Abstract
Background In the United States, people 85 years of age or older have a growing number of strokes each year, and this age group is most at risk for disability. Inpatient rehabilitation facilities (IRFs) adhere closest to post-acute stroke rehabilitation guidelines and have the most desirable outcomes compared with skilled nursing facilities. As stroke is one of the leading causes of disability, knowledge of postrehabilitation outcomes is needed for this age group, although at present such information is limited.
Objective The purpose of this study was to describe functional and discharge outcomes after IRF rehabilitation in people with stroke aged 85 years or older.
Design A serial, cross-sectional design was used.
Methods Inpatient Rehabilitation Facility–Patient Assessment Instrument data were analyzed beginning in 2002 for the first 5.5 years after implementation of the prospective payment system and included 71,652 cases. Discharge function, measured using the Functional Independence Measure (FIM), and community discharge were the discharge outcome measures. Sample description used frequencies and means. Generalized estimating equations (GEEs) with post hoc testing were used to analyze the annual trends for discharge FIM and community discharge by age group (85–89, 90–94, 95–99, and ≥100 years). Risk-adjusted linear and logistic GEE models, with control for cluster, were used to analyze the association between both outcome measures and age group.
Results Over 5.5 years, mean discharge FIM scores decreased by 3.6 points, and mean achievement of community discharge decreased 5.5%. Approximately 54% of the sample achieved community discharge. Continuous and logistic GEEs revealed factors associated with discharge outcomes.
Limitations Results obtained using an observational design should not be viewed as indicating causation. The lack of control for a caregiver may have altered results.
Conclusions The very elderly people admitted to IRF stroke rehabilitation made functional gains, and most were able to return to the community.
There is much interest in understanding how well very elderly people survive and recover from stroke. Interest stems from the rapid growth in the age group 85 years or older, which became the fastest growing segment of the population in the United States during the decade 2000–2010.1 In 2010, there were nearly 5.5 million people in this age group,1 and 17% of strokes occur within this group each year. Projections are that by 2050, total stroke incidence will more than double to nearly 1.5 million occurrences each year, and those aged 85 years or older will account for the largest increase in incidence.2 The increase also raises concern because stroke is currently one of the leading causes of long-term disability in the United States, with annual costs of more than $40 billion.2 With many stakeholders in the health care system interested in reducing the cost of such disability, evidence of the outcomes in the very elderly population may begin the discussion of the care needed to ameliorate impairments associated with stroke.
Outcomes of elderly people following stroke have been described in several studies that most commonly defined elderly people as being aged over 65 years. Increasing age has been associated with greater mortality,3–5 greater stroke severity,4–6 and discharge to nursing homes.4,5 Pre-existing disability prior to stroke has been found to increase with age and is associated with worse functional outcomes following stroke.4,5 An investigation of acute stroke units that offered “organized care,” defined as access to stroke team assessment and to physical therapy or occupational therapy, showed that providing such services to patients was associated with fewer deaths at any age, and findings were strongest in the eldest group (over 80 years of age).3 In addition, there is promise for rehabilitation in those over 80 years of age, as gains comparable to those of younger patients have been made during stroke rehabilitation.7 Although advanced age may increase vulnerability to adverse outcomes following acute stroke, there is evidence that response to rehabilitative measures is still possible. Unfortunately, most studies have not differentiated outcomes for people aged 85 years or older; therefore, understanding of stroke rehabilitation outcomes in this age group is currently lacking.
