Stroke Rehabilitation Patients, Practice, and Outcomes: Is Earlier and More Aggressive Therapy Better?

Article Outline

• Abstract
• Methods
  • Subsets of Patients With Moderate and Severe Strokes
  • Analysis Methods
• Results
  • Patient, Process, and Outcomes Characteristics
  • Regression Results for All Patients With Moderate and Severe Stroke
  • Demographic Variables
  • Health and Functional Status Variables
    • Stroke location
    • Admission FIM score
    • Severity of illness
    • Time of onset of symptoms to rehabilitation admission
  • Process Variables
    • Length of stay
    • Therapy
    • Medications
    • Tube feeding
  • Regression Results for Patients Admitted at FIM Locomotion Level 1 or Toilet Transfer Level 1
• Discussion
• Limitations
• Conclusions
• Acknowledgments
• Appendix 1. Variables allowed to enter and leave regression models
• References
• Copyright

Abstract 

Horn SD, DeJong G, Smout RJ, Gassaway J, James R, Conroy B. Stroke rehabilitation patients, practice, and outcomes: is earlier and more aggressive therapy better?

Objective

To examine associations of patient characteristics, rehabilitation therapies, neurotropic medications, nutritional support, and timing of initiation of rehabilitation with functional outcomes and discharge destination for inpatient stroke rehabilitation patients.

Design

Prospective observational cohort study.

Setting

Five U.S. inpatient rehabilitation facilities.

Participants

Poststroke rehabilitation patients (N=830; age, >18y) with moderate or severe strokes, from the Post-Stroke Rehabilitation Outcomes Project database.

Interventions

Not applicable.

Main Outcome Measures

Discharge total, motor, and cognitive FIM scores and discharge destination.

Results

Controlling for patient differences, various activities and interventions were associated with better outcomes including earlier initiation of rehabilitation, more time spent per day in higher-level rehabilitation activities such as gait, upper-extremity control, and problem solving, use of newer psychiatric medications, and enteral feeding. Several findings part with conventional practice, such as starting gait training in the first 3 hours of physical therapy, even for low-level patients, was associated with better outcomes.

Conclusions

Specific therapy activities and interventions are associated with better outcomes. Earlier rehabilitation admission, higher-level activities early in the rehabilitation process, tube feeding, and newer medications are associated with better stroke rehabilitation outcomes.

Key Words:  Cerebrovascular accident , Outcome assessment , Rehabilitation , Severity of illness , Stroke

A MAJOR CHALLENGE in stroke rehabilitation practice is how best to customize available rehabilitation resources to meet the needs of patients to optimize outcomes.¹ Failure to optimize rehabilitation interventions and therapies can result in too little or too much care relative to a patient’s needs and preferred outcomes. The association of stroke rehabilitation outcomes with process of care and patient characteristics has not been studied comprehensively. Stroke rehabilitation studies typically have been limited to a single set or subset of interventions and rarely examine all the processes of care concurrently.² Other studies, most of which involved limited numbers of patients, have described physical therapy (PT),³, ⁴, ⁵, ⁶, ⁷ occupational therapy (OT),⁸, ⁹, ¹⁰, ¹¹, ¹² and speech and language pathology (SLP)¹³, ¹⁴, ¹⁵, ¹⁶ in terms of duration or frequency but have rarely described specific activities preformed during therapy sessions.

The introductory article in this series presents the motivation, purpose, and scope of this study, as well as an extended literature review that establishes the case for the multicenter Post-Stroke Rehabilitation Outcomes Project (PSROP)¹ on which the findings presented in this article are based. Other articles in this series have documented the nature, scope, and variation of stroke rehabilitation practice as uncovered in the PSROP.¹⁷, ¹⁸, ¹⁹, ²⁰, ²¹, ²², ²³, ²⁴

Building on previous articles in this series that identified individual links between stroke rehabilitation patient characteristics, practices, and outcomes, this article seeks to put all of these together and describe the most significant associations between patient characteristics, PT, OT, SLP, neurotropic medications, nutritional support, and timing of initiation of rehabilitation with motor and cognitive functional outcomes and discharge destination. In short, we want to determine how specific rehabilitation therapies relate to outcomes, taking into account patient covariates.

One suggestion that emerged in previous PSROP articles is that challenging patients to perform higher-order tasks as early as possible in their rehabilitation stay, even when they may not appear ready to take on such activities, is associated with better outcomes. In other words, stroke rehabilitation patients may be able to leap-frog over lower-level activities prescribed by current traditional practice. This article further tests the hypothesis that earlier and more aggressive therapies (such as earlier rehabilitation, newer medications, enteral feeding, and higher-level therapies from physical, occupational, and speech and language therapists) are associated with better outcomes, taking into account each patient’s demographic, health, and functional profile. The leap-frog hypothesis challenges conventional wisdom in rehabilitation that patients should move incrementally through the rehabilitation process and that patients should be challenged to perform activities that are only a notch above their previous level of performance in the rehabilitation process. Conventional wisdom is based, in part, on the human development axiom that one must learn to crawl before one can walk and on the notion that the patient should not be challenged excessively for fear that it may induce a sense of failure or stress, if not depression, and thus compromise outcome.

The PSROP is well equipped to evaluate associations between stroke rehabilitation patients, processes, and outcomes.¹⁷ It provides detailed, comprehensive data on stroke patient characteristics, rehabilitation treatments and interventions, and outcomes. It allows clinicians and researchers to drill down to the most meaningful level of resolution regarding the types of care rendered. Previous studies, as noted in the introductory article of this supplement, do not provide this level of resolution,¹ nor do they provide the data required to determine how various sequences of services or activities may prove more efficient and effective than others in achieving better functional outcomes and more independent postdischarge living arrangements. This article presents promising insights that sometimes contradict conventional wisdom in stroke rehabilitation and suggest further exploration that is beyond the immediate scope of this article.

Back to Article Outline

Methods 

The methodology governing the full PSROP is provided by Gassaway et al¹⁷; Gassaway provides a detailed description of the larger study’s participating facilities, patient selection criteria, data collection instruments including their validity and reliability, and a detailed description of the project’s final study group. The methodology is summarized in Maulden et al.²³ The institutional review boards at Boston University and at each participating inpatient rehabilitation facility (IRF) approved the study.

Subsets of Patients With Moderate and Severe Strokes 

We examined a subset (n=1079) of the total 1161 patients in the U.S. PSROP database who had FIM scores available to categorize into case-mix groups (CMGs). Because we wanted to analyze the effects of the 3 primary rehabilitation therapies (PT, OT, SLP) and 1 site provided almost no SLP information to the PSROP database, we deleted all patients from that 1 site. To maintain sample sizes large enough to detect small effects, CMGs were combined into moderate (CMGs 104−107, 389 patients) and severe (CMGs 108−114, 441 patients) patient groups. We focused regression analyses on patients with moderate and severe strokes; there were too few patients with mild stroke to be analyzed at this time (CMGs 101−103, 62 patients).

Here we briefly define the variables found to be significant in the multivariate analyses that follow.

Patient variables (table 1) include demographic characteristics, health and functional status characteristics (type and location of stroke, body mass index [BMI], admission functional status [FIM score], admission severity of illness [Comprehensive Severity Index (CSI) and its components]), indications of neurobehavioral impairments, and prerehabilitation health care information. BMI on admission was categorized as underweight (<18.5kg/m²), normal (18.5−24.9kg/m²), overweight (25.0−29.0kg/m²), and obese (≥30.0kg/m²). Time from stroke symptom onset to rehabilitation admission was calculated from the number of days from first symptom onset to admission to a dedicated rehabilitation unit.

