Abstract
Background and purpose:
In Ontario (Canada’s most populous province), it has been suggested that mildly impaired stroke patients are being admitted to inpatient rehabilitation unnecessarily due to a lack of alternative options in the community. This ecological study aimed to formally test this hypothesis.
Methods:
Patients admitted to an inpatient rehabilitation bed in Ontario’s most highly functioning patient classification group (Rehabilitation Patient Group 1160) were retrospectively identified as potentially avoidable admissions, and the proportion of such patients was calculated for each Local Health Integration Network every year between 2006/2007 and 2010/2011. Five indicators of community-based rehabilitation availability were used to test the relationships between avoidable admissions and resource indicators.
Results:
Of the 25 correlations tested, 21 agreed with the hypothesized direction of effect and 4 reached statistical significance. Logistic-linear regressions on combined data from each of the 5 years demonstrated statistically significant associations between all 5 resource indicators and the proportion of potentially avoidable admissions.
Conclusions:
This study confirms the suggestion of variation in the proportion of mildly impaired patients admitted to inpatient rehabilitation across Ontario’s Local Health Integration Networks. It also adds evidence to support the concern that a lack of community-based rehabilitation is contributing to these potentially avoidable admissions.
Introduction
Approximately 40% of strokes and transient ischemic attacks (TIAs) in Canada occur in the province of Ontario 1 leading to roughly 20 000 patients arriving at emergency departments across the province annually. 2 Post-stroke rehabilitation in Ontario is provided in 3 settings: in hospital as an inpatient, in hospital as an outpatient, or in the community (usually in the patient’s residence). Ideally, care provided in each setting should align with best-practice recommendations including interdisciplinary team care and regular team communication. However, programs vary across the province in the services they provide, how well the teams work together, and the goals of care in each setting (which are generally left to the discretion of the care providers and patients). No formal criteria for admission to one setting over another have been established across the province and there is no routine evaluation of team performance.
In 2010, an Ontario Stroke Evaluation Report noted that the proportion of mild stroke patients being admitted to inpatient rehabilitation was increasing and suggested that this indicator be monitored in the future. 2 The report also noted wide variation in inpatient rehabilitation admission practices by region and suggested that this may be the result of discrepancies in resource availability (including therapy staff availability) across programs and regions. These suggestions are troubling given that for moderate-to-mildly impaired patients, interdisciplinary post-stroke rehabilitation at home has been demonstrated to improve recovery of functional independence at less cost compared with rehabilitation of similar patients in hospital.3,4
In Canada, the most commonly used measure of function after stroke is the Functional Independence Measure (FIM), and a score of 100 has been suggested as a clinically meaningful cut-point for discharge home from inpatient rehabilitation. 5 International studies have reported mean FIM scores well below 100 on discharge from inpatient rehabilitation (eg. 84.8 6 in the USA and 87.3 in Singapore). 7 However, the total FIM score does not distinguish between motor and cognitive impairment. Patients can achieve a high total FIM score in spite of significant cognitive impairment, which may affect their ability to return home safely and, therefore, justify a hospital admission.
In 2006, Ontario’s Ministry of Health and Long-Term Care endorsed development of the Rehabilitation Patient Group case-mix classification system, in part to account for the impact of cognitive function. 8 The Rehabilitation Patient Group system uses FIM sub-scores (motor and cognitive) separately along with age to stratify rehabilitation patients based on their anticipated length of stay in rehabilitation. Of the 7 Rehabilitation Patient Group categories developed for stroke, the group predicted to require the shortest length of stay is referred to as Rehabilitation Patient Group 1160. Patients admitted to inpatient rehabilitation in Rehabilitation Patient Group 1160 have an admission FIM >100 and a cognitive FIM score of 30 to 35 (indicating mild or no cognitive impairment). 8
In 2012, a group of stroke experts from across the province endorsed a recommendation that for the purpose of system evaluation, patients in Rehabilitation Patient Group 1160 are suitable candidates for outpatient or community-based rehabilitation and should not be admitted to inpatient rehabilitation. 9 In 2013, Health Quality Ontario also endorsed this recommendation in their Clinical Handbook for Stroke. 10 In both instances, these groups re-affirmed the notion that patients in Rehabilitation Patient Group 1160 are likely being admitted to rehabilitation because of a lack of community services. However, they also cautioned that avoiding these unnecessary admissions may not be possible until more community-based rehabilitation resources are available and that the relationship between community rehabilitation resources and mild stroke admissions needs further research.9,10
The objective of this study was to test the hypothesized association between the proportion of “potentially avoidable” mild admissions to inpatient rehabilitation (Rehabilitation Patient Group 1160) and the availability of in-home rehabilitation in Ontario, Canada. Our first hypothesis was that the proportion of potentially avoidable admissions to inpatient rehabilitation varies significantly across regions of Ontario. If confirmed, our second hypothesis was that a significant proportion of this variation could be explained by the availability of post-stroke, in-home rehabilitation.
