Disability Rehabilitation An international, multidisciplinary journal
http://informahealthcare.com/dre ISSN 0963-8288 print/ISSN 1464-5165 online
informa
Disabil Rehabil, 2014; 36(15): 1273-1278
healthcare
© 2 0 1 4 Informa UK Ltd. DOI: 10.3109/09638288.2013.845253
RESEARCH PAPER
Predicting goal achievement during stroke rehabilitation for Medicare beneficiaries Suzanne R. O'Brien1 and Ying Xue2 1Nazareth College, Department o f Health and Human Services, Rochester, NY, USA and 2University o f Rochester, School o f Nursing, Rochester, NY, USA
Abstract
Keywords
Purpose: Few studies have investigated the ability o f treatment teams to predict functional improvement and whether an association between predicted goals and discharge function in patients with stroke exists. This study investigated goal prediction during stroke rehabilitation delivered in inpatient rehabilitation facilities (IRF) and the factors associated w ith goal prediction. Methods: A serial, cross-sectional design analyzing the Medicare IRF Patient Assessment Instrument dataset. The sample included 179479 admissions for stroke aged over 65 years in 968 IRFs. Generalized estimating equations (GEE) controlled for facility cluster effects were used for analysis o f time trends for length of stay (LOS), predicted Functional Independence Measure (FIM) scores, discharge FIM scores and predicted-discharge difference FIM scores (goal FIM scores minus discharge FIM scores). GEE models were employed to determine the correlation between predicted FIM and discharge FIM scores and factors associated with goal achievement. Results: Mean LOS, predicted FIM scores and discharge FIM scores decreased 1.8d, 2.2 points and 3.6 points, respectively, while predicted-discharge difference FIM scores increased 1.3 points. Discharge goals were not met 78.9% o f the time. After controlling for patient characteristics, each predicted FIM point was associated with 0.6 discharge FIM points ( p < 0.0001). Factors associated with not meeting or exceeding goals were: age (odds ratio; OR = 0.997), African Americans (OR = 0.905), number of comorbidities (OR = 0.970), number of complications (OR = 0.932) and right brain stroke (OR = 0.869). Factors associated with meeting or exceeding goals were: LOS (OR = 1.03), admission FIM score (OR=1.02) and females (OR = 1.05). Conclusions: Trends for lower goals and lower discharge function occurred over time. A correlation existed between predicted FIM scores and discharge FIM scores. Patient factors were associated with goal achievement.
Discharge outcomes, functional independence measure, goal prediction, inpatient rehabilitation facilities, inpatient rehabilitation facility patient assessment instrument History Received 13 February 2013 Revised 9 September 2013 Accepted 12 September 2013 Published online 23 October 2013
>- implications for Rehabilitation • Using the Functional Independence Measure, rehabilitation teams set lower goals for stroke rehabilitation in inpatient rehabilitation facilities during first 5.5 years o f the IRF-PAI dataset. • Discharge FIM scores also trended lower and fell at faster rate than goal FIM scores. • Teams' goal FIM scores averaged nearly 12 points higher than discharge FIM scores, and over 75% o f patients did not reach goals for the rehabilitation stay. • Factors associated w ith meeting or exceeding goals were: length o f stay, admission FIM scores and being a female. Factors associated w ith not meeting or exceeding goals were: age, number of comorbidities and complications, having a right-brain stroke and being African American.