Table 1. Patient Variables for Moderate (CMGs 104–107) and Severe (CMGs 108–114) Stroke Groups for Multiple Regression Analyses

Patient Variables	CMGs 104–107 (n=389)	CMGs 108–114 (n=441)	P
Demographic and health plan characteristics
Mean age (y)	66.2	67.9	.092⁎
Female (%)	48.1	46.9	.781†
Race (%)			.611†
White	64.5	61.2
Black	16.7	18.6
Other	18.8	20.2
Payer (%)			.102†
Medicare	57.3	63.0
Other	42.7	37.0
Health and functional status characteristics
Type of stroke (%)			.068†
Hemorrhagic	22.9	28.6
Ischemic	77.1	71.4
Side of stroke (%)			.304†
Right	46.3	42.2
Left	42.9	43.3
Bilateral	9.3	11.6
Unknown	1.5	3.0
Location of stroke (%)			.030†
Brainstem/cerebellum	20.1	13.8
Subcortical	30.9	39.2
Brainstem + subcortical	6.2	4.1
Lobar	37.3	37.4
Unknown	5.7	5.4
BMI/weight (%)			.115†
Underweight BMI (<18.5kg/m2)	4.4	3.9
Normal BMI (18.5–24.9kg/m2)	44.5	36.7
Overweight BMI (25–29.9kg/m2)	33.4	37.6
Obese BMI (≥30kg/m2)	17.7	21.8
Mean admission total FIM ± SD	71.6±9.9	43.1±12.6	<.001⁎
Mean admission motor FIM ± SD	47.8±5.7	26.6±7.2	<.001⁎
Mean admission cognitive FIM ± SD	23.8±7.3	16.5±7.6	<.001⁎
Mean admission CSI ± SD	15.8±10.4	27.3±15.2	<.001⁎
Stroke symptoms
Dysphagia (%)			<.001†
Normal	54.2	21.5
Dysphagia not otherwise specified	32.7	44.9
Unable to swallow liquids or solids	13.1	33.6
Motor impairment (%)			<.001†
Mild motor impairment	10.0	7.5
Moderate motor impairment	86.1	76.0
Severe motor impairment	3.9	16.6
Aphasia (%)	14.9	32.9	<.001†
Neurobehavioral impairment			<.001†
Mood/behavior disturbances	38.3	33.6
Cognitive dysfunction	4.1	8.8
Both	8.0	16.8
Certain neurotropic medications, no mood/behavior or cognitive dysfunction	22.9	26.5
None	26.7	14.3
Prerehabilitation health care
Mean no. of days from stroke symptom onset to rehabilitation admission ± SD	11.4±12.7	18.5±29.5	<.001⁎

Abbreviation: SD, standard deviation.

CSI, a disease-specific severity assessment system, calculates severity scores using individual components of physical findings and laboratory results at specified levels of abnormality found in a resident’s chart based on diseases defined by International Classification of Diseases, 9th Revision (ICD-9),²⁵ coding. For stroke diagnosis, CSI components include degree of alertness, ataxia, aphasia, dysarthria, dyspnea, perceptual and sensation impairment, dysphagia, hemiplegia, lesion level, time postinjury, and acute confusion. The functional performance for each study patient on admission to and discharge from inpatient rehabilitation was obtained via retrospective chart review using the study site’s reporting of the FIM. We assumed all clinicians providing FIM data within IRFs as part of standard practice were FIM credentialed.¹⁷

A patient was defined as having neurobehavioral impairments if any of the following were present: (1) the patient had diagnoses associated with neurologic or behavioral impairment(s) documented in their chart (eg, major depression, ICD-9 codes 296.2 and 296.3); (2) mood or behavioral impairments were indicated in charted descriptors such as combative, agitated, restless, aggressive, anxious, depressed, emotionally labile, having hallucinations, flat affect, or impulsive; (3) cognitive impairments were indicated in charted descriptors such as decreased safety awareness, impaired or poor judgment or concentration, impaired memory, confused, disoriented, or lethargic; and (4) patients received certain neurotropic medication(s) but had no charted descriptions of mood/behavior or cognitive impairments.²¹ These neurotropic medications included antidepressants, benzodiazepines, anxiolytics, and antipsychotics. Process variables (table 2) included rehabilitation length of stay (LOS); details of PT, OT, and SLP activities derived from point-of-care intervention documentation forms; and use of specific treatments, including nutrition supplementation via tube feeding and neurotropic medications, obtained from postdischarge chart review.¹⁷, ¹⁸, ¹⁹, ²⁰, ²¹, ²², ²³

Table 2. Process Variables for Moderate (CMGs 104–107) and Severe (CMGs 108–114) Stroke Groups for Multiple Regression Analyses

Process Variables	CMGs 104–107 (n=389)	CMGs 108–114 (n=441)	P
Mean LOS	15.7±7.2	24.5±10.9	<.001⁎
PT (mean ± SD)
No. of min/d	43.5±13.6	41.4±13.9	.033⁎
Activities (no. of min/d)
Bed mobility	0.7±1.0	2.5±2.3	<.001⁎
Sitting	0.6±1.3	2.6±3.6	<.001⁎
Transfers	3.2±3.1	6.1±3.9	<.001⁎
Sit-to-stand	2.0±2.3	3.6±2.7	<.001⁎
Wheelchair mobility	0.5±0.9	1.5±1.4	<.001⁎
Pregait	3.1±3.1	3.3±3.0	.421⁎
Gait	16.5±7.9	10.4±7.5	<.001⁎
Advanced gait	2.9±3.5	1.0±1.7	<.001⁎
Community mobility	1.2±2.5	0.5±1.5	<.001⁎
OT (mean ± SD)
No. of min/d	40.9±15.3	39.1±15.6	.080⁎
Activities (no. of min/d)
Bathing	2.1±2.5	2.1±2.3	.430⁎
Dressing	5.5±4.7	7.1±5.3	<.001⁎
Grooming	1.6±1.8	2.7±2.6	<.001⁎
Toileting	1.2±1.7	1.5±1.9	.025⁎
Feeding/eating	0.8±2.7	1.4±3.4	.001⁎
Transfers	2.0±2.0	2.3±2.7	.044⁎
Bed mobility	0.1±0.4	0.4±0.8	<.001⁎
Functional mobility	3.5±3.8	1.6±2.0	<.001⁎
Home management	3.9±4.8	1.3±2.1	<.001⁎
Community integration	2.0±3.2	0.8±1.9	<.001⁎
Leisure performance	0.8±1.5	0.7±1.3	.580⁎
Upper-extremity control	9.3±8.4	9.3±6.6	.989⁎
Wheelchair management	0.3±0.7	0.5±1.1	.001⁎
Sitting balance/trunk control	0.6±1.2	1.5±2.1	<.001⁎
SLP (mean ± SD)
No. of min/d	25.6±16.2	31.5±15.2	<.001⁎
Activities (no. of min/d)
Swallowing	3.4±6.5	6.7±8.3	<.001⁎
Speech/intelligibility	2.1±4.2	2.3±4.0	.394⁎
Voice	0.4±1.5	0.8±2.2	.005⁎
Verbal expression	2.9±4.8	4.0±5.4	.002⁎
Alternative/nonverbal expression	0.3±1.3	0.6±1.8	.005⁎
Written expression	0.9±2.1	0.8±1.7	.408⁎
Auditory comprehension	1.6±2.8	3.1±4.1	<.001⁎
Reading comprehension	1.4±2.3	1.4±2.2	.890⁎
Problem solving/reasoning	3.8±5.6	3.3±4.5	.154⁎
Orientation	0.6±1.5	1.1±2.0	<.001⁎
Attention	0.9±2.2	1.7±2.9	<.001⁎
Memory	1.4±3.0	1.4±2.5	.743⁎
Pragmatics	0.1±0.5	0.1±0.5	.551⁎
Executive functional skills	0.7±1.5	0.4±1.2	.001⁎
Tube feeding use during rehabilitation (%)			<.001†
Use discontinued before discharge	5.4	24.9
Use continued on discharge	1.3	6.4
None	93.3	68.7
Neurotropic medications (%)
Old anticonvulsant	2.8	3.0	1.000†
Opioid analgesics	18.6	29.5	<.001†
Analgesic; muscle relaxant	5.4	9.5	.026†
New SSRIs	7.5	14.3	.002†
Old SSRIs	18.0	21.8	.192†
Atypical antipsychotics	4.6	12.9	<.001†
Anti-Parkinson’s	6.4	12.5	.003†
Anxiolytics	1.0	3.0	.083†
Modafinil	0.5	8.6	<.001†
Neurostimulants	3.3	13.2	<.001†
Old antinausea/antivomiting	10.3	17.7	.003†
Other antidepressants	18.0	35.4	<.001⁎
Benzodiazepines	11.3	13.2	.459†

Abbreviation: SSRIs, selective serotonin reuptake inhibitors.

