Hannah J. Wong, PhD

Hospitalizations of long‐term care (LTC) residents are known to be frequent, costly, often preventable,[1, 2, 3] and potentially associated with negative health outcomes.[4] Often, an advance directive (AD) is made at LTC admission and updated annually when residents are in relatively stable health. An AD is a document that helps to inform a substitute decision maker (SDM) about the consent process for life‐sustaining treatments and is a resource that supports advance care planning (ACP). ACP is a process that allows individuals to consider, express, and plan for future healthcare in the event that they lack capacity to make their own decisions. When an LTC resident's health deteriorates and hospitalization is required, there is an opportunity to update prognosis, discuss risks and benefits of previously held treatment preferences, as well as reassess goals of care (GOC).

Engaging in ACP discussions during relatively stable health can help ensure patient preferences are followed.[5, 6] These discussions, however, are often insufficient, as they involve decision making for hypothetical situations that may not cover all potential scenarios, and may not reflect a patient's reality at the time of health status decline. Discussions held in the moment more authentically reflect the decisions of patients and/or SDM based on the specific needs and clinical realities particular to the patient at that time.[7] GOC discussions, defined in this context as ACP discussions occurring during hospitalization, have the potential to better align patient wishes with care received,[6] improve quality of life and satisfaction,[8, 9, 10] and reduce unwanted extra care.[11, 12] Although in‐the‐moment GOC discussions are recommended for all hospitalized patients who are seriously ill with a high risk of dying,[13] research suggests that this occurs infrequently for elderly patients. A recent multicenter survey of seriously ill hospitalized elderly patients found that only 25% of patients and 32% of family members reported that they had been asked about prior ACP or AD.[14] Another study of hospitalized LTC residents found that resuscitation status and family discussion was documented in only 55% and 42% of admissions, respectively.[15]

Further investigation is required to determine how often LTC patients have GOC discussions, what prompts these discussions, and what are the outcomes. Previous studies have focused on barriers to performing GOC discussions, rather than the factors that are associated with them.[16] By understanding why these discussions currently happen, we can potentially improve how often they occur and the quality of their outcomes.

The objectives of this study were to determine the rate of documented GOC discussions among hospitalized LTC residents, identify factors that were associated with documentation, and examine the association between documentation and outcomes of care.

METHODS

RESULTS

We identified a total of 7084 hospitalizations to GIM between January 1, 2012 and December 31, 2012, of which 665 (9.4%) met inclusion criteria of residence in LTC and age 65 years. Of these 665 hospitalizations, 512 were unique patients. We randomly selected a convenience sample of 200 index hospitalizations of the 512 eligible hospitalizations (39%) to perform the chart review.

DISCUSSION

Our retrospective study of LTC residents admitted to the GIM service showed that these admissions comprised 9.4% of all admissions and that GOC discussions occurred infrequently (37.5%). Our study revealed no differences in baseline patient characteristics associated with discussions, whereas patient acuity at hospital presentation independently contributed to the likelihood of discussions. We found strong associations between documentation and certain outcomes of care, including orders for AD, LOS, in‐hospital death, and 1‐year mortality. No significant associations were found between documentation and subsequent hospital use. Lastly, we found that consistent communication back to the LTC home when there was a change in AD was very poor; only 26% of discharge summaries included this documentation.

Our finding of infrequent GOC discussions during hospitalization aligns with prior studies. A study that identified code status discussions in transcripts of audio‐recorded admission encounters found that code status was discussed in only 24% of seriously ill patient admissions.[17] Furthermore, in a study specific to LTC residents, only 42% of admissions longer than 48 hours had a documented GOC discussion.[15]

We found visit‐level, but not baseline, characteristics were associated with discussions. These findings are supported by a recent study that found that whether GOC discussions took place largely depended on the acute condition presented on admission.[15] Although these results suggest that clinicians are appropriately prioritizing sicker patients who might have the most pressing need for GOC discussions, they also highlight the gap in care for less‐sick patients and the need to broaden clinical practice and consider underlying conditions and functional status. Of note, although the GCS score was found to be significantly associated with discussions, patients in the lowest GCS range did not have significantly different odds of discussions compared to the reference level (highest GCS range). A recent study by You et al. may offer some insight into this finding. They found that patients lacking capacity to make GOC decisions was ranked fifth, whereas lack of SDM availability was eighth among 21 barriers to GOC discussions, as perceived by hospital‐based clinicians.[16]

A major finding of this study was that both in‐hospital and 1‐year mortality were strongly associated with having a GOC discussion, suggesting that patients at higher risk of dying are more likely to have discussions. This is reflected by illness severity measured at initial assessment and by persistence of the association between discussions and mortality after discharge back to LTC. To the best of our knowledge, no previous studies have reported these findings. There are likely some unmeasured clinical factors such as clinical deterioration during hospitalization that contributed to this strong association. Interestingly, in our logistic regression analysis for independent predictors of in‐hospital death, we found that having dementia was associated with lower odds of in‐hospital death. One interpretation of this finding is that perhaps only patients with mild dementia were hospitalized, and those with more advanced dementia had an AD established in LTC that allowed them to remain in their LTC home. This possibility is supported by a systematic review of factors associated with LTC home hospitalization, which found that dementia was shown to be associated with less hospitalization.[18]

For patients who survived hospitalization, we did not find an association between GOC discussions and hospital use in the year following index hospitalization. In both groups, nearly 30% of patients had 1 or more subsequent hospitalizations. This is relevant especially in light of the finding that among patients where GOC discussions resulted in an AD change, only 26% of discharge summaries back to LTC included this documentation. We can only speculate that had these discussions been properly documented, subsequent hospitalizations would have decreased in the GOC group. Previous research has found that omissions of critical information in discharge summaries were common. In a study of hip fracture and stroke patients discharged from a large Midwestern academic medical center in the United States, code status was included in the discharge summary only 7% of the time.[19] The discharge summary is the primary means of sharing patient information between the hospital and LTC home. If GOC discussions are not included in the discharge summary, it is very unlikely that this information will be subsequently updated in the LTC medical record and impact the care the patient receives. A key recommendation for hospital‐based providers is ensuring that GOC discussions are clearly, consistently, and completely documented in the discharge summary so that the care provided is based on the patients' wishes.

