Terrence O'Malley, MD

Recent studies have shown that a patient's discharge from the hospital is a vulnerable period for patient safety.14 With the reduction in length of stay (LOS) and the increase in patient acuity over the past decade, patients are discharged from acute care settings quicker and sicker, resulting in management of ongoing illness in a less‐monitored environment.5, 6 In addition, in teaching hospitals, residents are supervised by hospital‐based physicians who are rarely the primary care physician (PCP) for the residents' patients, which creates discontinuity of care.

One in 5 medical discharges is complicated by an adverse event believed, in part, to be due to poor communication between caregivers during this transition time.2 Discharge summaries, a key form of that communication, are not always done in a timely fashion and may lack key pieces of information.7, 8 For approximately 68% of patient discharges, the PCP will not have a discharge summary available for the patient's first follow‐up visit.911 In a survey of PCPs whose patients were in the hospital, only 23% reported direct communication with the hospital care team.12 This leaves PCPs unaware of pending test results or recommended follow‐up evaluations.10, 11, 13, 14 All of these factors are believed to contribute to adverse events, emergency department (ED) visits, and readmissions.

A recently published consensus statement on transitions of care by 6 major medical societies emphasizes the need for timely communication and transfer of information.15 These important processes are especially challenging to meet at academic medical centers, where discharge summaries and transition communication are done by residents in a hectic and challenging work environment, with multiple simultaneous and competing demands including outpatient clinic and required conferences.12 Residents have little formal training in how to write an effective discharge summary or how to systematically approach discharge planning. One study found higher error rates in discharge summaries written by residents compared with attending physicians.16 While the Accreditation Council for Graduate Medical Education (ACGME) limits the number of admissions per intern for both patient safety and educational reasons, the number of discharges per day is not limited despite the considerable amount of time required for appropriate discharge planning and communication.

Many interventions have been tried to improve the discharge process and reduce patient adverse events.17 Arranging early follow‐up appointments to reduce emergency department visits and readmissions has shown mixed results.13, 1820 Interventions that focus on specific populations, such as the elderly or patients with congestive heart failure, have been more successful.2123 Some interventions employed additional resources, such as a discharge form, transition coach, or discharge advocate, again with varying impact on results.18, 2427 A recent study by Jack et al. used nurse discharge advocates (DAs) to help with discharge planning and communication at an academic medical center.25 These DAs were independent of the care team, and focused on patient education and follow‐up plans, and reduced hospital reutilization in a selected population.

No studies have assessed the potential benefit of helping residents with the physician components of the discharge process. Prior studies have mainly focused on patient communication and follow‐up appointments, yet safe transitions also involve timely discharge summaries, physician‐to‐physician communication, physician‐to‐nurse communication, and medication reconciliation. Without support and time, these tasks can be very challenging for resident physicians with work‐hour limitations. We undertook a randomized, controlled trial to evaluate the impact on the discharge process of embedding a discharge facilitator in a resident medical team to help with the physician discharge process. We studied the effect for all the patients discharged from the resident team, rather than focusing on a select group or patients with a single diagnosis.

METHODS

RESULTS

Study Sample

During the 5‐month intervention (November 12, 2008 to April 14, 2009), a combined total of 999 patients were admitted to the intervention and control general medical teams. We excluded 96 patients who were not discharged but transferred to another service or intensive care units, and 24 patients who died. We also excluded 7 patients who were discharged from both teams the first day of the study, because the DF was not involved with the patients' discharge planning. That left 872 patients discharged to either home, a facility, or having left against medical advice (AMA) included in the study: 440 patients on the intervention team and 432 patients on the control team (Figure 1). Baseline patient demographic and clinical characteristics were similar across both teams with only gender being significantly different (Table 1). The mean age was 63 years (range, 1896) and the mean comorbidity score was 2.3 (range, 012). Of note, about a quarter of patients were discharged to facilities, about half were Medicare recipients, and approximately 80% had a PCP. The DF participated in the discharge process for nearly all of the intervention patients; she reported contributing approximately 50% of the content to the discharge summaries.

Figure 1

Table 1.Baseline Participant Characteristics

Characteristics	Intervention Team	Control Team
	n = 440	n = 432
Abbreviations: AMA, against medical advice; COPD, chronic obstructive pulmonary disease; PCP, primary care physician; SD, standard deviation. * P < 0.05; no other comparisons were statistically significant. Deyo Modification of the Charlson Comorbidity Index.
Mean age (SD), year	63 (18)	63 (18)
Women, n (%)*	181 (41)	207 (48)
Race, n (%)
White non‐Hispanic	267 (61)	243 (56)
Black non‐Hispanic	24 (5)	33 (8)
Hispanic	21 (5)	17 (4)
Unknown/other	128 (29)	139 (32)
Health insurance, n (%)
Medicare	213 (48)	226 (52)
Medicaid	85 (19)	81 (19)
Private	110 (25)	91 (21)
Other	32 (7)	34 (8)
PCP on admission, n (%)	370 (84)	356 (82)
Discharge disposition, n (%)
AMA	12 (3)	14 (3)
Home	305 (69)	315 (73)
Facility	123 (28)	103 (24)
Mean comorbidity index score (SD)	2.3 (2.4)	2.3 (2.4)
Diagnoses
Congestive heart failure	30 (6%)	27 (5%)
COPD/asthma	34 (7%)	47 (9%)
Cardiovascular disease	54 (11%)	50 (8%)
Alcohol/substance abuse	29 (6%)	34 (7%)
Gastrointestinal bleeds/ulcers	38 (8%)	41 (8%)
Hepatobiliary disease	30 (6%)	36 (7%)
Renal failure/kidney disease	33 (7%)	37 (7%)
Pneumonia	36 (7%)	22 (4%)
Musculoskeletal disease	26 (5%)	23 (5%)
Neurologic disease	22 (4%)	25 (5%)
Other	163 (33%)	172 (35%)

DISCUSSION

Our study embedded a nurse practitioner on a busy resident general medical team to help with all aspects of the discharge process for which physicians are responsible. Previous studies have been limited to patients with specific diagnoses, age, or disposition plans.1825 In this study, we included all general medical patients. Our intervention improved several important quality of care elements: the timeliness of completion of discharge summaries; and increased number of early follow‐up appointments, with more patients seen within 2 and 4 weeks after discharge. Patients reported better understanding of their follow‐up plans and more satisfaction with the discharge process. While not statistically significant, there was a trend towards better communication with nurses. For residents with work‐hour limitations, there was time savings with a trend towards finishing attending rounds on time and statistically significant earlier sign‐out rounds (46 minutes earlier). This intervention had no effect on patient length of stay, readmissions, or emergency department visits in the 30 days after discharge.

