Atrial Fibrillation (AF) affects approximately 2% of the UK population and is the cause of 20% of all strokes. Furthermore, strokes due to AF are more likely to result in death or disability. Although the risk of stroke in AF can be effectively managed by oral anticoagulation, treatment is known to be underused – for example a recent study on UK primary care (CPRD) data found 65.9% of high-risk newly diagnosed AF patients received anticoagulation in January 2016 (Lacoin et al., 2017).
Need and opportunity
There is an urgent need to identify patients who are not receiving stroke prophylaxis. Integrating care data across primary, secondary and community care is also vital to improve patient stratification and treatment planning beyond anticoagulation (e.g. ablation, cardioversion). Within HDRUK there is an unique opportunity to address this need, delivering real-world NLP capability to the NHS and a platform of validated NLP tools to the community.
Data and analysis
The three datasets used in this project cover primary, secondary and community care and do so at scale. 1. NHS Lothian (secondary), 1.2m patients. 2. KCH (secondary), 3m patients. 3. NE London Discovery (primary, secondary), 2.2m patients. Building on existing NLP and analytics capability within the HDR-UK applicants, we will apply NLP tools to the free text data across the three datasets to extract clinical features of interest (including medication, comorbidities, risk factors, brain imaging, treatment plans). The two key areas of analysis will be stratification, identifying at-risk populations and automatic risk scoring. Across all analysis we will focus on validating methods and performance across sites.
Benefit to community
This exemplar project delivers benefits to the NHS, the HDR-UK community and the wider research community. Through this project we will develop NLP tools for patient stratification and risk scoring which are validated across HDR-UK sites. These tools will be open source and available to the community as part of a broader NLP platform. Sharing technical expertise and real-world knowledge of EHR systems within the implementation project will accelerate the delivery of this capability to the NHS and greatly improve the generalisability of the methods.