Introduction In the United States, approximately 68 000 Medicare beneficiaries receive post-acute rehabilitation in an Inpatient Rehabilitation Facility (IRF) following a stroke each year [1], IRFs deliver comprehensive and multidisciplinary programs designed to improve cognitive and motor function following stroke. Care begins with assessment followed by setting appropriate goals that will accurately reflect the expected outcome of care [2], Setting Address for correspondence: Suzanne R. O’Brien, PhD, PT, Nazareth College, Department of Health and Human Services, 4245 East Avenue, Rochester, NY 14618, USA. E-mail: [email protected]
goals is an integral part of a rehabilitation program and reflects the treatment team’s expectations for patient progress during the rehabilitation stay, coordinates care, ensures comprehensive care and promotes program monitoring [3-5]. In addition, setting goals is considered a prudent part of discharge planning allowing for caregivers to efficiently plan for needs post discharge [6], Arriving at a goal requires a careful synthesis of patient goals, patient history, assessment results and clinical experience. Formal methods of goal setting exist, including goal attainment scaling [7], and the Canadian Occupational Performance Measure [8], which compare goals to results of care in a measureable manner. Proponents of measurable goal setting methods agree that patients
1274
S. R. O'Brien & Y. Xue
be involved and participate in goal setting for better motivation and adherence to the plan of care. Involving patients in goal setting, however, has been found to have inconsistent effects on functional outcomes [9], though difficult [10] and functional goals [11] have been found to motivate patients more effectively than easy or non-functional goals. Since 2002, the comprehensive patient assessment required in IRFs for Medicare admissions, the IRF Patient Assessment Instrument (IRF-PAI) has afforded the opportunity for treatment teams to set goals on the functional assessment portion of the tool. This portion of the IRF-PAI is called the Functional Independence Measure (FIM), which has been used in IRFs as the standard tool to measure functional improvement since the early 1990s [12], FIM items are functional in nature consisting of basic activities of daily living, physical and cognitive tasks. Scoring rules for the IRF-PAI allow that in addition to the actual admission FIM score, a predicted FIM score can also be reported [13], which allows for comparisons to the function achieved by discharge. No previous report has used the predicted FIM to investigate the relationship between goals of care and discharge function in IRFs following stroke. A certain amount of accuracy is required when setting goals to make them meaningful and motivating to patients. Achievement of such goals has not been commonly investigated. Inexperienced neurointensivists were found to correctly predict a six-month functional outcome in ventilator-dependent patients 80% (95% Cl: 72-86) of the time [14]. In addition, physical therapists’ predictions of three-month functional outcomes in patients with spinal cord injury were correlated with actual functional achieve ment (r = 0.53-0.92) [15], Despite these findings, there is a lack of understanding about how well rehabilitation teams can predict the achievement of discharge function. To date, only small studies have described goal setting in inpatient stroke settings and determined whether functional goals were achieved, with inconsistency noted between two systematic reviews [16,17], In addition, there is scant information about what factors are associated with accuracy of goal achievement. The purposes of this study were to examine the association between discharge goals set by the team and discharge function and to determine the factors associated with goal achievement in Medicare beneficiaries with stroke who received IRF rehabilitation. M ethods
The study was a secondary analysis of serial, cross-sectional Medicare IRF-PAI data from 1 January 2002 and 30 June 2007. Subjects were identified using admission group impairment codes that were unique to stoke. Subjects were aged 65 years and older, and only a first admission to an IRF for stroke rehabilitation was included. Subjects were excluded when a rehabilitation stay was not completed. This was deemed to occur when discharged against medical advice (n = 3840), expired in the IRF (n = 2213) or if admitted to the IRF for zero days (n = 605). The total sample consisted of 371211 cases in 1649 IRFs. The University of Rochester Research Subjects Review Board approved the study. V a ria b le s a n d m e asu res
Members of the treatment team score the FIM with the discipline responsible for scoring portions that relate to the tasks performed most often when under the care of that specific discipline [12,18], The FIM includes 18 items: eating, grooming, bathing, dressing upper body, dressing lower body, toileting, bladder management, bowel management, transferring to bed/chair, transferring to toilet, transferring to tub/shower, walk/wheelchair, stairs, com prehension, expression, social interaction, problem solving and memory. Total FIM scores were calculated by adding scores on