Despite improving many aspects of the discharge process and communication that have previously been raised as areas of concern for patient safety, there was no improvement in readmissions rates and ED utilization which are often used as the quality indicators for effective discharge planning. Similar types of interventions on general medical patients have generally also failed to show improvement in readmission rates.1820, 25 Weinberger et al. arranged follow‐up appointments within 1 week for patients discharged from a Veterans Administrative hospital; while patients were seen more often, the intervention actually increased readmission rates.20 Fitzgerald et al. had a case manager contact patients at home and encourage follow‐up, which increased follow‐up visits, but again had no effect on readmission.19 Einstadter et al. had a nurse case manager coordinate outpatient follow‐up on a resident team and also did not effect readmission rates or ED visits.18 Jack et al. in project reengineered discharge (RED) did show a significant reduction in combined hospital utilization measures. However, their study focused on a more limited patient population, and employed both a discharge advocate to arrange follow‐up and improve patient education, and a pharmacist to make postdischarge phone calls.25

So why did readmissions rates and ED visits not change in our study? It would be reasonable to think that having earlier follow‐up appointments, better and timely physician‐to‐physician communication, and a facilitator for patient questions should improve the quality of the discharge process. In a recent study, Jha et al. found there was no association between chart‐based measures of discharge quality and readmissions rates, and only a modest association for patient‐reported measures of discharge quality and readmission rates.28 The authors suggest readmission rates are driven by many factors beyond just improved discharge safety. Perhaps readmission rates are too complex a measure to use to assess discharge process improvement. For fiscal reasons, it is understandable that hospitals, insurance companies, and the Centers for Medicare and Medicaid want to reduce readmission rates and ED utilization. Jencks et al. noted the cost of readmissions in 2004 was 17.4 billion dollars.29 However, sweeping efforts to improve the discharge process for all general medical patients may not yield significant reductions in readmissions, as this study suggests. We may need to focus aggressive intervention on smaller target populations, as prior studies on focused groups suggest.2123

There are no evidence‐based studies to suggest when optimal follow‐up should occur after discharge.26 Several medical society guidelines recommend 2 weeks. More patients on the intervention team were seen within 2 weeks, but readmission rates were not affected. The University Health System Consortium recently reported that the majority of readmissions occurred within 6 days, with the average being about 2 to 3 days.30 In this study, the median days to readmit were 12 for the intervention team and 10 for the control. It is possible that even with our improved 2‐week follow‐up, this was not early enough to reduce readmissions. Follow‐up may need to be within 13 days of discharge for highly vulnerable patients, to significantly change readmission rates. Further studies focusing on this question would be helpful.

Finally, with ACGME limitation of work hours, many residency programs are looking for ways to reduce residents' workload and increase time for education. With a significant trend towards finishing attending rounds on time, it is likely that more residents on the intervention team were able to attend the noon‐time educational conferences. We speculate that this was due to fewer interruptions during rounds because the DF was available for nurses' questions. Sign‐out rounds occurred significantly earlier, possibly because of improved resident efficiency due to the DF's help with the discharge process. While residents may lose some educational experience from not performing all discharge tasks, they gain experience working in interdisciplinary teams, have increased time for education, and reduced work hours. Since the ACGME limits the number of residents per program and increasing the residency size is not an option, a DF should be considered as a possible solution to ACGME work‐hour restrictions.

This study had several limitations. First, the intervention team had 1 specific person embedded, and therefore the results of this study may have limited generalizability. Second, the limited number of residents working with the DF could have biased the intervention, as not all residents worked equally well with the DF. However, this may represent the real‐world experience on any teaching service, given variation in working styles and learning curves of residents over their training. Third, this study was done at 1 university‐affiliated urban Academic Medical Center, making it potentially less generalizable to resident teams in community hospitals. Fourth, we were not able to capture readmissions and ED visits at institutions outside the MGHPartners Healthcare System. However, given that patients were assigned at random to either team, this factor should have impacted both teams equally. Fifth, the study occurred during Massachusetts healthcare reform which requires everyone to have health insurance. This may have affected the rates of ED visits and readmission rates, especially with a shortage of primary care physicians and office visits. Finally, this intervention was not cost‐neutral. Paying for a nurse practitioner to help residents with the work of discharge and providing patients with additional services had many advantages, but this quality improvement project did not pay for itself through shorter LOS, or decreases in ED visits or readmissions.

While readmission rates and ED utilization are important patient outcomes, especially in the current healthcare climate, what determines readmissions and ED visits is likely complex and multifactorial. This study suggests that, in the nationwide effort to reduce readmissions, solely improving the discharge process for all general medical patients may not produce the hoped‐for financial savings. Improving the discharge process, however, is something valuable in its own right. Adding a DF to a resident team does improve some quality markers of the discharge process and decreases work hours for residents.

Recent studies have shown that a patient's discharge from the hospital is a vulnerable period for patient safety.14 With the reduction in length of stay (LOS) and the increase in patient acuity over the past decade, patients are discharged from acute care settings quicker and sicker, resulting in management of ongoing illness in a less‐monitored environment.5, 6 In addition, in teaching hospitals, residents are supervised by hospital‐based physicians who are rarely the primary care physician (PCP) for the residents' patients, which creates discontinuity of care.

One in 5 medical discharges is complicated by an adverse event believed, in part, to be due to poor communication between caregivers during this transition time.2 Discharge summaries, a key form of that communication, are not always done in a timely fashion and may lack key pieces of information.7, 8 For approximately 68% of patient discharges, the PCP will not have a discharge summary available for the patient's first follow‐up visit.911 In a survey of PCPs whose patients were in the hospital, only 23% reported direct communication with the hospital care team.12 This leaves PCPs unaware of pending test results or recommended follow‐up evaluations.10, 11, 13, 14 All of these factors are believed to contribute to adverse events, emergency department (ED) visits, and readmissions.

A recently published consensus statement on transitions of care by 6 major medical societies emphasizes the need for timely communication and transfer of information.15 These important processes are especially challenging to meet at academic medical centers, where discharge summaries and transition communication are done by residents in a hectic and challenging work environment, with multiple simultaneous and competing demands including outpatient clinic and required conferences.12 Residents have little formal training in how to write an effective discharge summary or how to systematically approach discharge planning. One study found higher error rates in discharge summaries written by residents compared with attending physicians.16 While the Accreditation Council for Graduate Medical Education (ACGME) limits the number of admissions per intern for both patient safety and educational reasons, the number of discharges per day is not limited despite the considerable amount of time required for appropriate discharge planning and communication.

Many interventions have been tried to improve the discharge process and reduce patient adverse events.17 Arranging early follow‐up appointments to reduce emergency department visits and readmissions has shown mixed results.13, 1820 Interventions that focus on specific populations, such as the elderly or patients with congestive heart failure, have been more successful.2123 Some interventions employed additional resources, such as a discharge form, transition coach, or discharge advocate, again with varying impact on results.18, 2427 A recent study by Jack et al. used nurse discharge advocates (DAs) to help with discharge planning and communication at an academic medical center.25 These DAs were independent of the care team, and focused on patient education and follow‐up plans, and reduced hospital reutilization in a selected population.

No studies have assessed the potential benefit of helping residents with the physician components of the discharge process. Prior studies have mainly focused on patient communication and follow‐up appointments, yet safe transitions also involve timely discharge summaries, physician‐to‐physician communication, physician‐to‐nurse communication, and medication reconciliation. Without support and time, these tasks can be very challenging for resident physicians with work‐hour limitations. We undertook a randomized, controlled trial to evaluate the impact on the discharge process of embedding a discharge facilitator in a resident medical team to help with the physician discharge process. We studied the effect for all the patients discharged from the resident team, rather than focusing on a select group or patients with a single diagnosis.

METHODS

RESULTS

Study Sample

During the 5‐month intervention (November 12, 2008 to April 14, 2009), a combined total of 999 patients were admitted to the intervention and control general medical teams. We excluded 96 patients who were not discharged but transferred to another service or intensive care units, and 24 patients who died. We also excluded 7 patients who were discharged from both teams the first day of the study, because the DF was not involved with the patients' discharge planning. That left 872 patients discharged to either home, a facility, or having left against medical advice (AMA) included in the study: 440 patients on the intervention team and 432 patients on the control team (Figure 1). Baseline patient demographic and clinical characteristics were similar across both teams with only gender being significantly different (Table 1). The mean age was 63 years (range, 1896) and the mean comorbidity score was 2.3 (range, 012). Of note, about a quarter of patients were discharged to facilities, about half were Medicare recipients, and approximately 80% had a PCP. The DF participated in the discharge process for nearly all of the intervention patients; she reported contributing approximately 50% of the content to the discharge summaries.

