EVIDENCE OF THE BENEFITS of medical rehabilitation

Transcription

1 100 Course of Functional Improvement After Stroke, Spinal Cord Injury, and Traumatic Brain Injury Rita K. Bode, PhD, Allen W. Heinemann, PhD ABSTRACT. Bode RK, Heinemann AW. Course of functional improvement after stroke, spinal cord injury, and traumatic brain injury. Arch Phys Med Rehabil 2002;83: Objective: To examine functional improvement patterns of persons with stroke, traumatic brain injury (TBI), and spinal cord injury (SCI). Design: Statistical analysis of data from a multisite study evaluating rehabilitation outcomes. Setting: Eight inpatient rehabilitation facilities. Participants: A total of 314 consecutive admissions of persons with stroke, SCI, and TBI who received acute medical rehabilitation between 1994 and Intervention: Calibration of motor and cognitive items from the FIM instrument, grouping of cases by number of weeks of rehabilitation (length of stay [LOS] groups), and plotting of weekly averages across time. Main Outcome Measures: Weekly motor and cognitive functional status. Results: With the exception of cognitive functioning for persons with SCI, LOS was related to initial functional status, with patients with greater disability having longer LOS (eg, initial motor status for persons with stroke was 48.3 for those with a 2-week stay, 36.8 for a 6-week stay, with the averages between decreasing monotonically). With the exception of cognitive gains for person with TBIs, the amount of functional gain during rehabilitation was essentially the same for all LOS groups (eg, the overall average total motor gain for persons with SCI is 22.3, with no patterns of increase or decrease across LOS groups); however, the rate of improvement in motor (but not cognitive) functioning differed across LOS groups, with patients with shorter stays having the greater rates of improvement (eg, the overall average weekly motor gain for persons with SCI was 3.6, with the averages by LOS group monotonically decreasing from 6.4 for those with 4-week stays to 2.7 for those with 9-week stays). Conclusions: When examined separately for persons grouped by LOS, functional status improved linearly during the rehabilitation stay, with differences in rate of improvement depending on initial functional status. Key Words: Brain injuries; Length of stay; Rehabilitation; Stroke; Spinal cord injuries; Treatment outcome. From the Department of Physical Medicine and Rehabilitation, Northwestern University Medical School, and the Rehabilitation Institute of Chicago, Chicago, IL. Accepted in revised form December 18, Supported by a Rehabilitation Research and Training Center Grant, National Institute of Disability and Rehabilitation Research, US Department of Education (grant no. H133B30041), the Ralph and Marian Falk Medical Research Trust, and the American Occupational Therapy Association. Presented in part at the American Congress for Rehabilitation Medicine s annual meetting, September 1998, Seattle, WA. No commercial party having a direct financial interest in the results of the research supporting this article has or will confer a benefit upon the author(s) or upon any organization with which the author(s) is/are associated. Reprint requests to Rita K. Bode, PhD, Rehabilitation Institute of Chicago, 345 E Superior St, Rm 1378, Chicago, IL /02/ $35.00/0 doi: /apmr by the American Congress of Rehabilitation Medicine and the American Academy of Physical Medicine and Rehabilitation EVIDENCE OF THE BENEFITS of medical rehabilitation is documented typically in terms of the reduction in disability of persons receiving inpatient rehabilitation. Services such as the Uniform Data System for Medical Rehabilitation (UDSMR) and Caredata.com (formerly MediRisk) not only provide an instrument to measure disability, the FIM instrument, 1 but also a means of comparing an institution s performance in changing disability level from admission to discharge with other, similar facilities. Such national data systems provide information on average admission status, discharge status, and admission-to-discharge gains across facilities. By using the UDSMR database, the relationship between functional status at admission and gains during rehabilitation for patients in various impairment groups has been examined. 2 The findings indicated that both discharge functional status and length of stay (LOS) were related to admission functional status. Discharge status was positively related to admission status (patients with greater disability at discharge had greater disability at admission); and LOS was negatively related to functional status at admission (patients with greater disability at admission had longer rehabilitation stays). These findings confirmed previous research that used functional status at admission to predict discharge status, functional gains, and LOS. 3-9 More recently, functional status at admission, patterns of functional gain, and LOS have been used to classify patients into function-related groups, 10,11 and to explore their use in a prospective payment system (PPS) Differences by impairment group in functional status at admission to inpatient rehabilitation, by using the FIM instrument and in lengths of rehabilitation stays, were found in multisite study of rehabilitation outcomes. 15 For persons with spinal cord injury (SCI), motor status at admission tended to be lower and cognitive status tended to be higher than for persons with either stroke or traumatic brain injury (TBI). Also, the rehabilitation stays for persons with SCIs were longer (approximately 8wk) than for persons with stroke and TBI (approximately 4wk). All of these analyses, however, only examined functional status at admission and discharge. Although admission-to-discharge gain provides information on the amount of overall improvement during rehabilitation, it does not describe patterns of change 16 or identify correlates of change. 17 To examine patterns of change, functional status needs to be assessed at least 3 times during the course of treatment. 18,19 Unlike other published studies in this area, periodic ratings were collected as part of the aforementioned multisite study. 15 By using these data, it was possible to examine weekly improvements in functional status and the rate of improvement over time. The ultimate goal of these analyses was to develop models of improvement that could predict functional status at discharge and to answer such questions as, For which groups of patients do initial status differ? and Is the rate of improvement the same for all patients?

