For efficient rehabilitation it is important to identify, as early as possible, the patients likely to be successfully returned to work after rehabilitation. The aim of this pilot study was to develop a statistical model for predicting this return as reliably as possible. The model uses only information readily available at the beginning of rehabilitation. A multiple regression analysis with backward elimination was used from a routine data base and identified 8 variables of prognostic value. The model offers a comfortable possibility to predict the probability of return to work of a patient on the basis of routinely registered data. The prognosis was found correct in 68% of those returning to work after rehabilitation (sensitivity) and in 80% of those who did not (specificity). Further work to improve the model for prognosis in rehabilitation research is considered reasonable.