Countries with well-developed health care systems use guidelines for the rehabilitation management of acute stroke. Rehabilitation guidelines for stroke developed in Europe8,9 and a separate US guideline10 specific to the delivery of stroke rehabilitation in a long-term care setting agree on many aspects of care, including the configuration of an interdisciplinary clinical team, the amount of time for therapy, and locations where services can be provided (hospital, home, or outpatient). Although these guidelines are fairly comprehensive with regard to process, none differentiate recommendations based on patient age. In the United States, inpatient rehabilitation facilities (IRFs) are best aligned with the type of stroke rehabilitation advocated for in the guidelines because of the presence of an interdisciplinary team. In addition, because IRFs are required to provide a minimum of 3 hours of therapy per day,11 therapy time recommendations made in the guidelines are exceeded. Although differentiation by age has not been investigated previously, stroke outcomes achieved during IRF care have been consistently associated with more functional improvements and more discharges to home than from skilled nursing facilities (SNFs), the US alternative to IRFs for inpatient postacute rehabilitation.12,13
By regulation, therapy dose in IRFs is fixed at 3 hours per day for all patients; therefore, length of stay (LOS) becomes the logical measure of intensity. Our previous study14 examined the immediate time period following implementation of the prospective payment system in IRFs and found a trend for shorter LOS for 5.5 years following implementation that began prior to the policy's implementation.15 As this payment policy is still in place today, and no previous study, to our knowledge, has investigated functional and discharge outcomes in patients with stroke aged 85 years or older treated in IRFs, an additional investigation was warranted. Obtaining evidence of the outcomes for this age group may facilitate consideration of the very elderly population for access to this most intensive form of postacute care (PAC). Poststroke functional and discharge outcomes associated with IRF care have not been described previously in this group. This study, therefore, was designed to describe functional and discharge outcomes after IRF rehabilitation in people with stroke aged 85 years or older.
Method
Data Source
Assessment data from the instrument specific to IRF care, the Inpatient Rehabilitation Facility–Patient Assessment Instrument (IRF-PAI), were obtained from the Centers for Medicaid & Medicare Services for this study, which spanned from January 2002 through June 2007. Data were drawn from all 50 states, the District of Columbia, and Puerto Rico. Missing data were not found because IRFs forego reimbursement for care unless all data are reported through the IRF-PAI.
Study Sample
The sample was drawn from Medicare Part A beneficiaries aged 85 years or older who were admitted to a US IRF for stroke rehabilitation during the study time frame. We included those beneficiaries who had 1 of 5 “admission group impairment codes” specific to stroke on the IRF-PAI (ie, right-brain, left-brain, bilateral-brain, no-paresis, or “other” stroke) and if this was their first admission to IRF care. Exclusion criteria included being discharged against medical advice (n=3,840), the person expired in the IRF (n=2,213), an LOS of 0 days in the IRF (n=47), and LOS was greater than 90 days (n=652). The rationale to exclude beneficiaries with an LOS greater than 90 days was that stays longer than 90 days were found to be rare, and we judged that issues related to discharge destination might have played a role in stays beyond that threshold. The study sample consisted of 71,652 individuals in 1,483 facilities.
Measures
Discharge functional ability was the first dependent variable and was measured using the Functional Independence Measure (FIM). The FIM is contained within the IRF-PAI assessment and is routinely scored at admission and discharge in IRFs. The FIM scores are added across 18 items (ie, eating, grooming, bathing, dressing upper body, dressing lower body, toileting, bladder management, bowel management, transferring to bed/chair/wheelchair, transferring to toilet, walk/wheelchair, stairs, comprehension, expression, social interaction, problem solving, and memory) that are measured on a Likert scale (0–7). Discharge FIM scores can vary from 7 to 126, with higher scores indicating better function. The reliability16,17 and validity18,19 of the tool have been established.
The second dependent variable was community discharge versus institutional discharge from the IRF. Discharge to independent living, with or without home care services, was deemed a community discharge, whereas institutional discharge was deemed when staff were present on the premises and discharge destination included assisted living facilities, SNFs, or nursing homes.
