29 and poor symptom management2, 4 at the end of life. Having many of the erroneous beliefs apparent in this study may be consistent with providing less pain medication than needed and appropriate. For instance, many physicians believed that developing addiction to opioids used for cancer pain is more likely to occur than it really is, according to research evidence. It is extremely rare for these patients to become addicted to opioids or other analgesics (fewer than 1 in 1000 patients).28 In addition, most physicians believed that complaints of increased pain among patients receiving opioid therapy for pain control meant tolerance to the medication, a belief consistent with physician reluctance to prescribe more medication because it would lead to tolerance.28 In reality, the increased pain experienced in these situations is typically not a result of tolerance to the pain medication but to the cancer getting worse.27 Additionally, many physicians mistakenly decreased the dose of morphine in converting the route of administration from PO to IV, as is often done in hospitals. Such an error may be a contributing factor to the unintended undertreatment of pain in hospitals. Given the variability of cancer pain4 and the difference in time to peak effect depending on the route of administration,5 it is critical for physicians to understand proper dosing in order to effectively treat cancer pain. Furthermore, many physicians were incorrect about the recommended medications for dyspnea and for agitation, 2 symptoms that are prevalent among patients at the end of life.

The hospitalists and residents reported having very positive views about hospice, as is consistent with the literature.10, 30 However, many respondents indicated that patients who would have benefited from hospice did not receive it at all or only late in their illness. Physicians indicated that better communication with patients and families about hospice, prognosis, and goals of care would enhance appropriate use of hospice. While hospitalists and residents are in a position to initiate such discussions, they reported that these discussions were difficult for them. The challenge is how to promote what is necessary and valuable conversation with patients and families despite their difficulty, so that a realistic plan of care can be designed for all involved. Providing hospitalists and residents with evidence about what approaches are most effective in such discussions would be helpful to better prepare them for their roles in caring for hospitalized patients with terminal illness.

The results of this study have substantiated the need to enhance the education of hospitalists and resident physicians, who can play a vital role in improving the transition from hospital to hospice. Such education could take place as part of the residency experience or be embedded in various continuing medical education requirements that most states now have. The results of a recent national survey of hospitalists31 indicates they consider their palliative care training inadequate and feel ill prepared to care for patients with terminal illness. Our findings are consistent with those of that survey, highlighting information that is poorly understood by both residents and hospitalists. As hospitalists continue to play key roles in linking hospital to posthospital care,21 including hospice, there is greater opportunity to improve end‐of‐life care by expanding hospitalists' understanding of these issues.

Our findings should be interpreted in light of the study's limitations. First, this was an exploratory study, and the sample was modest in size. Nevertheless, the response rate was high: 85.2%. Second, we conducted the study in a single location; results may differ in other geographical areas. Last, we were unable to link reported knowledge and attitudes to patient experiences including quality of care or adequacy of pain control. Inadequate knowledge likely limits the quality of clinical practices, but the magnitude of this effect remains unknown and worthy of future study.

Despite these limitations, this study has contributed to the literature by identifying a set of misunderstandings or myths that may be common among hospitalists and residents who frequently care for hospitalized patients with terminal illness. Many of these misunderstandings were related to pain and symptom management, although some misunderstandings related to logistical issues such as hospice eligibility rules. Previous studies have described interventions to improve physicians' knowledge about palliative and end‐of‐life care practices at the undergraduate, graduate, and postgraduate levels.13 Our findings identified specific gaps in physicians' knowledge. Interventions aimed at closing these gaps might emphasize both specific clinical information about pain management and medication recommendations, and more general information about eligibility for hospice and best practices for communicating early with patients and family is needed to promote more effective care for patients with terminal illness being cared for in acute care settings.

As the use of hospitalists has become a widely accepted model of hospital care,32 ensuring their increased training and education in the care of patients with terminal illness is an important step in improving end‐of‐life care. Larger comparison studies are needed to identify differences in the practices and perspectives of hospitalists and residents and to target educational interventions to meet their particular needs. Further, conducting these studies at additional sites including those with established palliative care programs would be useful for identifying needs among a more diverse set of physicians involved in delivering end‐of‐life care.