Our study has several limitations. Our analysis was based on chart review, and although our analyses take into account a number of patient characteristics, we did not capture other characteristics that might influence GOC discussions such as culture/religion, language barriers, SDM availability, or whether patients clinically deteriorated during the index admission. Additionally, provider‐level predictors, including seniority, previous GOC training, and time available to conduct these discussions, were not captured. We also did not capture the timing or number of occasions that GOC discussions took place during hospitalization. Due to the retrospective nature of our study, we were able to only look at documented GOC discussions. GOC discussions may have happened but were never documented. However, the standard of care is to document these discussions as part of the medical record, and if they are not documented, it can be considered not to have happened and indicates a lower quality of practice. A recent survey of Canadian hospital‐based healthcare providers identified standardized GOC documentation as an effective practice to improve GOC communication.[20] Finally, because our study was conducted in 2 academic hospitals, our results may be less generalizable to other community hospitals. However, our hospitals' catchment areas capture a diverse population, both culturally and in terms of their socioeconomic status.

CONCLUSION

GOC discussions occurred infrequently, appeared to be triggered by illness severity, and were poorly communicated back to LTC. Important outcomes of care, including in‐hospital death and 1‐year mortality, were associated with discussions. This study serves to identify gaps in who might benefit from GOC discussions and illustrates opportunities for improvement including implementing standardized documentation practices.

Hospitalizations of long‐term care (LTC) residents are known to be frequent, costly, often preventable,[1, 2, 3] and potentially associated with negative health outcomes.[4] Often, an advance directive (AD) is made at LTC admission and updated annually when residents are in relatively stable health. An AD is a document that helps to inform a substitute decision maker (SDM) about the consent process for life‐sustaining treatments and is a resource that supports advance care planning (ACP). ACP is a process that allows individuals to consider, express, and plan for future healthcare in the event that they lack capacity to make their own decisions. When an LTC resident's health deteriorates and hospitalization is required, there is an opportunity to update prognosis, discuss risks and benefits of previously held treatment preferences, as well as reassess goals of care (GOC).

Engaging in ACP discussions during relatively stable health can help ensure patient preferences are followed.[5, 6] These discussions, however, are often insufficient, as they involve decision making for hypothetical situations that may not cover all potential scenarios, and may not reflect a patient's reality at the time of health status decline. Discussions held in the moment more authentically reflect the decisions of patients and/or SDM based on the specific needs and clinical realities particular to the patient at that time.[7] GOC discussions, defined in this context as ACP discussions occurring during hospitalization, have the potential to better align patient wishes with care received,[6] improve quality of life and satisfaction,[8, 9, 10] and reduce unwanted extra care.[11, 12] Although in‐the‐moment GOC discussions are recommended for all hospitalized patients who are seriously ill with a high risk of dying,[13] research suggests that this occurs infrequently for elderly patients. A recent multicenter survey of seriously ill hospitalized elderly patients found that only 25% of patients and 32% of family members reported that they had been asked about prior ACP or AD.[14] Another study of hospitalized LTC residents found that resuscitation status and family discussion was documented in only 55% and 42% of admissions, respectively.[15]

Further investigation is required to determine how often LTC patients have GOC discussions, what prompts these discussions, and what are the outcomes. Previous studies have focused on barriers to performing GOC discussions, rather than the factors that are associated with them.[16] By understanding why these discussions currently happen, we can potentially improve how often they occur and the quality of their outcomes.

The objectives of this study were to determine the rate of documented GOC discussions among hospitalized LTC residents, identify factors that were associated with documentation, and examine the association between documentation and outcomes of care.

METHODS

RESULTS

We identified a total of 7084 hospitalizations to GIM between January 1, 2012 and December 31, 2012, of which 665 (9.4%) met inclusion criteria of residence in LTC and age 65 years. Of these 665 hospitalizations, 512 were unique patients. We randomly selected a convenience sample of 200 index hospitalizations of the 512 eligible hospitalizations (39%) to perform the chart review.

DISCUSSION

Our retrospective study of LTC residents admitted to the GIM service showed that these admissions comprised 9.4% of all admissions and that GOC discussions occurred infrequently (37.5%). Our study revealed no differences in baseline patient characteristics associated with discussions, whereas patient acuity at hospital presentation independently contributed to the likelihood of discussions. We found strong associations between documentation and certain outcomes of care, including orders for AD, LOS, in‐hospital death, and 1‐year mortality. No significant associations were found between documentation and subsequent hospital use. Lastly, we found that consistent communication back to the LTC home when there was a change in AD was very poor; only 26% of discharge summaries included this documentation.