2 COURSE OF FUNCTIONAL IMPROVEMENT, Bode 101 Table 1: Sample Demographics Stroke SCI TBI Gender (men) % Payer (private; non-medicare/medicaid) % No. of cases Mean SD Mean SD Mean SD Age (yr) LOS (d) Admission FIMmotor score* Admission FIMcognitive score* Discharge FIMmotor score* Discharge FIMcognitive score* Abbreviation: SD, standard deviation. * FIMmeasures are reported on a scale ranging from 0 (totally dependent in all activities) to 100 (completely independent in all activities). Unpublished results of analyses conducted on data collected early in this multisite study provided a clearer picture of the relationships between initial status, improvement, and LOS. We identified obstacles to the modeling of functional improvement, such as differences in initial status and LOS across impairment groups, and differences in rates of improvement by LOS. Our study continues this investigation by following motor and cognitive function as measured by the FIM instrument during inpatient rehabilitation and examining improvement patterns by LOS and impairment group. METHODS Sample The total sample consisted of 314 consecutive admissions to 8 inpatient medical rehabilitation programs between 1994 and The 8 sites consisted of facilities in the UDSMR database that agreed to participate in a study of rehabilitation outcomes; as such, they are not necessarily representative of all rehabilitation facilities. The sample consisted of 149 persons with stroke, 88 persons with SCI, and 77 persons with TBI who met the inclusion criteria (eg, 18yr of age, nonaphasic, first admission to rehabilitation) and were willing to participate in the study. Examination of the LOS distributions by impairment group showed atypical LOSs for small groups of patients ( 2wk or 6wk for persons with stroke, 4wk or 9wk for persons with SCI, 2wk or 5wk for persons with TBI). When data were aggregated by LOS group and their average weekly functional status ratings computed, the averages for patients with atypical LOS showed the same trend as for persons with typical LOS persons with longer LOSs tended to have lower functional status at admission. Although changes in functional status for the patients with atypical LOS are interesting, the number of persons with atypical stays was small. To avoid instability in the estimation of average functional status ratings, these persons were eliminated from the sample used in this study. As a result, the sample consisted of 129 persons with stroke, 52 persons with SCI, and 60 persons with TBI. The sample demographics are presented in table 1. Other than the calibrations of motor and cognitive function and the comparison of our sample composition to that in the UDSMR database analysis (table 2), all the results of this study are based on the sample excluding cases with atypical LOS. Instruments Motor and cognitive functional status was rated at admission and weekly until discharge by certified raters by using the FIM instrument. 1 Previous research has shown this instrument is a reliable and valid measure of disability 20 and one that can be used to measure reduction in disability. 21 Analysis The motor and cognitive functional status ratings were calibrated to produce equal-interval measures by using the rating scale (Rasch) model 22,23 and the WINSTEPS computer program. 24 Equal-interval measures are essential to measure change accurately. 25,26 Different raters and time intervals between ratings were sometimes used for the admission, discharge, and the weekly ratings. Admission and discharge status are typically rated by care managers or in team meetings as dictated by the outcomes management system at each of the participating facilities; admission ratings are conducted within the first 3 days and discharge ratings are conducted within 3 days of discharge. For the multisite study, individual therapists who participated in the study provided the weekly ratings. However, there were differences across facilities regarding when the first weekly ratings were conducted: weekly ratings were begun 1 week from admission in some facilities and 1 week from the admission rating in others. To avoid problems with different time intervals between ratings, admission and discharge ratings were not included in the calibration or subsequent analyses. Data from all weekly ratings, including the LOS outliers dropped from subsequent analyses, were used in a single calibration to improve the stability of the item difficulty estimates to avoid problems in the definition of the variable over time, 27 Table 2: Sample Composition UDSMR Database Study 2 Current Multisite Study 15 N Mean SD N Mean SD Stroke 10, LOS Initial motor status Initial cognitive status SCI LOS Initial motor status Initial cognitive status TBI LOS Initial motor status Initial cognitive status

