Efficiency, Effectiveness, and Duration of Stroke Rehabilitation

241 Efficiency, Effectiveness, and Duration of Stroke Rehabilitation Surya Shah, MEd OTR/L, Frank Vanclay, MSocSci, and Betty Cooper, BAppSc This prospective multicenter study identifies the variables significant in the prediction of efficiency, achievement of potential and duration of stay in 258 persons with a first stroke admitted to comprehensive inpatient in Brisbane, Australia, during 1984. All three dependent variables were poorly predicted, with only 17% of the variance in efficiency, only 30% of the variance in achievement of potential, and only 22% of the variance in duration of stay explained. Unlike other reports, we considered most of the major medical (side of paralysis, stroke etiology, site of the lesion, arterial distribution affected, etc.), rehabilitative (initial Barthel Index score, interval from stroke onset to acute-care hospital admission, interval from hospital admission to commencement, neurologic measures, etc.), and demographic (age, years of education, occupation, ethnicity, etc.) variables. The high proportion of unexplained variance is likely to be due to nonmedical factors influencing the selection of patients for. (Stroke 1990;21:241-246) A ttainment of maximal functional capacity and / \ subsequent improved quality of life are X A. important in stroke. 12 However, with limited numbers of beds and increasing costs, it is also important to apply available resources to patients who are likely to obtain the most benefit from comprehensive inpatient. 3 Since individual rates of efficiency and effectiveness vary considerably, the selection of inpatients is important, as is the provision for related community- or hospital-based welfare services for those unlikely to benefit from inpatient care. While the general principle of triage operates to select a 'middle-band' of patients for, 4 the proportion of stroke patients who receive comprehensive inpatient varies greatly. 5 Australia has one of the highest rates, with 43% being referred for inpatient, 6 compared with 19% in The Netherlands, 7 14% at the Mayo Clinic, 8 and 21% in other medical centers in the United States. 9 Due to methodologic flaws, there is considerable disagreement among the studies that have examined improvement in functional status, function on dis- From the Department of Occupational Therapy, University of Queensland, St. Lucia, Australia. Supported by a health services research and developments grant from the Commonwealth Department of Community Services and Health, Australia. Address for correspondence: Associate Professor Shah, Occupational Therapy, University of Queensland, St. Lucia, Queensland 4067, Australia. Received May 30, 1989; accepted September 19, 1989. charge, and variables predicting outcome measures. 1-2 - 4-8 - 10-14 Only a few studies have examined the efficiency of gains or the effectiveness of achieving potential. 1516 This prospective, multicenter, methodologically sound study examines these factors, as well as the duration of stay, using a population of all patients with a first stroke referred for comprehensive inpatient in Brisbane, Australia. Subjects and Methods We used a prospective pretest-posttest study design for patients with first stroke referred for comprehensive inpatient during 1984. Stroke was defined as rapidly developing clinical manifestation of a focal loss of cerebral function lasting >24 hours or leading to death within 24 hours. For 52% of the patients the diagnosis of stroke was based on neuroradiologic investigations and for the other 48%, on the clinical impression of the consulting physician and neurologist. Comprehensive inpatient was defined as the combined and coordinated use of a physical therapist, an occupational therapist, and a nurse in a multidisciplinary team, with the services of speech therapists, social workers, and psychologists available as required. The patients were primarily drawn from the population residing in Brisbane, Australia, which numbered slightly over 1 million in 1984. All seven public hospitals and the four private hospitals in Brisbane that admitted stroke patients participated in our study.

