(Stroke. 1997;28:550-556.)
© 1997 American Heart Association, Inc.
Articles |
A Prediction Model for Functional Recovery in Stroke
Correspondence to Margaret G. Stineman, MD, Ralston-Penn Center, Room 101, 3615 Chestnut St, Philadelphia, PA 19104-2676.
 |
Abstract |
|---|
 Top Abstract IntroductionSubjects and MethodsResultsDiscussionAppendix 1References |
|---|
Background and Purpose Stroke-related physical disability can diminish quality of daily living, place care burden on families, and increase need for long-term institutionalization. We developed a prognostic index for use in research and with potential for adaptation to clinical practice that establishes the likelihood of an individual achieving a specific stage of functional recovery after stroke rehabilitation.
Methods We constructed the index using logistic regression based on 3760 patient records from 96 rehabilitation facilities in 31 states. The stage, as measured by the Functional Independence Measure, includes achievement of the following: independence in eating, grooming, and dressing the upper body; continence in bowel and bladder; and transfer between a bed and chair with supervision only.
Results This stage was achieved by 26.1% of patients functioning below it at rehabilitation admission. Disability onset of less than 60 days was associated with more than a 3-fold increase in the likelihood of achieving the stage (adjusted odds ratio, 3.5; 95% confidence interval, 2.0 to 6.0). Each eight-point increase in an eight-item activities of daily living score, measured at admission to rehabilitation, increased the odds 2.5-fold (95% confidence interval, 2.3 to 2.8). For those living alone or employed before the stroke, the odds of achieving the stage increased by factors of 1.3 and 2.2, respectively. The index showed minimal shrinkage on cross validation. The achievement of this profile of function is important because 95.3% of stroke patients who achieved or exceeded it were discharged home, as opposed to only 66.8% of those who did not achieve it.
Conclusions The index can be used to establish prognoses for individual stroke patients at admission to rehabilitation with regard to achieving this stage. Achievement of the stage is associated with a high likelihood of discharge to home.
Key Words: activities of daily living • models, theoretical • prognosis • stroke outcome
|
Introduction |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionAppendix 1References |
|---|
Stroke is the leading cause of adult disability in the United States. Approximately 550 000 strokes occur each year, leaving 300 000 people with disability. A 1993 estimate placed the annual costs of stroke at $30 billion, of which $17 billion were direct medical costs and $13 billion were indirect costs due to lost productivity.1 The high prevalence of stroke and its high economic costs make the reduction of stroke-related disability a national healthcare priority and underscore the need for more accurate methods of identifying patients with different prognoses.
The prognosis for functional recovery in stroke is influenced by a broad array of neurological,2 3 4 functional,3 5 6 and psychosocial7 8 factors. This report presents an index that predicts the likelihood of stroke survivors recovering to or exceeding a specific stage of functional recovery based on a set of clinical characteristics known at rehabilitation admission. This index is called the Stroke Recovery of Activities of Daily Living and Mobility (RAM) Index because it predicts patient status in relation to these functions at discharge from rehabilitation. We chose to predict status at discharge because it determines the type of care families or others must be capable of providing if an individual is to return home.
|
Subjects and Methods |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionAppendix 1References |
|---|
Data were obtained from the UDSMR,9 10 which was developed as a uniform standard for the assessment of medical rehabilitation patients. The UDSMR collects data on each patient's primary impairment, functional status at admission and discharge, sex, race, living situation before hospitalization and at discharge, time since onset of disability, marital status, and previous life role. Additional data on each patient include date of birth, admission, discharge, and any transfers from rehabilitation to another hospital service.
The UDSMR uses the FIM10 to describe patients' functional status on 18 standardized items. The first 13 FIM items are similar in content to other functional status measures that describe physical disability.11 12 The remaining items are global ratings of an individual's ability to recall, solve problems, and communicate or interact with others. Each FIM item per- formance level is precisely defined and ranges in value from 1 to 7, with 1 indicating complete dependence in an activity and 7 complete independence. Patient performance on the FIM is assessed by rehabilitation clinicians, each of whom must pass a written examination certifying his or her proficiency in coding it.
