Dependent competing risks: a stochastic process model (Q1072949)

Traditionally, the problem of competing risks has been solved by assuming independence between hazard rates for each cause of death. This is largely so because of the unidentifiable aspect of the problem. In this paper, a competing risk model with a weaker assumption of conditional independence, given a stochastic covariate process, is developed and applied to a certain data set. Under the assumption that the conditional independence model is valid, the findings were that the standard model overestimates the effect on life expectancy at certain ages due to the hypothetical elimination of cancer and of cardiovascular/cerebrovascular disease. The overestimation was larger in higher age groups, suggesting the importance of avoiding the marginal independence assumption when appropriate data are available, especially when focusing on mortality at advanced ages.