IN STROKE REHABILITATION, it is common practice for. Selecting Patients for Rehabilitation After Acute Stroke: Are There Variations in Practice?

Transcription

1 788 ORIGINAL ARTICLE Selecting Patients for Rehabilitation After Acute Stroke: Are There Variations in Practice? Paul A. Ilett, BAppSc (Physio), Kim A. Brock, PhD, Christine J. Graven, PostGradDip (Health Research Methodology), Susan M. Cotton, PhD ABSTRACT. Ilett PA, Brock KA, Graven CJ, Cotton SM. Selecting patients for rehabilitation after acute stroke: are there variations in practice? Arch Phys Med Rehabil 2010;91: Objective: To investigate whether there were variations in practice in selection for rehabilitation after stroke, after adjustment for case mix. Design: Prospective multicenter audit. Setting: Seven acute stroke units in metropolitan and regional Victoria, Australia. Participants: Consecutive acute stroke admissions (N 616). Interventions: None. Main Outcome Measures: Mobility Scale for Acute Stroke Score and Modified Barthel Index (MBI) scores for continence at day 3 poststroke, discharge destination from the acute hospital. Results: Data were analyzed for 616 stroke survivors. Considerable variability in the percentage of cases accessing inpatient rehabilitation was observed in severe stroke (27% 67%) and mild stroke (27% 73%). To assess adjustment for case mix, a multinomial logistic regression was conducted with the outcome variable being discharge destination (home, rehabilitation, or nursing home), and the predictors being Mobility Scale for Acute Stroke Score, MBI continence scores, age, and social situation. The overall amount of variability explained in discharge destination by the predictors was 63% (Nagelkerke pseudo R 2 ). The regression analysis was repeated, adding unit code as a predictor. Unit code was a significant contributor to the model (P.01). Conclusion: The results of the study indicate that, after adjusting for case mix, there may be variations in practice in selection for rehabilitation leading to inequities of access. Key Words: Diagnosis-related groups; Health services accessibility; Prognosis; Rehabilitation; Stroke by the American Congress of Rehabilitation Medicine IN STROKE REHABILITATION, it is common practice for rehabilitation units to send an assessor to the acute hospitals to evaluate the potential of individual patients with stroke to benefit from rehabilitation and determine whether they should From St. Vincent s Hospital, Melbourne (Ilett, Brock, Graven), and Department of Psychiatry, University of Melbourne, Carlton, (Cotton), Australia. Presented to the Australasian Faculty of Rehabilitation Medicine, May 2, 2006, Cairns, Australia, and to the Stroke Society of Australasia, September 6 10, 2009, Cairns, Australia. No commercial party having a direct financial interest in the results of the research supporting this article has or will confer a benefit on the authors or on any organization with which the authors are associated. Correspondence to Kim A. Brock, PhD, Physiotherapy Dept, St Vincent s Hospital, PO Box 2900, Fitzroy 306, Victoria, Australia, kim.brock@svhm.org.au. Reprints are not available from the author /10/ $36.00/0 doi: /j.apmr be offered inpatient rehabilitation. While there is a considerable body of literature investigating prognostic factors for recovery poststroke, 1-3 there is no widely accepted model for objectively determining patients potential to improve and recover sufficient independence to return home and/or their potential to benefit from rehabilitation. 4 Key prognostic factors that have been identified include age; previous functional ability; initial disability poststroke; various neurologic impairments including conscious state, continence, sitting balance, and severity of paralysis; and social support. 2,,6 The decision of the rehabilitation assessor must be informed by the literature regarding prognostic factors but is necessarily multifactorial, taking into account many aspects of the patient s presentation. 4,7 In selection for rehabilitation, a balance should be maintained between people who will clearly benefit from rehabilitation and those who may benefit. 4,8 If only those who have a high likelihood of benefiting are admitted, then many who may have benefited will miss out. If every patient with stroke is admitted to rehabilitation, including all those with very severe stroke, then a significant number of episodes of rehabilitation will not deliver a functional benefit. Some studies investigating prediction of functional abilities or discharge home have demonstrated good sensitivity (ie, those who are predicted to have a good recovery were very likely to do so), but only moderate specificity (ie, those predicted not to have a good recovery or return home often did better than expected) Because it is sometimes difficult to discern at this early stage a patient s potential to improve, it is important to ensure that those patients with a less positive initial prognosis are not excluded by the system. 