In preparation for developing the Stroke RAM Index, we found that the 18-item FIM has three subscales relevant to stroke (see "Appendix"). The first two, referred to as the ADL and mobility subscales, respectively distinguish between functions that depend primarily on use of the upper and the lower extremities. The third subscale, referred to as cognitive, combines the more executive functions of cognition and communication. The ADL subscale includes the FIM items of eating, grooming, bathing, dressing the upper body, dressing the lower body, toileting, bladder management, and bowel management. When summed, the value ranges from 8 to 56. The mobility subscale (bed/chair/wheelchair transfer, toilet transfer, tub or shower transfer, walking or wheelchair, and stairs) and the cognitive subscale (comprehension, expression, social interaction, problem solving, and memory) both range in value from 5 to 35. A low score on any subscale indicates a more severe disability.
The study sample included patients discharged during 1990 from 45 freestanding rehabilitation facilities and 51 distinct part rehabilitation units within acute-care hospitals in 31 states. Patients with International Classification of Diseases, 9th Revision, Clinical Modification codes ranging from 430 through 438 (constituting various categories of cerebrovascular disease) were identified as having a diagnosis of stroke (n=6739). Because the objective of the study was to quantify the functional prognoses of adult stroke inpatients, nine patients younger than 17 years at rehabilitation admission were excluded. Patient records with missing admission or discharge data, coding inconsistencies, or out-of-range values were excluded (n=156 or 2.3% of the initial sample). These patients did not differ by age (t=1.2; P=.24), initial severity of physical disability (t=1.2; P=.22), or initial severity of cognitive disability (t=.97; P=.33). Patients whose FIM performance at admission was at or higher than the stage to be predicted were also excluded (n=312 cases or 4.75% of the 6574 usable records). The remaining 6262 patient records were then randomized, with a 60%/40% split, into model building (n=3760) and validation (n=2502) data sets. Univariate and multivariate analyses were limited to the model building data in which the mean age was 71.4 (SD, 11.8) years; 81.9% (3079/3760) of patients were white, 10.7% (402/3760) were black, and 7.4% (278/3760) were of a different racial origin. The validation data were held back until completion of the Index, when it was used to validate predictive utility.
The outcome predicted by the Stroke RAM Index is a stage of modified
(or partial) functional independence (Mod-FI), which specifies a
minimum level of performance on eight of the 18 FIM items. The stage is
referred to as modified functional independence because, for most
items, a minimum level of modified independence (FIM score=6) was
specified. Modified independence is the first performance level at
which physical assistance from another person is no longer required.
Lower performance levels were included only when too few patients
reached level 6 by discharge on any of the eight items and when the
clinical importance of the item warranted its inclusion, even at a
physically dependent level. The eight FIM items making up the Mod-FI
stage were selected for their clinical and statistical importance and
because of their potential physiological and sociological impact on
overall personal autonomy. Selection of these items combined
theoretical11 and clinical knowledge about expected
patterns of functional recovery with statistical rankings of patients'
performance on each item at discharge. The patient achieving or
exceeding the stage of Mod-FI is able to eat, groom, and dress the
upper body without assistance; manage bladder and bowel functions
without accidents; and manage toilet functions once set up. The patient
may still require supervision to transfer from bed to chair but no
longer requires lifting or contact assistance. Finally, the patient
must be able to propel a wheelchair 50 feet with no more than minimal
contact assistance and/or be able to walk the same distance with help
from another person. The explicit definitions of the eight FIM items
making up the stage of Mod-FI are shown in Table 1
.
|
We modeled the dichotomous outcome (achievement of the stage of Mod-FI
versus nonachievement) using the logistic regression procedure in
Statistical Analysis Software.13 In preparation for
developing the logistic model, we studied the underlying relationships
between explanatory variables and the likelihood of achieving the
stage, including the determination of unadjusted odds ratios. Based on
these descriptive analyses, appropriate mathematical transformations of
the independent variables were used in the logistic model. In addition
to searching for factors that directly affect recovery from stroke
disability, we examined the significance of pairwise statistical
interactions between age and each of the functional status
subscales.14 Primary effects and first-order interaction
terms were included if they were significant at P
.01
and/or if their inclusion changed the other model-estimated odds ratios
by at least 15%. The 15% criterion was used to reduce bias in the
estimates of the effects of specific variables arising from failure to
control for confounding variables.15
The relationship between patient performance on the mobility subscale and stage achievement demonstrated a quadratic effect, and therefore this variable was expressed as four categories in the subsequent logistic regression. Time since onset of disability showed a step function when analyzed as 20 strata (data not shown), in which case the log odds of recovery dropped sharply in those with onset of 60 days or longer. Thus, this variable was dichotomized. The remaining interval variables had approximately linear associations and were included as continuous variables. Categorical variables were entered as sets of indicator variables.
