|
Abstract:
|
Frailty-based competing risks model was introduced by Garfine to account for the situation in which subjects may experience an event other than the event of interest, and the heterogeneity between clusters needs to be considered at the same time. However, the frailty between each subject must also be considered in a competing risks setting. A specific model has yet to be introduced to evaluate this situation. In order to compare the performance of Cox, Parametric AFT, Fine-Gray and Shared Frailty models for the estimation of the covariate coefficient of event of interest for the competing risk data with univariate frailty, we conducted a sensitivity analysis. The competing risks datasets were generated by simulations that assumed time to competing events follow exponential distribution and univariate frailty processes. We found that the exponential AFT model showed better performance for estimating covariate coefficients for an event of interest as compared to the other models. When the distribution of time to events could not be identified, none of these models had good performance for prediction. We conclude that a univariate frailty based model for competing risks data is needed
|