Our finding of infrequent GOC discussions during hospitalization aligns with prior studies. A study that identified code status discussions in transcripts of audio‐recorded admission encounters found that code status was discussed in only 24% of seriously ill patient admissions.[17] Furthermore, in a study specific to LTC residents, only 42% of admissions longer than 48 hours had a documented GOC discussion.[15]

We found visit‐level, but not baseline, characteristics were associated with discussions. These findings are supported by a recent study that found that whether GOC discussions took place largely depended on the acute condition presented on admission.[15] Although these results suggest that clinicians are appropriately prioritizing sicker patients who might have the most pressing need for GOC discussions, they also highlight the gap in care for less‐sick patients and the need to broaden clinical practice and consider underlying conditions and functional status. Of note, although the GCS score was found to be significantly associated with discussions, patients in the lowest GCS range did not have significantly different odds of discussions compared to the reference level (highest GCS range). A recent study by You et al. may offer some insight into this finding. They found that patients lacking capacity to make GOC decisions was ranked fifth, whereas lack of SDM availability was eighth among 21 barriers to GOC discussions, as perceived by hospital‐based clinicians.[16]

A major finding of this study was that both in‐hospital and 1‐year mortality were strongly associated with having a GOC discussion, suggesting that patients at higher risk of dying are more likely to have discussions. This is reflected by illness severity measured at initial assessment and by persistence of the association between discussions and mortality after discharge back to LTC. To the best of our knowledge, no previous studies have reported these findings. There are likely some unmeasured clinical factors such as clinical deterioration during hospitalization that contributed to this strong association. Interestingly, in our logistic regression analysis for independent predictors of in‐hospital death, we found that having dementia was associated with lower odds of in‐hospital death. One interpretation of this finding is that perhaps only patients with mild dementia were hospitalized, and those with more advanced dementia had an AD established in LTC that allowed them to remain in their LTC home. This possibility is supported by a systematic review of factors associated with LTC home hospitalization, which found that dementia was shown to be associated with less hospitalization.[18]

For patients who survived hospitalization, we did not find an association between GOC discussions and hospital use in the year following index hospitalization. In both groups, nearly 30% of patients had 1 or more subsequent hospitalizations. This is relevant especially in light of the finding that among patients where GOC discussions resulted in an AD change, only 26% of discharge summaries back to LTC included this documentation. We can only speculate that had these discussions been properly documented, subsequent hospitalizations would have decreased in the GOC group. Previous research has found that omissions of critical information in discharge summaries were common. In a study of hip fracture and stroke patients discharged from a large Midwestern academic medical center in the United States, code status was included in the discharge summary only 7% of the time.[19] The discharge summary is the primary means of sharing patient information between the hospital and LTC home. If GOC discussions are not included in the discharge summary, it is very unlikely that this information will be subsequently updated in the LTC medical record and impact the care the patient receives. A key recommendation for hospital‐based providers is ensuring that GOC discussions are clearly, consistently, and completely documented in the discharge summary so that the care provided is based on the patients' wishes.

Our study has several limitations. Our analysis was based on chart review, and although our analyses take into account a number of patient characteristics, we did not capture other characteristics that might influence GOC discussions such as culture/religion, language barriers, SDM availability, or whether patients clinically deteriorated during the index admission. Additionally, provider‐level predictors, including seniority, previous GOC training, and time available to conduct these discussions, were not captured. We also did not capture the timing or number of occasions that GOC discussions took place during hospitalization. Due to the retrospective nature of our study, we were able to only look at documented GOC discussions. GOC discussions may have happened but were never documented. However, the standard of care is to document these discussions as part of the medical record, and if they are not documented, it can be considered not to have happened and indicates a lower quality of practice. A recent survey of Canadian hospital‐based healthcare providers identified standardized GOC documentation as an effective practice to improve GOC communication.[20] Finally, because our study was conducted in 2 academic hospitals, our results may be less generalizable to other community hospitals. However, our hospitals' catchment areas capture a diverse population, both culturally and in terms of their socioeconomic status.

CONCLUSION

GOC discussions occurred infrequently, appeared to be triggered by illness severity, and were poorly communicated back to LTC. Important outcomes of care, including in‐hospital death and 1‐year mortality, were associated with discussions. This study serves to identify gaps in who might benefit from GOC discussions and illustrates opportunities for improvement including implementing standardized documentation practices.

Hospitalizations of long‐term care (LTC) residents are known to be frequent, costly, often preventable,[1, 2, 3] and potentially associated with negative health outcomes.[4] Often, an advance directive (AD) is made at LTC admission and updated annually when residents are in relatively stable health. An AD is a document that helps to inform a substitute decision maker (SDM) about the consent process for life‐sustaining treatments and is a resource that supports advance care planning (ACP). ACP is a process that allows individuals to consider, express, and plan for future healthcare in the event that they lack capacity to make their own decisions. When an LTC resident's health deteriorates and hospitalization is required, there is an opportunity to update prognosis, discuss risks and benefits of previously held treatment preferences, as well as reassess goals of care (GOC).

Engaging in ACP discussions during relatively stable health can help ensure patient preferences are followed.[5, 6] These discussions, however, are often insufficient, as they involve decision making for hypothetical situations that may not cover all potential scenarios, and may not reflect a patient's reality at the time of health status decline. Discussions held in the moment more authentically reflect the decisions of patients and/or SDM based on the specific needs and clinical realities particular to the patient at that time.[7] GOC discussions, defined in this context as ACP discussions occurring during hospitalization, have the potential to better align patient wishes with care received,[6] improve quality of life and satisfaction,[8, 9, 10] and reduce unwanted extra care.[11, 12] Although in‐the‐moment GOC discussions are recommended for all hospitalized patients who are seriously ill with a high risk of dying,[13] research suggests that this occurs infrequently for elderly patients. A recent multicenter survey of seriously ill hospitalized elderly patients found that only 25% of patients and 32% of family members reported that they had been asked about prior ACP or AD.[14] Another study of hospitalized LTC residents found that resuscitation status and family discussion was documented in only 55% and 42% of admissions, respectively.[15]

Further investigation is required to determine how often LTC patients have GOC discussions, what prompts these discussions, and what are the outcomes. Previous studies have focused on barriers to performing GOC discussions, rather than the factors that are associated with them.[16] By understanding why these discussions currently happen, we can potentially improve how often they occur and the quality of their outcomes.

The objectives of this study were to determine the rate of documented GOC discussions among hospitalized LTC residents, identify factors that were associated with documentation, and examine the association between documentation and outcomes of care.