Shortcomings in the quality of care of hospitalized patients at the end of life, especially in the final days, are well documented.1, 2 Recent studies have highlighted inadequate pain and symptom control for hospitalized terminally ill patients,24 poor communication about treatment preferences,57 and limited or delayed referral for hospice care.810 Efforts to improve the quality of end‐of‐life care have been diverse, including increased educational programs,1113 development of palliative care units in hospitals,14, 15 and greater exposure to palliative care for physicians during residency training.16 Despite these efforts, studies assessing the attitudes and knowledge of physicians about hospice and palliative care continue to show deficits in knowledge about managing pain17, 18 as well as hospice policies and services.9

Among the interventions aimed at improving hospital care, the hospitalist movement has emerged as a model of care for improving the quality and cost efficiency of hospital care.1922 Because hospitalists spend substantial time on inpatient services,23 they are often involved in the care of patients with terminal illness, with potential to improve the quality of care that these patients receive while hospitalized. However, little is known about what specific knowledge and perspectives hospitalists and residents have about the care of patients with terminal illness. Although many studies have been conducted among physicians in private practice,9, 10, 2426 they have not focused on the knowledge, reported practices, and attitudes of hospitalists and residents concerning key aspects of end‐of‐life care and hospice. Such information can help to identify potential areas for improving knowledge and addressing common barriers highlighted in linking hospital and posthospital hospice care.

METHODS

RESULTS

DISCUSSION

This study demonstrated that, among hospitalists and residents, there are several misconceptions about fundamental aspects of caring for terminally ill patients. Given the potential importance of the role hospitalists play in improving the quality of inpatient care,1922 it is critical to identify and address these misconceptions. Additionally, physicians in this study indicated that more and earlier communication with patients and families about prognosis and about the option of hospice would be beneficial, but they themselves did not feel knowledgeable enough to discuss hospice and palliative care with patients and their families.

The nature of the misconceptions identified in this study shed light on the well‐documented phenomena of inadequate pain control24, 29 and poor symptom management2, 4 at the end of life. Having many of the erroneous beliefs apparent in this study may be consistent with providing less pain medication than needed and appropriate. For instance, many physicians believed that developing addiction to opioids used for cancer pain is more likely to occur than it really is, according to research evidence. It is extremely rare for these patients to become addicted to opioids or other analgesics (fewer than 1 in 1000 patients).28 In addition, most physicians believed that complaints of increased pain among patients receiving opioid therapy for pain control meant tolerance to the medication, a belief consistent with physician reluctance to prescribe more medication because it would lead to tolerance.28 In reality, the increased pain experienced in these situations is typically not a result of tolerance to the pain medication but to the cancer getting worse.27 Additionally, many physicians mistakenly decreased the dose of morphine in converting the route of administration from PO to IV, as is often done in hospitals. Such an error may be a contributing factor to the unintended undertreatment of pain in hospitals. Given the variability of cancer pain4 and the difference in time to peak effect depending on the route of administration,5 it is critical for physicians to understand proper dosing in order to effectively treat cancer pain. Furthermore, many physicians were incorrect about the recommended medications for dyspnea and for agitation, 2 symptoms that are prevalent among patients at the end of life.

The hospitalists and residents reported having very positive views about hospice, as is consistent with the literature.10, 30 However, many respondents indicated that patients who would have benefited from hospice did not receive it at all or only late in their illness. Physicians indicated that better communication with patients and families about hospice, prognosis, and goals of care would enhance appropriate use of hospice. While hospitalists and residents are in a position to initiate such discussions, they reported that these discussions were difficult for them. The challenge is how to promote what is necessary and valuable conversation with patients and families despite their difficulty, so that a realistic plan of care can be designed for all involved. Providing hospitalists and residents with evidence about what approaches are most effective in such discussions would be helpful to better prepare them for their roles in caring for hospitalized patients with terminal illness.

