Atrial fibrillation (AF) is a major public health issue: it is increasingly common, incurs substantial healthcare expenditure, and is associated with a range of adverse outcomes. There is rationale for the early diagnosis of AF, before the first complication occurs. Previous AF screening research is limited by low yields of new cases and strokes prevented in the screened populations. For AF screening to be clinically and cost-effective, the efficiency of identification of newly diagnosed AF needs to be improved and the intervention offered may have to extend beyond oral anticoagulation for stroke prophylaxis. Previous prediction models for incident AF have been limited by their data sources and methodologies. An accurate model that utilises existing routinely-collected data is needed to inform clinicians of patient-level risk of AF, inform national screening policy and highlight opportunities to improve patient outcomes from AF screening beyond that of only stroke prevention. The investigators will use routinely-collected hospital-linked primary care data to develop and validate a model for prediction of incident AF within a short prediction horizon, incorporating both a machine learning and traditional regression method. They will also investigate how atrial fibrillation risk is associated with other diseases and death. Using only clinical factors readily accessible in the community, the investigators will provide a method for the identification of individuals in the community who are at risk of AF, thus accelerating research assessing whether atrial fibrillation screening is clinically effective when targeted to high-risk individuals.

Atrial fibrillation (AF) is a major public health issue: it is increasingly common, incurs substantial healthcare expenditure, and is associated with a range of adverse outcomes. There is rationale for the early diagnosis of AF, before the first complication occurs. Previous AF screening research is limited by low yields of new cases and strokes prevented in the screened populations. For AF screening to be clinically and cost-effective, the efficiency of identification of newly diagnosed AF needs to be improved and the intervention offered may have to extend beyond oral anticoagulation for stroke prophylaxis. Previous prediction models for incident AF have been limited by their data sources and methodologies. An accurate model that utilises existing routinely-collected data is needed to inform clinicians of patient-level risk of AF, inform national screening policy and highlight opportunities to improve patient outcomes from AF screening beyond that of only stroke prevention.

The application of Random Forest will be investigated and multivariable logistic regression to predict incident AF within a 6 months prediction horizon, that is a time-window consistent with conducting investigation for AF. The Clinical Practice Research Datalink (CPRD)-GOLD dataset will be used for derivation, and the Clalit Health Services dataset will be used for international external geographical validation. Both comprise a large representative population and include clinical outcomes across primary and secondary care. Analyses will include metrics of prediction performance and clinical utility. Only risk factors accessible in the community will be used and the model could thus enable passive screening for high-risk individuals in electronic health records that is updated with presentation of new data. The study aims to create a calculator from a parsimonious model. Kaplan-Meier plots for individuals identified as higher and lower predicted risk of AF will be calculated and derive the cumulative incidence rate for non-AF cardio-renal-metabolic diseases and death over the longer term to establish how predicted AF risk is associated with a range of new non-AF disease states.

To ascertain whether the prediction model is transportable to geographies outside of the UK, the model's performance will be externally validated in the Clalit Health Services database in Israel. The validation will include participants insured by Clalit with continuous membership for at least 1 year before 01/01/2019: 2,159,663 patients with 4,330 of them having a new incident of AF (Atrial fibrillation and/or atrial flutter) in the first half of 2019. The study population will comprise all available patients who have at least 1-year follow up. The outcome of interest is the first diagnosed AF after baseline and will be identified using Read codes and ICD-9/10 codes. Patients with less than one year of registration, who are under thirty years of age at point of study entry, or have a preceding diagnosis of atrial fibrillation, will be excluded.

Inclusion Criteria:

A least 1 year follow-up

Exclusion Criteria:

Diagnosed AF before study entry

The derivation dataset will be the Clinical Practice Research Datalink-GOLD (CPRD-GOLD) dataset. The extracted dataset, including linked data, comprises all patients for the period between 2nd January 1998 and 30th November 2018 from the snapshot of CPRD-GOLD in October 2019.

No individual participant data will be shared.