The study’s physicians, nurses, psychologists, social workers, and physical, occupational, recreational, and speech therapists each completed point-of-care intervention documentation forms for each patient treatment session. We calculated the total number of minutes per patient per day spent in each therapy (PT, OT, SLP) and in each therapy activity by dividing the total (full stay) number of minutes in each therapy activity by the LOS.¹⁸, ¹⁹, ²⁰

Tube-feeding data included date, type, and reason a tube was placed and start and stop times of enteral formulas.²² Based on these data, we divided patients into 3 tube-feeding groups: (1) no tube feeding during rehabilitation (n=666), (2) tube feeding at any time during rehabilitation but discontinued before discharge (n=131), and (3) tube feeding for 100% of rehabilitation stay and discharged on tube feeding support (n=33). “Discharged on tube feeding support” was defined as (1) the patient’s last ordered diet type was nothing by mouth (no oral intake) or a speech and language pathologist was supervising all oral intake, (2) the CSI discharge severity indicator of the patient’s dysphagia status 24 hours before discharge indicated that the patient was unable to swallow liquids or solids, (3) a percutaneous endoscopic gastrostomy or other gastrostomy tube was in place, and (4) the patient was discharged to a skilled nursing facility (SNF) or home health. Group 3 patients—patients who received tube feeding for their entire rehabilitation stay and were discharged with tube feeding (n=33)—were excluded from regression analyses. They were sicker patients overall (higher CSI scores) but had similar motor and cognitive abilities on admission as other tube-fed patients. However, they had significantly lower abilities at the time of discharge, showing lack of progress during rehabilitation, which is supported by their short LOSs and discharge to institutional care.²² The project team determined they were an outlier group for whom the severity of dysphasia and subsequent recovery time frame were well outside usual recovery patterns.

Neurotropic medication information was collected at the drug level (including details about dosing and timing) and then grouped into categories by consensus of prescribing physicians of the PSROP clinical team based on similarity of drug content and effects on patients. Medications contained in drug categories used in these analyses, structured around medication groupings found in ePocrates,²⁶ are listed elsewhere.²¹

Outcome variables (table 3) included discharge function, severity of illness, and discharge destination. Function, as measured by the FIM, was captured as recorded at discharge, and change in FIM score (total, motor, cognitive) from admission to discharge was calculated. We captured the maximum CSI score and calculated increases in severity during rehabilitation from admission CSI to maximum CSI scores, which includes the most aberrant signs and symptoms regardless of when they occur.

Table 3. Outcome Variables for Moderate (CMGs 104–107) and Severe (CMGs 108–114) Stroke Groups for Multiple Regression Analyses

Outcome Variables	CMGs 104–107 (n=389)	CMGs 108–114 (n=441)	P
Severity (CSI) during rehabilitation
Mean maximum CSI ± SD	23.0±14.7	41.4±23.3	<.001⁎
Increase in severity (maximum − admission)	7.2±8.3	14.1±13.3	<.001⁎
Mean discharge CSI ± SD	6.4±7.1	14.8±14.1	<.001⁎
FIM
Mean discharge total FIM ± SD	97.7±12.8	72.3±21.7	<.001⁎
Mean increase in total FIM ± SD (discharge − admission)	26.2±10.8	29.1±16.9	.003⁎
Mean discharge motor FIM	69.9±10.0	50.4±16.8	<.001⁎
Mean increase in motor FIM ± SD (discharge − admission)	22.2±9.2	23.8±14.3	.051⁎
Mean discharge cognitive FIM ± SD	27.7±6.1	21.8±7.6	<.001⁎
Mean increase in cognitive FIM ± SD (discharge − admission)	4.0±3.6	5.2±4.5	<.001⁎
Discharge destination (%)			<.001†
Home/community	93.3	67.1
SNF	4.9	24.5
Hospital or other rehab	1.8	8.4

Analysis Methods 

Descriptive statistics were used to compare patient characteristics, therapy interventions, and outcomes for patients with moderate and severe strokes (see Table 1, Table 2, Table 3). Chi-square tests were used for categoric data and t tests or analysis of variance for continuous data.

We used ordinary least squares (OLS) multiple regression to examine associations between “onset days” (days from symptom onset to rehabilitation admission), medications used, nutritional support, and minutes of PT, OT, and SLP activity per patient per day with each patient’s functional outcome at discharge, controlling for patient characteristics, stroke symptoms, neurobehavioral impairment, and rehabilitation LOS. We used logistic regression analyses to determine associations of the same patient characteristics and treatments with the outcomes of discharge destination to home or community or achieving specified increases in FIM components.

Variables entering regression models were checked for multicollinearity; no correlations were greater than .60. Stepwise R² selection procedure for OLS regressions allowed independent variables to enter and leave each model. The importance of each predictor was determined by its F value (or Wald chi square in logistic regression). We created the most parsimonious model for each outcome by allowing only significant (P<.05) variables to remain in the model. Variables that were allowed to enter models are listed in appendix 1. All analyses were performed with SAS statistical software.^a

Analyses were performed within moderate and severe (CMG) stroke subpopulations. For analyses involving FIM outcomes, we excluded patients who were discharged to acute care or to another rehabilitation facility (7 patients with moderate and 37 with severe stroke) because we did not have access to FIM data scored by other facilities on discharge to home or SNF. Missing continuous data resulted in exclusion of those subjects from analyses.

For both the moderate and severe stroke CMG groupings, we performed separate regression analyses that included (1) variables based on therapy activities during the entire rehabilitation stay (Table 4, Table 5) and (2) variables based on therapy activities during the first block of therapy only (Table 6, Table 7, Table 8). First block of therapy is defined as receiving PT, OT, and SLP for at least 3 hours each (4h for OT) and includes activity time during only the first 3-hour (4h for OT) block. Thus, sample sizes are smaller in the first block of therapy only (see Table 6, Table 7, Table 8), because some rehabilitation patients received therapy for less than the defined first block period. Reasons for defining these blocks of time are presented elsewhere.¹⁸, ¹⁹, ²⁰

Table 4. Full Regressions for Moderate Stroke Patients CMGs 104–107

Independent Variables	Discharge FIM (R2=.604, n=372)⁎†			Discharge Motor FIM (R2=.511, n=373)⁎			Discharge Cognitive FIM (R2=.793, n=376)†
	Coeff	F	P	Coeff	F	P	Coeff	F	P
Patient variables
Age	−0.109	15.01	<.001	−0.073	9.04	.003
Female				−1.641	5.00	.026
BMI underweight	−4.795	4.42	.036
Stroke location: brainstem/cerebellum				−2.382	7.03	.008
Admission FIM motor score	0.411	24.10	<.001	0.403	31.39	<.001
Admission FIM cognitive score	0.666	86.25	<.001				0.668	732.56	<.001
Aphasia							−1.147	6.01	.015
Moderate motor impairment				−3.000	6.96	.009
No. of days from stroke symptom onset to rehabilitation admission	−0.135	16.68	<.001	−0.090	9.97	.002
Process variables
LOS	−0.153	5.65	.018	−0.231	15.59	<.001
PT: formal assessment (min per patient per d)				−0.374	10.41	.002
PT: bed mobility (min per patient per d)	−1.503	8.14	.005	−1.300	8.75	.003
PT: sitting (min per patient per d)	−0.781	4.98	.026
PT: transfers (min per patient per d)	−0.373	5.26	.022	−0.395	7.78	.006
PT: gait (min per patient per d)				0.129	6.48	.011
PT: advanced gait (min per patient per d)				0.247	4.86	.028
OT: feeding/eating (min per patient per d)							−0.139	6.50	.011
OT: bathing (min per patient per d)	0.595	9.41	.002
OT: toileting (min per patient per d)	−1.356	22.99	<.001	−0.803	12.26	.001
OT: sitting balance (min per patient per d)	−0.971	7.76	.006
OT: transfers (min per patient per d)				−0.426	4.67	.031
OT: home management (min per patient per d)	0.249	7.25	.007	0.261	10.16	.002
OT: upper-extremity control (min per patient per d)	0.189	12.16	<.001	0.130	7.32	.007
SLP: speech/intelligibility (min per patient per d)	−0.223	5.14	.024	−0.193	5.16	.024
SLP: voice (min per patient per d)	1.064	12.76	<.001	0.627	6.54	.011
SLP: auditory comprehension (min per patient per d)	−0.570	11.14	<.001	−0.397	8.73	.003	−0.162	7.59	.006
SLP: reading comprehension (min per patient per d)				0.341	4.22	.041
SLP: problem solving (min per patient per d)	0.208	7.00	.009	0.137	4.41	.037	0.054	4.18	.042
SLP: attention (min per patient per d)	0.529	6.44	.012
SLP: executive functioning (min per patient per d)	−0.696	5.92	.016
Medications
Oploid analgesics	3.140	8.68	.003	2.227	5.73	.017	0.725	3.90	.049
Atypical antipsychotics	6.127	9.28	.003	4.625	7.29	.007
New SSRIs							1.260	4.96	.027
Anti-Parkinson’s	−7.641	19.65	<.001	−4.577	9.60	.002	−1.894	9.91	.002