242 Stroke Vol 21, No 2, February 1990 In 1984 2,676 patients were admitted with a provisional diagnosis of cerebrovascular disease (International Classification of Diseases [ICD] codes 430-437). 17 Of these 2,676 patients, 1,342 had a diagnosis of stroke (ICD codes 431, 434, 436, or 437). Of the 1,342 stroke patients, 265 (20%) died within the first 2 weeks, 98 (7%) with persistent altered consciousness state were transferred directly to extended-care facilities, and 359 (27%) who were independent in the activities of daily living (ADL) were sent directly home. To minimize potential confounding, we did not consider the 68 stroke patients (5%) with neurosurgical intervention or the 291 patients (22%) with a second or multiple stroke. There remained 261 patients (19%) who survived their first stroke in 1984 and were referred for comprehensive inpatient following their discharge from acute-care hospitals. The time from stroke onset to commencement ranged from 1 to 79 (mean±sd 17 ±14.2, median 13) days. Patients were assessed both on admission to and on discharge from. Three patients died during, leaving 258 patients, the total population of survivors of first stroke who underwent comprehensive inpatient during 1984. The patients were discharged from when the multidisciplinary team considered that they had achieved their maximum benefit. Rehabilitation stay ranged from 6 to 276 (mean±sd 61±45.1, median 49) days. Functional recovery was measured as improvement in the ability to perform ADL and was scored according to the Barthel Index. Neurologic recovery was measured using the Brunnstrom motor recovery scales. 18 A small modification of the Barthel Index was incorporated to improve its sensitivity to small increments. 19 The Barthel Index measures functional independence at one point in time, and the change in Barthel Index score from commencement to discharge can be calculated as one measure of functional improvement. Rehabilitation efficiency 15 is the amount of improvement divided by the duration of stay and can be regarded as the average increase in Barthel Index score per day: efficiency=(discharge Barthel Index score-initial Barthel Index score)-^duration of stay. Since the potential improvement for patients with high initial scores is lower than that for those with low initial scores, another way to measure functional improvement is achievement of potential improvement. This is expressed as actual improvement divided by potential improvement, 15 a percentage reflecting the proportion of potential improvement actually achieved during : achievement of potential = (discharge Barthel Index score-initial Barthel Index score)-^(100-initial Barthel Index score) x-» 100%. Duration of stay, measured in days, is the time from commencement to discharge, not the time from stroke onset or admission to the acute-care hospital. TABLE 1. Ordinal/Interval Variables Considered in Predicting Rehabilitation Outcome in 256 Patients With First Stroke Variable Mean SD Median Dependent Rehabilitation efficiency Achievement of Duration of stay (days) Independent Initial Barthel Index score Discharge Barthel Index score Age (years) Education (years schooling) Interval from stroke onset to acute-care hospital admission (days) Interval from hospital admission to commencement (days) Bladder control (Barthel Index subscore) (0-10) Brunnstrom arm recovery level (1-16) Brunnstrom hand recovery level (1-6) Upper-extremity motor control (1-3) 0.84 66.0 60.7 44.6 77.7 69.6 8.4 6.5 10.9 7.0 2.8 2.6 1.5 0.84 35.8 45.1 24.8 26.5 10.8 2.0 13.9 9.5 4.0 1.6 1.6 0.7 0.60 76 49 41 90 71 8 We analyzed efficiency, achievement of potential, and duration of stay (each measured as a ratio) as the dependent variables and all medical, rehabilitative, demographic, and attribute variables available at commencement as independent variables. The data were coded and entered into a DEC-10 mainframe computer using the 1022 database system. Analysis was performed using SPSS-X 20 on a VAX 8550 mainframe computer. We obtained both descriptive statistics and initial correlations, and we performed statistical analysis using one-way and multi-way analysis of variance (ANOVA) and stepwise regression (forward selection). Results We excluded data from two patients because their initial Barthel Index scores of 100 prevented our calculating values for achievement of potential and gave nonsensical values for efficiency. Descriptive statistics for the ordinal/interval variables are presented in Table 1, Table 2 details the frequencies of the attribute variables, and Table 3 presents the results of one-way ANOVAs between the nominal variables and the three dependent variables. Marital status was significantly related to the achievement of potential (/><0.05). Side of paralysis had a significant bearing on efficiency (/><0.05) and achievement of potential (/?<0.01). Site of the lesion showed a significant association with duration of reha- 0 8 10 3 2 1