Risk Adjustment
We controlled for variation among individuals in the study sample. Unless otherwise stated, data were categorized using IRF-PAI categories or collapsed further to create larger numbers within categories. For example, race/ethnicity was collapsed from 8 to 2 groups (white versus nonwhite) because many subgroups were too small for accurate analysis. Age was collapsed into 4 groups (85–89, 90–94, 95–99, and ≥100 years), with patients aged 85 to 89 years comprising the reference age group. The remaining demographic variable was sex. Stroke impairment group was identified at admission, and individuals were categorized into 1 of 5 groups by type of stroke (right-brain, left-brain, bilateral-brain, no-paresis, or “other”). Patients who were identified as having “other” stroke were the reference group so that differences could be ascertained among the better-understood types of stroke. Several control variables were analyzed as continuous variables. Length of stay was defined as the number of days spent in the IRF. Admission stroke severity was measured using the admission FIM score (0–126). Described by International Classification of Diseases, ninth edition (ICD-9) codes, the number of comorbidities (up to 10) and the number of complications (up to 6) were included to control for pre-existing conditions and complications to the primary diagnosis or comorbidities that began after admission, respectively. Lastly, control for year of study was present in both models because each year represented a separate cohort with associated unmeasured characteristics.
Data Analysis
Descriptive statistics were used to characterize the patient-level factors. Annual trends for discharge FIM scores and the achievement of community discharge were analyzed using generalized estimating equations (GEEs). This method was necessary due to the nested structure of the data of patients within IRFs. Post hoc testing for differences across years was achieved using the Tukey-Kramer test.
A continuous GEE model was used to describe the factors associated with discharge FIM scores, and a logistic GEE model was used to describe the factors associated with the achievement of community discharge. All covariates for risk adjustment were included in the models. Both models were analyzed using an exchangeable correlation structure. Discharge FIM scores were an additional control variable in the logistic model examining the achievement of community discharge. This variable was added to control for discharge function because studies have shown an association between discharge function and discharge destination.14,20,21 The identity link function was used in the continuous model, and a logit link function was used in the dichotomous model. Alpha level was set at .05 for 2-tailed tests. Statistical analysis was conducted using SAS software (version 9.2, SAS Institute Inc, Cary, North Carolina).
Role of the Funding Source
Financial support was received from the American Heart Association Predoctoral Fellowship, 2010, and American Physical Therapy Association Health Policy and Administration Section Research Award, 2010.
Results
Characteristics of the sample are shown in Table 1. Nearly three-fourths of the sample were in the age group 85–89 years, and approximately 23% were in the age group 90–94 years. The age group 95–99 years comprised nearly 4% of the sample, and patients aged 100 years or older comprised <1% of the sample. The sample was predominantly female (65.5%) and white (85.5%). The number of individuals with right- or left-brain strokes was nearly equal and comprised just over 80% of the sample. Stroke without paresis was the next most common form at nearly 11%, and bilateral stroke comprised nearly 3% of the sample. A clear definition of “other” stroke was lacking, although it comprised more than 6% of the sample. More than 92% of the sample had at least 4 comorbidities noted, of which 70% had 7 or more comorbidities. Complications were rare in this sample, with more than 70% not having a complication while in the IRF. However, nearly 21% had at least 1 complication, and a small minority had between 4 and 6 complications during the IRF stay. The mean admission FIM score was 53.2 points, the mean discharge FIM score improved to 74.1 points, and the mean LOS was 16.0 days.
Patient Characteristics (N=71,652)
The eTable presents the discharge destinations for the sample. Nearly 54% of the sample achieved a community-based discharge. Figure 1 presents the mean LOS and FIM change scores by age group. The FIM scores changed approximately 1.2 to 1.3 points per day across all age groups.
Mean length of stay and mean Functional Independence Measure (FIM) change scores by age group after rehabilitation for stroke in inpatient rehabilitation facilities among adults aged 85 years or older. FIM change score=admission FIM score – discharge FIM score.
Annual Trends
The trends for mean annual discharge FIM scores and mean achievement of community discharge for those patients treated in an IRF who were 85 years of age or older are presented in Figures 2 and 3, respectively. Mean discharge FIM scores decreased 3.6 points, from 76.8 to 73.2 points (P<.0001), and achievement of community discharge decreased 5.5%, from 63.6% to 58.1% (P<.0001). Mean trends according to age group for discharge FIM scores and achievement of community discharge are shown in eFigures 1 and 2.
