A note on the prediction of frailties with misspecified shared frailty models

ABSTRACT

In this research note, we investigate the impact of misspecification of frailty distribution on the prediction of frailties by using shared frailty models with various estimation methods. We review the specification of a number of candidate frailty distributions and three estimation methods for the prediction of frailties in shared frailty models. We conducted statistical simulations to evaluate the performance of prediction of frailties in Cox proportional hazard shared frailty models with a working gamma frailty distribution and different estimation methods with misspecified frailty distributions. The goal of the statistical simulations is to examine the extent to which the frailty prediction is sensitive to misspecification of the frailty distributions and various estimation methods. An example of predicting the frailties of hospital readmissions of colorectal cancer patients after surgery or hospital discharge is presented to demonstrate the impact of such misspecification in survival analysis.

Keywords:

• EM algorithm
• frailty
• misspecification
• prediction
• penalized partial likelihood
• penalized likelihood

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

Dr. Xuejun Jiang's research was partially supported by National Natural Science Foundation of China (Grant No. 11871263) and Natural Science Foundation of Guangdong Province of China (Grant No 2017A030313012). Dr. Wei Liu's research was partially supported by the National Natural Science Foundation of China (Grant No. 11601106).