Methods
In 2004, Ontario began regionalization of health care by dividing the province into 14 Local Health Integration Networks (LHIN), which officially assumed responsibility for regional planning, coordination, and funding of health care services in 2007 (Figure 1). 11 Each Local Health Integration Network has a well-defined boundary and each has its own home care organization (Community-Care Access Centre [CCAC]) that coordinates in-home rehabilitation services. These LHINs formed the unit of analysis in this study with the proportion of patients with stroke admitted to inpatient rehabilitation in Rehabilitation Patient Group 1160 as the primary variable of interest.
Ontario, Canada’s Local Health Integration Networks.
All data used in this study were aggregated and no personal health information was seen by our research team at any time. Therefore, Research Ethics Board approval was deemed unnecessary by the Health Sciences Research Ethics Board at Western University London, Ontario. All data used in this study were compiled by the Institute for Clinical Evaluative Sciences (ICES) in Toronto, Ontario in accordance with their ethics protocols and privacy standards. Most of the data used here are publicly available 12 with the exception of data used to calculate the proportion of “potentially avoidable” admissions (Rehabilitation Patient Group 1160) by Local Health Integration Network. Potentially avoidable admissions were calculated by an analyst at ICES using data from the National Rehabilitation Reporting System (NRS), and aggregate information was provided for this study.
The NRS contains information on patients admitted to registered inpatient rehabilitation beds across Canada, and reporting is mandatory in Ontario. NRS records include demographic, clinical, and procedural information such as age, sex, birth date, FIM score, and discharge destination. 13 Patients with primary diagnosis of stroke admitted to inpatient rehabilitation in fiscal years 2006/2007 to 2010/2011 were identified and retrospectively assigned to a Rehabilitation Patient Group. Patients in Rehabilitation Patient Group 1160 were labeled as potentially avoidable admissions, and the proportion of potentially avoidable admissions, relative to all admissions, was calculated for each Local Health Integration Network in each year.
Explanatory variables were compiled from publicly available ICES reports 12 and used to derive 5 indicators of regional in-home rehabilitation availability. Indicators of in-home rehabilitation availability were designed to represent 2 constructs: access and provision. Access indicators included (1) the proportion of stroke survivors in each region who received in-home rehabilitation services and (2) the mean number of days between acute discharge and first in-home visit (wait time). Provision indicators included the mean number of visits per client for each of physiotherapy (PT), occupational therapy (OT), and speech language pathology (SLP).
Variation across LHINs in the proportion of potentially avoidable admissions in each of the 5 years was assessed using a χ2 test. Variation in each of the resource indicators across the 5 years was tested using a Kruskal-Wallis test. Correlations between the proportion of potentially avoidable admissions and each of the 5 indicators of resource availability were estimated for each year separately using Spearman’s rho. It was hypothesized that 4 of the 5 resource indicators (all except wait times) would be negatively correlated with potentially avoidable admissions (ie, fewer in-home clients, fewer therapy visits per client, and longer wait times would each be associated with more avoidable admissions). Significance in the number of tests whose direction of correlation agreed with the hypothesized direction of effect was tested using a Sign Test.
For each of the 5 variables, data from all 5 years were then entered into a logistic-linear model with the proportion of potentially avoidable admissions as the dependent variable and year, indicator, and the interaction term (year × indicator) as independent variables. The interaction term was removed if not found to be statistically significant, but year was left in all final models. All analyses were performed using SPSS version 21.0.
