Predicting time-to-event outcomes under different intervention strategies: methods and applications

Prosepe, I. (2025)

Doctoral Thesis

This dissertation focuses on developing and applying estimation methods to make “what if” predictions in clinical research. The primary aim of these methods is to estimate risks for time-to-event outcomes using observational data in a manner that supports valid prediction under different intervention strategies, potentially conditional on patient and treatment characteristics. To this end, we formulate the identifiability assumptions required to ensure the
validity of these estimates.

• Causal inference
• Predictions under interventions
• Inverse probability weighting
• Survival analysis
• Observational data
• Multi-state models

• All authors: Prosepe, I.
• Supervisor: Putter, H.
• Co-supervisor: Geloven, N. van
• Committee: Fiocco, M.; Keogh, R.H.; Post, R.A.J.; Amsterdam, W. van
• Qualification: Doctor (dr.)
• Awarding Institution: Faculty of Medicine, Leiden University Medical Center (LUMC), Leiden University
• Date: 2025-12-03
• ISBN (print): 9789465228068
• DOI: doi:10.60602/arjs-b248