Figure 1

Table 1.Baseline Participant Characteristics

Characteristics	Intervention Team	Control Team
	n = 440	n = 432
Abbreviations: AMA, against medical advice; COPD, chronic obstructive pulmonary disease; PCP, primary care physician; SD, standard deviation. * P < 0.05; no other comparisons were statistically significant. Deyo Modification of the Charlson Comorbidity Index.
Mean age (SD), year	63 (18)	63 (18)
Women, n (%)*	181 (41)	207 (48)
Race, n (%)
White non‐Hispanic	267 (61)	243 (56)
Black non‐Hispanic	24 (5)	33 (8)
Hispanic	21 (5)	17 (4)
Unknown/other	128 (29)	139 (32)
Health insurance, n (%)
Medicare	213 (48)	226 (52)
Medicaid	85 (19)	81 (19)
Private	110 (25)	91 (21)
Other	32 (7)	34 (8)
PCP on admission, n (%)	370 (84)	356 (82)
Discharge disposition, n (%)
AMA	12 (3)	14 (3)
Home	305 (69)	315 (73)
Facility	123 (28)	103 (24)
Mean comorbidity index score (SD)	2.3 (2.4)	2.3 (2.4)
Diagnoses
Congestive heart failure	30 (6%)	27 (5%)
COPD/asthma	34 (7%)	47 (9%)
Cardiovascular disease	54 (11%)	50 (8%)
Alcohol/substance abuse	29 (6%)	34 (7%)
Gastrointestinal bleeds/ulcers	38 (8%)	41 (8%)
Hepatobiliary disease	30 (6%)	36 (7%)
Renal failure/kidney disease	33 (7%)	37 (7%)
Pneumonia	36 (7%)	22 (4%)
Musculoskeletal disease	26 (5%)	23 (5%)
Neurologic disease	22 (4%)	25 (5%)
Other	163 (33%)	172 (35%)

DISCUSSION

Our study embedded a nurse practitioner on a busy resident general medical team to help with all aspects of the discharge process for which physicians are responsible. Previous studies have been limited to patients with specific diagnoses, age, or disposition plans.1825 In this study, we included all general medical patients. Our intervention improved several important quality of care elements: the timeliness of completion of discharge summaries; and increased number of early follow‐up appointments, with more patients seen within 2 and 4 weeks after discharge. Patients reported better understanding of their follow‐up plans and more satisfaction with the discharge process. While not statistically significant, there was a trend towards better communication with nurses. For residents with work‐hour limitations, there was time savings with a trend towards finishing attending rounds on time and statistically significant earlier sign‐out rounds (46 minutes earlier). This intervention had no effect on patient length of stay, readmissions, or emergency department visits in the 30 days after discharge.

Despite improving many aspects of the discharge process and communication that have previously been raised as areas of concern for patient safety, there was no improvement in readmissions rates and ED utilization which are often used as the quality indicators for effective discharge planning. Similar types of interventions on general medical patients have generally also failed to show improvement in readmission rates.1820, 25 Weinberger et al. arranged follow‐up appointments within 1 week for patients discharged from a Veterans Administrative hospital; while patients were seen more often, the intervention actually increased readmission rates.20 Fitzgerald et al. had a case manager contact patients at home and encourage follow‐up, which increased follow‐up visits, but again had no effect on readmission.19 Einstadter et al. had a nurse case manager coordinate outpatient follow‐up on a resident team and also did not effect readmission rates or ED visits.18 Jack et al. in project reengineered discharge (RED) did show a significant reduction in combined hospital utilization measures. However, their study focused on a more limited patient population, and employed both a discharge advocate to arrange follow‐up and improve patient education, and a pharmacist to make postdischarge phone calls.25

So why did readmissions rates and ED visits not change in our study? It would be reasonable to think that having earlier follow‐up appointments, better and timely physician‐to‐physician communication, and a facilitator for patient questions should improve the quality of the discharge process. In a recent study, Jha et al. found there was no association between chart‐based measures of discharge quality and readmissions rates, and only a modest association for patient‐reported measures of discharge quality and readmission rates.28 The authors suggest readmission rates are driven by many factors beyond just improved discharge safety. Perhaps readmission rates are too complex a measure to use to assess discharge process improvement. For fiscal reasons, it is understandable that hospitals, insurance companies, and the Centers for Medicare and Medicaid want to reduce readmission rates and ED utilization. Jencks et al. noted the cost of readmissions in 2004 was 17.4 billion dollars.29 However, sweeping efforts to improve the discharge process for all general medical patients may not yield significant reductions in readmissions, as this study suggests. We may need to focus aggressive intervention on smaller target populations, as prior studies on focused groups suggest.2123

There are no evidence‐based studies to suggest when optimal follow‐up should occur after discharge.26 Several medical society guidelines recommend 2 weeks. More patients on the intervention team were seen within 2 weeks, but readmission rates were not affected. The University Health System Consortium recently reported that the majority of readmissions occurred within 6 days, with the average being about 2 to 3 days.30 In this study, the median days to readmit were 12 for the intervention team and 10 for the control. It is possible that even with our improved 2‐week follow‐up, this was not early enough to reduce readmissions. Follow‐up may need to be within 13 days of discharge for highly vulnerable patients, to significantly change readmission rates. Further studies focusing on this question would be helpful.

Finally, with ACGME limitation of work hours, many residency programs are looking for ways to reduce residents' workload and increase time for education. With a significant trend towards finishing attending rounds on time, it is likely that more residents on the intervention team were able to attend the noon‐time educational conferences. We speculate that this was due to fewer interruptions during rounds because the DF was available for nurses' questions. Sign‐out rounds occurred significantly earlier, possibly because of improved resident efficiency due to the DF's help with the discharge process. While residents may lose some educational experience from not performing all discharge tasks, they gain experience working in interdisciplinary teams, have increased time for education, and reduced work hours. Since the ACGME limits the number of residents per program and increasing the residency size is not an option, a DF should be considered as a possible solution to ACGME work‐hour restrictions.

This study had several limitations. First, the intervention team had 1 specific person embedded, and therefore the results of this study may have limited generalizability. Second, the limited number of residents working with the DF could have biased the intervention, as not all residents worked equally well with the DF. However, this may represent the real‐world experience on any teaching service, given variation in working styles and learning curves of residents over their training. Third, this study was done at 1 university‐affiliated urban Academic Medical Center, making it potentially less generalizable to resident teams in community hospitals. Fourth, we were not able to capture readmissions and ED visits at institutions outside the MGHPartners Healthcare System. However, given that patients were assigned at random to either team, this factor should have impacted both teams equally. Fifth, the study occurred during Massachusetts healthcare reform which requires everyone to have health insurance. This may have affected the rates of ED visits and readmission rates, especially with a shortage of primary care physicians and office visits. Finally, this intervention was not cost‐neutral. Paying for a nurse practitioner to help residents with the work of discharge and providing patients with additional services had many advantages, but this quality improvement project did not pay for itself through shorter LOS, or decreases in ED visits or readmissions.

While readmission rates and ED utilization are important patient outcomes, especially in the current healthcare climate, what determines readmissions and ED visits is likely complex and multifactorial. This study suggests that, in the nationwide effort to reduce readmissions, solely improving the discharge process for all general medical patients may not produce the hoped‐for financial savings. Improving the discharge process, however, is something valuable in its own right. Adding a DF to a resident team does improve some quality markers of the discharge process and decreases work hours for residents.