3 102 COURSE OF FUNCTIONAL IMPROVEMENT, Bode confidence interval (CI) for the means were added. To test for significant differences between LOS groups in total and average weekly gains, analyses of variance (ANOVAs) with LOS group as the independent variable were used. Fig 1. Patterns of improvement in motor functioning for persons with stroke, by LOS group. Lines connect the average ratings for each LOS group across weeks of the rehabilitation stay. Error bars represent the 95% CIs for the means. and to provide a stable frame of reference when making comparisons over time 28 so that measures of change have unambiguous numeric representation and substantive meaning. 29 These data were treated as separate cases to obtain the most stable estimate of item calibrations and, once these values were fixed, they were used to estimate separate person measures for each time point. The item and step calibrations were anchored to values obtained in the 1994 calibration of the UDSMR database to allow for a comparison to the previous results. 20 For all analyses, additive and multiplicative weights were used to transform the logits obtained in the analysis to a measure ranging from 0 (totally dependent on all items) to 100 (completely independent on all items). Table 2 compares the sample composition for these 2 studies. This table shows: (1) shorter LOS for persons with stroke and TBIs and longer LOS for persons with SCI in the current multisite sample than in the UDSMR sample, and (2) lower initial motor status and higher initial cognitive status, which may reflect recent changes in rehabilitation admissions criteria. The original intent of our study was to describe the actual changes in functional status; significance tests were added once status was plotted over time and the patterns of change over time emerged. An initial analysis of the calibrated weekly ratings was conducted by using data from the complete sample. In this analysis, average weekly motor and cognitive ratings were plotted separately for each impairment group to examine their improvement patterns. To examine LOS differences across impairment groups, histograms of the LOS for each group were produced. In the subsequent analyses, patients were grouped by the number of weeks in their rehabilitation stay (LOS groups). Weekly ratings of motor and cognitive status were aggregated by LOS group and their averages calculated. For example, motor function ratings after the first week of rehabilitation were averaged for persons with stroke who stayed 2 weeks, 3 weeks, and so forth; ratings after the second week of rehabilitation were averaged for persons with stroke who stayed 2 weeks, 3 weeks, and so forth. These weekly average ratings were then plotted for each LOS group and error bars representing the 95% RESULTS The results of the analysis of improvement patterns by LOS and impairment group are presented in figures 1 through 6, separately for motor and cognitive functioning for persons with stroke, SCI, and TBI. In these plots, the week of rehabilitation stay is on the horizontal axis, the FIM (motor or cognitive) measure is on the vertical axis, and each line represents 1 LOS group. Lines connect the average ratings for that LOS group across weeks of the rehabilitation stay and error bars represent the 95% CIs for the means. Several patterns are apparent by examining these plots. First, with the exception of cognitive function in persons with SCI, persons with lower functional status at admission have a longer LOS. These results are consistent with the findings of an unpublished analysis of data collected during early stages of the study. For example, the average motor functional status at week 1 for persons with stroke with stays of 2 to 3 weeks was approximately 47, and for persons with stays of 4 to 6 weeks, it was approximately 36. Second, regardless of LOS, the amount of gain from admission to discharge was essentially the same. With the exception of cognitive gains for persons with TBI, the ANOVA results (presented in tables 3 5) show no significant overall differences between LOS groups in total motor or cognitive gains from the first to last week of rehabilitation. Examination of the plots, however, does suggest that the rates of improvement differ across LOS groups. Analysis of the average weekly gains for each LOS group, shows significant overall differences between LOS groups in the rate of improvement for motor functioning but not for cognitive functioning (tables 3 5). Although the rates of improvement in cognitive functioning are not significantly different, they tend to decrease with increases in LOS. Perhaps with larger samples for each impairment group, the differences would be significant. Thus, one can conclude that persons with greater rates of motor improvement have shorter stays, in Fig 2. Patterns of improvement in motor functioning for persons with SCI, by LOS group.