Results
Between 2006 and 2010, an average of 253 patients per year (7% of all inpatient rehabilitation admissions) were retrospectively identified as potentially avoidable admissions (Table 1). The proportion of potentially avoidable admissions per Local Health Integration Network in a given year ranged from a low of 1.6% in Local Health Integration Network 12 in 2007 to a high of 17.9% in Local Health Integration Network 14 in 2007. Regional comparison of the proportion of potentially avoidable admissions demonstrated significant variation for every year. Variation in 4 of the 5 resource indicators was also noted across the 5-year period (all P < .001, Table 2) with the exception of in-home rehabilitation clients per acute discharge. The mean number of visits per client by all 3 therapy disciplines (PT, OT, and SLP) generally increased over time, while days from acute discharge to in-home rehabilitation decreased in the last 3 years compared with the first 2 years.
The proportion of “potentially avoidable” admissions (Rehabilitation Patient Group 1160) to inpatient rehabilitation across Ontario’s Local Health Integration Networks between 2006/2007 and 2010/2011.
Abbreviations: LHIN, Local Health Integration Networks; RPG, Rehabilitation Patient Group.
P values derived from Pearson chi-squared test.
Summary data on in-home rehabilitation indicators across Ontario’s Local Health Integration Networks between 2006/2007 and 2010/2011.
Abbreviations: OT, occupational therapy; PT, physiotherapy; SLP, speech language pathology.
P values derived from Kruskal-Wallis test.
Spearman’s rho correlation coefficients for the proportion of potentially avoidable admissions to inpatient rehabilitation and each of the resource indicators are presented in Table 3. While only 4 of the 25 correlations reached statistical significance individually, 21 (84%) demonstrated an association that agreed with the hypothesized direction of effect (P value = .001). Wait times (days to service) demonstrated the weakest association with potentially avoidable admissions, as only 3 of the 5 correlations aligned with the hypothesized direction of effect. Provision indicators generally demonstrated stronger correlation with potentially avoidable admissions than access indicators.
Spearman rho correlations between resource indicators and the proportion of potentially avoidable admissions to inpatient rehabilitation (Rehabilitation Patient Group 1160) across Ontario’s Local Health Integration Networks for fiscal years 2006-2010.
Abbreviations: OT, occupational therapy; PT, physiotherapy; SLP, speech language pathology.
Significant at P value <.05.
Logistic-linear regressions of the frequency of potentially avoidable admissions to inpatient rehabilitation on resource availability by Local Health Integration Network were performed for each variable separately. The interaction term (year × indicator) was statistically significant in the models of OT and SLP visits per client and was removed from the other models. All coefficients agreed with the hypothesized direction of effect, and statistically significant correlations at P value <.05 were noted for each variable (Table 4).
Regressions of the frequency of potentially avoidable admission to inpatient rehabilitation on Local Health Integration Network resource variables for fiscal years 2006-2010 combined, adjusting for year.
Abbreviations: OT, occupational therapy; PT, physiotherapy; SLP, speech language pathology.
Adjusted for interaction term (year × indicator).
Discussion
In Ontario, it has been suggested that a lack of community-based rehabilitation services may contribute to patients being admitted to inpatient rehabilitation unnecessarily after stroke. 2 This ecological study was designed to formally test this hypothesis. To do so, Rehabilitation Patient Group 1160 was used to approximate the proportion of “potentially avoidable” admissions to inpatient rehabilitation across Ontario’s LHINs, which was then compared with 5 LHIN-level in-home rehabilitation resource indicators. Correlations between these resource indicators and the proportion of potentially avoidable admissions agreed with the hypothesized direction of effect in 21 out of 25 tests (84%, P value = .001), and significant associations were noted 4 times. Furthermore, estimates from logistic-linear regressions on 5-year data were statistically significant for all resource indicator variables. In combination, these results support the hypothesis that at the Local Health Integration Network level, a lack of in-home rehabilitation resources and long wait times for access to services are associated with higher rates of potentially avoidable admissions of mildly impaired patients to inpatient rehabilitation.
