Various statistical procedures have been applied to the analysis of factors associated with the tendency of patients to drop out of long-term clinical trials. Thus far, however, no analytical tool has been developed specifically for the analysis of these factors. A model is described which incorporates regression analysis into a two-compartment model in order to analyze covariate effects on patient transitions between active and inactive states. An application is made to a clinical trial in the treatment of hypertension. © 1980.