Evaluation of a multivariate model predicting noncompliance with medication regimens among renal transplant patients

Background, Because noncompliance with medication regimens is a major cause of renal allograft failure, we evaluated the stability over time of two lo gistic regression models (sets of variables) that predict noncompliance wit h immunosuppressive regimens. Methods. Models were based on questionnaire data from 1402 patients (all ov er 18, receiving cyclosporine or a cyclosporine-like replacement drug, and with a functioning renal graft). The same questionnaire was completed by a subset of 548 (39.1%) patients approximately 18 months later. The goodness of fit of each model to the new data set was tested. Results. The noncompliance logistic regression model including patient beli efs as well as patient and transplant characteristics was an excellent fit to the second data set. A noncompliance model composed of only patient and transplant characteristics fit the new data set less well. Conclusions. Clinicians and educators need to take explicit account of rena l transplant patients' attitudes when evaluating risks of noncompliance and when developing interventions and educational programs to minimize noncomp liance.