12 At the same time, it is important to avoid admitting patients with little likelihood of benefit (eg, those with very poor prognosis and minimal potential for improvement and those who have very mild stroke and are likely to recover full function with outpatient-based interventions only). 8 It has been recognized that the process of selection for rehabilitation may lead to inequities of access for patients with stroke. 13 This is particularly pertinent where funding models for rehabilitation are case mix based and likely to favor certain patient groups over others 14 (eg, mild or moderate stroke compared with severe stroke). Given the complexity of the decision-making process in selection for rehabilitation, it is possible that access to rehabilitation may be variable across health services. Evidence for variation in rate of discharge from the acute hospital setting to nursing home care after stroke, after adjusting for case mix, has been demonstrated by Portelli et al. 1 Evidence for significant variability in the case mix of patients with stroke admitted to rehabilitation units has been shown by Putman et al. 16 In this latter international study, nonpatient-related factors such as funding BI MBI List of Abbreviations Barthel Index Modified Barthel Index

2 SELECTING PATIENTS FOR REHABILITATION, Ilett 789 models played a significant role. Our interest was to examine selection for rehabilitation where the funding model was uniform across all units. Most stroke rehabilitation in Victoria is government-funded, using a funding model with 2 categories for strokebased on-admission BI score (0 9, ), with higher per-day payment for shorter stay patients and a lower rate of payment for those staying beyond specified periods. The funding model is designed to reward efficient practice financially, resulting in shorter length of stay, while still providing a lower level of reimbursement to allow a longer length of stay for those patients whose circumstances (clinical or social) require it. When investigating variations in practice between health services, it is necessary to control for patient case mix, because simply considering descriptive data may lead to invalid conclusions. 1,17,18 We reviewed the literature for studies predicting discharge destination from the hospital, using patient characteristics indicative of prognosis after stroke. Three studies have investigated prediction of discharge destination (home, rehabilitation, or nursing home) from the acute setting. Mauthe et al 19 used 6 items from the FIM (bathing, bowel management, toileting, social interaction, dressing lower body, eating), scored at 3 days poststroke, with 70% accuracy of prediction. Rieck and Moreland 20 achieved 8% accuracy of prediction of discharge destination using the Orpington Prognostic Scale (measuring motor function and cognition) scored at 7 days poststroke and 6% accuracy when scored at 14 days poststroke. Unsworth 8 demonstrated 74.9% accuracy of prediction using FIM items (bowel management, stairs, dressing upper body, expression, social interaction) scored 3 days prior to discharge. Inclusion of social situation, instrumental activities of daily living, and premorbid housing and cognitive status improved accuracy to 79.4%. A number of studies have investigated prediction of discharge destination (home vs residential care) from inpatient rehabilitation, with accuracy of prediction levels of 7% to 87%, 10,11,21,22 and/or the amount of variation explained ranging between 21% and 67%. 10,23,24 All models used either a functional status measure (FIM) or motor function measure (Chedoke McMaster Stroke Assessment, Motor Assessment Scale, Berg Balance Scale), and most measured social support. In a preliminary study for this project, the accuracy of prediction of discharge destination from the acute hospital was investigated using 3 measures: the Mobility Scale (Mobility Scale for Acute Stroke), 2 the BI, 26 and the MBI, 27 assessed at day 3 poststroke. 