Background:
Routinely collected data stored in electronic health records (EHRs) provide a powerful resource for evaluating interventions for real-world patient populations; however, internal validity is frequently impacted by confounding. The trial emulation framework proposed by Hernan and colleagues mimics a target trial using observational data and crucially requires addressing confounding bias. The recently developed Foresight transformer converts text information into timelines of events and thus might provide a convenient data format for improving confounding control.
Novelty & Importance:
Using Foresight generated timelines in trial emulation studies requires the development of statistical methodology. This project aims to develop a matching methodology for the estimation of causal treatment effects. To this end a tool/program will be developed that creates synthetic controls with similar pre-treatment timelines for comparison with treated patients, and which can be used on any EHR processed with Foresight.
Aims & Objectives:
• To extract data sets for target trial emulations from EHRs using the Foresight data generator.
• To develop statistical methodology for estimating causal effects of treatments (synthetic control matching).
• To develop exemplar target trial emulation studies using the KCL-CVD registry.
The KCL-CVD registry is a de-identified database of approximately 100,000 individuals derived from EHR data from two large university hospitals – Guy’s & St Thomas’ and King’s College Hospital between 2012 and 2022. It provides a rich resource to support research into cardiovascular diseases and their care. Case studies might include staged discontinuation of post-myocardial infarction pharmacological therapies or contemporary pharmacological therapies in heart failure.
References:
Kraljevic Z., Bean D., Sheck et al. (2024) Foresight – a generative pretrained transformer for modelling of patient timelines using electronic health records: a retrospective modelling study. Lancet Digital Health 6 (4): e282-e290. DOI: 10.1016/S2589-7500(24)00025-6
McGarvey M., Lam L.T., Razak M.A., et al. (2024) Impact of lesion morphology on stent elongation during bifurcation PCI: an in vivo OCT study. EuroIntervention 20 (18): e1184-e1194. DOI: 10.4244/EIJ-D-23-00663
Scola G., Chis-Ster A., Bean D., Pareek N., Emsley E. & Landau S. (2023) Implementation of the trial emulation approach in medical research: a scoping review, BMC Medical Research Methodology 23: 186; DOI: 10.1186/s12874-023-02000-9