Disabil Rehabil, 2014; 36(15): 1273-1278
a scale from 0 to 7, with higher scores indicating better function. The FIM has been found to have internal consistency and to be sensitive to change over time across disparate diagnosis groups [19,20]. Inter-rater reliability has been demonstrated in scorers from different professions [21-23], concurrent [24], predictive [18,25] and construct validity [19] have all been established. The FIM was modified in 2002 when it was added to the IRF-PAI, and the internal consistency has remained high [26]. Discharge function was measured by the discharge FIM score. The predicted FIM score measured the expected level of function by discharge and was scored at admission. The outcome variable was measured by the difference between the predicted and discharge FIM scores. Predicted-discharge FIM difference scores greater than zero indicated that the predicted FIM scores were greater than the discharge FIM scores; therefore, goals were not met for the rehabilitation stay. Scores less than zero indicated that discharge FIM scores were greater than predicted FIM scores; therefore, goals were exceeded for the rehabilitation stay. When the difference score equaled zero, goals were met as planned. In d e p e n d e n t v a ria b le s
Length of stay (LOS) was measured by the period between the date of admission and date of discharge. Admission FIM scores measured the patients’ function upon admission to rehabilitation. We included the admission FIM score in the models because admission FIM scores have been found to predict discharge function [27,28]. Demographic variables controlled for in the model included gender, age and race/ethnicity. This study compared women’s outcomes to the reference group men, and age was include:, as a continuous variable. The following races/ethnicities were included in all models: (a) American Indian/Alaska Native, (b) Asian, (c) Black/African American, (d) Hispanic/Latino, (e) Native Hawaiian/Pacific Islander and (f) White, (g) mixed race and (h) unknown race. The addition of the categories for mixed race and unknown race was made because the IRF-PAI allows for greater than one choice of race/ ethnicity. The reference group for this category was Whites. The number of co-morbidities and complications to account for variation in patients’ LOS and outcomes and weighted each equally [29]. Stroke impairment groups included: left brain, right brain, bilateral brain, no paresis or other, stroke. The reference group was “ other stroke” . Finally, control for the year in which the admission occurred was added to models to identify cases within a specific year. The year in which admission to IRF occurred determined the year in which the subject was included. S ta tis tic a l analysis
Missing and incomplete data were found in 51.7% of predicted FIM scores in the sample data because this variable was not part of the mandatory reporting requirement for participating organizations [13]. Missing data occurred when no scores for any predicted FIM items were reported or when missing scores were present for some items leading to inaccurate predicted scores. Before analysis was undertaken, a comparison of characteristics was conducted between those with and without predicted FIM scores to examine whether differences existed between subjects. Analysis found no differences for any factors in the model, except for the parameter estimate for discharge FIM scores (0.2; p < 0.0001; Table 1). Despite statistical significance, the differ ence in discharge FIM scores between the groups was too small to be clinically meaningful [30]. Since analysis showed that the characteristics of the subsample remained similar to the full sample, analyses using predicted FIM scores was conducted using the subsample with complete predicted FIM scores, which consisted of 179479 subjects.
Goal prediction during stroke rehabilitation
DOI: 10.3109/09638288.2013.845253
Table 1. Characteristics of sample (n — 179479). Characteristics Mean Age (SD) Male (%) Race/ethnicity (%) White African American Hispanic Asian NH/PI AI/AN Unknown Mixed race Stroke impairment group (%) Right brain Left brain Bilateral brain No paresis Other Mean comorbidities (SD) Mean complications (SD)
78.6 (6.7) 44.1 82.3 11.9 3.7 1.5 0.3 0.2 0.1 0.1 40.9 39.8 2.7 12.1 4.5 7.9 (0.08) 0.72 (0.02)
FIM = functional independence measure.
