Cookies on this website

We use cookies to ensure that we give you the best experience on our website. If you click 'Accept all cookies' we'll assume that you are happy to receive all cookies and you won't see this message again. If you click 'Reject all non-essential cookies' only necessary cookies providing core functionality such as security, network management, and accessibility will be enabled. Click 'Find out more' for information on how to change your cookie settings.

external supervisor

Dr Davide Morelli, Department of Engineering Science


Time-to-event (also called "survival") analysis is widely used in clinical research, where the Cox model remains popular. Recent advances in machine learning (ML) have provided unprecedented research opportunities for furthering survival analysis. One particularly promising method called "dynamic time-to-event analysis" allows the ML model to dynamically update its predictions based on real time events that occur at multiple time points.

This project will produce a novel methodology of dynamic time-to-event analysis using longitudinal data. The aim is to complement and improve the classical survival analysis, by utilising the information embedded in longitudinal data. Candidates will first implement state-of-the-art models for dynamic time-to-event, and then improve their methodology (e.g. applying normalizing flows or variational Bayes models to survival analysis).

The developed ML methods can be applied to a wide range of clinical settings, and this project will investigate time-to-normalisation of patient physiology and biomarkers in the treatment of infections in hospitals. The overall aim is to determine different recovery phenotypes from infection (or any other condition, e.g. gastrointestinal bleed, major surgery, etc.) which might help plan therapy and discharge, but also identify individuals who are not recovering as expected.

The strength of dynamic time-to-event analysis lies in its ability to dynamically take account of multiple events occurring at multiple time points, including updated vital sign measurements, laboratory tests and events of interest, e.g. a bloodstream infection, a list of operations, an Intensive Care Unit (ICU) stay, and particular diagnostic/procedure codes.


Candidates will acquire research skills through regular supervisory meetings, and by attending relevant seminars, courses, workshops. The student will learn how to handle and analyse large-scale epidemiological data, and how to interpret ongoing findings and subsequently explore potential methodologies. 

Candidates will be able to access the Infections in Oxfordshire Research Database (IORD) datasets that cover approximately 1% of England, providing comprehensive electronic healthcare data from patients attending hospitals in Oxfordshire.


There may be opportunities to work with external partners and/or on different datasets. For example, candidates can access the publicly-available datasets via the "Physionet" resource, if needed.