NOTE. Blank cells refer to variables and the coefficient, F, and P values that did not enter the model significantly.

Abbreviation: Coeff, coefficient.

Table 5. Full Regressions for Severe Stroke Patients CMGs 108–114

Independent Variables	Discharge FIM (R2=.728, n=372)⁎			Discharge Motor FIM (R2=.676, n=372)⁎			Discharge Cognitive FIM (R2=.796, n=376)			Discharge Home and Assisted Living (c=.836, n=413)
	Coeff	F	P	Coeff	F	P	Coeff	F	P	Coeff	Wald	P
Patient variables
Age	−0.219	25.80	<.001	−0.202	31.94	<.001
Race: black	−3.253	4.74	.030	−2.981	5.58	.019
Admission FIM motor score	0.430	12.96	<.001	0.378	15.48	<.001				0.075	10.39	.001
Admission FIM cognitive score	0.880	81.82	<.001	0.297	14.99	<.001	0.575	345.80	<.001
Aphasia							−1.429	8.48	.004
Mild motor impairment	6.041	7.24	.008	5.010	7.14	.008
Neurobehavioral impairment: mood and cognitive disturbances (both)	−4.185	6.64	.010				−2.389	24.54	<.001
No. of days from stroke symptom onset to rehabilitation admission	−0.085	18.01	<.001	−0.080	22.81	<.001
Process variables
LOS	0.180	7.38	.007				0.091	30.42	<.001	0.037	7.29	.007
PT: formal assessment (min per patient per d)	−0.880	8.97	.003	−0.948	17.01	<.001				−0.210	7.41	.007
PT: bed mobility (min per patient per d)	−1.547	22.80	<.001	−1.469	33.71	<.001
PT: transfers (min per patient per d)										0.167	16.42	<.001
PT: gait (min per patient per d)	0.527	31.19	<.001	0.497	39.60	<.001				0.065	8.12	.004
PT: advanced gait (min per patient per d)	2.010	29.02	<.001	1.845	35.87	<.001	0.427	16.8	<.001	0.364	9.57	.002
OT: dressing (min per patient per d)										−0.094	11.39	<.001
OT: grooming (min per patient per d)	−0.701	6.37	.012				−0.225	8.09	.005
OT: bed mobility (min per patient per d)							−0.567	4.85	.028
OT: functional mobility (min per patient per d)							0.234	7.17	.008
OT: community integration (min per patient per d)							0.310	11.73	<.001
OT: home management (min per patient per d)	1.189	17.30	<.001	0.998	17.20	<.001				0.372	10.77	.001
OT: wheelchair (min per patient per d)										−0.361	8.05	.005
SLP: swallowing (min per patient per d)				−0.179	6.76	.010
SLP: verbal expression (min per patient per d)							0.129	7.79	.006
SLP: auditory comprehension (min per patient per d)							−0.282	19.99	<.001
SLP: reading comprehension (min per patient per d)				0.470	4.59	.033
SLP: problem solving (min per patient per d)	0.437	10.76	.001				0.192	22.06	<.001
SLP: orientation (min per patient per d)	−0.985	9.03	.003	−0.692	6.49	.011	−0.457	21.69	<.001
SLP: attention (min per patient per d)										−0.102	5.60	.018
Tube feeding during rehabilitation	3.651	5.60	.019	4.172	9.80	.002
Medications
Old SSRIs	−3.867	7.50	.007	−3.447	8.35	.004
Modafinil	−10.70	16.77	<.001	−8.730	16.32	<.001
Anti-Parkinson’s	−6.388	12.40	<.001	−4.908	10.44	.001

NOTE. Blank cells refer to variables and their coefficient, F, and P values that did not enter the model significantly.

Table 6. Full Regressions for Moderate Stroke Patients CMGs 104–107, First Therapy Block Only

Independent Variables	Discharge FIM (R2=.546, n=283)⁎			Discharge Motor FIM (R2=.480, n=283)⁎			Discharge Cognitive FIM (R2=.772, n=287)
	Coeff	F	P	Coeff	F	P	Coeff	F	P
Patient variables
Age	−0.154	19.81	<.001	−0.143	23.95	<.001	−0.024	4.69	.031
Stroke location: brainstem/cerebellum				−2.943	7.73	.006
Stroke location: subcortical	2.318	4.40	.037
Admission FIM motor score	0.481	23.95	<.001	0.457	30.55	<.001
Admission FIM cognitive score	0.613	52.83	<.001				0.688	619.09	<.001
Aphasia							−1.181	5.52	.020
Moderate motor impairment	−4.853	9.69	.002	−5.598	18.09	<.001
No. of days from stroke symptom onset to rehabilitation admission	−0.094	6.09	.014	−0.078	5.72	.018
Process variables
LOS	−0.205	7.57	.006	−0.189	8.74	.003
PT: sitting (min in first 3h of therapy)				−0.161	5.34	.022
PT: transfers (min in first 3h of therapy)				−0.095	8.30	.004
PT: gait (min in first 3h of therapy)	0.059	10.23	.002	0.046	8.15	.005
OT: bathing (min in first 4h of therapy)				−0.060	5.27	.022
OT: feeding/eating (min in first 4h of therapy)	−0.060	6.10	.014	−0.050	6.12	.014
OT: toileting (min in first 4h of therapy)							0.038	5.75	.017
SLP: voice (min in first 3h of therapy)	0.093	4.79	.030	0.115	10.16	.002
SLP: auditory comprehension (min in first 3h of therapy)	−0.162	19.64	<.001	−0.091	9.60	.002	−0.028	5.27	.023
Medications
Old anticonvulsants	6.655	4.67	.032	5.173	3.87	.050
Analgesics; muscle relaxant	−7.560	11.39	<.001	−7.446	15.00	<.001
Opioid analgesics	2.972	5.27	.022	2.388	4.65	.032
Neurostimulants	−7.076	6.32	.013
Anti-Parkinson’s	−4.962	5.61	.019				−2.320	11.17	.001
Old antinausea/antivomiting							1.180	3.97	.047

NOTE. Blank cells refer to variables and their coefficient, F, and P values that did not enter the model significantly.