Annual trend: unadjusted mean discharge Functional Independence Measure (FIM) scores after rehabilitation for stroke in inpatient rehabilitation facilities among adults aged 85 years or older.
Annual trend: unadjusted mean frequency of community discharge after rehabilitation for stroke in inpatient rehabilitation facilities among adults aged 85 years or older.
Discharge Function
Table 2 displays the estimates, 95% confidence intervals, and P values for the model examining the factors associated with discharge FIM scores. Controlling for all other factors, LOS, admission FIM score, being aged 85–89 or 90–94 years, and having a no-paresis stroke were associated with increased discharge FIM scores (LOS=0.52 points, P<.0001; admission FIM score=1.10 points, P<.0001; age 85–89 years: 3.45 points higher than age ≥100 years, P<.0001; age 90–94 years: 2.59 points higher than age ≥100 years, P<.0001; no-paresis stroke: 1.17 points higher than “other” stroke, P<.0001).
Factors Associated With Discharge FIM Scores in Patients With Stroke Aged 85 Years or Older, With Model Covariatesa
Other factors were associated with reductions in discharge FIM scores, including being nonwhite, having a right-brain stroke, having a bilateral-brain stroke, increasing number of comorbidities, and increasing number of complications (0.84 points lower than white group, P<.0001; right-brain stroke: 1.53 points lower than “other” stroke, P<.0001; bilateral-brain stroke: 1.07 points lower than “other” stroke, P<.003; number of comorbidities=0.23 points, P<.0001; number of complications=0.70 points, P<.0001).
Factors that did not significantly contribute to the model explaining discharge FIM scores were being aged 95–99 years (P<.06), sex (P<.17), and having a left-brain stroke (P<.69). Finally, only year 2003 contributed to the model explaining discharge FIM scores (2.15 points lower from the reference year 2002, P<.003).
Community Discharge
Table 3 displays the odds ratios (ORs), 95% confidence intervals, and P values for the logistic model examining factors associated with achieving community discharge. Patient factors that were associated with reduced odds of achieving community discharge were LOS (OR=0.992, P<.01), higher admission FIM score (OR=0.985, P<.0001), being female (OR=0.885, P<.0001), having a right-brain stroke (OR=0.847, P<.0001), and having complications in the IRF (OR=0.935, P<.0001).
Factors Associated With Community Discharge in Patients With Stroke Aged 85 Years or Older, With Model Covariatesa
Patient factors that were associated with increased odds of achieving community discharge were higher discharge FIM scores (OR=1.068, P<.0001) and being nonwhite (OR=2.140, P<.0001). Several factors had no effect in the model predicting achievement of community discharge, including age group (85–89 years, P=.43; 90–94 years, P=.73; 95–99 years, P=.66), left brain stroke (P=.29), bilateral brain stroke (P=.34), no-paresis stroke (P=.79), and number of comorbidities (P=.25). Study years 2005 (OR=0.66, P=.01), 2006 (OR=0.58, P=.001), and 2007 (OR=0.43, P<.0001) were the only years that contributed to the model predicting the achievement of community discharge. Years 2003 and 2004 did not contribute to the model.
Discussion
The growing number of elderly people with stroke who are aged 85 years or older has implications for society and the health care system. Stroke often leads to disability that creates ongoing functional impairments that last long after the event and even to the end of life. Post-acute stroke rehabilitation in IRFs meets or exceeds recommended guidelines for care, and the amelioration of functional loss is the best way to control the long-term disability associated with stroke. The possibility of reduced disability may benefit families and the health care system. This study, for the first time, to our knowledge, has described discharge outcomes over a 5.5-year period in people aged 85 years or older who received rehabilitation in IRFs, arguably the most vulnerable group for disability poststroke.
