Frailty Model

frailty-modelsurvival-analysisrandom-effectscox-modelpenalized-likelihoodrecurrent-eventsproportional-hazards

Definition

A frailty model is a survival (time-to-event) model with a random effect — the "frailty" — that multiplies the hazard to capture unobserved heterogeneity or within-group correlation. Each individual or group ii has a latent frailty ωi\omega_i acting on its hazard, λij(t)=ωiλ0(t)exijβ\lambda_{ij}(t)=\omega_i\,\lambda_0(t)\,e^{x_{ij}'\beta} (or additively on the log-hazard), so that "frailer" subjects fail earlier than their covariates alone would predict. Therneau-Grambsch-Pankratz (2003) show these models can be fit efficiently as penalized Cox/parametric regressions (Therneau-Grambsch-Pankratz 2003).

Key Ideas

Why It Matters

Open Questions

Related