Generalized estimating equations (GEE) were used to obtain annual trends for LOS, predicted and discharge FIM scores, and predicted-discharge FIM difference scores while controlling for the IRF clustering effect. Post-hoc testing using the TukeyKramer Test compared year-by-year results for differences [31]. The correlation between predicted FIM scores and discharge FIM scores was examined in three phases using GEE models controlling for IRF cluster effects. An unadjusted model regressed predicted FIM scores on discharge FIM scores, model 2 controlled for year of study and the last model was fully adjusted with all model covariates. The predicted-discharge difference FIM scores were categor ized into goals met, unmet or goals exceeded groups, and GEE binomial and multinomial models were employed to identify the factors associated discharge goals for the rehabilitation stay. For the binomial model, those that met goals were removed from analysis, leaving groups that either exceeded goals or did not meet goals. The probability that goals were exceeded was modeled in GEE analysis. Those that met goals were included in the multinomial analysis for comparison to instances of exceeded goals or not meeting goals [32,33]. The probability that goals were met or exceeded was modeled in GEE analysis. SAS 9.2 software was used for data analysis (SAS Institute Inc., Cary, NC) [34], R e s u lt s
Characteristics of the sample included are presented in Table 1. The average age was 78.6 (SD = 6.7) years, and the sample was 44.1% male and 82.3% White. The predominate forms of stroke impairments were right and left brain strokes. The sample had an average of eight comorbidities and relatively few complications. Figure 1 displays time trends across the 5.5-year study for LOS, predicted FIM scores, discharge FIM scores and the predicted-discharge FIM scores. Mean LOS decreased 1.8 d from 17.9 to 16.1 d, mean predicted FIM scores decreased 2.2 points from 92.9 to 90.7 points and mean discharge FIM scores decreased 3.6 points from 80.1 to 76.5 points (p
http://informahealthcare.com/dre ISSN 0963-8288 print/ISSN 1464-5165 online
informa
Disabil Rehabil, 2014; 36(15): 1273-1278
healthcare
© 2 0 1 4 Informa UK Ltd. DOI: 10.3109/09638288.2013.845253
RESEARCH PAPER
Predicting goal achievement during stroke rehabilitation for Medicare beneficiaries Suzanne R. O'Brien1 and Ying Xue2 1Nazareth College, Department o f Health and Human Services, Rochester, NY, USA and 2University o f Rochester, School o f Nursing, Rochester, NY, USA
Abstract
Keywords
Purpose: Few studies have investigated the ability o f treatment teams to predict functional improvement and whether an association between predicted goals and discharge function in patients with stroke exists. This study investigated goal prediction during stroke rehabilitation delivered in inpatient rehabilitation facilities (IRF) and the factors associated w ith goal prediction. Methods: A serial, cross-sectional design analyzing the Medicare IRF Patient Assessment Instrument dataset. The sample included 179479 admissions for stroke aged over 65 years in 968 IRFs. Generalized estimating equations (GEE) controlled for facility cluster effects were used for analysis o f time trends for length of stay (LOS), predicted Functional Independence Measure (FIM) scores, discharge FIM scores and predicted-discharge difference FIM scores (goal FIM scores minus discharge FIM scores). GEE models were employed to determine the correlation between predicted FIM and discharge FIM scores and factors associated with goal achievement. Results: Mean LOS, predicted FIM scores and discharge FIM scores decreased 1.8d, 2.2 points and 3.6 points, respectively, while predicted-discharge difference FIM scores increased 1.3 points. Discharge goals were not met 78.9% o f the time. After controlling for patient characteristics, each predicted FIM point was associated with 0.6 discharge FIM points ( p < 0.0001). Factors associated with not meeting or exceeding goals were: age (odds ratio; OR = 0.997), African Americans (OR = 0.905), number of comorbidities (OR = 0.970), number of complications (OR = 0.932) and right brain stroke (OR = 0.869). Factors associated with meeting or exceeding goals were: LOS (OR = 1.03), admission FIM score (OR=1.02) and females (OR = 1.05). Conclusions: Trends for lower goals and lower discharge function occurred over time. A correlation existed between predicted FIM scores and discharge FIM scores. Patient factors were associated with goal achievement.
Discharge outcomes, functional independence measure, goal prediction, inpatient rehabilitation facilities, inpatient rehabilitation facility patient assessment instrument History Received 13 February 2013 Revised 9 September 2013 Accepted 12 September 2013 Published online 23 October 2013
>- implications for Rehabilitation • Using the Functional Independence Measure, rehabilitation teams set lower goals for stroke rehabilitation in inpatient rehabilitation facilities during first 5.5 years o f the IRF-PAI dataset. • Discharge FIM scores also trended lower and fell at faster rate than goal FIM scores. • Teams' goal FIM scores averaged nearly 12 points higher than discharge FIM scores, and over 75% o f patients did not reach goals for the rehabilitation stay. • Factors associated w ith meeting or exceeding goals were: length o f stay, admission FIM scores and being a female. Factors associated w ith not meeting or exceeding goals were: age, number of comorbidities and complications, having a right-brain stroke and being African American.