4 COURSE OF FUNCTIONAL IMPROVEMENT, Bode 103 Fig 3. Patterns of improvement in motor functioning for persons with TBI, by LOS group Fig 5. Patterns of improvement in cognitive functioning for persons with SCI, by LOS group. general, than persons with lower rates of improvement. Finally, the improvement patterns for all 3 groups appear to be linear or essentially constant across the rehabilitation stay. In the figures, there do not appear to be periods of accelerated improvement or plateaus in functional status. DISCUSSION When the complete data were analyzed and the averages by week of rehabilitation stay were plotted over time (not shown), it appeared that functional status peaked early in the rehabilitation stay for persons with stroke and TBI and then declined until discharge. Examination of these data and the distribution of cases across LOS (not shown) showed that the cases at week 1 represented the total population, but at later weeks, the cases represented only persons with longer LOS. As documented in previous research, 2-9 longer LOS is related to lower admission status; thus, the remaining cases at later weeks represented patients with greater initial disability than those in earlier weeks. These results reinforced the need for examining functional improvement aggregated by LOS group. Likewise, when data for patients with atypically short or long LOS for their impairment group were included in the plots, the CIs were extremely large and functional status appeared to fluctuate greatly from 1 week to the next at each end of the LOS distribution. Instability in average functional status for persons with atypically short or long LOS resulted in dropping the LOS outliers from subsequent analyses. The result of this pruning was to focus on stable estimates of functional status over time that are typical of most persons receiving inpatient rehabilitation services. Such pruned estimates are valuable to rehabilitation hospitals as they prepare for Medicare s PPS. Fig 4. Patterns of improvement in cognitive functioning for persons with stroke, by LOS group Fig 6. Patterns of improvement in cognitive functioning for persons with TBI, by LOS group.

5 104 COURSE OF FUNCTIONAL IMPROVEMENT, Bode LOS Group (by wk) N Week 1 Mean Table 3: FIM Gains for Persons With Stroke by LOS Group Total Motor Gains Average Weekly Motor Gains w/3,.36 (NS) w3, 1.87 (NS) w/4, 1.59 (NS) w4,.35 (NS) w/5,.39 (NS) w5, 1.77 (NS) w/6,.23 (NS) w6,.92 (NS) Total (NS) (.001) Total Cognitive Gains Average Weekly Cognitive Gains w/3,.41 (NS) w3, 1.58 (NS) w/4,.62 (NS) w4,.10 (NS) w/5,.27 (NS) w5,.28 (NS) w/6,.91 (NS) w6,.34 (NS) Total (NS) (NS) Abbreviations: SE, standard error; NS, not significant. The results of this study show that functional improvement patterns need to be examined separately by LOS group to reflect accurately the changes that occur during medical rehabilitation. When LOS groups are not taken into account, it appears as if functional status plateaus or even declines with longer LOS; however, when it is taken into account, it is apparent that improvement continues at approximately the same pace during early and later stages of rehabilitation. That persons with greater disabilities improve at a slower rate will not surprise experienced clinicians, but even they may not be aware that the amount of improvement for these persons is essentially the same as for those with less disability. As in previous research, longer LOS is associated with greater disability. However, unlike Stineman et al, 10 who found that gains were largest for patients with the greatest disability, the results of this study indicate approximately equal gains regardless of impairment group and LOS. This discrepancy might reflect the patient composition at the 8 facilities included in this study and at rehabilitation facilities in general, or it may reflect secular trends in that Stineman s 10 data may have come from a time when people with greater disability were allowed to stay longer and attain a higher discharge status. LOS has decreased in recent years and the timing of the data collection could result in differences in the relationship between LOS and functional improvement. As summarized in table 2, differences in LOS and initial motor and cognitive status were found between this study and the Heinemann et al 2 study of UDSMR data. If, as suggested by these findings, rates of improvement are relatively consistent over the LOS, initial indicators of functional status should be sufficient for predicting both LOS and functional status at discharge. This study examined actual