Recent studies have shown that a patient's discharge from the hospital is a vulnerable period for patient safety.14 With the reduction in length of stay (LOS) and the increase in patient acuity over the past decade, patients are discharged from acute care settings quicker and sicker, resulting in management of ongoing illness in a less‐monitored environment.5, 6 In addition, in teaching hospitals, residents are supervised by hospital‐based physicians who are rarely the primary care physician (PCP) for the residents' patients, which creates discontinuity of care.

One in 5 medical discharges is complicated by an adverse event believed, in part, to be due to poor communication between caregivers during this transition time.2 Discharge summaries, a key form of that communication, are not always done in a timely fashion and may lack key pieces of information.7, 8 For approximately 68% of patient discharges, the PCP will not have a discharge summary available for the patient's first follow‐up visit.911 In a survey of PCPs whose patients were in the hospital, only 23% reported direct communication with the hospital care team.12 This leaves PCPs unaware of pending test results or recommended follow‐up evaluations.10, 11, 13, 14 All of these factors are believed to contribute to adverse events, emergency department (ED) visits, and readmissions.

A recently published consensus statement on transitions of care by 6 major medical societies emphasizes the need for timely communication and transfer of information.15 These important processes are especially challenging to meet at academic medical centers, where discharge summaries and transition communication are done by residents in a hectic and challenging work environment, with multiple simultaneous and competing demands including outpatient clinic and required conferences.12 Residents have little formal training in how to write an effective discharge summary or how to systematically approach discharge planning. One study found higher error rates in discharge summaries written by residents compared with attending physicians.16 While the Accreditation Council for Graduate Medical Education (ACGME) limits the number of admissions per intern for both patient safety and educational reasons, the number of discharges per day is not limited despite the considerable amount of time required for appropriate discharge planning and communication.

Many interventions have been tried to improve the discharge process and reduce patient adverse events.17 Arranging early follow‐up appointments to reduce emergency department visits and readmissions has shown mixed results.13, 1820 Interventions that focus on specific populations, such as the elderly or patients with congestive heart failure, have been more successful.2123 Some interventions employed additional resources, such as a discharge form, transition coach, or discharge advocate, again with varying impact on results.18, 2427 A recent study by Jack et al. used nurse discharge advocates (DAs) to help with discharge planning and communication at an academic medical center.25 These DAs were independent of the care team, and focused on patient education and follow‐up plans, and reduced hospital reutilization in a selected population.

No studies have assessed the potential benefit of helping residents with the physician components of the discharge process. Prior studies have mainly focused on patient communication and follow‐up appointments, yet safe transitions also involve timely discharge summaries, physician‐to‐physician communication, physician‐to‐nurse communication, and medication reconciliation. Without support and time, these tasks can be very challenging for resident physicians with work‐hour limitations. We undertook a randomized, controlled trial to evaluate the impact on the discharge process of embedding a discharge facilitator in a resident medical team to help with the physician discharge process. We studied the effect for all the patients discharged from the resident team, rather than focusing on a select group or patients with a single diagnosis.

METHODS

RESULTS

Study Sample

During the 5‐month intervention (November 12, 2008 to April 14, 2009), a combined total of 999 patients were admitted to the intervention and control general medical teams. We excluded 96 patients who were not discharged but transferred to another service or intensive care units, and 24 patients who died. We also excluded 7 patients who were discharged from both teams the first day of the study, because the DF was not involved with the patients' discharge planning. That left 872 patients discharged to either home, a facility, or having left against medical advice (AMA) included in the study: 440 patients on the intervention team and 432 patients on the control team (Figure 1). Baseline patient demographic and clinical characteristics were similar across both teams with only gender being significantly different (Table 1). The mean age was 63 years (range, 1896) and the mean comorbidity score was 2.3 (range, 012). Of note, about a quarter of patients were discharged to facilities, about half were Medicare recipients, and approximately 80% had a PCP. The DF participated in the discharge process for nearly all of the intervention patients; she reported contributing approximately 50% of the content to the discharge summaries.

Figure 1

Table 1.Baseline Participant Characteristics

Characteristics	Intervention Team	Control Team
	n = 440	n = 432
Abbreviations: AMA, against medical advice; COPD, chronic obstructive pulmonary disease; PCP, primary care physician; SD, standard deviation. * P < 0.05; no other comparisons were statistically significant. Deyo Modification of the Charlson Comorbidity Index.
Mean age (SD), year	63 (18)	63 (18)
Women, n (%)*	181 (41)	207 (48)
Race, n (%)
White non‐Hispanic	267 (61)	243 (56)
Black non‐Hispanic	24 (5)	33 (8)
Hispanic	21 (5)	17 (4)
Unknown/other	128 (29)	139 (32)
Health insurance, n (%)
Medicare	213 (48)	226 (52)
Medicaid	85 (19)	81 (19)
Private	110 (25)	91 (21)
Other	32 (7)	34 (8)
PCP on admission, n (%)	370 (84)	356 (82)
Discharge disposition, n (%)
AMA	12 (3)	14 (3)
Home	305 (69)	315 (73)
Facility	123 (28)	103 (24)
Mean comorbidity index score (SD)	2.3 (2.4)	2.3 (2.4)
Diagnoses
Congestive heart failure	30 (6%)	27 (5%)
COPD/asthma	34 (7%)	47 (9%)
Cardiovascular disease	54 (11%)	50 (8%)
Alcohol/substance abuse	29 (6%)	34 (7%)
Gastrointestinal bleeds/ulcers	38 (8%)	41 (8%)
Hepatobiliary disease	30 (6%)	36 (7%)
Renal failure/kidney disease	33 (7%)	37 (7%)
Pneumonia	36 (7%)	22 (4%)
Musculoskeletal disease	26 (5%)	23 (5%)
Neurologic disease	22 (4%)	25 (5%)
Other	163 (33%)	172 (35%)

DISCUSSION

Our study embedded a nurse practitioner on a busy resident general medical team to help with all aspects of the discharge process for which physicians are responsible. Previous studies have been limited to patients with specific diagnoses, age, or disposition plans.1825 In this study, we included all general medical patients. Our intervention improved several important quality of care elements: the timeliness of completion of discharge summaries; and increased number of early follow‐up appointments, with more patients seen within 2 and 4 weeks after discharge. Patients reported better understanding of their follow‐up plans and more satisfaction with the discharge process. While not statistically significant, there was a trend towards better communication with nurses. For residents with work‐hour limitations, there was time savings with a trend towards finishing attending rounds on time and statistically significant earlier sign‐out rounds (46 minutes earlier). This intervention had no effect on patient length of stay, readmissions, or emergency department visits in the 30 days after discharge.