Table 7. Full Regressions for Severe Stroke Patients CMGs 108–114, First Therapy Block Only

Independent Variables	Discharge FIM (R2=.559, n=331)⁎			Discharge Motor FIM (R2=.479, n=331)⁎			Discharge Cognitive FIM (R2=.742, n=335)			Discharge Home and Assisted Living (c=.745, n=365)
	Coeff	F	P	Coeff	F	P	Coeff	F	P	Coeff	Wald	P
Patient variables
Age	−0.330	30.98	<.001	−0.324	43.14	<.001
Race: other							−1.153	5.33	.022
Stroke side: right brain							1.289	9.39	.002
Admission FIM motor score	0.635	17.78	<.001	0.656	27.53	<.001				0.113	31.66	<.001
Admission FIM cognitive score	0.959	63.71	<.001	0.216	4.65	.032	0.666	362.28	<.001
Maximum severity score (CSI)	−0.081	3.92	.048				−0.031	8.83	.003
Aphasia							−1.063	4.23	.041
Severe motor impairment	−7.871	11.53	<.001	−6.094	9.73	.002	−1.482	6.08	.014
No dysphagla				3.405	4.53	.034
Neurobehavioral impairment: mood and cognitive disturbances (both)	−4.846	5.20	.023				−1.964	11.11	.001	−0.671	4.12	.042
Neurobehavioral impairment: neurotropic medications, no mood/behavior or cognitive dysfunction				3.364	5.04	.026
No. of days from stroke symptom onset to rehabilitation admission	−0.119	21.47	<.001	−0.114	27.62	<.001	−0.014	4.07	.045
Process variables
LOS	0.395	22.26	<.001	0.246	12.78	<.001	0.127	36.58	<.001	0.068	25.50	<.001
PT: bed mobility (min in first 3h of therapy)	−0.170	6.54	.011	−0.168	9.03	.003
PT: gait (min in first 3h of therapy)	0.121	13.06	<.001	0.106	13.95	<.001
PT: advanced gait (min in first 3h of therapy)	0.337	4.81	.029	0.268	4.27	.040	0.126	9.01	.003
OT: bed mobility (min in first 4h of therapy)							−0.070	4.37	.037
OT: home management (min in first 4h of therapy)	0.176	6.57	.011	0.159	7.48	.007
SLP: problem solving (min in first 3h of therapy)							0.023	5.16	.024
SLP: orientation (min in first 3h of therapy)							−0.041	4.87	.028
Tube feeding during rehabilitation	4.850	6.19	.013	4.700	7.93	.005
Medications
Old SSRIs	−5.346	8.27	.004	−4.616	8.73	.003
Other antidepressant	−3.663	4.95	.027	−4.206	9.03	.003				−0.593	5.22	.022

NOTE. Blank cells refer to variables and their coefficient, F, and P values that did not enter the model significantly.

Table 8. Significant Therapy Variables Predicting Discharge FIM Walk and Toilet Transfer Levels, First Therapy Block Only

NOTE. Patients in severe stroke CMGs 108–114. Blank cells are not significant.

For first block analyses we included only time in each activity (excluding time spent in assessment) during the first block of PT, OT, and SLP treatment time, regardless of the total number of therapy blocks a patient received during the entire rehabilitation stay. This ensured that patients were functioning at the identified FIM locomotion, transfer, and communication levels (as measured by admission FIM score), among others, at the time of receiving the therapy activities. Because we did not measure incremental increases in FIM scores during the rehabilitation stay, it was important to reduce the confounding effect of naturally improving function (natural recovery) over the course of rehabilitation. Associating outcomes at discharge with time in activities throughout the whole stay might be confounded by the natural recovery process. By using the first block of therapy, we hypothesized that patients would not have time to improve their functioning naturally as they might have if we included all therapy blocks in regression analyses.

Back to Article Outline

Results 

Patient, Process, and Outcomes Characteristics 

Patient, process, and outcome characteristics for the 830 patients with moderate and severe stroke are presented in Table 1, Table 2, Table 3, respectively. Demographically, the samples are similar, although the severe stroke group is slightly older. As expected, the severe stroke group differed significantly in many other patient characteristics (see table 1). They had significantly higher admission severity scores (CSI), and individual component scores of the CSI (dysphagia, complete hemiplegia or worse, aphasia, mood and cognitive disturbances) also were more severe. By definition, the severe stroke group also had significantly lower admission FIM scores (total, motor, cognitive). In addition, the severe stroke group had more time between onset of stroke symptoms and rehabilitation admission.

Many process variables also were significantly different between the moderate and severe stroke groups (see table 2). Rehabilitation LOS was significantly longer for the severe stroke group. Time spent in therapy activities varied among the 2 groups, often significantly. Significantly more time was spent on higher-level PT (gait, advanced gait, community mobility), OT (home management and community integration), and SLP activities (executive function skills) in the moderate stroke group. Tube feeding was used significantly more with patients with severe stroke. Medications administered to the 2 groups also were different for several classes of neurotropic medications: there was greater use of opioid analgesics, analgesic muscle relaxants, new selective serotonin reuptake inhibitors (SSRIs), atypical antipsychotics, anti-Parkinson’s medications, modafinil, neurostimulants, other antidepressants, and old antinausea and antivomiting medications in patients with severe stroke.

Outcome measures also varied significantly for patients with moderate and severe stroke. Discharge total, motor, and cognitive FIM scores were higher for patients with moderate stroke. However, patients with severe stroke achieved greater increases in total, motor, and cognitive FIM scores from admission to discharge. Discharge and maximum CSI scores were significantly higher (indicating sicker patients) for patients with severe stroke; patients with severe stroke also had a greater increase in severity during rehabilitation. Significantly more patients with moderate stroke were discharged to home or community (see table 3).

Regression Results for All Patients With Moderate and Severe Stroke 

We allowed many variables (eg, demographics; function at admission [FIM score]; medical severity of illness [maximum CSI score]; components of severity; stroke location; minutes per day spent on PT, OT, and SLP activities; medication class; nutritional support; LOS) (see Table 1, Table 2) to enter stepwise selection regression models to identify those variables associated with higher or lower functional outcome by discharge or more or less likelihood of being discharged to home or community versus institution (SNF, hospital, other rehabilitation center).

Table 4, Table 6 present 2 regression approaches for patients with moderate CMG (104−107) stroke, and Table 5, Table 7 present 2 regression approaches for patients with severe CMG (108−114) stroke. The first approach for each group contained information and interventions from the full rehabilitation stay (see table 4); the second approach (see table 6) used the amount of PT, OT, and SLP from the first block of therapy only. Outcomes included discharge total, motor, and cognitive FIM scores. In addition, for the severe stroke group (CMG 108−114) we included discharge destination as a fourth outcome. Discharge destination was not used as an outcome for patients with moderate stroke because almost all of these patients went home (see table 3).

Demographic Variables 

In each model, older patients were associated with lower discharge FIM scores for at least 2 specified outcomes. Race (ie, black) was associated with lower discharge total and motor FIM scores for patients with severe stroke.

Health and Functional Status Variables 

Patients in the moderate group with brainstem and cerebellar strokes were associated with lower discharge motor FIM scores.

Patients with higher admission motor and cognitive FIM scores were associated with higher discharge FIM scores and with more likelihood of being discharged home.

Maximum severity scores were associated with lower discharge total and cognitive FIM scores in the first block analyses of patients with severe stroke. The high correlation of maximum severity score and admission FIM score in patients with severe stroke (r =−.491, P<.001) partially explains the CSI’s overall lack of significance in regression models that include the entire rehabilitation stay. However, components of the CSI including aphasia, levels of motor impairment, neurobehavioral impairment, and dysphagia entered each model as indicated. Patients with an aphasia diagnosis were associated with lower discharge cognitive FIM scores during rehabilitation (all models).

When the CSI and its related components were not allowed to enter models by not including them in the variable selection list, the R² and c statistics changed little (between 0% and 4.3%). Also, none or very few other predictors changed. Hence, the models were stable. This indicates that it is important to control for the CSI and its components but that other detailed process predictor variables correlate sufficiently with the CSI to retain the overall explanatory power of the models. When detailed process data were not available, the CSI explained between 12% and 20% of additional variation in outcomes beyond patient demographic data.¹⁷

In all models, more time from onset of stroke symptoms to rehabilitation admission was associated with lower discharge total and motor FIM scores.

Process Variables 

Longer rehabilitation LOS was associated significantly with lower discharge total and motor FIM scores for patients with moderate stroke. In contrast, however, for patients with severe stroke, longer LOS as associated significantly with higher discharge total and cognitive FIM scores and greater likelihood of being discharged to home.

