Subscribe to our "Population in Perspective" Substack Here!
Applications are now open for four University of Oxford’s DPhil in Population Health projects supervised by researchers at the Leverhulme Centre for Demographic Science (LCDS).
The projects span demographic forecasting, conflict related mortality, multimorbidity and the safe use of large language models in population health research, bringing together demographic, statistical and computational approaches to major population health questions.
Applications for entry in October 2027 close at noon on 1 December 2026.
(1) The limits of demographic predictability: forecasting fertility, mortality and migration under uncertainty
Supervisors: Professor Jakub Bijak and Associate Professor Charles Rahal
Population projections inform decisions across public services, infrastructure, labour markets and health, but demographic change is not equally predictable across fertility, mortality and migration.
This project will develop a comparative framework for understanding demographic predictability under changing conditions. It will examine how different sources of uncertainty affect fertility, mortality and migration forecasts, and how that uncertainty propagates into projections of population size and structure.
The research will draw on publicly available demographic data and established time series and entropy based methods, with scope to extend into probabilistic, machine learning or individual level approaches depending on the candidate’s interests and findings.
Find out more about the project.
(2) Novel methods for estimating conflict related mortality
Supervisors: Professor Jakub Bijak and Associate Professor Charles Rahal
Estimating mortality caused by conflict remains challenging because records can be incomplete, duplicated, selectively reported and inconsistent across sources.
This project will develop and evaluate an integrated statistical and computational framework for estimating conflict related mortality from public data. It will combine established demographic methods with machine learning approaches to identify, reconcile and model casualty records and contextual information.
Potential methods include natural language processing, large language models, probabilistic record linkage and Bayesian modelling, with uncertainty carried through to the final mortality estimates.
Find out more about the project.
(3) Interpretable and transportable measures of multimorbidity from population scale health records
Supervisors: Associate Professor Charles Rahal and Jiani Yan
Multimorbidity is often measured using simple counts of conditions or indices developed for specific outcomes, which can obscure important differences in how combinations of diseases affect people’s health and daily lives.
This project will develop transparent, multidimensional measures of multimorbidity that distinguish between burden associated with shortened life, day to day impact and the management of multiple medicines.
Measures will be developed using UK Biobank and externally evaluated in China Kadoorie Biobank to assess whether they remain meaningful across different health systems and social groups.
Find out more about the project.
(4) Breaking the guardrails: stress testing, jailbreaking and ultimately securing LLMs for Population Health research
Supervisors: Associate Professor Charles Rahal, Dr Daniel Valdenegro and Jiani Yan
Large language models are increasingly being used across health research, from literature screening and evidence synthesis to coding and analysis. Their growing use also creates new safety risks when safeguards can be bypassed through adversarial prompts, documents or tool interactions.
This project will develop a health research specific framework for measuring and mitigating vulnerabilities in LLM systems. It will compare different forms of attack, distinguish harmless non refusal from consequential failure, and test safeguards while measuring their effect on legitimate research use.
The work will use synthetic tasks, canary data and sandboxed tools rather than real patient data or live clinical systems.
Find out more about the project.
How to apply
These projects form part of the University of Oxford DPhil in Population Health within Oxford Population Health and are listed under the Demographic Science Unit.
Before submitting an application for an advertised project, prospective students should contact the relevant supervisor using the University’s prospective supervisor contact form. Applicants must secure a supervisor’s agreement in principle before submitting their formal application.
Applications for October 2027 entry close at noon on Tuesday 1 December 2026.
Full details of the entry requirements for admission to the DPhil in Population Health are available on the DPhil in Population Health Graduate Admissions webpage.