To assess the clinical utility of the final logistic regression model, we used the area under the ROC curve16 comparing values computed in the estimation and validation samples. This procedure and its interpretation are described in the "Appendix." The Hosmer-Lemeshow17 goodness-of-fit statistic was used to test the hypothesis that model-produced estimates of the likelihood of a patient achieving Mod-FI adequately fit the data.
|
Results |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionAppendix 1References |
|---|
The Mod-FI stage was achieved or exceeded by 981 (26.1%) of the 3760 model-building patients initially performing below it. Among those who achieved the stage, 95.3% (935/981) were discharged to the community, just under 2% (19/981) were discharged to long-term care facilities, 1% (10/981) were transferred to acute hospital services, and the remainder (17/981) were discharged to other settings. In contrast, among those who did not achieve the stage of Mod-FI, only 66.8% (1856/2779) were discharged to the community, 20% (556/2779) were discharged to long-term care facilities, 10% (278/2779) were transferred to acute hospital services, 1.1% (31/2779) died, and the remainder (58/2779) were discharged to other settings.
The unadjusted odds of recovering to the stage for different clinical
characteristics are shown in Table 2. Patients with
predominantly right-sided or no paresis were more likely than those
with left hemiparesis to achieve or exceed the stage of Mod-FI, while
those with bilateral paresis were less likely. The likelihood of
achieving the stage was dramatically lower for patients who were older,
admitted from the community rather than from an acute hospital service,
unemployed, or who had longer times since stroke onset or lower scores
on the ADL, mobility, and cognitive subscales.
|
Six of the eight patient characteristics hypothesized to be associated
with stage achievement remained statistically significant after
multivariable adjustment (Table 3). The odds of recovery
to or exceeding the stage of Mod-FI more than tripled in patients for
whom time since onset of disability was less than 60 days compared with
those with longer times since onset. The odds of recovery more than
doubled for each 8-point increase in the ADL score. The associations
between recovery and the patient's admission mobility and cognitive
FIM scores were more complex because of statistical interactions with
age. The impact of disabilities in these functional areas was increased
in the elderly, and the effect of mobility was nonlinear. The location
of paresis and previous hospitalization were not statistically
significant after we accounted for other factors. Living alone before
stroke, not initially hypothesized to be a predictor of recovery, was
associated with increased likelihood.
|
The Hosmer-Lemeshow goodness-of-fit test confirmed that the model fit
the data set with P=.61.17 Table 4 shows the predictive capacity of the Stroke RAM Index
in the validation data. Here, patients are organized into 10 classes by
decile probability of recovery. The actual percentage of patients in
each class who achieved the stage of Mod-FI is compared with its decile
predictive range, showing that the model accurately estimated the
probability of recovery. The area under the ROC curve was .86, as
detailed in the "Appendix." This value is well within the range
considered appropriate for clinical use.
|
Table 5 provides the procedure for computing discrete
probabilities of achieving the stage of Mod-FI from the Stroke RAM
Index. First, the log odds of achieving the stage are computed as a
linear function of the patient's admission characteristics (Equation
1). Then, Equation 2 is used to convert the log odds to the probability
of stage achievement.
|
|
Discussion |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionAppendix 1References |
|---|
Clinicians can use the Stroke RAM Index to distinguish patient subgroups at admission with different probabilities of recovering to or beyond the stage of Mod-FI by discharge. To our knowledge, this is the first such index developed and tested in a national set of medical rehabilitation records. Many indexes, staging algorithms, and classification systems are used in general medical and surgical practice18 19 20 to summarize patient severity and to relate it to future mortality, morbidity, or resource use. The Stroke RAM Index is similar in design and applications, but rather than predicting mortality or morbidity, it predicts individuals' functional recovery.
