Extension of Cox PH Model When Hazards are Non-Proportional Applied to Residential Treatment for Drug Abuse

Prashant Narayan KC, Minnesota State University, Mankato

Full-text copy available through subscription to ProQuest Dissertations & Theses Global.


Cox proportional hazard (PH) model (1972) is one of the most common methods used in survival analysis. CoxPH model has a strong assumption that the covariates have proportional effect on the hazard function of the lifetime distribution of an individual which need to be carefully verified before interpretation of parameters estimates. What to do if the proportional assumption is violated

This thesis discusses and presents ways of testing if the assumption of proportional hazard is satisfied and modification of Cox PH regression model in case when hazards are not proportional. The results are illustrated by an analysis of Drug remission data from UMASS Aids Research Unit IMPACT Study (UISSURV). Comparisons of cox regression model and proposed methods in case when hazards are not proportional is discussed. In particular the methods of incorporating time dependent variable and stratification method.