Functional Gain After Inpatient Stroke Rehabilitation
Correlates and Impact on Long-Term Survival
Background and Purpose—Prediction of functional outcome after stroke rehabilitation (SR) is a growing field of interest. The association between SR and survival still remains elusive. We sought to investigate the factors associated with functional outcome after SR and whether the magnitude of functional improvement achieved with rehabilitation is associated with long-term mortality risk.
Methods—The study population consisted of 722 patients admitted for SR within 90 days of stroke onset, with an admission functional independence measure (FIM) score of <80 points. We used univariable and multivariable linear regression analyses to assess the association between baseline variables and FIM gain and univariable and multivariable Cox analyses to assess the association of FIM gain with long-term mortality.
Results—Age (P<0.001), marital status (P=0.003), time from stroke onset to rehabilitation admission (P<0.001), National Institutes of Health Stroke Scale score at rehabilitation admission (P<0.001), and aphasia (P=0.021) were independently associated with FIM gain. The R2 of the model was 0.275. During a median follow-up of 6.17 years, 36.9% of the patients died. At multivariable Cox analysis, age (P<0.0001), coronary heart disease (P=0.018), atrial fibrillation (P=0.042), total cholesterol (P=0.015), and total FIM gain (P<0.0001) were independently associated with mortality. The adjusted hazard ratio for death significantly decreased across tertiles of increasing FIM gain.
Conclusions—Several factors are independently associated with functional gain after SR. Our findings strongly suggest that the magnitude of functional improvement is a powerful predictor of long-term mortality in patients admitted for SR.
Stroke is a leading cause of death and long-term disability in developed countries.1 Because of the aging population and the declining early mortality after stroke,1 the demand for stroke rehabilitation (SR), as well as the economic, social, and family burden of stroke, is expected to increase in the coming years. Thus, prediction of functional outcome after SR has become a growing field of interest as it may have important implications for planning management strategies and informing patients and relatives.2,3 Age, initial stroke severity, and functional status at rehabilitation admission emerged as the most informative predictors of functional outcome in previous studies.4 In the recent large study of Brown et al,5 functional independence measure (FIM) motor score at rehabilitation hospital admission dominated prediction of outcome at discharge. The model, including age, admission FIM motor score, and distance walked, however, explained only 10.7% of the variance in overall FIM gain.5
Importantly, the possible association between SR and long-term survival still remains elusive.6 In a rehabilitation-based study, functional status at 6 months was a significant predictor of long-term mortality, generating the hypothesis that “Interventions that improve post-stroke functional status may have a protective effect on mortality.”7
We sought to identify the factors associated with functional outcome after SR and to investigate whether the magnitude of functional improvement achieved with rehabilitation is associated with long-term mortality risk.
The study population consisted of 1010 patients consecutively admitted for SR from January 2002 to October 2011. Patients were included in the study if they had been admitted within 90 days of onset of an ischemic or hemorrhagic stroke and had an admission FIM score of <80 points.8 Patients who were admitted >90 days after stroke (n=144; 14.3%), had an admission FIM score of ≥80 points (n=77; 7.6%), were transferred to an acute care facility (n=51; 5%) or discharged against medical advice (n=7; 0.7%), or died during in-hospital stay (n=9; 0.9%) were excluded. Thus, 722 patients were available for analysis. Patients’ data were deidentified. The setting was the inpatient neurological rehabilitation unit of the Maugeri Rehabilitation Institute, which has a regional user base and is certified (ISO9001, quality management systems) for activities of rehabilitation. Our interdisciplinary SR team comprises the following professionals with expertise in SR: neurologist, physiatrist, physiotherapist, occupational therapist, speech and language therapist, neuropsychologist, and nurse. The patients received physical and occupational therapy for 3 hours per day for 5 days and for 1 hour for 1 day of each week. Patients with more severe initial disability were managed with lower intensity programs. Admission and discharge FIM scores were recorded by trained therapists, as a part of our formal rehabilitation program. Postdischarge vital status was ascertained in 686 residents by linking with the regional health information systems. FIM gain was calculated as the difference in total FIM score before and after rehabilitation. This study was approved by the local institutional review board.
Data were 99.9% complete. We used univariable and multivariable linear regression analyses to assess the association between baseline variables and FIM gain and univariable and multivariable Cox regression analyses to assess the association of FIM gain with long-term postdischarge mortality risk. All continuous covariates were treated as such. Redundant variables (Pearson coefficient, >0.50) were dropped from the multivariable model, including that with the strongest effect size. Candidate baseline variables with P≤0.10 at univariable analysis were retained for inclusion in the multivariable analysis. Discrimination of the model was assessed by calculating the C-index. The statistical interaction between FIM gain and admission or discharge total FIM score in predicting mortality was evaluated by likelihood ratio test comparing models with and without interaction terms. Analyses were conducted using STATA software, version 12 (StataCorp LP, College Station, TX).
Baseline characteristics and changes in FIM scores from admission to discharge are reported in Table 1. The mean FIM gain was 29.7±18.1 points.