Admitting mildly impaired patients to inpatient rehabilitation is concerning on several fronts. At the patient level, evidence suggests that moderately to mildly impaired patients achieve better outcomes at home 14 and that home is their preferred setting for rehabilitation. 15 At the system level, it has been demonstrated that rehabilitation at home can be provided at lower cost to the health care system than in-hospital 16 and that reducing the number of admissions of high-functioning patients to inpatient rehabilitation beds may avail these services for more severely impaired patients. 17 One of the most pressing challenges in Ontario’s rehabilitation system is a lack of consistency in rehabilitation program funding, which has resulted in dramatic variation across the province in team composition, program goals, and referral patterns. This, along with no provincial oversight of program performance, has contributed to much of the variation between programs seen here.
International research suggests that for appropriate high-functioning patients, community-based rehabilitation is an effective method of meeting their rehabilitation needs. A meta-analysis performed by Cochrane’s Early Supported Discharge (ESD) Trialists noted that patients participating in ESD programs after stroke demonstrated decreased odds of death or institutionalization and were more likely to be living at home, independent in daily activities, and satisfied with their outpatient care than were similar controls. 3 In the only published Canadian study of ESD, high-functioning patients admitted to a 4-week home rehabilitation program demonstrated significantly greater improvements in physical function (Stroke Impact Scale), health-related quality of life (36-Item Short Form Health Survey [SF-36]), and independent activities of daily living (Older Americans Resource Scale for Independent Activities of Daily Living) compared with patients receiving usual care (including inpatient rehabilitation). 18 Unfortunately, the long wait lists for community-based rehabilitation seen in the Ontario setting mean that patients are not experiencing these benefits within Ontario’s current stroke system.
A 2006 health technology assessment concluded that ESD was a “dominant health intervention” because it improved patient outcomes at lower cost compared with usual stroke unit care. 4 Similarly, in a follow-up economic analysis to the only Canadian study of ESD, the home-based rehabilitation program was demonstrated to cost an average of $3281 CAD less per patient in the first 3 months after stroke compared with usual care. 16 Not surprisingly, the cost reductions in these 2 reports came largely from reductions in hospital length of stay that was evident in both the acute and rehabilitation settings.
In addition to direct benefits for high-functioning patients and the potential for cost savings, appropriately resourced community-based rehabilitation also holds tremendous potential for improving system-wide efficiency. A 2012 study in Southwestern Ontario identified 37% of patients being discharged from acute care hospitals as candidates for inpatient rehabilitation; yet only 75% of these were actually admitted. 17 The most frequently cited reason for candidates not being admitted to inpatient rehabilitation was the lack of an available bed. In 2010/2011, 246 patients in Rehabilitation Patient Group 1160 were admitted to inpatient rehabilitation across Ontario occupying 3715 rehabilitation bed days (10 bed equivalents). 12 If any or all of these admissions were avoided, the opportunity to improve rehabilitation access for more severely impaired patients would have been substantial.
The resource indicators used in this study were designed to reflect 2 distinct, but equally important domains of care: access and provision. As hypothesized, both demonstrated associations with the proportion of potentially avoidable admissions. When faced with a decision regarding discharge destination, it seems reasonable that the more readily available in-home rehabilitation is, the more likely a clinician will be to make a referral to that service. In all analyses, a consistent relationship between the number of patients admitted to in-home rehabilitation and the frequency of avoidable admissions was demonstrated. However, mixed results were noted for wait times. One possible explanation for this finding is that wait times for in-home rehabilitation are long in all regions. Although mean wait times were seen to drop between 2006 and 2010, the lowest regional wait time achieved was still 17 days. Wait times may be important when considering discharge to in-home care, but they may be of equal concern in all regions, limiting our ability to detect significant associations.
Compared with the access indicators, the provision indicators generally demonstrated stronger correlations with avoidable admissions and all were found to be statistically significant on regression. This may indicate that when clinicians are faced with a decision about discharge destination, they are more interested in the content of the programs than access to them. This effect was most pronounced in the SLP visits per patient indicator where annual correlations were statistically significant in 3 of the 5 years. Estimates suggest that at discharge from an acute hospital, up to 65% of stroke patients demonstrate functional cognitive impairments 19 and 35% symptoms of aphasia. 20 In an Ontario-based study of high-functioning stroke patients, FIM motor and cognitive sub-scales, Mini Mental State Examination scores, and 5 items assessing orientation, financial independence, and verbal, written, and auditory communication were all significant predictors of long length of stay in inpatient rehabilitation among patients admitted with a FIM greater than 100. 21 In Ontario, access to neuropsychology is limited, which means that interventions to address many of these items fall within the scope of practice of SLP (especially within the community setting). If clinicians are concerned with the availability of in-home therapy services, they may be more inclined to keep patients in inpatient rehabilitation where they are more likely to have access to neuropsychology and SLP services. Collectively, these provision indicators may point to areas where targeted investments could have a meaningful impact.
