Parametric survival modelling is a key ingredient in all health economic models. Some of the models we offer are:
- A host of flexible parametric survival models (click here for ISPOR article) that can cope with any shape of hazard curve (often needed in diseases such as oncology)
- Competing risk models where more than one health state can be directly reached from another. We can embed competing risks in flexible survival models incorporating time varying hazard ratios
- Joint models for longitudinal and time-to-event data (can account for the effect of an endogenous time-dependent covariate among other uses)
- Relative survival and Cure models – very pertinent to Oncology areas enabling more accurate survival models – particularly where “cure” is a potential outcome.
- Models to deal with treatment switching and appropriate sensitivity analysis – every technique has potential drawbacks/restrictive assumptions that need investigating.
- Machine learning algorithms such as Random Survival Forests (robust, nonlinear technique that can optimize predictive accuracy and deal effectively with many covariates)
- Mixed/Latent class models such as Weibull-Weibull. Frailly models. Four Parameter Generalised F models.
Developments in survival modelling are happening all the time. We monitor the output of certain academics who we admire in this field. We employ visual diagrams and bootstrapped cross-validation techniques to supplement more traditional AIC criteria in deciding between models. integration.