28 The study took place in an acute stroke unit of a tertiary teaching hospital, with 86 participants. The accuracy of prediction of discharge destination was above 7% for all scales, with the most accurate classification being 83% for the combination of Mobility Scale scores and the MBI bowel item. The primary purpose of the current study was to investigate whether there are variations in practice in selection for rehabilitation. The study focuses on both access to rehabilitation for patients with more severe stroke and use of inpatient rehabilitation resources for those with more mild stroke. In order to adjust for case mix, level of function poststroke was measured using the variables tested in the preliminary study (unpublished) described above (Mobility Scale and MBI continence). Prior to investigating variations in practice in selection for rehabilitation, we assessed the suitability of these variables to control for case mix in a multicenter study by examining the overall accuracy of prediction of discharge destination. The aims of the study were to (1) investigate accuracy of prediction of the indicators developed in the pilot study in a multicenter study, and (2) investigate whether there were variations in practice in selection for rehabilitation at different acute hospitals. There were 2 hypotheses: (1) that the variables used in the preliminary study would be accurate predictors of discharge destination in the multicenter study, and (2) that, after adjusting for case mix, the acute hospital unit would be a significant variable in prediction of discharge destination. METHODS Study Design This was a prospective, observational cohort study, conducted as a benchmarking project. Setting The setting was 7 acute stroke services in large, tertiary referral hospitals in metropolitan Melbourne and 1 stroke service in a large regional hospital in Victoria. All units admitted more than 100 patients with stroke a year. Ethics This project was conducted as a multisite benchmarking prospective audit in which deidentified data were provided to a central location by the facilities. Because all data in the main study were anonymous and obtained from routine clinical care, this was deemed by the ethics committees of the participating units not to require informed consent. Participants Data were collected from a cohort of consecutive patients admitted with the primary diagnosis of stroke (including cerebral infarct, intracerebral hemorrhage, intraventricular hemorrhage, and subarachnoid hemorrhage). Day of stroke was classed as day 0. Patients with previous strokes were included in this project. Patients were excluded in the following circumstances: admitted from residential care, admitted with another primary illness or incident, stroke was not the main cause of disability, had a stroke onset more than 3 days prior to admission, or symptoms had resolved by day 3. Only data from patients who survived were analyzed. Data from patients discharged from one acute hospital to another acute hospital were not included. Each participating center undertook to provide data on 100 prospective consecutive patients. Measures The scale used in the preliminary study, the Mobility Scale, 2 was selected as an indicator of motor function at the early poststroke. This scale rates the level of assistance required for the patient to do 6 simple mobility tasks: bridging, sitting up from lying, balanced sitting, sit to stand, balanced standing, and walking (appendix 1). The rating scale has 6 levels, from unable to do to independent. Minor changes were made to items of the Mobility Scale to facilitate ease of use in the very acute patient. Three tasks (bridging, sitting from supine, sit to stand) were performed once rather than 3 times, and the balanced sitting item was tested for 1 minute rather than 3 minutes. The sit to stand item was performed from the hospital bed rather than a chair. Intrarater, interrater, and test-retest reliability have been demonstrated for the Mobility Scale, with weighted kappa ratings of greater than.7, greater than.7, and.64 to.88, respectively, representing fair to excellent levels of agreement. 2 The Mobility Scale has been used at 2 weeks poststroke to predict length of stay in rehabilitation, explaining

3 790 SELECTING PATIENTS FOR REHABILITATION, Ilett 69% of the variance in length of stay. 29 The scale was found to have a high level of concurrent validity when the total score was correlated with the total scores of the Motor Assessment Scale, Functional Ambulation Classification, FIM, and BI. 30 The items of the Mobility Scale are routinely assessed in a basic physiotherapy assessment and do not involve extra tests beyond normal physiotherapy practice when getting the patient with early stroke out of bed. The Mobility Scale was preferred over other functional scales because the assessment could be performed readily at the bedside and did not require any additional assessment of the patient above usual care. Continence was measured using the MBI scores for bowel and bladder management. 