METHODS

RESULTS

We identified a total of 7084 hospitalizations to GIM between January 1, 2012 and December 31, 2012, of which 665 (9.4%) met inclusion criteria of residence in LTC and age 65 years. Of these 665 hospitalizations, 512 were unique patients. We randomly selected a convenience sample of 200 index hospitalizations of the 512 eligible hospitalizations (39%) to perform the chart review.

DISCUSSION

Our retrospective study of LTC residents admitted to the GIM service showed that these admissions comprised 9.4% of all admissions and that GOC discussions occurred infrequently (37.5%). Our study revealed no differences in baseline patient characteristics associated with discussions, whereas patient acuity at hospital presentation independently contributed to the likelihood of discussions. We found strong associations between documentation and certain outcomes of care, including orders for AD, LOS, in‐hospital death, and 1‐year mortality. No significant associations were found between documentation and subsequent hospital use. Lastly, we found that consistent communication back to the LTC home when there was a change in AD was very poor; only 26% of discharge summaries included this documentation.

Our finding of infrequent GOC discussions during hospitalization aligns with prior studies. A study that identified code status discussions in transcripts of audio‐recorded admission encounters found that code status was discussed in only 24% of seriously ill patient admissions.[17] Furthermore, in a study specific to LTC residents, only 42% of admissions longer than 48 hours had a documented GOC discussion.[15]

We found visit‐level, but not baseline, characteristics were associated with discussions. These findings are supported by a recent study that found that whether GOC discussions took place largely depended on the acute condition presented on admission.[15] Although these results suggest that clinicians are appropriately prioritizing sicker patients who might have the most pressing need for GOC discussions, they also highlight the gap in care for less‐sick patients and the need to broaden clinical practice and consider underlying conditions and functional status. Of note, although the GCS score was found to be significantly associated with discussions, patients in the lowest GCS range did not have significantly different odds of discussions compared to the reference level (highest GCS range). A recent study by You et al. may offer some insight into this finding. They found that patients lacking capacity to make GOC decisions was ranked fifth, whereas lack of SDM availability was eighth among 21 barriers to GOC discussions, as perceived by hospital‐based clinicians.[16]

A major finding of this study was that both in‐hospital and 1‐year mortality were strongly associated with having a GOC discussion, suggesting that patients at higher risk of dying are more likely to have discussions. This is reflected by illness severity measured at initial assessment and by persistence of the association between discussions and mortality after discharge back to LTC. To the best of our knowledge, no previous studies have reported these findings. There are likely some unmeasured clinical factors such as clinical deterioration during hospitalization that contributed to this strong association. Interestingly, in our logistic regression analysis for independent predictors of in‐hospital death, we found that having dementia was associated with lower odds of in‐hospital death. One interpretation of this finding is that perhaps only patients with mild dementia were hospitalized, and those with more advanced dementia had an AD established in LTC that allowed them to remain in their LTC home. This possibility is supported by a systematic review of factors associated with LTC home hospitalization, which found that dementia was shown to be associated with less hospitalization.[18]

For patients who survived hospitalization, we did not find an association between GOC discussions and hospital use in the year following index hospitalization. In both groups, nearly 30% of patients had 1 or more subsequent hospitalizations. This is relevant especially in light of the finding that among patients where GOC discussions resulted in an AD change, only 26% of discharge summaries back to LTC included this documentation. We can only speculate that had these discussions been properly documented, subsequent hospitalizations would have decreased in the GOC group. Previous research has found that omissions of critical information in discharge summaries were common. In a study of hip fracture and stroke patients discharged from a large Midwestern academic medical center in the United States, code status was included in the discharge summary only 7% of the time.[19] The discharge summary is the primary means of sharing patient information between the hospital and LTC home. If GOC discussions are not included in the discharge summary, it is very unlikely that this information will be subsequently updated in the LTC medical record and impact the care the patient receives. A key recommendation for hospital‐based providers is ensuring that GOC discussions are clearly, consistently, and completely documented in the discharge summary so that the care provided is based on the patients' wishes.

Our study has several limitations. Our analysis was based on chart review, and although our analyses take into account a number of patient characteristics, we did not capture other characteristics that might influence GOC discussions such as culture/religion, language barriers, SDM availability, or whether patients clinically deteriorated during the index admission. Additionally, provider‐level predictors, including seniority, previous GOC training, and time available to conduct these discussions, were not captured. We also did not capture the timing or number of occasions that GOC discussions took place during hospitalization. Due to the retrospective nature of our study, we were able to only look at documented GOC discussions. GOC discussions may have happened but were never documented. However, the standard of care is to document these discussions as part of the medical record, and if they are not documented, it can be considered not to have happened and indicates a lower quality of practice. A recent survey of Canadian hospital‐based healthcare providers identified standardized GOC documentation as an effective practice to improve GOC communication.[20] Finally, because our study was conducted in 2 academic hospitals, our results may be less generalizable to other community hospitals. However, our hospitals' catchment areas capture a diverse population, both culturally and in terms of their socioeconomic status.

CONCLUSION

GOC discussions occurred infrequently, appeared to be triggered by illness severity, and were poorly communicated back to LTC. Important outcomes of care, including in‐hospital death and 1‐year mortality, were associated with discussions. This study serves to identify gaps in who might benefit from GOC discussions and illustrates opportunities for improvement including implementing standardized documentation practices.

Smooth and timely hospital patient flow can have multiple positive effects including reduced wait times for services, decreased congestion in the Emergency Department (ED), and increased patient and staff satisfaction.14 One way to improve patient flow is to remove variation along the care pathway.57

For teaching hospitals that provide team‐based care, 1 significant source of variation involves the emergent admission process.8, 9 Typically, for services that admit the majority of their patients from the ED, 1 team is assigned to all admitting duties on a particular day; the on‐call team. While teams rotate between designations of on‐call, post‐call, and pre‐call over the course of the week, only the team designated on‐call accepts new admissions. This bolus call structure creates the need for extensive cross‐coverage, large variations in team admissions, and disparate team workloads.1012 Moreover, the effects of these variations may persist and extend along the care pathway, ultimately impacting timely patient discharge. Therefore, interventions aimed at improving the admission process may be candidates for improved patient flow.

