Prediction of cardiovascular disease in patients with unattributed chest pain in UK primary care

Conference:

SAPC ASM 2022 - UCLan

Talk Code:

1D.3

Presenter:

Kelvin Jordan

Co-authors:

Trishna Rathod-Mistry, James Bailey, Danielle van der Windt, Ying Chen, Lorna Clarson, Spiros Denaxas, Richard Hayward, Harry Hemingway, Theocharis Kyriacou, Mamas Mamas

Author institutions:

Keele University, University College London, Xi'an Jiaotong - Liverpool University

Problem

Most adults presenting in primary care with chest pain symptoms will not receive a specific diagnosis (“unattributed” chest pain). These patients are more likely to develop cardiovascular disease (CVD) than patients with chest pain attributed to a non-coronary reason. Determining which patients with unattributed chest pain have the greatest risk of CVD would allow for targeted intervention strategies. Current risk prediction algorithms for CVD (e.g. QRISK3) have been developed for the general population but may not be appropriate for this group of patients. The aim was to assess within patients with unattributed chest pain, whether those at greatest risk of CVD can be ascertained.

Approach

We used the CPRD Aurum database containing electronic health records from English general practices linked to admitted patient hospitalisations from the Hospital Episode Statistics database. The study population was patients aged 18 and over with a new primary care record of unattributed chest pain between 2002 and 2018, and no record of CVD up to six months (diagnostic window) afterwards. Outcomes were cardiovascular events starting from end of the diagnostic window.

Flexible parametric survival analyses were used to derive risk factors for future CVD over 10 years. Baseline candidate factors (n=23) were those included in general population cardiovascular risk algorithms, alternative explanations for chest pain, and other comorbidities predictive of CVD. We developed and validated a prediction model, with external validation in a second primary care database (CPRD GOLD) linked to admitted patient hospital data and compared performance to a risk prediction model (QRISK3) developed for use in the general population.

Findings

There were 374,917 patients with unattributed chest pain. Median follow-up was 6.1 years. Incidence of CVD was 19.3 per 1000 person-years. The strongest comorbidity risk factors for CVD included type I diabetes (adjusted hazard ratio 2.41; 95% CI 2.11, 2.76), atrial fibrillation (1.95; 1.85, 2.06), and hypertension (1.55; 1.50, 1.59). Socio-demographic risk factors included older age, male gender, greater deprivation, and Asian ethnicity.

Internal validation of the final model showed high predictive performance with c-statistic of 0.797. Discrimination and calibration performance were good when stratified by gender, neighbourhood deprivation, and geographical region.

In the external validation dataset, the c-statistic was 0.805 and calibration slope close to one. Calibration plots showed good agreement between observed and expected risk at all levels of risk. A reduced model using a subset of key risk factors for CVD gave nearly identical performance. QRISK3 performed less well in this population.

Consequences

Patients presenting to primary care with unattributed chest pain are at increased risk of cardiovascular events, but it is feasible to ascertain those most at risk using routinely recorded information in the primary care record. These patients could then be targeted for preventative measures.

Submitted by:

Kelvin Jordan

Funding acknowledgement:

Study funded by the British Heart Foundation, reference PG/19/46/34307. KJ also supported by matched funding awarded to the NIHR Applied Research Collaboration (West Midlands). This study is based in part on data from the Clinical Practice Research Datalink obtained under licence from the UK Medicines and Healthcare products Regulatory Agency. The data is provided by patients and collected by the NHS as part of their care and support. The Office for National Statistics (ONS) is the provider of the ONS data contained within the linked CPRD data used for this study. ONS Data and Hospital Episode Statistics (HES) Data: copyright © (2020), re-used with the permission of The Health and Social Care Information Centre; all rights reserved. The interpretation and conclusions contained in this study are those of the authors alone and not necessarily the views of The British Heart Foundation, the NHS, the NIHR or the Department of Health and Social Care.