27 Basic demographic and medical details including patient s age, sex, pathology (infarct or hemorrhage), social situation (lived alone, lived with family, hostel), length of acute hospital stay, and discharge destination were collected. All data were collected on day 3 poststroke, except where this fell on a weekend. In the latter case, data were collected on day4or. Data Analysis Multinomial logistic regression was used to develop a model for predicting discharge destination. Descriptive analysis of discharge destination used data grouped according to Mobility Scale score. The data were divided into 3 groups (mild, moderate, severe) with each group approximately one third of the total cases. Finally, to investigate variations in practice, unit code was entered into the regression equation. RESULTS Data were collected on 700 consecutive cases. After excluding those who did not survive or were discharged to another acute hospital, 616 cases remained. The mean age SD of the cohort was years, ranging from 22 to 98 years. Fifty-three percent (n 326) of the patients were men. Seventy-nine percent (n 489) had an infarct, and 18% (n 113) had a hemorrhage. Left-sided lesions were present in 49% (n 297) of cases, right-sided lesions in 46% (n 277), and bilateral lesions in % (n 31). Prior to their stroke, 29% (n 174) had lived at home alone, 6% (n 399) at home with family, and 6% (n 37) in supported accommodation or hostel. Mobility Scale scores were categorized into Table 2: Accuracy of Prediction of Discharge Destination Observed Predicted Home Rehabilitation Nursing Home Correct (%) Home (n) Rehabilitation (n) Nursing home (n) Overall (%) mild, moderate, and severe physical disability, with approximately one third of the total sample in each group, yielding the following cutoff scores: mild (31 36), moderate (16 30), and severe (6 1). Discharge destination from the acute hospital was home for 21.3% (n 128), rehabilitation for 60.3% (n 363), and nursing home (or interim care awaiting nursing home) for 18.4% (n 111). Predictive Validity Multinomial logistic regression was conducted with the outcome variable discharge destination (home, rehabilitation [reference category], nursing home) and the predictors Mobility Scale, social situation prior to stroke (lived alone, lived with family, hostel), MBI bowel, MBI bladder, and age. Of these, Mobility Scale score was the only significant predictor of discharge home versus rehabilitation. Mobility Scale score, MBI bowel score, age, and living with family contributed significantly to prediction of rehabilitation versus nursing home. The overall amount of variability in discharge destination explained by the predictors was 63% (Nagelkerke pseudo R 2 )(table 1). Overall the model accurately predicted 61.4% of those discharged home, 81.9% of those discharged to rehabilitation, and 74.6% discharged to a nursing home (table 2). The overall percentage of accurate prediction using the model was 7.6%. Variations in Practice Figures 1 to 3 show the percentage of patients discharged to home, rehabilitation, or nursing home for each hospital, according to severity of disability. Considerable variability was Table 1: Logistic Regression Predicting Discharge Destination Variable B Significance* Nursing home versus rehabilitation Intercept Mobility Scale Score Age MBI bowel MBI bladder Lives with family Lives alone Rehabilitation versus home Intercept Mobility Scale Score Age MBI bowel MBI bladder Lives with family Lives alone *Derived from the Wald test. Count (n=30) 2 (n=22) 3 (n=30) 4 (n=40) Unit Code (n=20) 6 (n=30) 7 (n=23) Discharge destination Home Rehabilitation Nursing Home Fig 1. Discharge destination for patients with severe disability (total mobility score, 6 1).

4 SELECTING PATIENTS FOR REHABILITATION, Ilett 791 Count (n=3) 2 (n=3) 3 (n=22) 4 (n=27) Unit Code (n=31) 6 (n=27) 7 (n=28) Discharge destination Home Rehabilitation Nursing Home Fig 2. Discharge destination for patients with moderate disability (16 30). observed for severe cases (see fig 1), with the percentage discharged to rehabilitation in the 7 centers varying from 27% to 70%, with a similar pattern for discharge to nursing home (27% 67%). Average length of stay for the severe group varied from 20 to 38 days, with the overall mean SD days. Figure 2 shows the pattern was more consistent in moderate stroke with between 79% and 97% going to rehabilitation. This group had an average length of stay between 10 and 1 days, with the overall mean SD days. The variability was also marked for mild stroke, with the percentage discharged to rehabilitation varying from 27% to 73% (see fig 3), with almost all of the remaining cases discharged home. Average length of stay for the mild group varied between 7 and 11 days, with the overall mean SD 9. days. The multinomial logistic regression was repeated, adding the variable of unit code. The overall amount of variability in outcome discharge destination explained by the predictors was 67% (Nagelkerke pseudo R 2 ). Discharge to home versus rehabilitation was predicted by total Mobility Score and unit codes 2 and 3 (less likely to discharge home, P.01). Differences between discharge to a nursing home versus rehabilitation was predicted by age, Mobility Scale score, MBI bowel, unit code 1 (less likely to discharge to nursing home, P.01), and living with family. The overall percentage