Functional outcome measures in stroke rehabilitation.

I examine statistical considerations in the analysis of functional outcome following stroke and discuss the mathematical relation between improvement in function and discharge functional score. I demonstrate mathematically that the predictor variables of improvement and discharge functional score are the same and that the regression coefficients for improvement and discharge functional score will be equal, except for the admission functional score, for which a mathematically defined relation exists. I argue that the relation between admission functional score and discharge functional score must be positive and strong and that the relation between admission functional score and improvement must be negative for the stroke population. I believe that an ignorance of statistical concepts, especially confounding, and of the differences between raw correlations, partial correlations, and predictors have led to much confusion in functional outcome research.