Despite improving many aspects of the discharge process and communication that have previously been raised as areas of concern for patient safety, there was no improvement in readmissions rates and ED utilization which are often used as the quality indicators for effective discharge planning. Similar types of interventions on general medical patients have generally also failed to show improvement in readmission rates.1820, 25 Weinberger et al. arranged follow‐up appointments within 1 week for patients discharged from a Veterans Administrative hospital; while patients were seen more often, the intervention actually increased readmission rates.20 Fitzgerald et al. had a case manager contact patients at home and encourage follow‐up, which increased follow‐up visits, but again had no effect on readmission.19 Einstadter et al. had a nurse case manager coordinate outpatient follow‐up on a resident team and also did not effect readmission rates or ED visits.18 Jack et al. in project reengineered discharge (RED) did show a significant reduction in combined hospital utilization measures. However, their study focused on a more limited patient population, and employed both a discharge advocate to arrange follow‐up and improve patient education, and a pharmacist to make postdischarge phone calls.25

So why did readmissions rates and ED visits not change in our study? It would be reasonable to think that having earlier follow‐up appointments, better and timely physician‐to‐physician communication, and a facilitator for patient questions should improve the quality of the discharge process. In a recent study, Jha et al. found there was no association between chart‐based measures of discharge quality and readmissions rates, and only a modest association for patient‐reported measures of discharge quality and readmission rates.28 The authors suggest readmission rates are driven by many factors beyond just improved discharge safety. Perhaps readmission rates are too complex a measure to use to assess discharge process improvement. For fiscal reasons, it is understandable that hospitals, insurance companies, and the Centers for Medicare and Medicaid want to reduce readmission rates and ED utilization. Jencks et al. noted the cost of readmissions in 2004 was 17.4 billion dollars.29 However, sweeping efforts to improve the discharge process for all general medical patients may not yield significant reductions in readmissions, as this study suggests. We may need to focus aggressive intervention on smaller target populations, as prior studies on focused groups suggest.2123

There are no evidence‐based studies to suggest when optimal follow‐up should occur after discharge.26 Several medical society guidelines recommend 2 weeks. More patients on the intervention team were seen within 2 weeks, but readmission rates were not affected. The University Health System Consortium recently reported that the majority of readmissions occurred within 6 days, with the average being about 2 to 3 days.30 In this study, the median days to readmit were 12 for the intervention team and 10 for the control. It is possible that even with our improved 2‐week follow‐up, this was not early enough to reduce readmissions. Follow‐up may need to be within 13 days of discharge for highly vulnerable patients, to significantly change readmission rates. Further studies focusing on this question would be helpful.

Finally, with ACGME limitation of work hours, many residency programs are looking for ways to reduce residents' workload and increase time for education. With a significant trend towards finishing attending rounds on time, it is likely that more residents on the intervention team were able to attend the noon‐time educational conferences. We speculate that this was due to fewer interruptions during rounds because the DF was available for nurses' questions. Sign‐out rounds occurred significantly earlier, possibly because of improved resident efficiency due to the DF's help with the discharge process. While residents may lose some educational experience from not performing all discharge tasks, they gain experience working in interdisciplinary teams, have increased time for education, and reduced work hours. Since the ACGME limits the number of residents per program and increasing the residency size is not an option, a DF should be considered as a possible solution to ACGME work‐hour restrictions.

This study had several limitations. First, the intervention team had 1 specific person embedded, and therefore the results of this study may have limited generalizability. Second, the limited number of residents working with the DF could have biased the intervention, as not all residents worked equally well with the DF. However, this may represent the real‐world experience on any teaching service, given variation in working styles and learning curves of residents over their training. Third, this study was done at 1 university‐affiliated urban Academic Medical Center, making it potentially less generalizable to resident teams in community hospitals. Fourth, we were not able to capture readmissions and ED visits at institutions outside the MGHPartners Healthcare System. However, given that patients were assigned at random to either team, this factor should have impacted both teams equally. Fifth, the study occurred during Massachusetts healthcare reform which requires everyone to have health insurance. This may have affected the rates of ED visits and readmission rates, especially with a shortage of primary care physicians and office visits. Finally, this intervention was not cost‐neutral. Paying for a nurse practitioner to help residents with the work of discharge and providing patients with additional services had many advantages, but this quality improvement project did not pay for itself through shorter LOS, or decreases in ED visits or readmissions.

While readmission rates and ED utilization are important patient outcomes, especially in the current healthcare climate, what determines readmissions and ED visits is likely complex and multifactorial. This study suggests that, in the nationwide effort to reduce readmissions, solely improving the discharge process for all general medical patients may not produce the hoped‐for financial savings. Improving the discharge process, however, is something valuable in its own right. Adding a DF to a resident team does improve some quality markers of the discharge process and decreases work hours for residents.

Effective communication among physicians during the hospital discharge process is critical to patient care. Patients are at high risk of having an adverse drug event,1 readmission, or death2 during the transition from hospital to home.3 Ineffective communication between inpatient and outpatient providers has been implicated as a leading cause of adverse events.35 Conversely, efforts to improve communication have been shown to improve compliance with follow‐up tests and decrease readmission rates.6, 7 Recently, the absence of several specific data elements in discharge documentation have been shown to be common and to have potential for patient harm, including test results that are pending at the time of discharge.8, 9 Unexplained discrepancies between preadmission and discharge medication regimens are also common and potentially dangerous.1

According to the Joint Commission for Accreditation of Healthcare Organizations (TJC), the following elements should be included in discharge summaries: the reason for hospitalization; significant findings; procedures performed and care, treatment, and services provided; the patient's condition at discharge; and information provided to the patient and family, as appropriate.10 TJC also advocates medication reconciliation, a process of identifying the most accurate list of all medications a patient is takingincluding name, dosage, frequency, and routeand using this list to provide correct medications for patients anywhere within the health care system.11

Despite the importance of complete communication among providers at hospital discharge, a recent systematic review showed that discharge summaries often lacked important information such as diagnostic test results (missing from 33%‐63%), treatment or hospital course (7%‐22%), discharge medications (2%‐40%), test results pending at discharge (65%), patient or family counseling (90%‐92%), and follow‐up plans (2%‐43%).1

Most of the studies addressing this issue have evaluated communication pitfalls between acute care hospitals and primary care physicians among patients discharged home.17 In contrast, the quality of discharge documentation among patients discharged to rehabilitation centers and other subacute care facilities has been less well studied, perhaps due to relatively smaller numbers of patients discharged to such facilities. This communication is as or more important because these patients are potentially more vulnerable and their medical conditions more active than for patients discharged home.12 Furthermore, discharge information from acute care hospitals will often form the basis for admission orders at subacute facilities. Last, these patients will have a second transition in care (from subacute facility to home) whose quality is dependent at least in part on the quality of communication during the first transition.

The aim of this study was to evaluate the quality of information transfer among patients discharged from acute hospitals to subacute facilities across an integrated healthcare delivery system. The long‐term goals of this effort were to determine the areas most in need of improvement, to guide interventions to address these problems, and to track improvements in these measures over time as interventions are implemented and refined.

Methods

This observational study was conducted as part of a quality improvement project evaluating the quality of information provided during the discharge process across Partners Health Care System. The institutional review boards of the participating institutions approved the study.

Results

A total of 1501 discharge documentation packets were reviewed, including 980 patients (65%) from a medical unit and 521 patients (35%) from a surgical unit. Based on 2007 data, these packets represent approximately 4% of all eligible discharges to subacute facilities. Patients discharged from 1 of the 2 academic medical centers represented 44% of the sample. A total of 644 discharge packets (43%) were reviewed at acute sites and 814 packets (54%) were reviewed at subacute sites. Information about reviewer site was missing in 43 discharge packets (3%). For the 29 charts independently reviewed by 2 reviewers, there was complete agreement for 331 out of 348 data elements (95.1%).

Only 1055 (70%) discharge summaries had all the information required by TJC (Table 2). Physical examination at admission (a component of significant findings, as noted above) and condition at discharge were the 2 elements most often missing. The defect‐free rate varied by site, with a range of 61% to 76% across the 5 acute care hospitals (data not shown).