The objective of this study is to evaluate the effect of changing the admission process from a bolus admission system to a trickle system that evenly distributes newly admitted patients to each of the physician‐led care teams. We hypothesize that by removing variation within the team admission process, team workload will be smoothed and ultimately result in patients being discharged by the team in a more uniform pattern. We evaluate this hypothesis by measuring length of stay and daily discharge rate.

METHODS

Setting

This retrospective study was conducted on the General Internal Medicine clinical teaching unit (GIM CTU) at a large academic tertiary care center in Toronto, Canada. GIM provides acute, nonsurgical care to a patient population composed primarily of elderly patients with complex chronic illnesses. GIM receives 98% of its inpatient admissions from the ED. On a daily basis, the ED sees approximately 100 patients, of which nearly 20% are admitted to hospital. GIM constitutes the single largest admitting service in the ED, admitting nearly half of all emergent admissions. Surgical and specialized medical services (eg, Cardiology, Oncology, Nephrology) admit the remaining half.

On March 2, 2009, the GIM CTU underwent a structural change from a bolus admission system to a trickle system of admissions to each care team. Figure 1 depicts a typical pre‐change admission pattern where each of the 4 care teams would admit a bolus of patients on a given day (left panel), and a typical post‐change admission pattern where the variation in daily admissions is smoothed out as a result of the trickle admission system (right panel). No change was made to care team members; each team consisted of an attending physician, 1 senior resident, 2 to 3 junior residents, 1 social worker, 1 physiotherapist, 1 occupational therapist, and 1 pharmacist. The Appendix provides a detailed description of the structural change.

Figure 1

RESULTS

During the 2 study periods, a total of 2734 patients were discharged, 1446 in the pre‐change period (1535 admitted), and 1288 in the post‐change period (1363 admitted). Table 1 presents mean age and primary CMG diagnosis.

Figure 2 shows the estimated average team‐specific DDR's according to call schedule status, along with 95% credible intervals. With the exception of the postpost‐call day, each black point (2009, post‐change period) is closer to the overall average DDR of 9.9% than each corresponding gray point (2008, pre‐change period). In our Bayesian model, there was a 96.9% probability that the variability across call schedule status was reduced in the post‐change period, substantial evidence of smoother discharge rates across different types of call days.

Figure 2

Summary statistics for the LOS for both groups can be seen in Table 2. The median LOS in the post‐change period was statistically significantly shorter than in the pre‐change period (4.8 days vs 5.1 days, P < 0.001).

Figure 3 shows the estimated KaplanMeier curves of time to discharge (LOS) in both time periods. Differences between the 2 study periods in the proportion of patients that had been discharged at each time point (the vertical distance between the curves) can be observed, particularly in the shorter LOS times.

Figure 3

DISCUSSION

Previous studies have suggested that systems become more efficient when every day runs the same way.15 Achieving this for the number of daily discharges from the ward should have a positive effect on the flow of patients through the GIM service.16 Wong et al. showed how the on call schedule of medical personnel had a strong effect on the variation in daily discharges.17 A more recent study by the same authors demonstrated, through a computer simulation model, that smoothing patient discharges over the course of the week decreases the number of ED beds occupied by admitted patients.18 After introducing a structural change to our admission system that made the daily admissions of patients to each care team uniform, we showed a significant reduction in the variation of discharge rates from day to day, and the expected improvement in patient flow as shown by a decrease in the median LOS.

This intervention changed only 1 component of a complex patient care process, of which the resident on‐call schedule is only a small part. Nevertheless, this small change, designed to optimize the doctors' contribution to patient flow, was sufficient in effecting a significant reduction in the variation of the DDR. Inpatients follow a usual course in the hospital, requiring an average LOS of 4 to 5 days. In the bolus system of admissions, we observed what was essentially a cohort effect where the same bolus of patients was discharged on roughly the same day, an average of 4 to 5 days after admission. If the daily variation in discharges were only dependent on the daily variation in admissions, by making the influx of inpatients constant, we should have eliminated this cohort effect. Although the variation in discharges was reduced, it was not completely eliminated, suggesting that elements of the old system are retained. It is possible that the senior resident's management of the patients on the team has a stronger influence than that of other members of the team, and the flow of patients may still be affected by their call schedule.

We observed a significant reduction (0.3 days) in median LOS. By making each day look the same for admissions to each care team, and by making each day look more uniform for discharges from each care team, we were able to improve our unit's operational efficiency. Other benefits of the new system included: less cross‐coverage, since after‐hours there was always a member of each team to look after their own patients; the elimination of the post‐call day for the entire team; and the relatively decreased average daily workload.

The bulk of the reduction in median LOS was attributed to short‐stay patients. The flow of very sick patients who require prolonged inpatient treatment, or those waiting for post‐acute care beds (rehabilitation, long‐term care, convalescence, etc) may be less sensitive to improvements in internal efficiencies.

Although the improvement in LOS was modest, it was certainly no worse than in the older system, and the change was accompanied by the many other benefits already mentioned. In fact, ours is not the only hospital in the city that has made this change. Early results of a qualitative study exploring the perceptions of attending staff, residents, and students of the new systemparticularly its effects on the educational experienceare encouraging, showing overall positive opinions about the change. Further studies aimed at analyzing the barriers to efficient patient discharges may help identify important factors, such as those already mentioned, that this change in structure did not address. Policymakers could address other components of the discharge process, particularly the chronic shortage of post‐acute care beds. Finally, an economic analysis could provide insights about the potential savings that such structural changes could represent.