improvement patterns rather than those predicted from initial status. Several techniques are well suited for modeling functional status during rehabilitation to predict improvement patterns. Hierarchical linear modeling is a technique that estimates initial status and rates of improvement and examines group differences within each. By using this technique, data for persons with stroke collected in early stages of the multisite study were analyzed to model LOS Group (by wk) N Week 1 Mean Table 4: FIM Gains for Persons With SCI by LOS Group Total Motor Gains Average Weekly Motor Gains w5, 1.35 (NS) w5, 3.42 (.001) w6,.68 (NS) w6,.46 (NS) w7,.03 (NS) w7,.79 (NS) w8,.08 (NS) w8,.45 (NS) w9,.48 (NS) w9,.12 (NS) Total (NS) (.001) Total Cognitive Gains Average Weekly Cognitive Gains w5,.07 (NS) w5,.21 (NS) w6,.87 (NS) w6, 1.21 (NS) w7, 1.26 (NS) w7,.99 (NS) w8, 1.02 (NS) w8,.89 (NS) w9, 1.25 (NS) w9,.77 (NS) Total (NS) (NS)

6 COURSE OF FUNCTIONAL IMPROVEMENT, Bode 105 LOS Group (by wk) N Week 1 Mean Table 5: FIM Gains for Persons With TBI by LOS Group Total Motor Gains Average Weekly Motor Gains Mean SE SE F/Contrasts (P) Mean SD SE F/Contrasts (P) w3,.08 (NS) w3, 2.04 (NS) w4,.13 (NS) w4,.98 (NS) w5,.68 (NS) w5,.19 (NS) Total (NS) (.05) Total Cognitive Gains Average Weekly Cognitive Gains w3, 2.32 (.05) w3,.97 (NS) w4, 1.18 (NS) w4, 2.24 (NS) w5, 2.28 (.05) w5,.04 (NS) Total (.01) (NS) functional improvement. The same decline in functional status over weeks of rehabilitation stay in the plots of actual averages for all LOS groups from this study are found in the plots of predicted functional status (not shown). The differences between the initial plots of actual averages in which data were not aggregated by LOS group, and subsequent plots in which they were, show the need for adding parameters representing LOS groups in the modeling of functional improvement. With a larger sample, an analysis using hierarchical linear modeling could also be used to determine if improvement is linear or curvilinear and to identify other patient characteristics that could explain variations in rates of improvement. Modeling improvements in functional status during rehabilitation, however, cannot proceed until methods are developed for controlling for varying LOSs. Variation in LOS may reflect differences in decision making regarding the point at which a patient is discharged. Factors such as the ability of family or other caregivers to care for patients at home and the availability of outpatient rehabilitation options may also contribute to the timing of discharge. Patients do not receive therapy for a fixed period of time, nor will that or should that change. Understanding LOS requires a better understanding of the considerations that go into deciding when a patient should be discharged from rehabilitation. One rationale is that patients are discharged when they have attained the rehabilitation goals set at admission. However, this is not always the case. When the prognosis for continued improvement is poor, for example, when a patient whose original discharge destination was to home but the lack of progress results in discharge to a nursing home, the rehabilitation goals may be modified and the stays shortened to reflect these more modest goals. Alternatively, stays may be extended when patients are making good progress or when clinical evidence supports additional time to reach therapy goals. For example, a patient may experience a medical complication that slows the rate of recovery and necessitates additional time to meet rehabilitation goals. In general, a patient s discharge from rehabilitation can be considered as a positive or a negative event. It is positive when the patient has met goals and is able to return to the community; it is negative when the patient is judged as not being able to benefit from more rehabilitation or, more frequently than in the past, medical benefits have been capitated. Whatever decisions are made concerning LOS, discharge is treated as the point at which the patient is deemed to have progressed as far as possible within these parameters set for the level of rehabilitative care. It is thought that patients may leave inpatient status not reaching all goals, but will continue in outpatient (or home health) care to achieve them. The categorization of reasons for discharge may represent the missing variable in controlling for varying LOSs in the modeling of functional improvements. Identifying patterns of improvement has policy implications. Being able to predict functional outcomes allows facilities to estimate and manage their