Just over 95% (935/981) of patients who achieved the stage of Mod-FI were able to be discharged to the community after inpatient rehabilitation. Specifically, these patients were, at a minimum, able to eat, groom, and dress the upper body without help; manage bowel and bladder functions without accidents; accomplish toilet functions with setup; transfer from bed to chair without physical assistance; and either propel a wheelchair or walk (with or without help) 50 feet. This profile represents a reasonable "high end" set of functional goals for inpatient stroke rehabilitation. Because only 26.1% were able to achieve these goals, there is need for more tailored programs of treatment and community support services to assist patients who do not achieve them.
The strongest predictors of stage achievement were patients' performance of ADL, mobility, and cognitive functions at rehabilitation admission. Although the association between high initial ADL performance and the achievement was direct, the associations between cognitive and mobility functioning and stage achievement were complex, since they depended on age. The consequences of dependency in mobility and cognition were greater in older people than in younger people, suggesting that the effect of age on patients' prognosis depends on both severity and patterns of functional loss. The social predictors (life role and living alone) may approximate different motivations. Living alone before stroke may drive patients to higher achievement because they recognize fewer opportunities for receiving assistance. Similarly, those previously serving as homemakers or working outside the home, by virtue of these social roles, might be more accustomed to the type of goal-directed behavior necessary to successful rehabilitation.
The activities and performance levels making up the stage of Mod-FI were selected for their biological and sociological importance. For example, eating is necessary for survival and, in the absence of a caregiver, the inability to feed oneself is life-threatening.11 Bladder and bowel continence are particularly important to the personal dignity of stroke survivors and to those who care for them. The ability to transfer between bed and chair without hands-on assistance suggests that care can be provided by someone without significant physical strength. Therefore, patients who achieve the stage of Mod-FI are able to perform some fundamental self-care tasks without help. The importance of reaching this stage was further highlighted in our data by an associated reduced risk of nursing home placement.
Predictions from the Stroke RAM Index can be formulated as continuous probabilities, ranges, or an optimal cut point. When expressed as continuous probabilities, predictive values measure the likelihood of recovery along a continuum. In contrast, ranges distinguish subgroups of patients at admission with low (for example, <25%), intermediate (25% to 75%), and high (>75%) probabilities of recovery. Finally, the optimal cut point method provides a prognostic value (ie, predicted probability) above which Mod-FI would seem a reasonable goal for a patient. The optimal cut point method mirrors the approach taken in the development of diagnostic tests in which parameters such as sensitivity, specificity, and predictive value can be calculated to evaluate clinical utility. This approach is detailed in the "Appendix."
The probability estimates from the Stroke RAM Index can be used to establish functional prognosis at the beginning of rehabilitation, thus anticipating the likelihood of patients achieving the stage of Mod-FI by the time they are discharged back to the community. This knowledge is particularly important for family caregivers early in the course of rehabilitation, because it will enable them to better plan for the type of care they will need to provide. Clinicians can also use predictions from the Index at admission to rehabilitation to determine whether achievement of the stage of Mod-FI is a reasonable goal for individuals. Clinicians can use estimates after discharge in quality improvement efforts to compare services received by patients who had low predictive probabilities of achieving Mod-FI but achieved it to those who had high predictive probabilities but did not achieve it. Policy analysts can use predictions from the Index as a quality monitor, particularly following implementation of a new payment system for medical rehabilitation, such as one based on the Functional Independence Measure–Function Related Groups21 being considered by the Health Care Financing Administration. We believe that predictions from the Stroke RAM Index should not be used to decide which patients receive rehabilitation services. Although the Index fits the data well, there is always uncertainty in predicting the outcomes of individuals, and a broader array of clinical attributes than those in the model should be considered in making clinical decisions. In making the decision about whether inpatient rehabilitation is appropriate, the Index can provide one more piece of quantified information that might be applied in a fashion similar to the use of diagnostic tests.