Across all age groups, functional gains on the FIM averaged just over 20 points during the rehabilitation admission, although gains in the 2 youngest age groups (85–89 and 90–94) averaged somewhat higher than gains in patients older than 95 years. Multivariate analysis confirmed the association between age and discharge function, although with differences in FIM scores averaging between 1 and 3 FIM points for each group, the threshold for clinical difference among groups was not met.22 According to Beninato et al,22 the minimum clinically important difference (MCID) for the FIM is 22 points; therefore, all age groups in this study could be considered to have clinically similar discharge FIM scores. Of greater concern, the average change in FIM score did not reach the MCID for any age group, which may indicate that average LOS was too short for all patients aged 85 years or older. Support for the association between longer LOS and higher discharge FIM scores was provided by the multivariate analysis, which may indicate that extending the LOS for people aged 85 years or older may improve discharge FIM scores above this threshold. One advantage of this study's serial design was a longer time frame in which to view IRF stroke rehabilitation practice. Length of stay grew shorter during this study, which continued a trend of shorter LOS beginning in the 1990s.15 Since that time, FIM scores have changed less during the rehabilitation stay, and total discharge FIM scores have trended lower, which leaves patients of all ages at risk for not achieving a community level discharge.15
Functional improvement has been found associated with community discharge,14,20,21 and approximately 60% of this sample was able to achieve this goal. Age was not found to be associated with the achievement of community discharge, and one possibility related to the goal of achieving community discharge could be a lack of an adequate caregiver.23 Limited information about caregivers is available through the IRF-PAI, which left a potentially important variable unmeasured. Results showed that the strongest predictor for achieving community discharge was being nonwhite; therefore, a cultural component that mediates the discharge outcome may be likely. Although this study controlled for race/ethnicity, characteristics specific to culture that may have led to these differences were left unmeasured. People of cultures who tend to live multigenerationally may be a good risk for admission to IRF for stroke rehabilitation because they may be less likely to consider nursing home placement. New investigations into the interplay among age, race/ethnicity, LOS, and having an adequate caregiver are needed. Clinical decision making may be enhanced with better understanding of the relationship among these factors, as well as the potential benefit of returning patients to the community and preventing institutional long-term care.
One potentially counterintuitive finding was the association between longer LOS and reduced odds of achieving community discharge. Although this relationship was found to be significant, the OR value was nearly equivocal, with very small odds (0.08%, P<.01) of LOS playing any role in achieving community discharge. For this study, all patients were included in the analysis if they were in the IRF up to 90 days. We included patients for this time frame because our purpose was to describe a complete picture of IRF stroke rehabilitation practice, although researchers in future studies may want to shorten this view to approach a more typical LOS. Having less variability in LOS may alter the relationship between LOS and achievement of community discharge found in this study.
Sex differences also were found in the achievement of community discharge. Despite having discharge FIM scores similar to those of men, women had lower odds of returning to the community after stroke IRF rehabilitation. The lack of data collected on the IRF-PAI about caregivers again limited the possibility to find an association, which may explain why women did not achieve community discharge as often as men. It is possible that fewer caregivers were present in the lives of women in this sample, making a community discharge impossible, although we could not identify this as a possibility from the IRF-PAI data.
Based on our findings, the average discharge FIM scores for the entire sample indicated that a caregiver was likely needed for many aspects of mobility, activities of daily living, and cognition. There is evidence that a discharge FIM score should reach 78 before community discharge is more likely.24 This very elderly sample was not discharged at that threshold at any point during this study, although the discharge FIM scores of the youngest age group (85–89 years) approximated it in 2002. By 2007, all of the age groups averaged at least 4 points under the threshold, and the age group 95–99 years averaged 14 points under the threshold. Although the optimal LOS for IRF care poststroke of any age is not known, additional study is warranted to determine the optimal LOS for IRF rehabilitation based on age and stroke severity. The present health care climate in the United States will demand that any potential higher costs associated with longer LOS in an IRF will have to be justified. A Canadian research team25 developed benchmarks for LOS for poststroke inpatient rehabilitation, but with the pronounced differences between the Canadian and US health care systems, new research within the United States is needed. In the meantime, we recommend that plans of care developed in IRFs should ensure the approximation of 78/126 discharge FIM points to best ensure the possibility of community discharge. Further study examining the validity of this threshold is warranted.