Introduction In the United States, approximately 68 000 Medicare beneficiaries receive post-acute rehabilitation in an Inpatient Rehabilitation Facility (IRF) following a stroke each year [1], IRFs deliver comprehensive and multidisciplinary programs designed to improve cognitive and motor function following stroke. Care begins with assessment followed by setting appropriate goals that will accurately reflect the expected outcome of care [2], Setting Address for correspondence: Suzanne R. O’Brien, PhD, PT, Nazareth College, Department of Health and Human Services, 4245 East Avenue, Rochester, NY 14618, USA. E-mail: [email protected]
goals is an integral part of a rehabilitation program and reflects the treatment team’s expectations for patient progress during the rehabilitation stay, coordinates care, ensures comprehensive care and promotes program monitoring [3-5]. In addition, setting goals is considered a prudent part of discharge planning allowing for caregivers to efficiently plan for needs post discharge [6], Arriving at a goal requires a careful synthesis of patient goals, patient history, assessment results and clinical experience. Formal methods of goal setting exist, including goal attainment scaling [7], and the Canadian Occupational Performance Measure [8], which compare goals to results of care in a measureable manner. Proponents of measurable goal setting methods agree that patients
1274
S. R. O'Brien & Y. Xue
be involved and participate in goal setting for better motivation and adherence to the plan of care. Involving patients in goal setting, however, has been found to have inconsistent effects on functional outcomes [9], though difficult [10] and functional goals [11] have been found to motivate patients more effectively than easy or non-functional goals. Since 2002, the comprehensive patient assessment required in IRFs for Medicare admissions, the IRF Patient Assessment Instrument (IRF-PAI) has afforded the opportunity for treatment teams to set goals on the functional assessment portion of the tool. This portion of the IRF-PAI is called the Functional Independence Measure (FIM), which has been used in IRFs as the standard tool to measure functional improvement since the early 1990s [12], FIM items are functional in nature consisting of basic activities of daily living, physical and cognitive tasks. Scoring rules for the IRF-PAI allow that in addition to the actual admission FIM score, a predicted FIM score can also be reported [13], which allows for comparisons to the function achieved by discharge. No previous report has used the predicted FIM to investigate the relationship between goals of care and discharge function in IRFs following stroke. A certain amount of accuracy is required when setting goals to make them meaningful and motivating to patients. Achievement of such goals has not been commonly investigated. Inexperienced neurointensivists were found to correctly predict a six-month functional outcome in ventilator-dependent patients 80% (95% Cl: 72-86) of the time [14]. In addition, physical therapists’ predictions of three-month functional outcomes in patients with spinal cord injury were correlated with actual functional achieve ment (r = 0.53-0.92) [15], Despite these findings, there is a lack of understanding about how well rehabilitation teams can predict the achievement of discharge function. To date, only small studies have described goal setting in inpatient stroke settings and determined whether functional goals were achieved, with inconsistency noted between two systematic reviews [16,17], In addition, there is scant information about what factors are associated with accuracy of goal achievement. The purposes of this study were to examine the association between discharge goals set by the team and discharge function and to determine the factors associated with goal achievement in Medicare beneficiaries with stroke who received IRF rehabilitation. M ethods
The study was a secondary analysis of serial, cross-sectional Medicare IRF-PAI data from 1 January 2002 and 30 June 2007. Subjects were identified using admission group impairment codes that were unique to stoke. Subjects were aged 65 years and older, and only a first admission to an IRF for stroke rehabilitation was included. Subjects were excluded when a rehabilitation stay was not completed. This was deemed to occur when discharged against medical advice (n = 3840), expired in the IRF (n = 2213) or if admitted to the IRF for zero days (n = 605). The total sample consisted of 371211 cases in 1649 IRFs. The University of Rochester Research Subjects Review Board approved the study. V a ria b le s a n d m e asu res
Members of the treatment team score the FIM with the discipline responsible for scoring portions that relate to the tasks performed most often when under the care of that specific discipline [12,18], The FIM includes 18 items: eating, grooming, bathing, dressing upper body, dressing lower body, toileting, bladder management, bowel management, transferring to bed/chair, transferring to toilet, transferring to tub/shower, walk/wheelchair, stairs, com prehension, expression, social interaction, problem solving and memory. Total FIM scores were calculated by adding scores on
Disabil Rehabil, 2014; 36(15): 1273-1278