This study has several limitations. It took place in a single teaching hospital in Canada and, therefore, may not be generalizable to community hospitals or to settings that do not provide single‐payer free public healthcare. Nevertheless, most hospital units are subject to the effects of medical personnel scheduling, and the variation in patient flow processes that this produces. The current resident association collective agreement in Ontario still allows trainees to be scheduled for continuous 24‐hour duty periods. An exact replication of our structure would not be possible in settings with more stringent duty‐hour restrictions. Nevertheless, the goal of the structural change was to make the influx of patients to each care team constant, and this is achievable regardless of the length of the trainee call period. Although there is no reason to suspect a systematic difference in the mix of patients from 2008 to 2009, it would have been preferable to use a propensity score to compare clinical characteristics of the 2 patient groups. We used a relatively new metric, DDR, which was created in our institution and already has been used in several studies. However, it has not yet been validated in other centers.

One of the limitations of a before‐and‐after analysis is our inability to adjust for other changes that may have occurred during the study periods. These known and unknown factors may have had effects on the findings.

CONCLUSIONS

A new admission structure was introduced to the GIM CTU in March 2009, with the intention of changing the admissions to each care team from a bolus to a trickle system. This study was a real‐world demonstration of a concept that had, until this point, only been observed in robust simulation models. When the daily influx of patients to a care team becomes constant, the number of discharges from that team experience less daily variation, and the overall efficiency of the team improves, as measured by a reduction in the median LOS. Standardizing the care processes on the GIM inpatient ward improves overall efficiency and capacity.

Smooth and timely hospital patient flow can have multiple positive effects including reduced wait times for services, decreased congestion in the Emergency Department (ED), and increased patient and staff satisfaction.14 One way to improve patient flow is to remove variation along the care pathway.57

For teaching hospitals that provide team‐based care, 1 significant source of variation involves the emergent admission process.8, 9 Typically, for services that admit the majority of their patients from the ED, 1 team is assigned to all admitting duties on a particular day; the on‐call team. While teams rotate between designations of on‐call, post‐call, and pre‐call over the course of the week, only the team designated on‐call accepts new admissions. This bolus call structure creates the need for extensive cross‐coverage, large variations in team admissions, and disparate team workloads.1012 Moreover, the effects of these variations may persist and extend along the care pathway, ultimately impacting timely patient discharge. Therefore, interventions aimed at improving the admission process may be candidates for improved patient flow.

The objective of this study is to evaluate the effect of changing the admission process from a bolus admission system to a trickle system that evenly distributes newly admitted patients to each of the physician‐led care teams. We hypothesize that by removing variation within the team admission process, team workload will be smoothed and ultimately result in patients being discharged by the team in a more uniform pattern. We evaluate this hypothesis by measuring length of stay and daily discharge rate.

METHODS

Setting

This retrospective study was conducted on the General Internal Medicine clinical teaching unit (GIM CTU) at a large academic tertiary care center in Toronto, Canada. GIM provides acute, nonsurgical care to a patient population composed primarily of elderly patients with complex chronic illnesses. GIM receives 98% of its inpatient admissions from the ED. On a daily basis, the ED sees approximately 100 patients, of which nearly 20% are admitted to hospital. GIM constitutes the single largest admitting service in the ED, admitting nearly half of all emergent admissions. Surgical and specialized medical services (eg, Cardiology, Oncology, Nephrology) admit the remaining half.

On March 2, 2009, the GIM CTU underwent a structural change from a bolus admission system to a trickle system of admissions to each care team. Figure 1 depicts a typical pre‐change admission pattern where each of the 4 care teams would admit a bolus of patients on a given day (left panel), and a typical post‐change admission pattern where the variation in daily admissions is smoothed out as a result of the trickle admission system (right panel). No change was made to care team members; each team consisted of an attending physician, 1 senior resident, 2 to 3 junior residents, 1 social worker, 1 physiotherapist, 1 occupational therapist, and 1 pharmacist. The Appendix provides a detailed description of the structural change.

Figure 1

RESULTS

During the 2 study periods, a total of 2734 patients were discharged, 1446 in the pre‐change period (1535 admitted), and 1288 in the post‐change period (1363 admitted). Table 1 presents mean age and primary CMG diagnosis.

Figure 2 shows the estimated average team‐specific DDR's according to call schedule status, along with 95% credible intervals. With the exception of the postpost‐call day, each black point (2009, post‐change period) is closer to the overall average DDR of 9.9% than each corresponding gray point (2008, pre‐change period). In our Bayesian model, there was a 96.9% probability that the variability across call schedule status was reduced in the post‐change period, substantial evidence of smoother discharge rates across different types of call days.

Figure 2

Summary statistics for the LOS for both groups can be seen in Table 2. The median LOS in the post‐change period was statistically significantly shorter than in the pre‐change period (4.8 days vs 5.1 days, P < 0.001).

Figure 3 shows the estimated KaplanMeier curves of time to discharge (LOS) in both time periods. Differences between the 2 study periods in the proportion of patients that had been discharged at each time point (the vertical distance between the curves) can be observed, particularly in the shorter LOS times.

Figure 3

DISCUSSION

Previous studies have suggested that systems become more efficient when every day runs the same way.15 Achieving this for the number of daily discharges from the ward should have a positive effect on the flow of patients through the GIM service.16 Wong et al. showed how the on call schedule of medical personnel had a strong effect on the variation in daily discharges.17 A more recent study by the same authors demonstrated, through a computer simulation model, that smoothing patient discharges over the course of the week decreases the number of ED beds occupied by admitted patients.18 After introducing a structural change to our admission system that made the daily admissions of patients to each care team uniform, we showed a significant reduction in the variation of discharge rates from day to day, and the expected improvement in patient flow as shown by a decrease in the median LOS.