Shah et al Stroke Rehabilitation Efficiency 243 TABLE 2. Attribute Variables Considered in Predicting Rehabilitation Outcome in 256 Patients With First Stroke Variable % Male sex 56 Ethnicity 16 Visual confrontation affected 61 Hypertension 70 Myocardial infarction 29 Other ischemic diseases 29 Diabetes mellitus 16 Congestive cardiac failure 14 Peripheral vascular disease 8 Fractured femur 4 Other associated illness* 55 Includes any other medical condition not itemized, none of which have individual frequency of >4%. bilitation stay (/?<0.05). When these significant variables were incorporated into a multi-way ANOVA that controlled for age and initial Barthel Index score, the only variable to remain significant was site of the lesion. In stepwise regression with the other predictor variables, site of the lesion failed to significantly increase R 2, indicating that site of the lesion does not explain additional variance not already accounted for by the combination of the other predictor variables. Table 4 presents the initial correlations, while Table 5 presents the final results of regression for the three dependent variables. For this population study, side of paralysis, stroke etiology, site of the lesion, and the arterial distribution affected as indicators of pathology were not significant predictors of outcome measured as efficiency, achievement of potential, or duration of stay after controlling for other variables. Measures of impairment such as the initial Barthel Index score, the Brunnstrom arm recovery level, hand control, and the motor control score are associated with pathology and in the prediction of outcome are better than indicators of pathology. In statistical terms, the relation between pathology and TABLE 3. outcome is antecedent to the relation between the Barthel Index score and outcome. There is no independent effect of pathology that is not associated with impairment. Controlling for the variables measuring impairment reveals a nonsignificant relation between pathology and outcome. The relation between the initial and discharge Barthel Index scores is not linear due to the maximum discharge score of 100. Patients with high initial scores cannot improve beyond this boundary, so the ratio of improvement relative to the initial score diminishes. Furthermore, patients with very low initial scores tend not to improve greatly, whereas those exhibiting the most improvement appear to be the patients with initial scores of 40-80. 12-13 Discussion The major finding of our study is that all three dependent variables are poorly predicted, with only 17% of the variance in efficiency, only 30% of the variance in achievement of potential, and only 22% of the variance in duration of stay being explained (Table 5). Other research into these dependent variables reveals equally poor, or even worse, prediction, 2-15 - 16 which indicates that there is a lot of unexplained variance that cannot be attributed to the independent variables considered by this and other studies. It is possible that further research into medical,, and demographic factors not considered by this study may explain additional variance. However, unlike many other studies, ours is comprehensive in its determination of important independent variables. 1-3 - 8-10 ' 13 It is therefore likely that the unexplained variance is due to social, personal, and family factors that are unrelated to the process and that affect the decision to refer patients to, especially in a free hospital system. In Australia, any member of the medical team can recommend referral, in contrast to other countries in which recommendation for is the domain of physiatrists. In Australia, Nominal Variables Considered in Predicting Rehabilitation Outcome in 256 Patients With First Stroke Dependent variable Independent nominal variable Marital status Current employment Occupational class (previous) Occupational prestige (previous) Side of paralysis Stroke etiology Site of lesion Arterial distribution Handedness Rehabilitation efficiency 1.04 Achievement of Duration of stay (days) 3.13* 1.17 No discrimination, 86% pensioners 0.09 0.32 3.10* 0.42 0.65 1.31 5.17t 0.76 1.21 0.24 2.57 1.86 2.48* No discrimination, 78% middle cerebral artery territory No discrimination, 96% right 'tp<0.05, <0.01, respectively, different from 1.0. Data are values for F, result of one-way analysis of variance.