In addition to its clinical applications, the Stroke RAM Index can be used to stratify patients by prognosis or to control for statistical confounding in randomized trials in which functional recovery to the stage of Mod-FI is the end point. Currently, the extent to which functional recovery results from rehabilitation is not known, and trials on the benefits of stroke rehabilitation are inconclusive.22 23 24 25 The conflicting results of those trials may have been due to a failure to account for important prognostic factors. There is evidence that stratification of patients by initial level of disability is necessary when differences among alternative rehabilitation interventions are evaluated.23 The Stroke RAM Index provides a means to identify, stratify, and study patients who fall within particular prognostic ranges.
One potential criticism of the Stroke RAM Index is that substantial variations in rehabilitation LOS might invalidate estimates from it. To evaluate how LOS affects calculated probabilities, we rederived the Index but added rehabilitation LOS to the list of independent variables. LOS had little impact on predictive utility of the model (see "Appendix"). Based on previous findings,3 26 27 we assumed that this was because rehabilitation LOS was highly associated with variables already contained in the Stroke RAM Index. To test this assumption, we performed an ancillary analysis in which a multiple linear regression model was used to predict LOS. This model contained the same independent variables and structures as the Stroke RAM Index, but the dependent variable (Stage Mod-FI) was replaced with LOS transformed by its natural logarithm. The Stroke RAM Index variables explained 15% of the variance in the natural logarithm of LOS, thus confirming that LOS is related to the variables already included. Only large differences in LOS appear clinically important. Perhaps one reason that achieving the stage of Mod-FI was not strongly associated with LOS in the multivariable model is that rehabilitation clinicians control LOS based on patients' progression. Patients are discharged once their achievements have plateaued.
The Stroke RAM Index differs from other predictive models of stroke recovery3 6 28 29 30 31 because it predicts the likelihood of patients achieving a specific stage of functioning as represented by a specific profile of functional abilities rather than an aggregated functional status score. Because of this, the prediction avoids the loss of descriptive power that occurs when performance on multiple functional status items is aggregated into a summated score.32 The stage of Mod-FI is one of many potentially relevant profiles of functional recovery. Since its achievement is not a reasonable goal for all stroke survivors, it will be important to develop additional indexes to predict achievement of both more fundamental and more advanced stages of recovery. These indexes could establish prognoses for stroke survivors at various points of recovery and in various settings. Supplementing the FIM and other predictive attributes in the Index with more focused impairment measures, information on medical complications, and greater details about the patients' environmental circumstances could enhance our ability to study and contrast various dimensions of outcome across a variety of rehabilitative interventions.
In conclusion, we have developed a prognostic index that appears to have excellent predictive capabilities. This Stroke RAM Index is unique in that it predicts achievement of a stage of functional recovery which, based on high community discharge rates, appears biologically and sociologically essential to personal autonomy. The Index has numerous potential applications to clinical medicine and research and provides a means to evaluate the achievements of local and national programs for stroke survivors. It is a prototype for the development of a larger series of RAM indexes. These indexes will operate as tools to predict multidimensional outcomes, incorporating specific levels of functioning in the ADL, mobility, and cognitive areas along a continuum. Such indexes will provide a spectrum of data that will more closely define groups of individuals and their ability to function in the world after stroke. By splitting rather than lumping data (as is often done in determining outcome through a single functional status score), it is possible to add definition in the description of characteristics of patients who succeed with regard to specified profiles of functional achievement.
The Mod-FI stage represents one of many functional recovery stages. Future research will be designed to identify the probability of patients achieving higher-end and lower-end stages of functional recovery or the achievements of other outcomes, including community discharge and independent living.
|
Selected Abbreviations and Acronyms |
|---|
|
|
|
Acknowledgments |
|---|
This study was supported in part by grant RO1-HS07595 from the Agency for Health Care Policy and Research, National Institutes of Health grant KO8-AG00487 from the National Institute on Aging, and National Institutes of Health grant 1-RO1-HD-34101 from the National Institute of Child Health and Human Development. The opinions of the authors are not necessarily those of the sponsoring agencies. The authors thank John Lehr and Delores Foster-Kennedy for preparing and editing the manuscript.
|
Appendix 1 |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionAppendix 1References |
|---|
Development of Subscales
The three subscale dimensions of the FIM (ADL, mobility, and cognitive) were confirmed through factor analysis with the use of an orthogonal varimax rotation (which assumes independence among factors) followed by an oblique promax rotation (which does not assume independence).33 The factor structures were very similar, indicating that the independence assumption was reasonable. Thus, the simpler orthogonal solution was selected. These three factors explained 71% of variation in the total FIM score. Cronbach's
for the three
subscales ranged from .88 to .92, demonstrating good internal
consistency.