This intervention changed only 1 component of a complex patient care process, of which the resident on‐call schedule is only a small part. Nevertheless, this small change, designed to optimize the doctors' contribution to patient flow, was sufficient in effecting a significant reduction in the variation of the DDR. Inpatients follow a usual course in the hospital, requiring an average LOS of 4 to 5 days. In the bolus system of admissions, we observed what was essentially a cohort effect where the same bolus of patients was discharged on roughly the same day, an average of 4 to 5 days after admission. If the daily variation in discharges were only dependent on the daily variation in admissions, by making the influx of inpatients constant, we should have eliminated this cohort effect. Although the variation in discharges was reduced, it was not completely eliminated, suggesting that elements of the old system are retained. It is possible that the senior resident's management of the patients on the team has a stronger influence than that of other members of the team, and the flow of patients may still be affected by their call schedule.

We observed a significant reduction (0.3 days) in median LOS. By making each day look the same for admissions to each care team, and by making each day look more uniform for discharges from each care team, we were able to improve our unit's operational efficiency. Other benefits of the new system included: less cross‐coverage, since after‐hours there was always a member of each team to look after their own patients; the elimination of the post‐call day for the entire team; and the relatively decreased average daily workload.

The bulk of the reduction in median LOS was attributed to short‐stay patients. The flow of very sick patients who require prolonged inpatient treatment, or those waiting for post‐acute care beds (rehabilitation, long‐term care, convalescence, etc) may be less sensitive to improvements in internal efficiencies.

Although the improvement in LOS was modest, it was certainly no worse than in the older system, and the change was accompanied by the many other benefits already mentioned. In fact, ours is not the only hospital in the city that has made this change. Early results of a qualitative study exploring the perceptions of attending staff, residents, and students of the new systemparticularly its effects on the educational experienceare encouraging, showing overall positive opinions about the change. Further studies aimed at analyzing the barriers to efficient patient discharges may help identify important factors, such as those already mentioned, that this change in structure did not address. Policymakers could address other components of the discharge process, particularly the chronic shortage of post‐acute care beds. Finally, an economic analysis could provide insights about the potential savings that such structural changes could represent.

This study has several limitations. It took place in a single teaching hospital in Canada and, therefore, may not be generalizable to community hospitals or to settings that do not provide single‐payer free public healthcare. Nevertheless, most hospital units are subject to the effects of medical personnel scheduling, and the variation in patient flow processes that this produces. The current resident association collective agreement in Ontario still allows trainees to be scheduled for continuous 24‐hour duty periods. An exact replication of our structure would not be possible in settings with more stringent duty‐hour restrictions. Nevertheless, the goal of the structural change was to make the influx of patients to each care team constant, and this is achievable regardless of the length of the trainee call period. Although there is no reason to suspect a systematic difference in the mix of patients from 2008 to 2009, it would have been preferable to use a propensity score to compare clinical characteristics of the 2 patient groups. We used a relatively new metric, DDR, which was created in our institution and already has been used in several studies. However, it has not yet been validated in other centers.

One of the limitations of a before‐and‐after analysis is our inability to adjust for other changes that may have occurred during the study periods. These known and unknown factors may have had effects on the findings.

CONCLUSIONS

A new admission structure was introduced to the GIM CTU in March 2009, with the intention of changing the admissions to each care team from a bolus to a trickle system. This study was a real‐world demonstration of a concept that had, until this point, only been observed in robust simulation models. When the daily influx of patients to a care team becomes constant, the number of discharges from that team experience less daily variation, and the overall efficiency of the team improves, as measured by a reduction in the median LOS. Standardizing the care processes on the GIM inpatient ward improves overall efficiency and capacity.

Smooth and timely hospital patient flow can have multiple positive effects including reduced wait times for services, decreased congestion in the Emergency Department (ED), and increased patient and staff satisfaction.14 One way to improve patient flow is to remove variation along the care pathway.57

For teaching hospitals that provide team‐based care, 1 significant source of variation involves the emergent admission process.8, 9 Typically, for services that admit the majority of their patients from the ED, 1 team is assigned to all admitting duties on a particular day; the on‐call team. While teams rotate between designations of on‐call, post‐call, and pre‐call over the course of the week, only the team designated on‐call accepts new admissions. This bolus call structure creates the need for extensive cross‐coverage, large variations in team admissions, and disparate team workloads.1012 Moreover, the effects of these variations may persist and extend along the care pathway, ultimately impacting timely patient discharge. Therefore, interventions aimed at improving the admission process may be candidates for improved patient flow.

The objective of this study is to evaluate the effect of changing the admission process from a bolus admission system to a trickle system that evenly distributes newly admitted patients to each of the physician‐led care teams. We hypothesize that by removing variation within the team admission process, team workload will be smoothed and ultimately result in patients being discharged by the team in a more uniform pattern. We evaluate this hypothesis by measuring length of stay and daily discharge rate.

METHODS

Setting

This retrospective study was conducted on the General Internal Medicine clinical teaching unit (GIM CTU) at a large academic tertiary care center in Toronto, Canada. GIM provides acute, nonsurgical care to a patient population composed primarily of elderly patients with complex chronic illnesses. GIM receives 98% of its inpatient admissions from the ED. On a daily basis, the ED sees approximately 100 patients, of which nearly 20% are admitted to hospital. GIM constitutes the single largest admitting service in the ED, admitting nearly half of all emergent admissions. Surgical and specialized medical services (eg, Cardiology, Oncology, Nephrology) admit the remaining half.