The rates of inclusion of other (non‐TJC required) data elements are shown in Table 2. Most often missing were preadmission medication regimens, any documented reason for any difference between preadmission and discharge medications, pertinent laboratory results, and an adequate follow‐up plan (including who to follow up with, when to follow‐up, and a list of tasks to be accomplished at the follow‐up visit). Notation regarding significant test results that were pending at the time of transfer was missing in 160 of 341 applicable patients (47%), and in 162 patients (11%), physicians uncovered a significant condition that was not mentioned in the discharge documentation. Only 503 (33.5%) discharge documentation packets had all applicable measures present. In addition, the discharge summary was not received at all on the day of discharge according to the receiving site in 90 patients (6%).

Reviewers were asked in a separate question which missing data were necessary for patient care. Data elements most often cited were explanations for any medication discrepancies and test results pending at the time of the hospital discharge.

Community hospitals had a higher rate of inclusion of TJC‐required data elements when compared to academic medical centers (Table 3). Also, among non‐TJC required data elements, inclusion rates were higher among the community hospitals, especially regarding information about medication discrepancies, pending test results, and follow‐up information (Table 3).

Although no differences were found between medical and surgical services regarding compliance with TJC requirements, a difference was noted in documentation of explanations of medication discrepancies and pending test results, with medical services performing better in both measures (Table 3).

In general, reviewers at subacute sites more often evaluated packets as deficient than reviewers at acute sites, up to an absolute difference of 33% in the proportion of missing data, depending on the data element (see Appendix, Table 1).

Discussion

Our study evaluated the completeness of documentation in the discharge summaries of patients discharged from acute care to subacute care facilities. Our results for the inclusion of TJC‐required data elements were similar to those quoted in the literature for patients discharged home.6 Our results also demonstrated a high rate of other missing data elements that are arguably of equal or greater importance, including reasons for discrepancies between preadmission and discharge medication regimens and tests that are pending at the time of discharge.1, 8, 9 Our results also demonstrated the relatively poorer performance of academic centers compared to community hospitals regarding inclusion of information about medication reconciliation, follow‐up, pending test results, and complete information required by TJC. Finally, we found that patients discharged from surgical services more often lacked documentation of medication discrepancies and pending test results compared with patients from medical services.

To our knowledge, this is one of the first studies looking at the quality of information transfer in patients discharged to subacute care facilities. The results of this study are not surprising given the known problems with general information transfer at hospital discharge.1 The fact that community hospitals provided more complete information than academic medical centers for certain data elements may be due to the difference between residents and more senior physicians preparing discharge documentation. Such differences could reflect differences in experience, training, and degree of appreciation for the importance of discharge documentation, and/or restrictions in work hours among residents (eg, resulting in time‐pressure to complete discharge summaries and/or summaries being written by residents who know the patients less well). These hypotheses deserve further exploration. The differences between medical and surgical services should also be validated and explored in other healthcare systems, including both academic and community settings.

The results of this study should be viewed in light of the study's limitations. Packets evaluated by reviewers at subacute facilities were chosen by the reviewers and may not have been representative of all patients received by that facility (in contrast to those reviewed at the acute sites, which were chosen at random and more likely to be representative, although we did not formally test for this). It is possible that reviewers at subacute sites selected the worst discharge documentation packets for evaluation. Second, evaluations by reviewers at subacute sites did not distinguish between information missing from discharge documentation and failure to receive the documentation at all from the acute care hospital (again in contrast to reviewers at acute sites, who always had access to the documentation). Lastly, reviewers at acute and subacute sites may have graded packets differently due to their different clinical perspectives. These 3 factors may explain the relatively poorer results of discharge packets reviewed by reviewers at subacute sites. Further study would be needed to distinguish among these possibilities (eg, having acute and subacute reviewers answer the same questions for the same discharge packets to allow us to measure interrater reliability between the different kinds of reviewers; explicitly asking subacute reviewers about receipt of each piece of documentation; comparing the distribution of diagnosis‐related group [DRG] codes and hospital length of stay in evaluated vs. total discharge packets as a measure of representativeness). We also cannot rule out the possibility of reviewer bias, but all reviewers were trained in a standardized fashion and we know that reliability of assessments were high, at least among reviewers at acute sites. Last, we did not measure actual or potential adverse events caused by these information deficits.

As part of a Partners‐wide initiative to improve transitions in care, the results were presented to the administrations of each of the 5 acute care hospitals. The Partners High Performance Medicine Transition team then began work with a steering committee (composed of representatives from each hospital) to address these deficiencies. Since then, the hospitals have taken several steps to improve the quality of information transfer for discharged patients, including the following:

• 1

  Technological improvements to the hospitals' discharge ordering systems to actively solicit and/or autoimport the required information into discharge documentation.

• 2

  Creation of discharge templates to record the required information on paper.

• 3

  Provision of feedback to clinicians and their service chiefs regarding the ongoing quality of their discharge documentation.

• 4

  Creation of an online Partners‐wide curriculum on discharge summary authorship, with a mandatory quiz to be taken by all incoming clinicians.

In conclusion, we found room for improvement in the inclusion of data elements required for the safe transfer of patients from acute hospitals to subacute facilities, especially in areas such as medication reconciliation, pending test results, and adequate follow‐up plans. We also found variation by site and type of service. For patients discharged to rehabilitation and other subacute facilities, improvement is needed in the communication of clinically relevant information to those providing continuing care.

Effective communication among physicians during the hospital discharge process is critical to patient care. Patients are at high risk of having an adverse drug event,1 readmission, or death2 during the transition from hospital to home.3 Ineffective communication between inpatient and outpatient providers has been implicated as a leading cause of adverse events.35 Conversely, efforts to improve communication have been shown to improve compliance with follow‐up tests and decrease readmission rates.6, 7 Recently, the absence of several specific data elements in discharge documentation have been shown to be common and to have potential for patient harm, including test results that are pending at the time of discharge.8, 9 Unexplained discrepancies between preadmission and discharge medication regimens are also common and potentially dangerous.1

According to the Joint Commission for Accreditation of Healthcare Organizations (TJC), the following elements should be included in discharge summaries: the reason for hospitalization; significant findings; procedures performed and care, treatment, and services provided; the patient's condition at discharge; and information provided to the patient and family, as appropriate.10 TJC also advocates medication reconciliation, a process of identifying the most accurate list of all medications a patient is takingincluding name, dosage, frequency, and routeand using this list to provide correct medications for patients anywhere within the health care system.11

Despite the importance of complete communication among providers at hospital discharge, a recent systematic review showed that discharge summaries often lacked important information such as diagnostic test results (missing from 33%‐63%), treatment or hospital course (7%‐22%), discharge medications (2%‐40%), test results pending at discharge (65%), patient or family counseling (90%‐92%), and follow‐up plans (2%‐43%).1

Most of the studies addressing this issue have evaluated communication pitfalls between acute care hospitals and primary care physicians among patients discharged home.17 In contrast, the quality of discharge documentation among patients discharged to rehabilitation centers and other subacute care facilities has been less well studied, perhaps due to relatively smaller numbers of patients discharged to such facilities. This communication is as or more important because these patients are potentially more vulnerable and their medical conditions more active than for patients discharged home.12 Furthermore, discharge information from acute care hospitals will often form the basis for admission orders at subacute facilities. Last, these patients will have a second transition in care (from subacute facility to home) whose quality is dependent at least in part on the quality of communication during the first transition.

The aim of this study was to evaluate the quality of information transfer among patients discharged from acute hospitals to subacute facilities across an integrated healthcare delivery system. The long‐term goals of this effort were to determine the areas most in need of improvement, to guide interventions to address these problems, and to track improvements in these measures over time as interventions are implemented and refined.

