Title: Personalised screening schedules for optimal prevention of cardiovascular disease
Authors: Francesca Gasperoni - MRC Biostatistics Unit, University of Cambridge (United Kingdom) [presenting]
Paul Newcombe - MRC Biostatistics Unit - University of Cambridge (United Kingdom)
Chris Jackson - MRC Biostatistics Unit -Cambridge (United Kingdom)
Angela Wood - University of Cambridge (United Kingdom)
Jessica Barrett - MRC Biostatistics Unit (United Kingdom)
Abstract: Cardiovascular disease (CVD) population screening strategies aim to identify and treat people at high risk of CVD. Current UK guidelines recommend screening adults over 40 years old every 5 years and prescribing statins for those with a predicted 10-year CVD risk greater than 10\%. We propose an incremental net benefit function to investigate a personalised screening schedule, considering personal CVD risk profile. This function is composed of benefit (event-free life years) and costs (of statins and visits provided by the health services). The prescription of statins is assumed to start at the first visit after the 5-year CVD risk exceeds the 5\% threshold. To assess this risk by adjusting for time-varying endogenous covariates, we use a two-stage dynamic landmark model. The first stage consists in fitting at each landmark age (i.e., 40,45,..,80 years) a multivariate linear mixed effect model with random intercepts and slopes. The second stage consists in predicting the CVD risk through a Cox model, adjusted for the risk factor values estimated at stage one. We apply the proposed model to data from the Clinical Practice Research Datalink (CPRD), comprising primary care Electronic Health Records from the UK. From preliminary analyses, baseline characteristics play a significant role in the optimal schedule. In particular, people labelled as high-risk seem to require more frequent visits, while low-risk people seem to require visits less frequently than every 5 years.