A variety of PT, OT, and SLP activities were associated significantly with higher or lower discharge FIM scores and discharge destination. Consistently, more minutes per day spent in PT gait activities, OT upper-extremity control and home management activities, and SLP problem-solving activities were associated significantly with higher discharge FIM scores and greater rates of discharge to home. Other therapy activities were associated consistently with lower discharge FIM scores: more minutes per day spent in PT bed mobility and sitting, OT bed mobility, and SLP auditory comprehension and orientation.

Use of anti-Parkinson medications (bromocriptine, pergolide, pramipexole, carbidopa/levodopa, amantadine) was associated significantly with lower discharge FIM scores. Interestingly, only 5 (0.6%) patients had a diagnosis of Parkinson’s disease. Use of new SSRI medications (citalopram, escitalopram), opioid analgesics (codeine, fentanyl, hydrocodone, hydromorphone, methadone, morphine, oxycodone, propoxyphene), and atypical antipsychotics (clozapine, olanzapine, quetiapine, risperidone) were associated with higher discharge FIM scores; however, use of older SSRI medications (fluoxetine, paroxetine, sertraline) had a significant association with lower discharge FIM scores.

Enteral tube feeding was associated significantly with higher discharge total and motor FIM scores for patients with severe stroke, even when controlling for degree of dysphagia and other variables. It was not a significant variable in regression models for patients with moderate stroke.

Regression Results for Patients Admitted at FIM Locomotion Level 1 or Toilet Transfer Level 1 

It could be argued that, in Table 4, Table 5, Table 6, Table 7, discharge FIM scores do not isolate adequately the effects of individual therapies to specific areas of function because we look at the impact of individual therapy activities on broad categories of function such as total FIM and motor FIM scores. In table 8, we take a more focused approach. We looked at patients in the severe stroke group who started at FIM locomotion/walk level 1 (n=177) and tried to determine which therapies in the early stages (first block only) made a difference in getting patients from locomotion/walk level 1 to a locomotion/walk level of 4 or higher. We also wanted to consider how important the first block of therapy was regardless of how many additional blocks of therapy a patient received in total. Here, we found that minutes of gait training in the first block of therapy was consistently the most important PT activity associated with better outcome, regardless of the total amount of PT rendered over the course of the rehabilitation stay, while controlling for other patient characteristics.

We also wanted to determine whether benefits of gait training generalized to other lower-level functional areas that one might focus on before gait training. In this case, we arbitrarily chose toilet transfer and considered those patients who started at toilet transfer level 1 and progressed to level 4 or higher (see table 8). Again, we found that amount of time spent on gait in the first block was the most important predictor in advancing from FIM toilet transfer level 1 to level 4 or higher, while controlling for other patient covariates. Early gait training appears to allow the patient to leap-frog over lower levels of toilet transfer.

Back to Article Outline

Discussion 

Many of the results in Table 4, Table 5, Table 6, Table 7, Table 8 are in the expected direction and are consistent with findings in other studies that have examined the relation between patient characteristics, functional status, LOS, and outcome. What is different here is the ability to examine how specific therapy activities, medications, and other interventions are associated with outcomes. There are 2 consistent findings across all regressions presented in this article. The first is that earlier is associated with better. We found a strong and consistent negative association between time of stroke symptom onset to rehabilitation admission and functional outcomes. In other words, the sooner a patient with stroke starts inpatient rehabilitation after his/her stroke, no matter how severe, the better the outcome. Moreover, we find that earlier gait activities, particularly in the first block of PT, have a significant association with outcome, regardless of how much additional therapy a patient receives or what his/her admission functioning level (FIM score) is. This second finding supports more aggressive therapy. That is, earlier participation in higher-order, more challenging therapy activities, even at the outset in the first block of therapy and even for low-functioning patients, is associated with better outcomes in general, and extended participation in lower-level activities often is associated with worse outcomes. Participation in higher-order or more difficult therapeutic activities appears to assist in the improvement of lower-level functional activities, even without direct attention to that activity. This last observation was most evident in examining how gait training during the first block of therapy was associated strongly with greater independence in toilet transfers (see table 8). Also, Hatfield et al²⁰ found that it may not be necessary to spend much time enhancing basic verbal expression skills. Instead, therapists should initiate problem-solving activities, and the verbal expression will come back in the process.

These findings challenge conventional wisdom in rehabilitation practice. It is important to understand the nature of this conventional wisdom and how it arises. Rehabilitation clinicians work with patients in particular ways based on how they were taught or based on therapeutic theories and approaches espoused by textbook authors. Although often unsupported by scientific evidence, the theories and approaches make a good deal of intuitive sense and become incorporated into conventional wisdom and practice.

A few examples may be helpful here. Consider Rood’s clinical maxim: “Proximal stability before distal mobility.” It suggests that a patient cannot learn to use their hands or feet if their trunk and proximal limbs are weak. Consider a clinical rule of thumb in rehabilitation: activities should be planned to allow a patient to be successful for 80% of trials, thus minimizing his/her frustration and risk of depression due to excessive experience of failure. Also consider the theory underlying neurodevelopmental treatment, developed from the pioneering pediatric rehabilitation work of Bobath. It is a therapeutic approach that can be described simplistically based on the observation that a child first learns to crawl and integrate spinal and brainstem reflexes before learning to walk. Finally, consider the theory underlying the Fugl-Meyer Assessment of motor recovery after stroke. It was developed as the first quantitative evaluative instrument for measuring sensorimotor stroke recovery, based on Twitchell and Brunnstrom’s observations and conceptualization of the “sequential stages of motor return” in hemiplegic patients with stroke.²⁷ Collectively, these theories and approaches advocate starting at a patient’s current level of functioning and then building gradually toward recovery of normal function. Some of our findings challenge these time-honored theories and approaches.

Other findings actually reinforce conventional wisdom. The finding that earlier is better supports the rehabilitation axiom that patients should start rehabilitation sooner rather than later and that delaying rehabilitation can have a deleterious effect on outcomes. Rehabilitation clinicians have long been concerned that, with extended stays in acute care, patients becomes progressively deconditioned and less able to partake fully in rehabilitation on transfer to a rehabilitation unit. Again, data analyses presented here suggest that the sooner patients with stroke, especially those with severe stroke, get to the rehabilitation setting, the more likely they are to have an optimal gain in FIM score and have the best chance of being discharged to home instead of being institutionalized. This may mean transferring patients who are not yet 100% stable (eg, may have a urinary tract infection or pneumonia) to a rehabilitation unit more quickly, rather than spending a few more days in the hospital waiting for complete stabilization.

Rehabilitation providers often wonder if the acute care hospital payment system encourages acute care providers to discharge patients to rehabilitation when they are not yet medically stable. The findings here suggest that “sicker and quicker” may in some cases be better. This inference is supported by the variables that have significant association with discharge total and motor FIM scores (see Table 4, Table 5, Table 6, Table 7). A longer time between onset of stroke symptoms and admission to inpatient rehabilitation was associated with reduced discharge FIM score, after controlling for overall severity of illness or its components. This suggests that earlier admission to rehabilitation, even if a patient’s severity of illness is increased according to a higher CSI score or its components, is associated with better outcomes. In any event, the findings should encourage more timely coordination in the handoff from acute care to rehabilitation for patients with stroke and more willingness of rehabilitation facilities to admit medically challenging, sicker, patients.

Once in rehabilitation, patients appear to have different responses to LOS. For patients in the moderate stroke CMGs (104−107), our findings indicate that there is a negative association of longer LOS with outcomes. However, for patients in the severe CMGs (108−114), our findings indicate that there is a strong positive association of longer LOS with outcomes. At the risk of overinterpreting these findings, one could conclude that patients with moderate stroke do better with shorter and more intense rehabilitation stays, whereas patients with severe stroke do better with a more extended rehabilitation process. More data analyses are needed to determine accurately the relation between rehabilitation LOS and outcomes among various subgroups of rehabilitation patients to identify more clearly the rehabilitation patients who would benefit from shorter or longer rehabilitation stays.