On March 2, 2009, the GIM CTU underwent a structural change from a bolus admission system to a trickle system of admissions to each care team. Figure 1 depicts a typical pre‐change admission pattern where each of the 4 care teams would admit a bolus of patients on a given day (left panel), and a typical post‐change admission pattern where the variation in daily admissions is smoothed out as a result of the trickle admission system (right panel). No change was made to care team members; each team consisted of an attending physician, 1 senior resident, 2 to 3 junior residents, 1 social worker, 1 physiotherapist, 1 occupational therapist, and 1 pharmacist. The Appendix provides a detailed description of the structural change.

Figure 1

RESULTS

During the 2 study periods, a total of 2734 patients were discharged, 1446 in the pre‐change period (1535 admitted), and 1288 in the post‐change period (1363 admitted). Table 1 presents mean age and primary CMG diagnosis.

Figure 2 shows the estimated average team‐specific DDR's according to call schedule status, along with 95% credible intervals. With the exception of the postpost‐call day, each black point (2009, post‐change period) is closer to the overall average DDR of 9.9% than each corresponding gray point (2008, pre‐change period). In our Bayesian model, there was a 96.9% probability that the variability across call schedule status was reduced in the post‐change period, substantial evidence of smoother discharge rates across different types of call days.

Figure 2

Summary statistics for the LOS for both groups can be seen in Table 2. The median LOS in the post‐change period was statistically significantly shorter than in the pre‐change period (4.8 days vs 5.1 days, P < 0.001).

Figure 3 shows the estimated KaplanMeier curves of time to discharge (LOS) in both time periods. Differences between the 2 study periods in the proportion of patients that had been discharged at each time point (the vertical distance between the curves) can be observed, particularly in the shorter LOS times.

Figure 3

DISCUSSION

Previous studies have suggested that systems become more efficient when every day runs the same way.15 Achieving this for the number of daily discharges from the ward should have a positive effect on the flow of patients through the GIM service.16 Wong et al. showed how the on call schedule of medical personnel had a strong effect on the variation in daily discharges.17 A more recent study by the same authors demonstrated, through a computer simulation model, that smoothing patient discharges over the course of the week decreases the number of ED beds occupied by admitted patients.18 After introducing a structural change to our admission system that made the daily admissions of patients to each care team uniform, we showed a significant reduction in the variation of discharge rates from day to day, and the expected improvement in patient flow as shown by a decrease in the median LOS.

This intervention changed only 1 component of a complex patient care process, of which the resident on‐call schedule is only a small part. Nevertheless, this small change, designed to optimize the doctors' contribution to patient flow, was sufficient in effecting a significant reduction in the variation of the DDR. Inpatients follow a usual course in the hospital, requiring an average LOS of 4 to 5 days. In the bolus system of admissions, we observed what was essentially a cohort effect where the same bolus of patients was discharged on roughly the same day, an average of 4 to 5 days after admission. If the daily variation in discharges were only dependent on the daily variation in admissions, by making the influx of inpatients constant, we should have eliminated this cohort effect. Although the variation in discharges was reduced, it was not completely eliminated, suggesting that elements of the old system are retained. It is possible that the senior resident's management of the patients on the team has a stronger influence than that of other members of the team, and the flow of patients may still be affected by their call schedule.

We observed a significant reduction (0.3 days) in median LOS. By making each day look the same for admissions to each care team, and by making each day look more uniform for discharges from each care team, we were able to improve our unit's operational efficiency. Other benefits of the new system included: less cross‐coverage, since after‐hours there was always a member of each team to look after their own patients; the elimination of the post‐call day for the entire team; and the relatively decreased average daily workload.

The bulk of the reduction in median LOS was attributed to short‐stay patients. The flow of very sick patients who require prolonged inpatient treatment, or those waiting for post‐acute care beds (rehabilitation, long‐term care, convalescence, etc) may be less sensitive to improvements in internal efficiencies.

Although the improvement in LOS was modest, it was certainly no worse than in the older system, and the change was accompanied by the many other benefits already mentioned. In fact, ours is not the only hospital in the city that has made this change. Early results of a qualitative study exploring the perceptions of attending staff, residents, and students of the new systemparticularly its effects on the educational experienceare encouraging, showing overall positive opinions about the change. Further studies aimed at analyzing the barriers to efficient patient discharges may help identify important factors, such as those already mentioned, that this change in structure did not address. Policymakers could address other components of the discharge process, particularly the chronic shortage of post‐acute care beds. Finally, an economic analysis could provide insights about the potential savings that such structural changes could represent.

This study has several limitations. It took place in a single teaching hospital in Canada and, therefore, may not be generalizable to community hospitals or to settings that do not provide single‐payer free public healthcare. Nevertheless, most hospital units are subject to the effects of medical personnel scheduling, and the variation in patient flow processes that this produces. The current resident association collective agreement in Ontario still allows trainees to be scheduled for continuous 24‐hour duty periods. An exact replication of our structure would not be possible in settings with more stringent duty‐hour restrictions. Nevertheless, the goal of the structural change was to make the influx of patients to each care team constant, and this is achievable regardless of the length of the trainee call period. Although there is no reason to suspect a systematic difference in the mix of patients from 2008 to 2009, it would have been preferable to use a propensity score to compare clinical characteristics of the 2 patient groups. We used a relatively new metric, DDR, which was created in our institution and already has been used in several studies. However, it has not yet been validated in other centers.

One of the limitations of a before‐and‐after analysis is our inability to adjust for other changes that may have occurred during the study periods. These known and unknown factors may have had effects on the findings.

CONCLUSIONS

A new admission structure was introduced to the GIM CTU in March 2009, with the intention of changing the admissions to each care team from a bolus to a trickle system. This study was a real‐world demonstration of a concept that had, until this point, only been observed in robust simulation models. When the daily influx of patients to a care team becomes constant, the number of discharges from that team experience less daily variation, and the overall efficiency of the team improves, as measured by a reduction in the median LOS. Standardizing the care processes on the GIM inpatient ward improves overall efficiency and capacity.