a scale from 0 to 7, with higher scores indicating better function. The FIM has been found to have internal consistency and to be sensitive to change over time across disparate diagnosis groups [19,20]. Inter-rater reliability has been demonstrated in scorers from different professions [21-23], concurrent [24], predictive [18,25] and construct validity [19] have all been established. The FIM was modified in 2002 when it was added to the IRF-PAI, and the internal consistency has remained high [26]. Discharge function was measured by the discharge FIM score. The predicted FIM score measured the expected level of function by discharge and was scored at admission. The outcome variable was measured by the difference between the predicted and discharge FIM scores. Predicted-discharge FIM difference scores greater than zero indicated that the predicted FIM scores were greater than the discharge FIM scores; therefore, goals were not met for the rehabilitation stay. Scores less than zero indicated that discharge FIM scores were greater than predicted FIM scores; therefore, goals were exceeded for the rehabilitation stay. When the difference score equaled zero, goals were met as planned. In d e p e n d e n t v a ria b le s
Length of stay (LOS) was measured by the period between the date of admission and date of discharge. Admission FIM scores measured the patients’ function upon admission to rehabilitation. We included the admission FIM score in the models because admission FIM scores have been found to predict discharge function [27,28]. Demographic variables controlled for in the model included gender, age and race/ethnicity. This study compared women’s outcomes to the reference group men, and age was include:, as a continuous variable. The following races/ethnicities were included in all models: (a) American Indian/Alaska Native, (b) Asian, (c) Black/African American, (d) Hispanic/Latino, (e) Native Hawaiian/Pacific Islander and (f) White, (g) mixed race and (h) unknown race. The addition of the categories for mixed race and unknown race was made because the IRF-PAI allows for greater than one choice of race/ ethnicity. The reference group for this category was Whites. The number of co-morbidities and complications to account for variation in patients’ LOS and outcomes and weighted each equally [29]. Stroke impairment groups included: left brain, right brain, bilateral brain, no paresis or other, stroke. The reference group was “ other stroke” . Finally, control for the year in which the admission occurred was added to models to identify cases within a specific year. The year in which admission to IRF occurred determined the year in which the subject was included. S ta tis tic a l analysis
Missing and incomplete data were found in 51.7% of predicted FIM scores in the sample data because this variable was not part of the mandatory reporting requirement for participating organizations [13]. Missing data occurred when no scores for any predicted FIM items were reported or when missing scores were present for some items leading to inaccurate predicted scores. Before analysis was undertaken, a comparison of characteristics was conducted between those with and without predicted FIM scores to examine whether differences existed between subjects. Analysis found no differences for any factors in the model, except for the parameter estimate for discharge FIM scores (0.2; p < 0.0001; Table 1). Despite statistical significance, the differ ence in discharge FIM scores between the groups was too small to be clinically meaningful [30]. Since analysis showed that the characteristics of the subsample remained similar to the full sample, analyses using predicted FIM scores was conducted using the subsample with complete predicted FIM scores, which consisted of 179479 subjects.
Goal prediction during stroke rehabilitation
DOI: 10.3109/09638288.2013.845253
Table 1. Characteristics of sample (n — 179479). Characteristics Mean Age (SD) Male (%) Race/ethnicity (%) White African American Hispanic Asian NH/PI AI/AN Unknown Mixed race Stroke impairment group (%) Right brain Left brain Bilateral brain No paresis Other Mean comorbidities (SD) Mean complications (SD)
78.6 (6.7) 44.1 82.3 11.9 3.7 1.5 0.3 0.2 0.1 0.1 40.9 39.8 2.7 12.1 4.5 7.9 (0.08) 0.72 (0.02)
FIM = functional independence measure.
Generalized estimating equations (GEE) were used to obtain annual trends for LOS, predicted and discharge FIM scores, and predicted-discharge FIM difference scores while controlling for the IRF clustering effect. Post-hoc testing using the TukeyKramer Test compared year-by-year results for differences [31]. The correlation between predicted FIM scores and discharge FIM scores was examined in three phases using GEE models controlling for IRF cluster effects. An unadjusted model regressed predicted FIM scores on discharge FIM scores, model 2 controlled for year of study and the last model was fully adjusted with all model covariates. The predicted-discharge difference FIM scores were categor ized into goals met, unmet or goals exceeded groups, and GEE binomial and multinomial models were employed to identify the factors associated discharge goals for the rehabilitation stay. For the binomial model, those that met goals were removed from analysis, leaving groups that either exceeded goals or did not meet goals. The probability that goals were exceeded was modeled in GEE analysis. Those that met goals were included in the multinomial analysis for comparison to instances of exceeded goals or not meeting goals [32,33]. The probability that goals were met or exceeded was modeled in GEE analysis. SAS 9.2 software was used for data analysis (SAS Institute Inc., Cary, NC) [34], R e s u lt s
Characteristics of the sample included are presented in Table 1. The average age was 78.6 (SD = 6.7) years, and the sample was 44.1% male and 82.3% White. The predominate forms of stroke impairments were right and left brain strokes. The sample had an average of eight comorbidities and relatively few complications. Figure 1 displays time trends across the 5.5-year study for LOS, predicted FIM scores, discharge FIM scores and the predicted-discharge FIM scores. Mean LOS decreased 1.8 d from 17.9 to 16.1 d, mean predicted FIM scores decreased 2.2 points from 92.9 to 90.7 points and mean discharge FIM scores decreased 3.6 points from 80.1 to 76.5 points (p