Common clinical practice also has a powerful sway in the choice of medications. A few years ago, stroke rehabilitation physicians were happy to adopt the use of SSRI medications for patients with depressed mood because of remarkably low side effect profiles compared with tricyclic antidepressants, which were notorious for side effects. The first generation of SSRIs included fluoxetine, sertraline, and paroxetine. A newer generation of SSRIs, including citalopram and escitalopram, has been developed and adopted into use by psychiatric physicians; however, rehabilitation physicians have been slower to adopt them. Given that there are few side effects from the first generation SSRIs and little to no research on the merits or side effects of newer SSRIs on patients undergoing stroke rehabilitation, there is no strong reason to shift to the unknown from the well established. However, our analyses indicate that stroke rehabilitation patients might benefit from such a shift.

It is also common clinical practice to avoid the use of antipsychotic medications, based on beliefs extrapolated from animal research and psychiatric literature that the antidopaminergic and anticholinergic effects of chlorpromazine and haloperidol, among others, could reduce alertness and learning capacity in stroke survivors. A new family of medications referred to as atypical antipsychotic medications (olanzepine, quetiapine, risperidone, ziprasidone) has seen little use in stroke rehabilitation because its newness, a long-standing bias against antipsychotic medications as a group, and the lack of randomized studies in the stroke population. This persists despite the growing literature showing nootropic effects of this family of medications.²¹ Our analyses indicate that patients with stroke might benefit from greater use of atypical antipsychotics. Conroy et al²¹ found that new atypical antipsychotic medications and second generation SSRIs appear to have a positive association with stroke rehabilitation outcomes.

Another family of medications for which there exists a long-standing bias against use in stroke rehabilitation is narcotic pain medications. Narcotics are understood to sedate patients, dull cognition, cause depression, and reduce respiratory drive and, therefore, are expected to diminish outcomes if used in stroke rehabilitation. A lack of specific research examining medication use in stroke rehabilitation allows common clinical practice to prevail. Our data suggest otherwise—that narcotic pain medications are effective in reducing pain and that patients make greater improvements in motor FIM scores with them than without, despite their sedating and cognitive dulling effects.²¹

In summary, the PSROP database is large enough that we can locate narrow subpopulations where actual clinical activities and interventions went against common clinical practice: patients given narcotic or atypical antipsychotic medications consistently, low-functioning patients (admission FIM scores of 1 for locomotion or toilet transfer) who participated in PT sessions in their first 3 hours of PT where they practiced gait activities, and patients requiring total assistance for toileting who participated in PT where a therapist practiced gait in the first 3 hours of therapy. Results from these analyses indicate a strong and consistent association of rehabilitation activities that challenge patients and stress them well beyond their current level with better outcomes; that is, they move quickly to practicing upper-extremity functional activities rather than focusing on trunk strengthening. (The trunk will strengthen secondarily out of necessity.) There seem to be positive benefits for patients to jump ahead in the established sequence of activities and leap-frog into activities that might seem excessively challenging for them according to common clinical practice.

These findings challenge rehabilitation providers to rethink how they approach patients. They suggest that many current strategies about how to help a patient improve may not be optimal. Work carried out in the PSROP is not intended to reduce the value of rehabilitation but to discover its best aspects. Rehabilitation clinicians will continue to work on trunk stability, make sure a patient can move in bed, and choose to use fluoxetine at times. The difference may be in the timing and knowing when is the best opportunity to use each technique with specific types of patients.

The findings presented here are based on findings from facilities in the United States. The larger study also included a rehabilitation facility in New Zealand. Overall, the findings here are consistent with findings arising from our comparison of practice and outcomes between facilities in the United States and New Zealand, presented in this supplement by McNaughton et al.²⁸ This comparison notes that U.S. stroke rehabilitation patients received earlier and more intense rehabilitation and had better outcomes despite presenting a more severe clinical profile on admission.

Back to Article Outline

Limitations 

Several of the PSROP’s limitations are noted in the study’s baseline methods article by Gassaway et al¹⁷ and in other articles in this supplement.¹⁸, ¹⁹, ²⁰, ²¹, ²², ²³ Observational studies such as this, however, naturally raise several concerns. We want to address 3 of them: (1) controlling for patient differences, (2) selection bias, and (3) association versus causation. The first 2 are closely related.

The strength of an observational study depends on the study’s ability to control for patient differences that would otherwise be addressed through randomization. In the absence of randomization, it is critical that important patient covariates be addressed adequately. As noted in the study’s baseline methods article,¹⁷ the study’s use of the admission FIM and the CSI scores provides a comprehensive patient functional and severity profile, although there is always the chance that some unknown critical variable may have been overlooked.

Selection bias is a concern when patients are not randomly assigned to certain treatment arms or when some patients fail to enroll in the study or drop out. In this study, there was no treatment arm, sham treatment, or placebo; the study examined only existing practice. Moreover, in this study, patients entered the study consecutively as they were admitted to the facility. There was no formal enrollment or informed consent because no new intervention was being introduced—and thus there were no dropouts that might otherwise bias the study sample.

The chief criticism of any observational study of this genre is that association is not causation. We agree. But when associations remain consistent regardless of how the study group is partitioned or when the findings are tested from other vantage points, the evidence becomes increasingly persuasive and needs to be taken seriously despite the exploratory nature of the study. One of the next steps is to determine the predictive validity of the study’s findings. One way this can be done is to implement the findings as suggested here and then evaluate whether the outcomes observed are those that were predicted. The field also could conduct 1 or more randomized clinical trials to test these findings to determine more conclusively the predictive validity of the findings. A more formal trial of the study’s leap-frog hypothesis would be particularly compelling.

The analyses presented here examine the relation between rehabilitation activities and interventions and outcomes on discharge from rehabilitation. These findings would be even more compelling if they were also found to be true for longer-term outcomes (eg, 6 and 12mo postonset). Funding limitations simply did not allow the research team to look beyond the patients’ discharge statuses.

Back to Article Outline

Conclusions 

The PSROP’s database allows researchers and clinicians to examine a rich array of associations between rehabilitation patients, processes, and outcomes. The database enables investigators to discover treatment practices that are associated with better outcomes for patients with stroke, taking into account their demographic and clinical profiles. A key finding is that earlier and more aggressive therapy is better. We find that starting therapy sooner after a stroke and starting higher-order or more challenging activities sooner are associated with better outcomes, even with lower-level functioning patients. We find this to be the case for each of 3 rehabilitation sentinel therapies—PT (early gait activities), OT (early community mobility activities), and SLP (early problem solving activities). In the area of medication use, the analyses suggest that making the jump to newer SSRI antidepressant and atypical antipsychotic medications is associated with greater ability to benefit from inpatient rehabilitation for our patients.

These findings have significant implications for future research. Our findings suggest that health care providers need to shorten the duration from onset of stroke to onset of rehabilitation and to move patients as quickly as possible to higher-level, more difficult therapy activities and that rehabilitation providers may be able to shorten the LOS for some patients but increase the LOS for others. Validation studies may lead to changes in clinical practice and health policy as it relates to rehabilitation. These findings suggest continued study to reconsider target LOSs and payment weights associated with various CMGs in the IRF prospective payment system. They provide us the opportunity to develop more creative stroke rehabilitation “products” that could better coordinate each patient’s care from stroke onset to rehabilitation and to discharge.

A strength of the clinical practice improvement (CPI) approach is the ability to uncover best practices more quickly than conventional studies. Such practices can later be vetted in validation studies or through controlled trials. A dilemma we have now is determining what therapeutic activities and interventions are truly ready for prime-time controlled studies. By focusing exclusively on randomized studies, we risk wasting valuable rehabilitation research resources on studies that may show no or minimal differences. Through use of CPI studies, therapeutic activities and interventions can be identified and unproductive activities and interventions can be weeded out before such confirmatory studies. The results here require us to validate the findings (predictive validity) through actual implementation and perhaps clinical trials and to rethink aspects of stroke rehabilitation practice and policy.

