LAUSR

We thank Dijkland et al for their interest in our article1 on prediction of functional outcome after stroke rehabilitation. Dr Dijkland raises some issues about the outcome measures and the predictive model for the secondary outcome. In general, models predictive of dichotomous clinical outcomes are most relevant in medical applications. Consistently, dichotomized outcome measures, including the Barthel index and the modified Rankin Scale, have been used to assess the effects of stroke treatment in clinical trials. For patients undergoing rehabilitation after a moderate-to-severe stroke, regaining minimal levels of disability or physical independence are clinically highly relevant outcomes. According to the article of Altman cited by Dijkland, we did not convert continuous explanatory variables to 2 (dichotomization) or multiple (categorization) groups, but …