Amid growing concern about limited in-home rehabilitation services, several LHINs have initiated programs to address this issue. In 2009, the South East Local Health Integration Network implemented an enhanced Community Care Access Centre program that allowed for greater provision of in-home rehabilitation to stroke patients. 22 Interestingly, the proportion of potentially avoidable admissions to inpatient rehabilitation in this Local Health Integration Network went from being above the provincial average between 2006 and 2008 to below average for 2009 and 2010. Similarly, the South West Local Health Integration Network established community stroke rehabilitation teams in January of 2009 and were below the provincial average for potentially avoidable admissions in fiscal year 2010. While anecdotal, each of these projects demonstrate the potential impacts of targeted investment and allow the opportunity for more detailed exploration going forward.
Limitations
One of the largest limitations facing this study is the small number of LHINs in Ontario, which represents a challenge to statistical power and raises concern about type II error in all analyses. Combining 5 years of data, as was done in the logistic regressions, is one way to overcome this limitation; however, it is not perfect and assumes that each Local Health Integration Network in each year is statistically independent. Although year was adjusted for in each model (along with an interaction term between year and indicator when significant), results must be interpreted with caution.
The definition of avoidable admissions operationalized in this study was the best available, but it should not be interpreted as ideal. Despite general consensus among Ontario’s Stroke Reference Group on the use of Rehabilitation Patient Group 1160 as a system-level proxy, it is not possible to confirm that all patients admitted to inpatient rehabilitation in Rehabilitation Patient Group 1160 could have been cared for at home. The Rehabilitation Patient Group only reflects functional independence and age. Additional considerations such as the patient’s living situation, family support, and safety issues are also frequently factored into the decision about where patients should receive rehabilitation. For instance, very mild communication difficulties are not necessarily identified by the FIM, but can be extremely problematic for patients who live alone and are not able to use a telephone effectively. 23 Conversely, there was also agreement among stroke reference group members that some patients in Rehabilitation Patient Group 1150 and possibly 1140 might also be able to receive services at home; although there is currently no way of identifying such patients retrospectively. Ongoing research into the clinical characteristics that best predict suitability for community-based rehabilitation is warranted, and collection of these data at the system level in Ontario will help to better inform future system-level evaluations.
Finally, this study focused exclusively on in-home rehabilitation resources and did not account for the availability of hospital-based outpatient rehabilitation services. Currently, there is no central database for outpatient rehabilitation in Ontario. At point of discharge from an acute hospital, patients returning to the community may be referred for outpatient rehabilitation as an alternative to in-home Community Care Access Centre services. It is possible that some of the LHINs lacking in-home resources have invested in outpatient services instead. However, one would anticipate the same relationship between outpatient rehabilitation and potentially avoidable admissions as that demonstrated here for in-home rehabilitation. Better data collection on outpatient rehabilitation is important and future studies should aim to evaluate the combined impact of outpatient and in-home rehabilitation resources on potentially avoidable admissions.
Conclusion
This ecological study confirms variation in the proportion of mild stroke patients admitted to inpatient rehabilitation across Ontario’s LHINs and adds further evidence that this variation is due, in part, to a lack of community-based rehabilitation resources. Across LHINs, correlations that agreed with this hypothesis were consistently found between fiscal years 2006 and 2010. Furthermore, regression of combined data demonstrated statistically significant associations between all indicators of in-home rehabilitation resource availability and potentially avoidable admissions to inpatient rehabilitation. Future research is required to test for similar associations with outpatient rehabilitation resources and to adjust for differences in patient characteristics between regions.
Footnotes
Funding:
This work was made possible by a Vanier Canada Graduate Scholarship provided by the Canadian Institute for Health Research and by funding support provided by the Ontario Stroke Network.
Declaration of conflicting interests:
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Author Contributions
Dr. Meyer was the lead author of this manuscript and performed all statistical analyses. All other authors were involved in hypothesis generation, oversite of research methods, manuscript creation, and review.
