Modeling survival times using frailty models

Abstract

Traditional survival models, including Kaplan Meier, Nelson Aalen and Cox regression assume a homogeneous population; however, these are inappropriate in the presence of heterogeneity. The introduction of frailty models four decades ago addressed this limitation. Fundamentally, frailty models apply the same principles of survival theory, however, they incorporate a multiplicative term in the distribution to address the impact of frailty and cater for any underlying unobserved heterogeneity. These frailty models will be used to relate survival durations for censored data to a number of pre-operative, operative and post-operative patient related variables to identify risks factors. The study is mainly focused on fitting shared and unshared frailty models to account for unobserved frailty within the data and simultaneously identify the risk factors that best predict the hazard of death.