Methods

This observational study was conducted as part of a quality improvement project evaluating the quality of information provided during the discharge process across Partners Health Care System. The institutional review boards of the participating institutions approved the study.

Results

A total of 1501 discharge documentation packets were reviewed, including 980 patients (65%) from a medical unit and 521 patients (35%) from a surgical unit. Based on 2007 data, these packets represent approximately 4% of all eligible discharges to subacute facilities. Patients discharged from 1 of the 2 academic medical centers represented 44% of the sample. A total of 644 discharge packets (43%) were reviewed at acute sites and 814 packets (54%) were reviewed at subacute sites. Information about reviewer site was missing in 43 discharge packets (3%). For the 29 charts independently reviewed by 2 reviewers, there was complete agreement for 331 out of 348 data elements (95.1%).

Only 1055 (70%) discharge summaries had all the information required by TJC (Table 2). Physical examination at admission (a component of significant findings, as noted above) and condition at discharge were the 2 elements most often missing. The defect‐free rate varied by site, with a range of 61% to 76% across the 5 acute care hospitals (data not shown).

The rates of inclusion of other (non‐TJC required) data elements are shown in Table 2. Most often missing were preadmission medication regimens, any documented reason for any difference between preadmission and discharge medications, pertinent laboratory results, and an adequate follow‐up plan (including who to follow up with, when to follow‐up, and a list of tasks to be accomplished at the follow‐up visit). Notation regarding significant test results that were pending at the time of transfer was missing in 160 of 341 applicable patients (47%), and in 162 patients (11%), physicians uncovered a significant condition that was not mentioned in the discharge documentation. Only 503 (33.5%) discharge documentation packets had all applicable measures present. In addition, the discharge summary was not received at all on the day of discharge according to the receiving site in 90 patients (6%).

Reviewers were asked in a separate question which missing data were necessary for patient care. Data elements most often cited were explanations for any medication discrepancies and test results pending at the time of the hospital discharge.

Community hospitals had a higher rate of inclusion of TJC‐required data elements when compared to academic medical centers (Table 3). Also, among non‐TJC required data elements, inclusion rates were higher among the community hospitals, especially regarding information about medication discrepancies, pending test results, and follow‐up information (Table 3).

Although no differences were found between medical and surgical services regarding compliance with TJC requirements, a difference was noted in documentation of explanations of medication discrepancies and pending test results, with medical services performing better in both measures (Table 3).

In general, reviewers at subacute sites more often evaluated packets as deficient than reviewers at acute sites, up to an absolute difference of 33% in the proportion of missing data, depending on the data element (see Appendix, Table 1).

Discussion

Our study evaluated the completeness of documentation in the discharge summaries of patients discharged from acute care to subacute care facilities. Our results for the inclusion of TJC‐required data elements were similar to those quoted in the literature for patients discharged home.6 Our results also demonstrated a high rate of other missing data elements that are arguably of equal or greater importance, including reasons for discrepancies between preadmission and discharge medication regimens and tests that are pending at the time of discharge.1, 8, 9 Our results also demonstrated the relatively poorer performance of academic centers compared to community hospitals regarding inclusion of information about medication reconciliation, follow‐up, pending test results, and complete information required by TJC. Finally, we found that patients discharged from surgical services more often lacked documentation of medication discrepancies and pending test results compared with patients from medical services.

To our knowledge, this is one of the first studies looking at the quality of information transfer in patients discharged to subacute care facilities. The results of this study are not surprising given the known problems with general information transfer at hospital discharge.1 The fact that community hospitals provided more complete information than academic medical centers for certain data elements may be due to the difference between residents and more senior physicians preparing discharge documentation. Such differences could reflect differences in experience, training, and degree of appreciation for the importance of discharge documentation, and/or restrictions in work hours among residents (eg, resulting in time‐pressure to complete discharge summaries and/or summaries being written by residents who know the patients less well). These hypotheses deserve further exploration. The differences between medical and surgical services should also be validated and explored in other healthcare systems, including both academic and community settings.

The results of this study should be viewed in light of the study's limitations. Packets evaluated by reviewers at subacute facilities were chosen by the reviewers and may not have been representative of all patients received by that facility (in contrast to those reviewed at the acute sites, which were chosen at random and more likely to be representative, although we did not formally test for this). It is possible that reviewers at subacute sites selected the worst discharge documentation packets for evaluation. Second, evaluations by reviewers at subacute sites did not distinguish between information missing from discharge documentation and failure to receive the documentation at all from the acute care hospital (again in contrast to reviewers at acute sites, who always had access to the documentation). Lastly, reviewers at acute and subacute sites may have graded packets differently due to their different clinical perspectives. These 3 factors may explain the relatively poorer results of discharge packets reviewed by reviewers at subacute sites. Further study would be needed to distinguish among these possibilities (eg, having acute and subacute reviewers answer the same questions for the same discharge packets to allow us to measure interrater reliability between the different kinds of reviewers; explicitly asking subacute reviewers about receipt of each piece of documentation; comparing the distribution of diagnosis‐related group [DRG] codes and hospital length of stay in evaluated vs. total discharge packets as a measure of representativeness). We also cannot rule out the possibility of reviewer bias, but all reviewers were trained in a standardized fashion and we know that reliability of assessments were high, at least among reviewers at acute sites. Last, we did not measure actual or potential adverse events caused by these information deficits.

As part of a Partners‐wide initiative to improve transitions in care, the results were presented to the administrations of each of the 5 acute care hospitals. The Partners High Performance Medicine Transition team then began work with a steering committee (composed of representatives from each hospital) to address these deficiencies. Since then, the hospitals have taken several steps to improve the quality of information transfer for discharged patients, including the following:

• 1

  Technological improvements to the hospitals' discharge ordering systems to actively solicit and/or autoimport the required information into discharge documentation.

• 2

  Creation of discharge templates to record the required information on paper.

• 3

  Provision of feedback to clinicians and their service chiefs regarding the ongoing quality of their discharge documentation.

• 4

  Creation of an online Partners‐wide curriculum on discharge summary authorship, with a mandatory quiz to be taken by all incoming clinicians.

In conclusion, we found room for improvement in the inclusion of data elements required for the safe transfer of patients from acute hospitals to subacute facilities, especially in areas such as medication reconciliation, pending test results, and adequate follow‐up plans. We also found variation by site and type of service. For patients discharged to rehabilitation and other subacute facilities, improvement is needed in the communication of clinically relevant information to those providing continuing care.

Effective communication among physicians during the hospital discharge process is critical to patient care. Patients are at high risk of having an adverse drug event,1 readmission, or death2 during the transition from hospital to home.3 Ineffective communication between inpatient and outpatient providers has been implicated as a leading cause of adverse events.35 Conversely, efforts to improve communication have been shown to improve compliance with follow‐up tests and decrease readmission rates.6, 7 Recently, the absence of several specific data elements in discharge documentation have been shown to be common and to have potential for patient harm, including test results that are pending at the time of discharge.8, 9 Unexplained discrepancies between preadmission and discharge medication regimens are also common and potentially dangerous.1

According to the Joint Commission for Accreditation of Healthcare Organizations (TJC), the following elements should be included in discharge summaries: the reason for hospitalization; significant findings; procedures performed and care, treatment, and services provided; the patient's condition at discharge; and information provided to the patient and family, as appropriate.10 TJC also advocates medication reconciliation, a process of identifying the most accurate list of all medications a patient is takingincluding name, dosage, frequency, and routeand using this list to provide correct medications for patients anywhere within the health care system.11

Despite the importance of complete communication among providers at hospital discharge, a recent systematic review showed that discharge summaries often lacked important information such as diagnostic test results (missing from 33%‐63%), treatment or hospital course (7%‐22%), discharge medications (2%‐40%), test results pending at discharge (65%), patient or family counseling (90%‐92%), and follow‐up plans (2%‐43%).1

Most of the studies addressing this issue have evaluated communication pitfalls between acute care hospitals and primary care physicians among patients discharged home.17 In contrast, the quality of discharge documentation among patients discharged to rehabilitation centers and other subacute care facilities has been less well studied, perhaps due to relatively smaller numbers of patients discharged to such facilities. This communication is as or more important because these patients are potentially more vulnerable and their medical conditions more active than for patients discharged home.12 Furthermore, discharge information from acute care hospitals will often form the basis for admission orders at subacute facilities. Last, these patients will have a second transition in care (from subacute facility to home) whose quality is dependent at least in part on the quality of communication during the first transition.

