Prostate cancer (PrCA) is the most common malignancy in men and a leading cause of cancer mortality among males in the United States. Large geographical variation and racial disparities exist in both the incidence of PrCA and the survival rate after diagnosis. In this population-based study, a joint spatial survival model is constructed to investigate factors that affect the age at diagnosis of PrCA and the subsequent survival. The joint model for these two time-to-event outcomes is specified through parametric models for age at diagnosis and survival time conditional on diagnosis age. To account for possible correlation in these outcomes among men from the same geographical region, frailty terms are included in the survival model. Both spatially correlated and uncorrelated frailties are incorporated in each model considered. The deviance information criterion is used to select a best-fitting model within the Bayesian framework. The results from our final best-fitting model indicate that race, marital status at diagnosis, and cancer stage are significantly associated with both of the two time-to-event outcomes. No pattern emerged in the geographical distribution of age at PrCA diagnosis. In contrast, a spatially clustered pattern was observed in the geographic distribution of survival experience post diagnosis.
2008 John Wiley & Sons, Ltd