244 Stroke Vol 21, No 2, February 1990 TABLE 4. Initial Correlations in Predicting Rehabilitation Outcome in 256 Patients With First Stroke Dependent variable Independent variable Achievement of potential Duration of stay Initial Barthel Index score Age Education Onset-admission interval Admission- interval Bladder control Brunnstrom arm recovery level Brunnstrom hand recovery level Upper-extremity motor control Male sex Ethnicity Visual confrontation Hypertension Myocardial infarction Other ischemic diseases Diabetes mellitus Congestive cardiac failure Peripheral vascular disease Fractured femur Other associated illness Minimum significant correlation is 0.12. Rehabilitation efficiency 0.50-0.46-0.09-0.12-0.11 0.12 0.21 0.19 0.09-0.07-0.13-0.14 0.01-0.07 - the allocation of funding to units on a patient-day basis makes it in the interests of the units not to refuse referrals. Consequently, Australia has one of the highest comprehensive inpatient rates in the world (43%), more than double that in other Western countries. 5 " 9 The mass referral to and extended duration of comprehensive inpatient TABLE 5. Achievement of -0.25 0.09-0.12-0.14 0.29 0.37 0.31-0.11-0.10-0.29-0.19 0.02 Duration of stay (days) -0.37-0.10-0.01 0.10 0.02-0.15-0.31-0.30-0.30 0.11 0.07-0.07-0.00 0.08-0.05 Initial Barthel Index score -0.37 - -0.18 0.04 0.02 0.01 0.68 0.64 0.58 0.63-0.12-0.09-0.45-0.19-0.02 0.10-0.05 0.05 0.05 are constraints on the development of more appropriate health care facilities and overburden the available and nursing services. The objective of comprehensive inpatient implicit in our research is that should attempt to provide the maximum functional and neurologic recovery for all individuals indepen- Results of Regression Analyses Predicting Rehabilitation Outcome in 256 Patients With First Stroke Raw regression coefficients Age Brunnstrom arm recovery level Initial Barthel Index score (Initial Barthel Index score) 2 Myocardial infarction Onset-admission internal Admission- interval Bladder control Peripheral vascular disease Diabetes mellitus Constant Multiple R R 2, not significant. Rehabilitation efficiency -0.01 0.19-0.0004 0.78 0.42 0.17 Achievement of -0.91 6.9-10 -0.49-0.64 1.2 115 0.54 0.30 Duration of stay (days) -0.74-1.04 2.5 22-16 142 0.46 0.22

dent of nonmedical factors such as ethnicity, family support, and architectural barriers. However, medical factors, especially level of impairment, age, and associated illnesses, as shown by this and other research, 12-13 ' 15 limit the expected outcome and success. While nonmedical and/or social factors do contribute to discharge disposition decision, referral to comprehensive inpatient should not be influenced by these factors. These unmeasured and often unmeasurable social factors reduce the ability of a regression equation to explain a large proportion of the variance in each dependent variable. However, the equations still serve to identify the independent variables that are significant in predicting efficiency, achievement of potential, or duration of stay. The large proportion of unexplained variance due to confounding variables could mask the potential impact of other variables, and analysis of data from another population in which the impact of these confounding variables is less may produce different results. Of the considered independent variables (Tables 1, 2, and 3), the only four significant predictors of efficiency are age, Brunnstrom arm recovery level, 18 initial Barthel Index score, and initial Barthel Index score squared (Table 5). Each 10 years of age reduces the efficiency (daily increase in Barthel Index score) by 0.1 unit. Since the average stay is 61 days, this amounts to 6 units during. However, older people are likely to have shorter stays, with each additional 10 years of age reducing the duration of stay by 7 days. The Brunnstrom arm recovery scale comprises six levels. 18 The greater the initial level (degree of arm control), the higher the efficiency, with each higher level increasing efficiency by 0.19 units. A quadratic term was required to model the effect of the initial Barthel Index score. 12 Therefore, it is not possible to state the linear increment in efficiency due to changes in initial score. Since the increment varies with the initial score, the effect must be calculated for each individual score. Six variables significantly contributed to the prediction of achievement of potential. Each 10 years of age reduced the achieved potential by 9%. Every day's delay from stroke onset to admission to the acute-care hospital reduced the achieved potential by almost 0.5%, and each day's delay from admission to commencement reduced it by 0.64%. Each level in the Brunnstrom arm recovery scale was worth 7%, with the presence of myocardial infarction reducing achieved potential by 10%. Bladder control, an item in the Barthel Index and measured in this study on a 5-point scale, also contributed to the prediction, with full bladder control (score of 10 points) increasing achieved potential by 12% compared with a similar patient with no bladder control (score of 0). 