The results of this study have substantiated the need to enhance the education of hospitalists and resident physicians, who can play a vital role in improving the transition from hospital to hospice. Such education could take place as part of the residency experience or be embedded in various continuing medical education requirements that most states now have. The results of a recent national survey of hospitalists31 indicates they consider their palliative care training inadequate and feel ill prepared to care for patients with terminal illness. Our findings are consistent with those of that survey, highlighting information that is poorly understood by both residents and hospitalists. As hospitalists continue to play key roles in linking hospital to posthospital care,21 including hospice, there is greater opportunity to improve end‐of‐life care by expanding hospitalists' understanding of these issues.

Our findings should be interpreted in light of the study's limitations. First, this was an exploratory study, and the sample was modest in size. Nevertheless, the response rate was high: 85.2%. Second, we conducted the study in a single location; results may differ in other geographical areas. Last, we were unable to link reported knowledge and attitudes to patient experiences including quality of care or adequacy of pain control. Inadequate knowledge likely limits the quality of clinical practices, but the magnitude of this effect remains unknown and worthy of future study.

Despite these limitations, this study has contributed to the literature by identifying a set of misunderstandings or myths that may be common among hospitalists and residents who frequently care for hospitalized patients with terminal illness. Many of these misunderstandings were related to pain and symptom management, although some misunderstandings related to logistical issues such as hospice eligibility rules. Previous studies have described interventions to improve physicians' knowledge about palliative and end‐of‐life care practices at the undergraduate, graduate, and postgraduate levels.13 Our findings identified specific gaps in physicians' knowledge. Interventions aimed at closing these gaps might emphasize both specific clinical information about pain management and medication recommendations, and more general information about eligibility for hospice and best practices for communicating early with patients and family is needed to promote more effective care for patients with terminal illness being cared for in acute care settings.

As the use of hospitalists has become a widely accepted model of hospital care,32 ensuring their increased training and education in the care of patients with terminal illness is an important step in improving end‐of‐life care. Larger comparison studies are needed to identify differences in the practices and perspectives of hospitalists and residents and to target educational interventions to meet their particular needs. Further, conducting these studies at additional sites including those with established palliative care programs would be useful for identifying needs among a more diverse set of physicians involved in delivering end‐of‐life care.

Shortcomings in the quality of care of hospitalized patients at the end of life, especially in the final days, are well documented.1, 2 Recent studies have highlighted inadequate pain and symptom control for hospitalized terminally ill patients,24 poor communication about treatment preferences,57 and limited or delayed referral for hospice care.810 Efforts to improve the quality of end‐of‐life care have been diverse, including increased educational programs,1113 development of palliative care units in hospitals,14, 15 and greater exposure to palliative care for physicians during residency training.16 Despite these efforts, studies assessing the attitudes and knowledge of physicians about hospice and palliative care continue to show deficits in knowledge about managing pain17, 18 as well as hospice policies and services.9

Among the interventions aimed at improving hospital care, the hospitalist movement has emerged as a model of care for improving the quality and cost efficiency of hospital care.1922 Because hospitalists spend substantial time on inpatient services,23 they are often involved in the care of patients with terminal illness, with potential to improve the quality of care that these patients receive while hospitalized. However, little is known about what specific knowledge and perspectives hospitalists and residents have about the care of patients with terminal illness. Although many studies have been conducted among physicians in private practice,9, 10, 2426 they have not focused on the knowledge, reported practices, and attitudes of hospitalists and residents concerning key aspects of end‐of‐life care and hospice. Such information can help to identify potential areas for improving knowledge and addressing common barriers highlighted in linking hospital and posthospital hospice care.

METHODS

RESULTS

DISCUSSION

This study demonstrated that, among hospitalists and residents, there are several misconceptions about fundamental aspects of caring for terminally ill patients. Given the potential importance of the role hospitalists play in improving the quality of inpatient care,1922 it is critical to identify and address these misconceptions. Additionally, physicians in this study indicated that more and earlier communication with patients and families about prognosis and about the option of hospice would be beneficial, but they themselves did not feel knowledgeable enough to discuss hospice and palliative care with patients and their families.