of accurate prediction using the model was 76.7%. As unit code was a significant predictor of discharge destination, the results suggest that there may be variations in practice in selection for rehabilitation. DISCUSSION Access to rehabilitation is a very important issue for optimizing outcomes for people with stroke. If patients who may significantly benefit from rehabilitation are denied rehabilitation, this may potentially diminish the patient s outcome and result in costly higher care needs for the remainder of the patient s life. This study has demonstrated that there may be differences in practice in selection for rehabilitation, affecting equity of access. The results of the study support our initial hypothesis, demonstrating that, after adjusting for case mix, the acute hospital unit was a significant variable in prediction of discharge destination. This also suggests that there may be case mix differences in the rehabilitation services, with some services taking patients with more severe strokes and others having a higher number of patients with milder strokes. These findings are in agreement with the findings of Portelli 1 and Putman. 16 It is possible that the funding model for stroke rehabilitation may have influenced the process of selection for rehabilitation. This effect is known as skimming, 14 in which those patients likely to be financially rewarding to the health care provider are preferentially selected. The case mix based funding model in Victoria has 2 categories, 1 for moderate and severe stroke (BI score 60) and 1 for mild stroke. Within the model, there are incentives for achieving shorter lengths of stay for each category (higher per diem payment for short stay cases, lower per diem payment for long stay cases). This model may have resulted in facilities limiting the number of patients with severe stroke, who may perhaps be difficult to discharge in a shorter time. In this case, those who may have benefited from rehabilitation may not be offered this option. Equally, centers may be admitting those with mild stroke who perhaps could have been treated as outpatients, achieving short lengths of stay and enhanced financial performance. Another facility specific factor that may influence selection for rehabilitation is the availability of rehabilitation resources, including the number of inpatient rehabilitation beds and outpatient based resources, such as early supported discharge programs, rehabilitation in the home, or intensive therapy in ambulatory settings. With severe stroke, the availability of interim care for those patients who have a trial of rehabilitation without significant benefit can greatly affect the ability of a rehabilitation unit to accept these patients. 4 In summary, selection for rehabilitation may have been influenced by 1 or more of these facility-based factors: concern for financial performance of the rehabilitation facility, availability of either inpatient-based or outpatient-based rehabilitation resources, and/or availability of alternative discharge destinations from rehabilitation when return home is not possible. There may also be other facility-based factors not considered in this article. The results of the study also support our hypothesis that Mobility Scale scores, age, continence, and social support are accurate predictors of discharge destination. The overall accu- Count (n=41) 2 (n=31) 3 (n=33) 4 (n=19) Unit Code (n=23) 6 (n=29) 7 (n=40) Discharge destination Home Rehabilitation Nursing Home Fig 3. Discharge destination for patients with mild disability (31 36).

5 792 SELECTING PATIENTS FOR REHABILITATION, Ilett racy of prediction of the model is higher than described in previous studies predicting discharge destination from the acute hospital in the first week after stroke. 19,20 These results support the validity of using the model to adjust for case mix with regard to discharge destination. The Mobility Scale, a simple measure of motor function, can be scored from a routine physiotherapy assessment during the first few days poststroke. Age, continence, and social support can be obtained from the medical record. This data can be readily collected in the acute setting to enable adjustment for case mix in ongoing benchmarking activities. Study Limitations The major limitation in this study is the degree to which patient characteristics can be accounted for when comparing one service with another. While the amount of variation in discharge destination explained is relatively high (63%) compared with other studies, 10,23,24 a large amount of the variation remains unexplained. It is possible that the differences between discharge practices could be explained by other patient characteristics such as cognition, communication, or premorbid function. However, with regard to cognition, the accuracy of prediction achieved in this study