resources when patients are admitted to rehabilitation. With the advent of Medicare s PPS, this information becomes even more crucial. The findings confirm the need to consider both the impairment group and the initial functional status level when setting rehabilitation goals. Because goals are typically based on initial functional status and the anticipated LOS, both need to be considered when selecting goals that can be met in a specific period of time. All else being equal, the findings from this study can be used to project outcomes after specific LOSs for persons are established, predicated on specific admission functional status. This evidence can be useful in working with payers to determine if an increase in the rehabilitation stay would be warranted and what effect an extended stay might be anticipated to have. It can also be useful in showing the negative effect that an earlier than anticipated discharge might have on functional status at discharge. To strengthen this evidence, factors such as age, delayed rehabilitation admission, the occurrence of comorbidities and medical complications, and interruptions in care could be used to predict functional status at discharge, as performed in the MediRisk calculator. 30 Further research would also be required to answer policy questions such as: Should selected patients with low functional status at admission go directly to skilled nursing facilities? And, if so, how low? Exceptions would need to be made for some low-functioning persons, such as SCI patients who are ventilator-dependent or soon to be weaned, or other patients with higher medical needs, who cannot be managed in skilled nursing facilities and require acute rehabilitation. Should other patients with high functional status at admission go directly to outpatient rehabilitation? Exceptions would need to be made for some high-functioning people, such as new diabetics, new dialysis patients, burn patients, and Parkinson patients on drug holidays, who may have medical needs that require acute hospital and rehabilitation management. If so, are there enough programs to manage the increased outpatient caseloads? How will coordination of services be maintained in outpatient settings where à la carte programming prevails (single services in disjointed schedules)?

7 106 COURSE OF FUNCTIONAL IMPROVEMENT, Bode Several limitations to this study must be considered. First, weekly functional status was rated by the same clinicians who provided therapy. Although this familiarity provides useful information for treatment planning, it reduces the independence of the individual ratings. However, this is a typical practice and apt to continue regardless of payment system. Second, the patient selection criteria used at these 8 facilities and the study s inclusion criteria may result in reducing the generalizability of these results to other rehabilitation settings. Third, and most crucial, although the overall sample is adequate, the sample sizes by impairment group within LOS groups tend to be small, hence, the error bars are relatively large. As a result, these findings should be considered tentative until this study can be replicated with a larger sample. CONCLUSION This study shows the effect that initial functional status and LOS has on functional improvement. Not only do persons with higher initial status have shorter LOS, but also their rates of improvement are greater. Persons with lower initial status improve essentially as much as those with higher initial status; however, their rates of improvement are slower and it takes them longer to achieve these gains. Improvement patterns that do not take LOS into account confound the modeling of improvement and can lead to incorrect conclusions of the amount of improvement achieved during medical rehabilitation. The setting of rehabilitation goals and the decision making that goes into discharge planning must take these factors into account. Acknowledgments: The authors acknowledge the contributions of Robin Mieli, Jennifer Jo Kim, and Christa Ambrose, who collected data at Rehabilitation Institute of Chicago; Joyce Santora, PhD (Erie County Medical Center), Donna Selsor (Long Beach VA Medical Center), Kathy Thomas (Baylor Institute for Rehabilitation), Mary Pat Welc and Mary Thorson (Lutheran General Hospital), Dan Graves (The Institute for Rehabilitation and Research), and Susan Carty and Teri Jorgensen (Mid-America Rehabilitation Hospital), who supervised the data collection at their sites; Peggy Kirk and Camille O Reilly, who coordinated the data collection across sites; Patrick