The aim of this study was to evaluate the quality of information transfer among patients discharged from acute hospitals to subacute facilities across an integrated healthcare delivery system. The long‐term goals of this effort were to determine the areas most in need of improvement, to guide interventions to address these problems, and to track improvements in these measures over time as interventions are implemented and refined.

Methods

This observational study was conducted as part of a quality improvement project evaluating the quality of information provided during the discharge process across Partners Health Care System. The institutional review boards of the participating institutions approved the study.

Results

A total of 1501 discharge documentation packets were reviewed, including 980 patients (65%) from a medical unit and 521 patients (35%) from a surgical unit. Based on 2007 data, these packets represent approximately 4% of all eligible discharges to subacute facilities. Patients discharged from 1 of the 2 academic medical centers represented 44% of the sample. A total of 644 discharge packets (43%) were reviewed at acute sites and 814 packets (54%) were reviewed at subacute sites. Information about reviewer site was missing in 43 discharge packets (3%). For the 29 charts independently reviewed by 2 reviewers, there was complete agreement for 331 out of 348 data elements (95.1%).

Only 1055 (70%) discharge summaries had all the information required by TJC (Table 2). Physical examination at admission (a component of significant findings, as noted above) and condition at discharge were the 2 elements most often missing. The defect‐free rate varied by site, with a range of 61% to 76% across the 5 acute care hospitals (data not shown).

The rates of inclusion of other (non‐TJC required) data elements are shown in Table 2. Most often missing were preadmission medication regimens, any documented reason for any difference between preadmission and discharge medications, pertinent laboratory results, and an adequate follow‐up plan (including who to follow up with, when to follow‐up, and a list of tasks to be accomplished at the follow‐up visit). Notation regarding significant test results that were pending at the time of transfer was missing in 160 of 341 applicable patients (47%), and in 162 patients (11%), physicians uncovered a significant condition that was not mentioned in the discharge documentation. Only 503 (33.5%) discharge documentation packets had all applicable measures present. In addition, the discharge summary was not received at all on the day of discharge according to the receiving site in 90 patients (6%).

Reviewers were asked in a separate question which missing data were necessary for patient care. Data elements most often cited were explanations for any medication discrepancies and test results pending at the time of the hospital discharge.

Community hospitals had a higher rate of inclusion of TJC‐required data elements when compared to academic medical centers (Table 3). Also, among non‐TJC required data elements, inclusion rates were higher among the community hospitals, especially regarding information about medication discrepancies, pending test results, and follow‐up information (Table 3).

Although no differences were found between medical and surgical services regarding compliance with TJC requirements, a difference was noted in documentation of explanations of medication discrepancies and pending test results, with medical services performing better in both measures (Table 3).

In general, reviewers at subacute sites more often evaluated packets as deficient than reviewers at acute sites, up to an absolute difference of 33% in the proportion of missing data, depending on the data element (see Appendix, Table 1).

Discussion

Our study evaluated the completeness of documentation in the discharge summaries of patients discharged from acute care to subacute care facilities. Our results for the inclusion of TJC‐required data elements were similar to those quoted in the literature for patients discharged home.6 Our results also demonstrated a high rate of other missing data elements that are arguably of equal or greater importance, including reasons for discrepancies between preadmission and discharge medication regimens and tests that are pending at the time of discharge.1, 8, 9 Our results also demonstrated the relatively poorer performance of academic centers compared to community hospitals regarding inclusion of information about medication reconciliation, follow‐up, pending test results, and complete information required by TJC. Finally, we found that patients discharged from surgical services more often lacked documentation of medication discrepancies and pending test results compared with patients from medical services.

To our knowledge, this is one of the first studies looking at the quality of information transfer in patients discharged to subacute care facilities. The results of this study are not surprising given the known problems with general information transfer at hospital discharge.1 The fact that community hospitals provided more complete information than academic medical centers for certain data elements may be due to the difference between residents and more senior physicians preparing discharge documentation. Such differences could reflect differences in experience, training, and degree of appreciation for the importance of discharge documentation, and/or restrictions in work hours among residents (eg, resulting in time‐pressure to complete discharge summaries and/or summaries being written by residents who know the patients less well). These hypotheses deserve further exploration. The differences between medical and surgical services should also be validated and explored in other healthcare systems, including both academic and community settings.

The results of this study should be viewed in light of the study's limitations. Packets evaluated by reviewers at subacute facilities were chosen by the reviewers and may not have been representative of all patients received by that facility (in contrast to those reviewed at the acute sites, which were chosen at random and more likely to be representative, although we did not formally test for this). It is possible that reviewers at subacute sites selected the worst discharge documentation packets for evaluation. Second, evaluations by reviewers at subacute sites did not distinguish between information missing from discharge documentation and failure to receive the documentation at all from the acute care hospital (again in contrast to reviewers at acute sites, who always had access to the documentation). Lastly, reviewers at acute and subacute sites may have graded packets differently due to their different clinical perspectives. These 3 factors may explain the relatively poorer results of discharge packets reviewed by reviewers at subacute sites. Further study would be needed to distinguish among these possibilities (eg, having acute and subacute reviewers answer the same questions for the same discharge packets to allow us to measure interrater reliability between the different kinds of reviewers; explicitly asking subacute reviewers about receipt of each piece of documentation; comparing the distribution of diagnosis‐related group [DRG] codes and hospital length of stay in evaluated vs. total discharge packets as a measure of representativeness). We also cannot rule out the possibility of reviewer bias, but all reviewers were trained in a standardized fashion and we know that reliability of assessments were high, at least among reviewers at acute sites. Last, we did not measure actual or potential adverse events caused by these information deficits.

As part of a Partners‐wide initiative to improve transitions in care, the results were presented to the administrations of each of the 5 acute care hospitals. The Partners High Performance Medicine Transition team then began work with a steering committee (composed of representatives from each hospital) to address these deficiencies. Since then, the hospitals have taken several steps to improve the quality of information transfer for discharged patients, including the following:

• 1

  Technological improvements to the hospitals' discharge ordering systems to actively solicit and/or autoimport the required information into discharge documentation.

• 2

  Creation of discharge templates to record the required information on paper.

• 3

  Provision of feedback to clinicians and their service chiefs regarding the ongoing quality of their discharge documentation.

• 4

  Creation of an online Partners‐wide curriculum on discharge summary authorship, with a mandatory quiz to be taken by all incoming clinicians.

In conclusion, we found room for improvement in the inclusion of data elements required for the safe transfer of patients from acute hospitals to subacute facilities, especially in areas such as medication reconciliation, pending test results, and adequate follow‐up plans. We also found variation by site and type of service. For patients discharged to rehabilitation and other subacute facilities, improvement is needed in the communication of clinically relevant information to those providing continuing care.