is higher than in the study by Rieck and Moreland, 20 which included cognition. Other factors not investigated include availability of home support and patient and caregiver choice. Two patient groups are likely to be underrepresented in this study. Patients with very severe stroke, either in intensive care or for whom the decision was made very early to take a palliative approach, may not have had routine neurologic physiotherapy assessment at day 3. Patients with very minimal or no mobility dysfunction may be underrepresented in some units. There was considerable variability in the number of cases with perfect Mobility Scale scores in each facility. Discussion with the participating units revealed that there were variations in practice regarding referral to physiotherapy. Some units had blanket referral, in which all patients admitted with stroke were screened by the physiotherapist, resulting in more cases with perfect Mobility Scale scores. In some units, physiotherapists assessed only those patients directly referred by medical or nursing staff. However, both the patients with very severe stroke and the patients with very mild stroke are generally less likely to be selected for inpatient rehabilitation. We do not have data about the achievements of the inpatient rehabilitation episode to comment on the appropriateness or otherwise of the selection process. This study can identify only that there was a difference and that this difference is worthy of further exploration. While this study has produced another model for predicting discharge destination, we join with others in the literature 4,7,1,20 in cautioning against using this model in a deterministic manner to decide who should be offered a place in rehabilitation. We may never have a model sufficiently robust to measure accurately factors such as motivation, the degree of family support, and the level of function prior to the stroke. Two patients could present with similar stroke deficits and identical scores in the predictive model. However, if one is highly motivated, exercises regularly, and has extensive family support, this patient may achieve much more than a patient with similar scores with lower previous activity levels and a less positive attitude. Rather than using this model to determine who should have rehabilitation, we recommend that it is used to benchmark clinical practice, to monitor equity of access, and to identify variations that require further evaluation. If a rehabilitation unit is successfully treating those with severe stroke who may not have been accepted in another unit, this then should become the benchmark for rehabilitation of severe stroke. Likewise, if high-quality community resources can meet the needs of patients with mild stroke in their homes to the patient s and carer s satisfaction and at a lower cost, then this should be identified and appropriately resourced. CONCLUSION This project has demonstrated that, in acute stroke, discharge destination from the acute hospital (home, rehabilitation, or nursing home) can be predicted with reasonable accuracy, using data that are routinely available from simple clinical assessments. In this benchmarking project, the facility the person was admitted to had a significant effect on the discharge destination, suggesting variations in practice in selection for rehabilitation that require further exploration. Acknowledgments: We thank the Neurology Special Interest Group of the Australian Physiotherapy Association, Victorian Branch, for their active support of this project. We also thank the dedicated physiotherapy staff of the stroke services of the participating hospitals for agreeing to take part in this study. APPENDIX 1: MOBILITY SCALE FOR ACUTE STROKES (MODIFIED VERSION) Items 1. Bridging from supine: bend knees, buttocks clear of the bed, return to supine. 2. Sitting from supine: let patient choose side, legs over edge of bed. 3. Balanced sitting for 1 minute. Maximum base of support, defined as thighs in contact with the bed, legs flexed at right angles to thighs, feet supported on stool/floor at right angles to leg. 4. Sitting to vertical stand: with full extension of hips and knees from a hospital bed.. Balanced standing with full extension of hips and knees for 1 minute. Only assess standing, not sit to stand. Record overall assistance provided for the duration of the activity. 6. Gait: walking 10m indoors on level ground, with or without a gait aid. Record overall assistance provided for the duration of the activity. Rating 1. Unable to do the activity: patient makes no contribution to the activity or is unable to complete the activity. 2. Maximum assistance of 1 or 2 people: patient makes minimal contribution to the activity. 3. Moderate assistance of 1 person: hands-on for most of the activity. Patient is able to perform part of the activity independently. 4. Minimal assistance: hands on for part of the activity.. Supervised: verbal input, no hands on, physiotherapist prepared to give assistance. 6. Independent and safe: no verbal input. References 1. Counsel C, Dennis M. Systematic review of prognostic models in patients with acute stroke. Cerebrovasc Dis 2001;12: Kwakkel G, Wagenaar RC, Kollen B, Lankhorst GJ. Predicting disability in stroke a critical review of the literature. Age Ageing 1996:2; Huybrechts KF, Caro JJ. The Barthel Index and modified Rankin Scale as prognostic tools for long-term outcomes after stroke: a