Semik, who set up and maintained the database; and Marcel Dijkers PhD, who provided feedback on the manuscript. References 1. Hamilton BB, Granger CV, Sherwin FS, Zielezny M, Tashman JS. A uniform national data system for medical rehabilitation. In: Fuhrer MJ, editor. Rehabilitation outcomes: analysis and measurement. Baltimore: Brookes; p Heinemann AW, Linacre JM, Wright BD, Hamilton BB, Granger C. Prediction of rehabilitation outcomes with disability measures. Arch Phys Med Rehabil 1994;75: Carey RG, Seibert JH, Posavac EJ. Who makes the most progress in inpatient rehabilitation: an analysis of functional gain. Arch Phys Med Rehabil 1988;69: Yarkony G, Roth E, Heinemann AW, Lovell L, Wu Y. Functional skills after spinal cord injury rehabilitation: three-year longitudinal follow-up. Arch Phys Med Rehabil 1988;69: Granger CV, Hamilton BB, Gresham GE, Kramer AA. The stroke rehabilitation outcome study Part I: general description. Arch Phys Med Rehabil 1988;69: Granger CV, Hamilton BB, Gresham GE. The stroke rehabilitation outcome study Part II: relative merits of the total Barthel Index score and a four-item subscore in predicting patient outcomes. Arch Phys Med Rehabil 1989;70: Heinemann AW, Roth E, Kaplan P, Betts H. Multivariate analysis of improvement and outcome following stroke rehabilitation. Arch Neurol 1987;44: Heinemann AW, Yarkony G, Roth E, et al. Functional outcome following spinal cord injury: a comparison of spinal cord injury center vs general hospital short-term care. Arch Neurol 1989;46: Stineman MG, Williams SV. Predicting inpatient rehabilitation length of stay. Arch Phys Med Rehabil 1990;71: Stineman MG, Escarce JJ, Goin JE, Hamilton BB, Granger CV, Williams SV. A case-mix classification system for medical rehabilitation. Med Care 1994;32: Stineman MG, Hamilton BB, Granger CV, Goin JE, Escarce JJ, Williams SV. Four methods for characterizing disability in the formation of function-related groups. Arch Phys Med Rehabil 1994;75: Stineman MG, Goin JE, Hamilton BB, Granger CV. Efficiency patterns analysis for medical rehabilitation. Am J Med Qual 1995;10: Stineman MG, Hamilton BB, Goin JE, Granger CV, Fiedler RC. Functional gain and length of stay for major rehabilitation impairment categories. Patterns revealed by function related groups. Am J Phys Med Rehabil 1996;75: Stineman MG, Escarce JJ, Tassoni CJ, Goin JE, Granger CV, Williams SV. Diagnostic coding and medical rehabilitation length of stay: their relationship. Arch Phys Med Rehabil 1998;79: Heinemann AW, Hamilton BB, Johnston M. Relation of rehabilitation intervention to functional outcome. J Rehabil Outcomes Meas 2000;4: Willett JB. Questions and answers in the measurement of change. In: Rothkopf E, editor. Review of research in education. Vol 15. Washington (DC): American Educational Research Association; p Rogosa D, Willett JB. Understanding correlates of change by modeling individual differences in growth. Psychometrika 1985; 50: Bryk AS, Raudenbush SW. Application of hierarchical linear models to assessing change. Psychol Bull 1987;101: Bryk AS, Raudenbush SW. Hierarchical linear models: applications and data analysis methods. Newbury Park (CA): Sage; Heinemann AW, Linacre JM, Wright BD, Granger CV. Measurement characteristics of the Functional Independence Measure. Top Stroke Rehabil 1994;1: Linacre JM, Heinemann AW, Wright BF, Granger CV, Hamilton BB. The structure and stability of the Functional Independence Measure. Arch Phys Med Rehabil 1994;75: Rasch G. Probabilistic models for some intelligence and attainment tests. Chicago: MESA Pr; Wright BD, Masters GN. Rating scale analysis: Rasch measurement. Chicago: MESA Pr; Wright BD, Linacre JM. WINSTEPS: computer program for Rasch analysis. Chicago: MESA Pr; Schulz EM, Perlman C, Rice WC, Jr, Wright BD. Vertically equated reading tests: an example from the Chicago Public Schools. In: Wilson M, editor. Objective measurement: theory into practice, Vol 1. Norwood (NJ): Ablex; p Setlzer MH, Frank KA, Bryk AS. The metric matters: the sensitivity of conclusions regarding growth in student achievement to choice of metric. Educ Eval Policy Analysis 1994;16: Smith RM. Pre-post comparisons in Rasch measurement. In: Wilson M, Engelhard G, Draney K, editors. Objective measurement: theory into practice. Vol 4. Greenwich (CT): Ablex; p Wright BD. Comparisons require stability. Rasch Meas Trans 1996;10: Wright BD. Time 1 to time 2 comparisons. Rasch Meas Trans 1996;10: Leiter P. Measuring casemix, severity, and complexity in geriatric patients undergoing rehabilitation. Small Group I report. Med Care 1997;35 Suppl 6:JS113-7.