Assessing Clinical Utility Through ROC Curve Analyses
ROC curves measure predictive utility by demonstrating the
trade-off between the true-positive rate (appearing on the vertical
axis) and the false-positive rate (appearing on the horizontal axis)
(Figure
) inherent in selecting specific thresholds on
which predictions might be based. The area under the ROC curve
represents the chance that a randomly selected patient who achieved
Mod-FI has a greater predicted probability than a randomly selected
patient who did not. The area under the ROC curve was .864 (95% CI,
.851 to .877) in the model building data and .855 (95% CI, .839 to
.871) in the validation data, thus demonstrating negligible bias on
cross validation in our estimate of the model's predictive
capability.
The ROC curve can be used to define an optimal cut point based on the outcome prevalence. Here recovery to the stage of Mod-FI is assumed a reasonable goal for all patients whose probabilities of achieving the stage are above a certain value. For the Stroke RAM Index, there was no compelling reason to assume that the repercussions of falsely predicting that a patient will achieve the stage are any worse than falsely predicting a patient will not achieve it. Thus, we selected the cut point that minimizes total prediction errors, given the proportion of patients actually achieving the stage in our sample. Adjustments for unequal costs or for differences in outcome prevalence may be made.16 34 Our cut point is located where the slope of the line tangent to the ROC curve is equal to the reciprocal of the odds of achieving the stage. This point corresponds to a prognostic index value of approximately .50 in both the model building and validation samples.
Positive and negative predictive values and sensitivity and specificity were then determined for the optimal cut point of .50. The estimated positive and negative predicted values (95% CI) of this cut point are .72 (.68 to .76) and .84 (.82 to .86), respectively. The estimated sensitivity and specificity of this cut point are .58 (.55 to .61) and .91 (.89 to .92), respectively. These parameters appear consistent with clinical utility. When the optimal cut point prediction method is used, if a patient's prognostic value (ie, predicted probability) is at least .50, recovery to the stage of Mod-FI would seem a reasonable goal. This optimal cut point designates a probability threshold above which all patients are assumed relatively likely to achieve or exceed the stage. In our full sample, only 26.1% achieved this level of Mod-FI. If the predicted value was larger than the optimal cut point of .50, the percentage that achieved Mod-FI increased to 72%.
To evaluate how LOS affects calculated probabilities, we rederived the index but added rehabilitation LOS to the list of independent variables. This addition increased the area under the ROC curve only from .855 to .857 in the validation data, showing that LOS had little impact on the ability of the index to predict.
Received July 22, 1996; revision received December 3, 1996; accepted December 3, 1996.
|
References |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionAppendix 1References |
|---|
1. Dobkin B. The economic impact of stroke. Neurology. 1995;45(suppl 1):S6-S9.
2.
Duncan PW, Goldstein LB, Matcher D, Divine GW,
Feussner J. Measurement of motor recovery after stroke: outcome
assessment and sample size requirements. Stroke. 1992;23:1084-1089.
3. Hosek S, Kane R, Carney M, Hartman J, Reboussin D, Serrato C, Melvin J. Charges and Outcomes for Rehabilitative Care. Santa Monica, Calif: Rand Corp; 1986. R-3424-HCFA.
4.
Sandin KJ, Smith BS. The measure of balance in
sitting in stroke rehabilitation prognosis. Stroke. 1990;21:82-86.
5. Alexander MP. Stroke rehabilitation outcome: a potential use of predictive variables to establish levels of care. Stroke. 1994;25:128-134. [Abstract]
6. Heinemann AW, Linacre JM, Wright BD, Hamilton BB, Granger CV. Prediction of rehabilitation outcomes with disability measures. Arch Phys Med Rehabil. 1994;75:133-143. [Medline] [Order article via Infotrieve]
7. Hyman MD. Social psychological determinants of patients' performance in stroke rehabilitation. Arch Phys Med Rehabil. 1972;53:217-226. [Medline] [Order article via Infotrieve]
8. Novack TA, Satterfield WT, Connor M. Stroke onset and rehabilitation: time lag as a factor in treatment outcome. Arch Phys Med Rehabil. 1984;65:316-319. [Medline] [Order article via Infotrieve]
9. Granger CV, Hamilton BB, Keith RA, Zielezny M, Sherwin FS. Advances in functional assessment for medical rehabilitation. Topics Geriatr Rehabil. 1986;1:59-74.
