Problems of methodological heterogeneity in studies predicting disability after stroke.

Abstract

Comparison among studies is hindered by methodological heterogeneity. Differences among patient samples mean that predictor variables identified with one sample will not necessarily be accurate for other samples. Studies conducted across treatment centers would help ensure that the sample is representative of the population to which the predictor measure is to be applied in the future. The criterion by which functional outcome has been measured has varied. Many studies have measured function on discharge, while others have measured improvement in function. If interest lies with improvement in function, it is important to identify factors that predict which patients have greater (or less) than predicted recovery. If interest lies with the prediction of function at a particular point after the stroke, function should not be measured at discharge (since length of hospital stay varies enormously) but at a set time poststroke. Future studies should use scales whose reliability and validity is well established.