This analysis included approximately 90% of the initial hospital sample from our larger study,14 leaving open the possibility that not all IRFs admitted patients in this age group to their programs. Clinicians who make IRF admission decisions for patients may believe that having multiple comorbidities and a stroke at an extreme age would limit the ability to tolerate IRF-level rehabilitation and return to the community; however, we found this was not the case. Institutional-level factors, such as whether a hospital owned a nursing home, have been found to be associated with using that facility for PAC.26,27 Two studies have reported contradictory evidence over whether the availability of the types of PAC in a local market, which includes IRF, SNF, or home care agencies, drives admission.27,28 Decisions made at discharge from acute care to select patients for access to IRF-level rehabilitation may be influenced by the hospital system structure, local availability of PAC, and patient age. With few studies addressing access to IRF care poststroke, many questions remain about the type of patient admitted to IRF care. Although this issue warrants further attention, this study examined only those patients admitted to IRF care poststroke, so selection to IRF care preceded this study's time span. Ensuring that no age bias exists in IRF admission criteria is needed, and further research is recommended.
Many of the limitations of this study are related to the use of a large existing database where researchers are not in control of the types of variables collected. The group aged ≥100 years was found to have most variable results (eFigs. 1 and 2), which was most likely related to the small number of patients in the group (n=247). In addition, the extreme age of these patients may have led to additional complexity left unmeasured in this study, such as learning, motivation, and brain plasticity. Research on how extreme age may affect rehabilitation following stroke is warranted because it is expected that the very elderly group will grow in the future, and clinicians should be prepared for such encounters.
Although the sample size in this study was large, the results should be viewed cautiously because the observational design lacks the statistical control of an experimental design, and the associations found should not be viewed as indicating causation. Not knowing with certainty that a caregiver was available for discharge may have altered results. New studies should ensure that information about a caregiver is obtained to further control for potential race/ethnicity and sex differences. Ceiling effects for the variables of comorbidities and complications may be present because the IRF-PAI limits these to 10 and 6, respectively. In addition, the vague definition of “other” stroke leaves questions as to whether there is overlap with other known types of stroke.
Other error present in this study is at the facility level, which is not collected on the IRF-PAI. Facility characteristics would have added further control and information about what factors might drive discharge outcomes. The selection of patients for admission to IRFs may lead to differences from those not admitted. Therefore, caution should be taken when generalizing these results to those not treated in IRFs. Despite the probable selection bias, this study described functional discharge outcomes in people aged 85 years or older, which were previously unknown. Although little age disparity among those aged 85 years or older was found in this study, the potential for age disparities remains, and the possibility for such should be investigated.
In conclusion, this study showed that once admitted to an IRF for postacute care for stroke, the very elderly patients (85 years or older) made functional gains and returned to the community approximately 54% of the time. During the time frame of this study, however, worrisome trends were found that may limit the ability of very elderly people to return to the community poststroke. By the end of the study, none of the groups achieved discharge FIM scores that approximated a threshold where risk of institutional discharge is reduced. In addition, the MCID for the FIM was not reached at any point in time for any age group, indicating that functional change during the IRF stay may not have been adequate to achieve community discharge. There may be a benefit to patients and the health care system to extend the LOS for very elderly patients to better approximate a level of function that may allow a community discharge. However, further study into the optimal LOS for the elderly population poststroke is needed.
Footnotes
Both authors provided concept/idea/research design and data analysis. Dr O'Brien provided writing, project management, and fund procurement. Dr Xue provided consultation (including review of manuscript before submission).
The University of Rochester Human Subjects Review Committee granted approval for this study.
Financial support was received from the American Heart Association Predoctoral Fellowship, 2010, and American Physical Therapy Association Health Policy and Administration Section Research Award, 2010.
- Received June 30, 2015.
- Accepted February 13, 2016.
- © 2016 American Physical Therapy Association
References
Issue highlights