10. Guide for the Uniform Data System for Medical Rehabilitation, Version 4.0. Buffalo, NY: State University of New York at Buffalo, 1990.
11. Katz S, Ford AB, Moskowitz RW, Jackson BA, Jaff MW. Studies of illness in the aged—the index of ADL: a standardized measure of biological and psychosocial function. JAMA. 1963;185:914-919.
12. Mahoney FI, Barthel DW. Functional evaluation: the Barthel Index. Md Med J. 1965;14:61-65.
13. SAS Institute Inc. SAS Procedures Guide, Release 6.03 Edition. Cary, NC: SAS Institute Inc; 1991.
14. Kleinbaum DG, Kupper LL, Muller KE. Applied Regression Analysis and Other Multivariable Methods. 2nd ed. Boston, Mass: PWS-Kent; 1988.
15.
Mickey R, Greenland S. The impact of confounder
selection criteria on effect estimation. Am J
Epidemiol. 1989;129:125-133.
16.
Hanley JA, McNeil BJ. The meaning and use of the
area under a receiver operating characteristic (ROC) curve.
Radiology. 1982;143:29-36.
17. Hosmer DW, Lemeshow S. Applied Logistic Regression. New York, NY: John Wiley & Sons; 1989.
18. Horn SD, Horn RA. The computerized severity index: a new tool for case-mix management. J Med Syst. 1986;10:73-78. [Medline] [Order article via Infotrieve]
19. Knaus WA, Draper EA, Wagner DP, Zimmerman JE. APACHE II: a severity of disease classification system. Crit Care Med. 1985;13:818-829. [Medline] [Order article via Infotrieve]
20.
Le Gall JR, Lemeshow S, Saulnier F. A new
simplified acute physiology score (SAPS II) based on a European/North
American multicenter study. JAMA. 1993;270:2957-2963.
21. Stineman MG, Escarce JJ, Goin J, Hamilton BB, Granger CV, Williams SV. A case mix classification system for medical rehabilitation. Med Care. 1994;32:366-379. [Medline] [Order article via Infotrieve]
22.
Indredavik B, Bakke F, Solberg R, Rokseth R, Haaheim
LL, Holme I. Benefit of a stroke unit: a randomized controlled
trial. Stroke. 1991;22:1026-1031.
23. Kalra L, Crome P. The role of prognostic scores in targeting stroke rehabilitation in elderly patients. J Am Geriatr Soc. 1993;41:396-400. [Medline] [Order article via Infotrieve]
24. Langehorne P, O'Williams B, Gilchrist W, Howie K. Do stroke units save lives? Lancet. 1993;342:395-398. [Medline] [Order article via Infotrieve]
25. Smith ME, Garraway WM, Smith DL, Akhtar AJ. Therapy impact on functional outcome in a controlled trial of stroke rehabilitation. Arch Phys Med Rehabil. 1982;63:21-24. [Medline] [Order article via Infotrieve]
26. Stineman MG, Escarce JJ, Goin JE, Hamilton BB, Granger CV, Williams SV. A case mix classification system for medical rehabilitation. Med Care. 1994;32:366-379.
27. Stineman MG, Williams SV. Predicting inpatient rehabilitation length of stay. Arch Phys Med Rehabil. 1990;71:881-887. [Medline] [Order article via Infotrieve]
28. Granger CV, Hamilton BB, Gresham G, Kramer A. The stroke rehabilitation outcome study, part II: relative merits of the Barthel Index score and a four-item subscore in predicting patient outcomes. Arch Phys Med Rehabil. 1989;70:100-103. [Medline] [Order article via Infotrieve]
29. Harada N, Sofaer S, Kominski G. Functional status outcomes in rehabilitation: implications for prospective payment. Med Care. 1993;31:345-357. [Medline] [Order article via Infotrieve]
30.
Heinemann AW, Roth EJ, Cichowski K, Betts HB.
