Use of a standardized assessment to predict rehabilitation care after acute stroke

Objective: To pilot a program of formal assessment of rehabilitation needs and predictors of referral to rehabilitation.

Design: A prospective pilot project to collect standardized measures of stroke severity and function: National Institutes of Health Stroke Scale, premorbid modified Rankin scale, Short Portable Mental Status Questionnaire, and Barthel Index (BI). These were collected in addition to routine data in the Get With The Guidelines-Stroke registry. Logistic regression was used to examine predictors of referral to any institution-based rehabilitation versus discharge home and referral to an inpatient rehabilitation facility (IRF) versus a skilled nursing facility (SNF).

Setting: Twenty-two hospitals within the Northeast Cerebrovascular Consortium (located in the northeastern United States).

Participants: Data were collected on individuals with acute ischemic and hemorrhagic stroke (N=736).

Interventions: Not applicable.

Main outcome measures: Discharge disposition location.

Results: The BI score was recorded in 736 (81%) patients. In multivariable analyses, a higher BI score (85-100) was the only factor associated with return home versus need for institution-based rehabilitation (P<.001). Among patients discharged to IRF versus SNF, discharge to IRF was less likely in older patients (odds ratio [OR], .96; confidence interval [CI], .94-.98; P<.001) and in those with prestroke disability (modified Rankin scale score, 2-5) (OR, .47; CI, .28-.78; P=.004) and more likely in those with moderate-severe (BI score, 25-40; OR, 3.26; CI, 1.45-7.30; P=.004) or moderate (BI score, 45-60; OR, 2.47; CI, 1.17-5.21; P=.018) activities of daily living (ADL) impairment.

Conclusions: Formal standardized assessment of rehabilitation needs was feasible in this pilot project. Patients' sociodemographic characteristics, premorbid function, and ADL impairment discriminated better between discharge home and institution-based rehabilitation than between IRF and SNF. Selection of IRF versus SNF appears to be influenced either by unmeasured clinical characteristics of individuals with stroke or by nonclinical factors, such as cost, geography, referral relationships, or IRF availability.