Combining data and AI to predict heart problems following Covid

Electronic health records contain a wealth of information that has the potential to be used in research. However, these kinds of datasets are vast and messy. Longitudinal studies take place over many years with participants being followed up throughout their lives. These studies have very rich datasets with information not found in health records. Secure data environments (SDEs) are a secure and efficient way of enabling trusted researchers to safely access data for approved projects. The UK Longitudinal Linkage Collaboration (UKLLC) is an SDE which brings together both kinds of data. Bringing both these types of data together has the potential to be very powerful. Electronic health records only represent those who are seeking healthcare. Longitudinal datasets only represent their participants. Combining health records with longitudinal data could broaden the range of people in a study. This could mean that findings from research combining these two kinds of data would be more applicable across the population. Project aims We want to find out whether combining health record data with longitudinal data gives better results than using health record data alone. Artificial intelligence (AI) is a useful tool when using large volumes of data from these kinds of sources. We will investigate the risk of heart attack or stroke in the 12 months following a Covid-19 diagnosis. Clinicians use scores to predict a patient’s risk. We want to understand whether our AI risk prediction tool is better than standard clinical risk scores. What we hope to achieve Our use case has the potential to improve the prediction of a patient’s risk of heart attack or stroke following a Covid-19 diagnosis. We also hope this project can demonstrate the potential of combining large datasets in this way.