Predicting rehabilitation outcomes from clinical and statistical data: a probability model.

Abstract

This research was focused on the feasibility of a Markov Chain probability model for predicting rehabilitation outcome. A combination of clinical ratings from counselors and statistical data from client files was used to predict 'successful' and 'unsuccessful' outcomes. Predictions were made at three separate time periods: Time 1, during the first week after initial interview; Time 2, during the fourth week after initial interview; and Time 3, during the thirteenth week after initial interview. The final outcome was determined at the end of six months. The research sample consisted of seventy-one clients from two private rehabilitation agencies, all of whom suffered from work-related injuries or illnesses. The results of this research indicate that a Markov Chain probability model is quite feasible for predicting rehabilitation outcomes. In addition, it was observed that the predictions generally followed a progression over time toward stronger and more accurate predictions for both 'success' and 'failure'. With only one exception, occurring at Time 1, the clinical ratings from the counselors were shown to be stronger predictors than the statistical data from case files. This finding is inconsistent with previous studies. It is suggested that a model such as this can serve to alert counselors to the possible need for different or more innovative strategies for service to clients who show a poor probability for success.