Precision Drug Theraputics: Risk Prediction in Pharmacoepidemiology

Health Data Research UK (HDR UK) is an institute for health and biomedical informatics research. Part of its mission is to transform health care through the application of multidisciplinary team expertise.

This project aims to contribute to the HDR UK mission by applying advanced methodological approaches to build models which will enable outcomes to be predicted for appropriate groups of patients with distinct mechanisms of disease (cardiovascular disease, cancer) or particular responses to treatment (stratified medicine) and to support the development of precision therapeutics.

The objective of the project is to create Clinical Decision Support tools for routine use.

The project will develop statistical and machine learning algorithms to analyse NHS datasets to build predictive models to support precision therapeutics focusing on the following two areas:
1) Cancer chemotherapy
2) Cardiovascular therapies

HDR UK will provide the data repositories for linkage of established patient cohorts to further phenotypic resources. This will provide deeper phenotyping and should allow more precise prediction algorithms to be developed. Advanced statistical modelling (logistic regression, least absolute shrinkage and selection operator (LASSO) regression, Bayesian modelling) and machine learning approaches (neural nets, deep learning algorithms) will be used as part of this approach. The precision and incremental value of additional datasets will be assessed as will automated approaches to optimising algorithms. Existing research in the use of Bayesian hierarchical point-mass models, based around body-systems, to detect safety signals in clinical trials may be extended for use in observational studies. This approach is designed to use the additional information given by the relationships within and between body-systems in a statistical analysis.

The outcomes of the project, in this case the prediction models, will be translated into Clinical Decision Support (CDS) tools which will promote learning health systems. User and usability experience data will be captured during CDS design and development. The output will be a prototype intervention suitable for evaluation through the MRC Complex Intervention Framework to support clinicians in making informed choices and highlighting the risks in the different possible prescribing choices.