Predicting long-term functional outcome of stroke using multivariate analysis

Abstract

The purpose of the present study was to investigate long-term prognosis of patients with stroke and to identify predictive factors for functional outcome. Subjects were consecutively hospitalized 1,056 patients with stroke due to unilateral hemispheric lesions at the Department of Neurology, Kitasato University Hospital from 1986 to 1997. One hundred and eighty-three patients died during hospital stay. The remaining 873 surviving patients were followed for 5 years after onset of stroke. Locomotive function (mobility status) of surviving patients was assessed by questionnaires mailed at 2 years and at 5 years after stroke. Outcome of locomotive function was classified into the following 5 categories: normal, walk alone, walk with device, using wheelchair, and bedridden. The following factors were tested by means of multiple regression analysis: age, sex, level of consciousness, location and size of lesion, history of previous stroke and 2 major risk factors for stroke (hypertension and diabetes mellitus). The results of multivariate analysis revealed that in patients with intracerebral hemorrhage, although level of consciousness on admission was the strongest predictor at discharge, locomotive function at discharge was the most important predictor at 2 years and at 5 years after stroke. Moreover, age was more strongly related to functional outcome than level of consciousness at 2 years and at 5 years after stroke. This trend was also found in patients with cerebral infarction. Multivariate analysis revealed that a linear combination of these predictors accounted for about 50% of the variance in the estimate of functional outcome at 2 years and at 5 years after stroke. The results of the present study suggest that the long-term outcome of stroke patients is more deeply influenced by unidentified factors which may not be present during the hospital stay.