The nature of the misconceptions identified in this study shed light on the well‐documented phenomena of inadequate pain control24, 29 and poor symptom management2, 4 at the end of life. Having many of the erroneous beliefs apparent in this study may be consistent with providing less pain medication than needed and appropriate. For instance, many physicians believed that developing addiction to opioids used for cancer pain is more likely to occur than it really is, according to research evidence. It is extremely rare for these patients to become addicted to opioids or other analgesics (fewer than 1 in 1000 patients).28 In addition, most physicians believed that complaints of increased pain among patients receiving opioid therapy for pain control meant tolerance to the medication, a belief consistent with physician reluctance to prescribe more medication because it would lead to tolerance.28 In reality, the increased pain experienced in these situations is typically not a result of tolerance to the pain medication but to the cancer getting worse.27 Additionally, many physicians mistakenly decreased the dose of morphine in converting the route of administration from PO to IV, as is often done in hospitals. Such an error may be a contributing factor to the unintended undertreatment of pain in hospitals. Given the variability of cancer pain4 and the difference in time to peak effect depending on the route of administration,5 it is critical for physicians to understand proper dosing in order to effectively treat cancer pain. Furthermore, many physicians were incorrect about the recommended medications for dyspnea and for agitation, 2 symptoms that are prevalent among patients at the end of life.

The hospitalists and residents reported having very positive views about hospice, as is consistent with the literature.10, 30 However, many respondents indicated that patients who would have benefited from hospice did not receive it at all or only late in their illness. Physicians indicated that better communication with patients and families about hospice, prognosis, and goals of care would enhance appropriate use of hospice. While hospitalists and residents are in a position to initiate such discussions, they reported that these discussions were difficult for them. The challenge is how to promote what is necessary and valuable conversation with patients and families despite their difficulty, so that a realistic plan of care can be designed for all involved. Providing hospitalists and residents with evidence about what approaches are most effective in such discussions would be helpful to better prepare them for their roles in caring for hospitalized patients with terminal illness.

The results of this study have substantiated the need to enhance the education of hospitalists and resident physicians, who can play a vital role in improving the transition from hospital to hospice. Such education could take place as part of the residency experience or be embedded in various continuing medical education requirements that most states now have. The results of a recent national survey of hospitalists31 indicates they consider their palliative care training inadequate and feel ill prepared to care for patients with terminal illness. Our findings are consistent with those of that survey, highlighting information that is poorly understood by both residents and hospitalists. As hospitalists continue to play key roles in linking hospital to posthospital care,21 including hospice, there is greater opportunity to improve end‐of‐life care by expanding hospitalists' understanding of these issues.

Our findings should be interpreted in light of the study's limitations. First, this was an exploratory study, and the sample was modest in size. Nevertheless, the response rate was high: 85.2%. Second, we conducted the study in a single location; results may differ in other geographical areas. Last, we were unable to link reported knowledge and attitudes to patient experiences including quality of care or adequacy of pain control. Inadequate knowledge likely limits the quality of clinical practices, but the magnitude of this effect remains unknown and worthy of future study.

Despite these limitations, this study has contributed to the literature by identifying a set of misunderstandings or myths that may be common among hospitalists and residents who frequently care for hospitalized patients with terminal illness. Many of these misunderstandings were related to pain and symptom management, although some misunderstandings related to logistical issues such as hospice eligibility rules. Previous studies have described interventions to improve physicians' knowledge about palliative and end‐of‐life care practices at the undergraduate, graduate, and postgraduate levels.13 Our findings identified specific gaps in physicians' knowledge. Interventions aimed at closing these gaps might emphasize both specific clinical information about pain management and medication recommendations, and more general information about eligibility for hospice and best practices for communicating early with patients and family is needed to promote more effective care for patients with terminal illness being cared for in acute care settings.

As the use of hospitalists has become a widely accepted model of hospital care,32 ensuring their increased training and education in the care of patients with terminal illness is an important step in improving end‐of‐life care. Larger comparison studies are needed to identify differences in the practices and perspectives of hospitalists and residents and to target educational interventions to meet their particular needs. Further, conducting these studies at additional sites including those with established palliative care programs would be useful for identifying needs among a more diverse set of physicians involved in delivering end‐of‐life care.