Multivariate analysis of improvement and outcome following stroke
rehabilitation. Arch Neurol. 1987;44:1167-1172.
31.
Loewen SC, Anderson BA. Predictors of stroke
outcome using objective measurement scales. Stroke. 1990;21:78-81.
32. Feinstein AR, Josephy BR, Wells CK. Scientific and clinical problems in indexes of functional disability. Ann Intern Med. 1986;105:413-420.
33. Kim J, Mueller CW. Factor Analysis: Statistical Methods and Practical Issues. Newbury Park, Calif: SAGE Publications, Inc; 1978.
34. McNeil BJ, Keeler E, Adelstein SJ. Primer on certain elements of medical decision making. N Engl J Med. 1975;293:211-215.[Abstract]
This article has been cited by other articles:
 |
|
 |
|
  M. I. Dallas, S. Rone-Adams, J. L. Echternach, L. M. Brass, and D. M. Bravata Dependence in Prestroke Mobility Predicts Adverse Outcomes Among Patients With Acute Ischemic Stroke Stroke, August 1, 2008; 39(8): 2298 - 2303. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 G. V. Ostir, I.-M. Berges, M. E. Ottenbacher, A. Clow, and K. J. Ottenbacher Associations Between Positive Emotion and Recovery of Functional Status Following Stroke Psychosom Med, May 1, 2008; 70(4): 404 - 409. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. L Fritz, S. Z George, S. L Wolf, and K. E Light Participant Perception of Recovery as Criterion to Establish Importance of Improvement for Constraint-Induced Movement Therapy Outcome Measures: A Preliminary Study Physical Therapy, February 1, 2007; 87(2): 170 - 178. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. P. Harriss Granulocyte colony-stimulating factor. Can. Med. Assoc. J., October 24, 2006; 175(9): 1095 - 1095. [Full Text] [PDF]
|
|
|
|
|
|
S. K. Schiemanck, G. Kwakkel, M. W.M. Post, L. J. Kappelle, and A. J.H. Prevo Predicting Long-Term Independency in Activities of Daily Living After Middle Cerebral Artery Stroke: Does Information From MRI Have Added Predictive Value Compared With Clinical Information? Stroke, April 1, 2006; 37(4): 1050 - 1054. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. Koyama, K. Matsumoto, T. Okuno, and K. Domen A new method for predicting functional recovery of stroke patients with hemiplegia: logarithmic modelling Clinical Rehabilitation, July 1, 2005; 19(7): 779 - 789. [Abstract] [PDF]
|
|
|
|
 |
|
 M. Stuart, C. Ryser, A. Levitt, S. Beer, J. Kesselring, S. Chard, and M. Weinrich Stroke Rehabilitation in Switzerland versus the United States: A Preliminary Comparison Neurorehabil Neural Repair, June 1, 2005; 19(2): 139 - 147. [Abstract] [PDF]
|
|
|
|
|
|
L. B. Goldstein, G. P. Samsa, D. B. Matchar, and R. D. Horner Charlson Index Comorbidity Adjustment for Ischemic Stroke Outcome Studies Stroke, August 1, 2004; 35(8): 1941 - 1945. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
I-P. Hsueh, W.-C. Wang, C.-F. Sheu, and C.-L. Hsieh Rasch Analysis of Combining Two Indices to Assess Comprehensive ADL Function in Stroke Patients Stroke, March 1, 2004; 35(3): 721 - 726. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. F. Gubrium, M. R. Rittman, C. Williams, M. E. Young, and C. A. Boylstein Benchmarking as Everyday Functional Assessment in Stroke Recovery J. Gerontol. B. Psychol. Sci. Soc. Sci., July 1, 2003; 58(4): S203 - 211. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C.-L. Hsieh, C.-F. Sheu, I-P. Hsueh, and C.-H. Wang Trunk Control as an Early Predictor of Comprehensive Activities of Daily Living Function in Stroke Patients Stroke, November 1, 2002; 33(11): 2626 - 2630. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
B. Han and W. E. Haley Family Caregiving for Patients With Stroke : Review and Analysis Stroke, July 1, 1999; 30(7): 1478 - 1485. [Abstract] [Full Text] [PDF]
|
|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||