
05
Jun
Flexible parametric survival models, competing risks, and non-proportionality of hazards
This seminar explores flexible parametric survival models using cubic splines to model complex hazard functions. Applications include competing risks, cumulative incidence functions (CIF), and non-proportional hazards. A cancer registry dataset will illustrate how these models reveal sex differences in comorbidity and long-term mortality.
10
Apr
An introduction to prognosis and prediction research
Prognosis research quality has been questioned due to conflation with causal research. The PROGRESS framework (2013) outlined four key themes: prognosis description, prognostic factors, prognostic models, and stratified medicine. This talk will highlight their clinical importance, discuss appropriate research methods, and differentiate variable selection approaches for prognostic and causal research. The insights also extend to broader prediction research.
06
Feb
Target trial emulation: a framework for causal inference from observational data
Non-randomised studies should aim to mimic randomised trial designs to reduce bias. The target trial emulation framework helps researchers achieve this, but reporting remains poor. Harrison will introduce the framework, review its use in medical literature, and present TARGET, the first reporting guideline for these studies.
10
Oct