Cross-validation of a model predicting discharge home after stroke rehabilitation - Validating stroke discharge predictors

A new sample of 116 stroke patients was collected in order to validate a lo gistic regression model, predicting the chances of severely affected stroke patients being discharged home to independent living. The model was found to be accurate in the new sample, especially for those patients who had a h igh estimated probability of being discharged home. When the dividing line for the predicted probability for discharge home was set at a value of grea ter than or equal to 0.5, the positive and negative predictive values were 74 and 73%, respectively. Further modelling resulted in a new extended mode l including the variables postural stability on admission, cohabiting, age and perceptual impairment on admission that formed the basis for an index p redicting discharge home. This index was then validated in the sample of 93 patients that the first developed model was derived from and showed positi ve and negative predictive values of 85 and 77%, respectively. Copyright (C ) 2000 S. Karger AG. Basel.