Evidence synthesis and evaluation of health policies

Health policies are commonly informed using models that combine evidence from multiple sources. For example,

• models to inform policies for prevention of chronic diseases, e.g. routine mid-life health checks (paper)
• decision models used by NICE to recommend whether a new treatment should be funded by the health service

Chris's work involves statistical methods to ensure that the decisions based on these kind of models accurately reflect the available evidence. For example,

• how to specify Bayesian models that allow direct data, indirect data and expert belief to be combined (paper)
• methods to combine short-term data from clinical trials with longer-term data and assumptions ("survival extrapolation") (paper)
• methods to compare alternative model assumptions ("structural uncertainty") (paper|paper)

He has also made developments to Value of Information methods for quantifying uncertainties and prioritising further research in these models (paper)(review paper).

Survival and multistate modelling

Chris has developed methods and software for survival and multistate modelling, in particular parametric survival models (paper) and multi-state models for intermittently-observed data (paper). He recently developed a novel framework for multistate modelling, which was applied to hospital admissions for COVID-19 (paper).   Chris is an Associate Editor of the journal Biometrics.

Statistical computing

Developing and maintaining several R packages, including

• msm for multistate modelling of intermittently-observed data
• flexsurv for parametric survival modelling and multistate modelling of time-to-event data
• voi for calculating the Expected Value of Information
• disbayes for estimation of chronic disease burden from indirect data
• denstrip for illustrating distributions
• fic for focused model comparison
• survextrap for Bayesian survival extrapolation

Previously contributed to the OpenBUGS and BRugs software for Bayesian analysis.  

Teaching

• Public short course at MRC Biostatistics Unit: Bayesian Statistics (to be advertised soon)
• University of Cambridge, MPhil in Population Health Sciences, Bayesian Statistics module
• University of Cambridge, Statistics in Mathematical Tripos Part III: coordinating and teaching on Statistics in Medical Practice
• Multi-state modelling with msm: freely-available course material.

Recent presentations

• Armitage Lecture workshop 2021: Bayesian multi-state modelling of incomplete chronic disease data for health impact models

Google Scholar profile