Correlates of FIM Gain
Table 2 shows the results of univariable and multivariable linear regression analyses. At multivariable analysis, age (P<0.001), marital status (P=0.003), time from stroke onset to rehabilitation admission (P<0.001), National Institutes of Health Stroke Scale score at rehabilitation admission (P<0.001), and aphasia (P=0.021) were independently associated with FIM gain. The R2 of the model was 0.275.
During a median follow-up of 6.17 years, 253 of the 686 residents (36.9%) died. Table3 shows the results of univariable and multivariable Cox regression analyses. FIM gain had a univariable C-index of 0.667. At multivariable analysis, age (P<0.0001), coronary heart disease (P=0.018), atrial fibrillation (P=0.042), total cholesterol (P=0.015), and FIM gain (P<0.0001) were independently associated with mortality. The C-index of the predictive model was 0.745. The Figure shows survival curves of the patients stratified by tertiles of FIM gain. The adjusted hazard ratio for death significantly decreased across tertiles of increasing FIM gain. No interaction between FIM gain and admission (P=0.800) or discharge (P=0.127) total FIM score was found, indicating that the association of FIM gain with mortality risk was independent of admission or discharge total FIM score.
There are 2 major findings of this study: (1) several variables were independently associated with total FIM gain; the model explained 27% of the variance in FIM gain and (2) after adjusting for established mortality risk markers in patients with stroke, FIM gain resulted to be a powerful predictor of long-term mortality risk.
Consistent with previous studies,4 age, a measure of stroke severity (National Institutes of Health Stroke Scale score), and a process of care indicator were significant predictors of functional gain. The finding that time from stroke onset to rehabilitation admission is positively associated with functional gain can be a relevant issue for clinicians involved in decision making about admission to rehabilitation hospital and policy makers.4 The significant association between married status and functional improvement is a novel finding,4 which is difficult to interpret. However, this finding is in line with the Oxford Vascular Study where nonpartnered patients were more likely to be disabled at 6 months after index stroke.9 However, the proportion of explained variance in FIM gain was modest, indicating that other unmeasured processes of care-related or patient-related factors contribute to functional outcome.
In a recent population-based study, early SR was associated with a lower risk of mortality, after accounting for age and sex.10 De Wit et al7 showed that functional status at 6 months predicts long-term mortality. To the best of our knowledge, this study is the first to report an independent, strong association between the magnitude of functional improvement achieved with rehabilitation and long-term mortality risk in stroke survivors. Adjusting for established markers of mortality risk in patients with stroke, such as age, sex, diabetes mellitus, coronary heart disease, renal function, and stroke severity,11 FIM gain emerged as a highly significant predictor of long-term mortality. There was a significant graded and independent decrease in the hazard ratio for death across tertiles of increasing FIM gain. The patients in the highest FIM gain tertile had a 60% lower likelihood of dying compared with their counterparts in the lowest FIM gain tertile. The finding that admission National Institutes of Health Stroke Scale was a highly significant independent predictor of FIM gain suggests that some of the relationship between FIM gain and mortality may really represent a relationship between stroke severity and mortality. Although a significant association does not prove a cause–effect relationship, our findings strongly suggest that continued efforts should be devoted to enhancing the effectiveness of SR to improve not only disability and quality of life but also, hopefully, life expectancy of stroke survivors.
The main limitation of this study is its retrospective nature. However, all consecutive patients admitted for SR during the index period were considered for inclusion in the study and selected according to prespecified criteria; moreover, the data set was >99% complete. We excluded patients with mild stroke. Ideally, a measure of functioning and disability should not be susceptible to ceiling effect.12 Ceiling effects, indeed, may limit the ability of a measure to accurately assess patient improvement and lead to type I error inflation.13,14 A ceiling effect for FIM score has been demonstrated.13 To minimize ceiling effect, we excluded patients with an FIM score of ≥80.8 This was a single-center study. Although this ensures the uniformity of data collection and treatment across the studied population, it may limit the generalizability of the results.
Although several predictors of functional improvement were identified, the explained variation in functional gain was modest, suggesting the need for continued research aimed at identifying additional process of care- and patient-related factors influencing the effectiveness of SR. Our findings strongly suggest that the magnitude of functional improvement is a powerful predictor of long-term mortality in patients admitted for SR. Studies are needed to investigate the pathophysiological mechanisms underlying this association.
We thank all physicians, therapists, and nurses who were involved in the care of the patients and Francesco Colucci for his assistance in acquiring the data.
- Received June 15, 2015.
- Revision received July 29, 2015.
- Accepted August 11, 2015.
- © 2015 American Heart Association, Inc.
- Mozaffarian D,
- Benjamin EJ,
- Go AS,
- Arnett DK,
- Blaha MJ,
- Cushman M,
- et al
- Meyer MJ,
- Pereira S,
- McClure A,
- Teasell R,
- Thind A,
- Koval J,
- et al
- Brown AW,
- Therneau TM,
- Schultz BA,
- Niewczyk PM,
- Granger CV
- Luengo-Fernandez R,
- Paul NL,
- Gray AM,
- Pendlebury ST,
- Bull LM,
- Welch SJ,
- et al
- Wijnhoud AD,
- Maasland L,
- Lingsma HF,
- Steyerberg EW,
- Koudstaal PJ,
- Dippel DW
- Kwon S,
- Hartzema AG,
- Duncan PW,
- Min-Lai S