VerifiedEvidence: highv1.0.0

Frailty Model

A survival model incorporating an unobserved random effect, called frailty, to represent risk heterogeneity not captured by observed covariates.

Last reviewedDarrin Baines IP Ltd

Concept Architecture

Concept

Theoretically, a Frailty Model is an extension of a survival model that incorporates unobserved heterogeneity between individuals or groups through a latent random effect known as frailty. The model recognises that individuals with identical observed characteristics may still differ in their underlying risk of experiencing an event because of unmeasured biological, behavioural or environmental factors. In health economics, frailty models are used to improve the estimation of survival, disease progression and treatment effects when unexplained variability influences long-term outcomes.

Mathematically, a frailty model modifies the hazard function by multiplying the baseline hazard by a positive random frailty term. The frailty variable is commonly assumed to follow a gamma or log-normal distribution, allowing the model to account for additional variation in hazard rates beyond that explained by observed covariates. Model parameters are estimated using maximum likelihood or Bayesian methods.

In practice, frailty models are fitted to individual patient survival data using specialised statistical software. They are particularly useful for analysing clustered survival data, recurrent events and multicentre clinical trials where correlated observations or unobserved patient characteristics influence outcomes. Health economic models use frailty-adjusted survival estimates to improve long-term projections of costs, quality-adjusted life-years and life expectancy.


Purpose

Used to account for unobserved heterogeneity in survival data, improve estimation of treatment effects and generate more realistic long-term survival projections for health economic evaluation.


Mathematical Formulae

Primary Formula

h(t | x, z) = z ? h?(t) ? exp(x??)

where:

  • h(t | x, z) = individual hazard function
  • z = frailty term
  • h?(t) = baseline hazard
  • x = vector of observed covariates
  • ? = regression coefficients

Supporting Formulae

For gamma frailty:

E(z) = 1

Var(z) = ?

where ? represents the degree of unobserved heterogeneity.

Model parameters are estimated using:

?? = arg max L(?)

Related Mathematical Methods

  • Cox proportional hazards model
  • Shared frailty models
  • Gamma frailty models
  • Log-normal frailty models
  • Maximum likelihood estimation
  • Bayesian survival modelling
  • Mixed-effects survival models

Example

A multinational oncology trial includes patients from multiple hospitals. Although age, tumour stage and treatment are included in the model, substantial unexplained differences in mortality remain between centres. A shared gamma frailty model is fitted to account for centre-level heterogeneity, resulting in improved estimates of long-term survival used to calculate lifetime QALYs and incremental cost-effectiveness ratios.


Excel Implementation

FunctionExample FormulaHealth Economics Application
EXP=EXP(SUMPRODUCT(B2:E2,$B$1:$E$1))Calculate the relative hazard from observed covariates.
GAMMA.DIST=GAMMA.DIST(A2,Shape,Scale,FALSE)Evaluate gamma frailty distributions during exploratory analyses.
SUMPRODUCT=SUMPRODUCT(Covariates,Coefficients)Calculate the linear predictor for the survival model.
LN=LN(A2)Calculate log-likelihood components during model estimation.

VBA (Optional)

Automate frailty model estimation across multiple datasets and compare alternative frailty distributions for survival modelling.


Sources

  • Hougaard P. Analysis of Multivariate Survival Data.
  • Duchateau L, Janssen P. The Frailty Model.
  • Klein JP, Moeschberger ML. Survival Analysis: Techniques for Censored and Truncated Data.
  • Briggs A, Claxton K, Sculpher M. Decision Modelling for Health Economic Evaluation.
  • NICE. Health Technology Evaluation Manual.

Library

Publications

1
  • Guidance

    NICE DSU Technical Support Document 21: Flexible methods for survival analysis — Rutherford, Lambert, Sweeting, Pennington, Crowther, Abrams & Latimer, TSD 21 ed., 2020 (NICE Decision Support Unit (University of Sheffield))

    Guidance extending standard survival analysis to flexible parametric methods — spline-based models, fractional polynomials, mixture and cure models, and relative-survival approaches — for capturing complex hazard functions in economic evaluation.

Frequently Asked Questions (6)

  • What is a frailty model?

    A survival model incorporating an unobserved random effect, called frailty, to represent risk heterogeneity not captured by observed covariates.

    Source: Vaupel, Manton & Stallard 1979

  • Why can unobserved differences between patients distort survival analysis?

    Patients differ in ways that affect their risk but are not recorded, such as unmeasured genetic or lifestyle factors, and ignoring this hidden variation can bias a survival analysis, for instance by making the population's hazard appear to fall over time simply because the frailest die first. A frailty model adds an unobserved random term representing each individual's or group's extra risk, so the analysis accounts for variation the measured covariates miss. This corrects the distortion that unmeasured heterogeneity introduces. Klein and Moeschberger (2003) describe frailty models.

    Source: Klein & Moeschberger 2003

  • How does a frailty model work?

    A frailty model works by including a random frailty term, usually multiplying the hazard, that varies across individuals or groups according to a distribution, representing unobserved risk factors. Individuals with higher frailty have a proportionally higher hazard. The model estimates the effect of observed covariates alongside the variance of the frailty distribution, which captures the extent of unobserved heterogeneity. Frailty can be at the individual level or shared within groups, such as families or centres, to model correlated survival times.

    Source: Vaupel, Manton & Stallard 1979

  • Why are frailty models used?

    Frailty models are used to account for unobserved heterogeneity in survival risk, which, if ignored, can bias estimates and produce misleading hazard patterns. For example, when a population contains individuals of differing unmeasured risk, the frailer die earlier, so the observed population hazard can appear to decline even if each individual's hazard rises, an artefact frailty models can explain. They are also used to model correlated survival times within groups. By representing unobserved variation, frailty models give a more accurate account of survival.

    Source: Vaupel, Manton & Stallard 1979

  • What is the frailty term?

    The frailty term is the unobserved random effect in a frailty model that represents an individual's or group's susceptibility to the event beyond what observed covariates explain. Typically it multiplies the baseline hazard, so a frailty above the average raises the hazard and below it lowers the hazard, and it follows an assumed distribution, such as the gamma distribution, whose variance measures the amount of unobserved heterogeneity. The frailty term thus captures the hidden variation in risk among individuals or within groups.

    Source: Kalbfleisch & Prentice 2002

  • How does a frailty model account for unobserved heterogeneity in survival analysis?

    Unobserved heterogeneity affects survival analysis because a population of individuals with differing unmeasured risk behaves differently from a homogeneous one: the higher-risk individuals experience the event earlier, leaving lower-risk survivors, so the observed population hazard can fall over time even when each individual's hazard is constant or rising. Ignoring this can misrepresent the hazard and bias covariate effects. Frailty models address it by explicitly representing the heterogeneity, distinguishing the individual hazard pattern from the artefact created by selective survival of the less frail.

    Source: Vaupel, Manton & Stallard 1979

Trust Record

Verified by Dr Darrin Baines

British health economist

Professional identity: darrinbaines.org

Verification date: 20 Oct 2025

Content version: 1.0.0

Canonical Identity

Term code
HE-EM-SM-025

Stable URI · Machine-readable · Resolvable · CC BY 4.0