19 Shah et al Stroke Rehabilitation Efficiency 245 Five items significantly predicted the duration of stay. Each additional 10 years of age reduced the stay by 7 days. This unexpected finding is related to the fact that older people make fewer gains, so that the benefit from prolonged is less. Older people are also more likely to be recommended for on the basis of nonmedical and non reasons. The initial Barthel Index score was linearly related to the duration of stay, with every additional point reducing the stay by 1 day. Bladder control significantly predicted length of stay; a patient with full bladder control stays 25 days longer than a similar individual with no bladder control. Patients with limited bladder control are also likely to have low initial Barthel Index scores and low achievement of potential. Their stays are brief because their small gains are made in a relatively short time, and they are Likely to benefit little from further. Such patients are often discharged to partial or full care. The presence of peripheral vascular disease and associated complications affected 8% of the study population (Table 2) and increased stay by 22 days due to the increased effort required for maximum benefit to be obtained. The reverse was true for patients with insulin-dependent diabetes. These results differ considerably from those of other studies. 14-16 Nevertheless, identification of variables significant in the prediction of efficiency, achievement of potential, and duration of stay are of paramount importance in promoting the efficient use of existing resources and in developing alternative strategies to maximize quality of life for all stroke patients. References 1. Grcsham GE: Stroke outcome research. Stroke 1986; 17:358-360 2. Lind K; A synthesis of studies on stroke. J Chronic Dis 1982;35:133-149 3. Korner-Bitensy N, Mayo N, Cabot R, Becker R, Coopersmith H: Motor and functional recovery after stroke: Accuracy of physical therapists' predictions. Arch Phys Med Rehabil 1989; 70:95-99 4. Prescott RJ, Garraway WM, Akhtar AJ: Predicting functional outcome following acute stroke using a standard clinical examination. Stroke 1982;13:641-647 5. Dombovy ML, Basford JR, Whisnant JP, Bergstralh EJ: Disability and use of services following stroke in Rochester, Minnesota, 1975-1979. Stroke 1987;18:83O-836 6. Shah S, Bain C: Admissions, patterns of utilization and disposition of acute strokes in Brisbane hospitals. Med J Aust 1989;150:256-260 7. Herman B, Schultz B, Van Lrijk M, Leyton A, Frenken C: Epidemiology of stroke in Tilburg, The Netherlands. Stroke 1980;ll:162-165 8. Dombovy ML, Sandok BA, Basford JR: Rehabilitation for stroke: A review. Stroke 1986;17:363-369 9. Cochrane A: Burden of cerebrovascular disease. Br Med J 1970;2:165-171 10. Alder MK, Brown CC, Acton P: Stroke Is age a determinant? J Am Geriatr Soc 1980;28:499-503 11. Osberg JS, DeJong G, Haley SM, Seward M, McGinnis GE, Germaine G: Predicting long term outcome among post stroke patients. Am J Phys Med 1988;68:94-103

246 Stroke Vol 21, No 2, February 1990 12. Shah S, Vanclay F, Cooper B: Predicting discharge status at commencement of stroke. Stroke 1989; 20:766-769 13. Jongbloed L: Prediction of function after stroke: A critical review. Stroke 1986;17:765-776 14. Vanclay F, Shah S: Validity of outcome following stroke. Arch Neurol (in press) 15. Heinemann AW, Roth EJ, Cichowski K, Betts HB: Multivariate analysis of improvement in outcome following stroke. Arch Neurol 1987;44:1167-1172 16. Granger CV, Hamilton BB, Gresham GE, Kramer AA: The stroke outcome study: Part 2. Relative merits of the total Barthel Index score and a four-item subscore in predicting patient outcomes. Arch Phys Med Rehabil 1989; 70:100-103 17. World Health Organization: International Classification of Diseases, Vol 1. Geneva, World Health Organization, 1977 18. Brunnstrom S: Movement Therapy in Hemiplegia. New York, Harper & Row Publishers Inc, 1970 19. Shah S, Vanclay F, Cooper B: Improving the sensitivity of the Barthel Index in stroke. J Clin Epidemiol 1989; 42:703-709 20. SPSS Inc: SPSS-X Users Guide, ed 3. Chicago, SPSS Inc, 1988 KEY WORDS Australia cerebrovascular disorders