Supplier

Back to Article Outline

Acknowledgments 

We acknowledge contributions of collaborators at each clinical site represented in the Post-Stroke Rehabilitation Outcomes Project: Brendan Conroy, MD (Stroke Recovery Program, National Rehabilitation Hospital, Washington, DC); Richard Zorowitz, MD (Department of Rehabilitation Medicine, University of Pennsylvania Medical Center, Philadelphia, PA); David Ryser, MD (Neuro Specialty Rehabilitation Unit, LDS Hospital, Salt Lake City, UT); Jeffrey Teraoka, MD (Division of Physical Medicine and Rehabilitation, Stanford University, Palo Alto, CA); Frank Wong, MD, and LeeAnn Sims, RN (Rehabilitation Institute of Oregon, Legacy Health Systems, Portland, OR); Murray Brandstater, MD (Loma Linda University Medical Center, Loma Linda, CA); and Harry McNaughton, MD (Wellington and Kenepuru Hospitals, Wellington, NZ). We also acknowledge the role of Alan Jette, PhD (Rehabilitation Research and Training Center on Medical Rehabilitation Outcomes, Boston University, Boston, MA).

Back to Article Outline

Appendix 1. Variables allowed to enter and leave regression models 

Abbreviation: SSRIs, selective serotonin reuptake inhibitors.

Back to Article Outline

References 

1. DeJong G , Horn SD , Conroy B , Nichols D , Healton EB . Opening the black box of poststroke rehabilitation (stroke rehabilitation patients, processes, and outcomes) . Arch Phys Med Rehabil . 2005;86(12 Suppl 2):S1–S7
   • View In Article
   • MEDLINE
2. Bode RK , Heinemann AW , Semik P , Mallinson T . Patterns of therapy activities across length of stay and impairment levels (peering inside the “black box” of inpatient stroke rehabilitation) . Arch Phys Med Rehabil . 2004;85:1901–1908
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (426 KB)
   • CrossRef
3. Sulch D , Perez I , Melbourn A , Kalra L . Randomized controlled trial of integrated (managed) care pathway for stroke rehabilitation . Stroke . 2000;31:1929–1934
   • View In Article
   • MEDLINE
   • CrossRef
4. Alexander H , Bugge C , Hagen S . What is the association between the different components of stroke rehabilitation and health outcomes? . Clin Rehabil . 2001;15:207–215
   • View In Article
   • MEDLINE
   • CrossRef
5. Ballinger C , Ashburn A , Low J , Roderick P . Unpacking the black box of therapy—a pilot study to describe occupational therapy and physiotherapy interventions for people with stroke . Clin Rehabil . 1999;13:301–309
   • View In Article
   • MEDLINE
   • CrossRef
6. vanVliet PM , Lincoln NB , Robinson E . Comparison of the content of two physiotherapy approaches for stroke . Clin Rehabil . 2001;15:398–414
   • View In Article
   • MEDLINE
   • CrossRef
7. de Weerdt W , Selz B , Nuyens G , et al.   Time use of stroke patients in an intensive rehabilitation unit (a comparison between a Belgian and a Swiss setting) . Disabil Rehabil . 2000;22:181–186
   • View In Article
   • MEDLINE
   • CrossRef
8. Rijken PM , Dekker J . Clinical experience of rehabilitation therapists with chronic diseases (a quantitative approach) . Clin Rehabil . 1998;12:143–150
   • View In Article
   • MEDLINE
   • CrossRef
9. Contract No. HHS-100-97-0010 Liu K , Gage B , Kramer A . Medicare post-acute care: quality measurement final report. Appendix A. 1998 reports . Washington (DC): US. Department of Health and Human Services; 1998;
   • View In Article
10. National Board for Certification in Occupational Therapy Inc . A practice analysis study of entry-level occupational therapists registered and certified occupational therapy assistant practice . Occup Ther J Res . 2004;24(Suppl 1):s7–s31
    • View In Article
11. American Occupational Therapy Association . Occupational therapy practice framework (domain and process) . Am J Occup Ther . 2002;56:609–639
    • View In Article
    • MEDLINE
    • CrossRef
12. Ma H , Trombly CA . A synthesis of the effects of occupational therapy for persons with stroke. Part II: remediation of impairments . Am J Occup Ther . 2002;56:260–274
    • View In Article
    • MEDLINE
    • CrossRef
13. Holland AL , Wertz RT . Measuring aphasia treatment effects (large-group, small-group, and single-subject studies) . Res Publ Assoc Res Nerv Ment Dis . 1988;66:267–273
    • View In Article
    • MEDLINE
14. Johnston MV , Findley TW , DeLuca J , Katz RT . Research in physical medicine and rehabilitation. XII. Measurement tools with application to brain injury . Am J Phys Med Rehabil . 1991;70:40–56
    • View In Article
    • MEDLINE
    • CrossRef
15. Bryant M . Biofeedback in the treatment of a selected dysphagia patient . Dysphagia . 1991;6:140–144
    • View In Article
    • MEDLINE
    • CrossRef
16. Sohlberg MM , Mateer CA . Effectiveness of an attention training program . J Clin Exp Neuropsychol . 1987;9:117–130
    • View In Article
    • MEDLINE
    • CrossRef
17. Gassaway J , Horn SD , DeJong G , Smout RJ , Clark C , James R . Applying the clinical practice improvement approach to stroke rehabilitation (methods used and baseline results) . Arch Phys Med Rehabil . 2005;86(12 Suppl 2):S16–S33
    • View In Article
    • MEDLINE
18. Latham NK , Jette DU , Slavin M , et al.   Physical therapy during stroke rehabilitation for people with different walking abilities . Arch Phys Med Rehabil . 2005;86(12 Suppl 2):S41–S50
    • View In Article
    • MEDLINE
19. Richards LG , Latham NK , Jette DU , Rosenberg L , Smout RJ , DeJong G . Characterizing occupational therapy practice in stroke rehabilitation . Arch Phys Med Rehabil . 2005;86(12 Suppl 2):S51–S60
    • View In Article
    • MEDLINE
20. Hatfield B , Millet D , Coles J , Gassaway J , Conroy B , Smout RJ . Characterizing speech and language pathology outcomes in stroke rehabilitation . Arch Phys Med Rehabil . 2005;86(12 Suppl 2):S61–S72
    • View In Article
    • MEDLINE
21. Conroy B , Zorowitz R , Horn SD , Ryser DK , Teraoka J , Smout RJ . An exploration of central nervous system medication use and outcomes in stroke rehabilitation . Arch Phys Med Rehabil . 2005;86(12 Suppl 2):S73–S81
    • View In Article
    • MEDLINE
22. James R , Gines D , Menlove A , Horn SD , Gassaway J , Smout RJ . Nutrition support (tube feeding) as a rehabilitation intervention . Arch Phys Med Rehabil . 2005;86(12 Suppl 2):S82–S92
    • View In Article
    • MEDLINE
23. Maulden SA , Gassaway J , Horn SD , Smout RJ , DeJong G . Timing of initiation of rehabilitation after stroke . Arch Phys Med Rehabil . 2005;86(12 Suppl 2):S34–S40
    • View In Article
    • MEDLINE
24. DeJong G , Horn SD , Gassaway J , Dijkers M , Slavin M . Toward a taxonomy of rehabilitation interventions (using an inductive approach to examine the ‘black box’ of rehabilitation) . Arch Phys Med Rehabil . 2004;55:678–686
    • View In Article
25. In:  Hart AC ,  Schmidt KM ,  Aaron WS editor. ICD-9-CM code book . Reston: St. Anthony’s Publishing; 1999;
    • View In Article
26. Epocrates. Available at: http://www2.epocrates.com. Accessed September 22, 2005.
    • View In Article
27. Gladstone DJ , Danells CJ , Black SE . The Fugl-Meyer assessment of motor recovery after stroke (a critical review of its measurement properties) . Neurorehabil Neural Repair . 2002;16:232–240
    • View In Article
    • MEDLINE
28. McNaughton H , DeJong G , Smout RJ , Melvin JL , Brandstater M . A comparison of stroke rehabilitation practice and outcomes between New Zealand and United States facilities . Arch Phys Med Rehabil . 2005;86(12 Suppl 2):S115–S120
    • View In Article
    • MEDLINE

• a Release 8.2; SAS Institute Inc, 100 SAS Campus Dr, Cary, NC 27513.