6 SELECTING PATIENTS FOR REHABILITATION, Ilett 793 qualitative review of the literature. Curr Med Res Opin 2007:23; Wade DT. Selection criteria for rehabilitation services. Clin Rehabil 2003;17: Konig IR, Ziegler A, Bluhmki E, et al. Predicting long-term outcome after acute ischemic stroke: a simple index works in patients from controlled clinical trials. Stroke 2008;39: Counsel C, Dennis M, McDowell M. Predicting functional outcome in acute stroke: comparisons of a simple six variable model with other predictive systems and informal clinical prediction. J Neurol Neurosurg Psychiatry 2004;7: Dennis M. Predictions models in acute stroke: potential uses and limitations. Stroke 2008;39: Unsworth CA. Selection for rehabilitation: acute care discharge patterns for stroke and orthopaedic patients. Int J Rehabil Res 2001;24: Kollen B, Kwakkel G, Lindeman E. Longitudinal robustness of variables predicting independent gait following severe middle cerebral artery stroke: a prospective cohort study. Clin Rehabil 2006;20: Brauer SG, Bew PG, Kuys SS, Lynch MR, Morrison G. Prediction of discharge destination after stroke using the Motor Assessment Scale on admission: a prospective, multisite study. Arch Phys Med Rehabil 2008;89: Wilson DB, Houle DM, Keith RA. Stroke rehabilitation: a model predicting return home. West J Med 1991;14: Teasell RW, Foley NC, Bhogal SK, Chakravertty R, Bluvol A. A rehabilitation program for patients recovering from severe stroke. Can J Neurosci 200;32: Stineman MG. Casemix measurement in medical rehabilitation. Arch Phys Med Rehabil 199;76: Duckett S, Gray L, Howe A. Designing a funding system for rehabilitation services, part 2: policy objectives and options for achieving efficiency and quality of care. Aust Health Rev 199; 18: Portelli R, Lowe D, Irwin P, Pearson M, Rudd AG. Institutionalization after stroke. Clin Rehabil 200;19: Putman K, De Wit L, Schupp W, et al. Inpatient stroke rehabilitation: a comparative study of admission criteria to stroke rehabilitation units in four European centres. J Rehabil Med 2007;39: Lingsma HF, Dippel DW, Hoeks SE, et al. Variation between hospitals in patient outcome after stroke is only partly explained by differences in quality of care: results from the Netherland Stroke Survey. J Neurol Neurosurg Psychiatry 2008;79: Weir N, Dennis MS. Towards a national system for monitoring the quality of hospital based stroke services. Stroke 2001;32: Mauthe RW, Haaf DC, Hayn P, Krall JM. Predicting discharge destination of stroke patients using a mathematical model based on six items from the Functional Independence Measure. Arch Phys Med Rehabil 1996;77: Rieck M, Moreland J. The Orpington Prognostic Scale for patients with stroke: reliability and pilot predictive data for discharge destination and therapeutic services. Disabil Rehabil 200;27: Oczkowski WJ, Barreca S. Neural network modeling accurately predicts the functional outcome of stroke survivors with moderate disability. Arch Phys Med Rehabil 1997;78: Smith PM, Ottenbacher KJ, Cranley M, et al. Predicting follow-up living setting in patients with stroke. Arch Phys Med Rehabil 2002;83: Agarwal V, Mc Rae MP, Bhardwaj A, Teasell RW. A model to aid in the prediction of discharge location for stroke rehabilitation patients. Arch Phys Med Rehabil 2003;84: Wee JY, Hopman WM. Stroke impairment predictors of discharge function, length of stay, and discharge destination in stroke rehabilitation. Am J Phys Med Rehabil 200;84: Simondson J, Goldie P, Brock K, Nosworthy J. The Mobility Scale for Acute Stroke Patients: intra-rater and inter-rater reliability. Clin Rehabil 1996;10: Mahoney F, Barthel D. Functional evaluation: the Barthel Index. Md State Med J 196;2: Shah S, Vanclay F, Cooper B. Improving the sensitivity of the Barthel Index for stroke rehabilitation. J Clin Epidemiol 1989;42: Graven C. The use of disability scales to predict discharge destination from acute hospital following stroke [graduate diploma thesis]. Melbourne: La Trobe University; Brock K, Robinson P, Simondson J, Goldie P, Nosworthy J, Greenwood K. Prediction of length of hospital stay following stroke. J Qual Clin Pract 1997;17: Simondson J, Goldie P, Greenwood KM. The Mobility Scale for acute stroke patients: concurrent validity. Clin Rehabil 2003;17: 8-64.