Abstract 4121039: Refining the American Heart Association Predicting Risk of Cardiovascular Disease EVENTs (PREVENT) equations using depression and anxiety disorder

Abstract

Background: Growing evidence has suggested associations of common mental health conditions including depression and anxiety disorder with CVD, but no studies have examined their incremental predictive values in PREVENT (AHA Predicting Risk of CVD EVENTs). Research questions: Does including anxiety and/or depression improve the PREVENT predictive performance? Aims: We used a range of measures of anxiety and depression and combinations to examine whether these conditions improved the prediction performance of PREVENT. Methods: New risk prediction models were developed and internally validated using 60% and 40% of the UK Biobank cohort data, respectively. Mental health predictors included: depressive symptom score based on the four-item Patient Health Questionnaire and self-reported and record-based anxiety and depression diagnoses before baseline. CVD events were ascertained through hospital admission and death certificate data after baseline. The Cox proportional hazards models were used to calculate coefficients and 95% confidence intervals. The incremental predictive values were examined by adding the mental health predictors to the PREVENT predictors based on Harrell’s C-indices, sensitivity, specificity, positive and negative predictive values, and net reclassification improvement indices (NRIs). Results: Of the 502,366 UK Biobank participants, 195,489 were included in the derivation set and 130,326 were included in the validation set. In the validation set, the inclusion of each mental health measure produced an increase in C-index and higher sensitivity, specificity, and positive and negative predictive values compared with the PREVENT equations, with the greatest improvements for depressive symptom score. Depressive symptom score produced the greatest improvements in net reclassification overall (NRI 0.84, 95% CI 0.46–1.20) and this was further improved by adding recorded anxiety and depression (NRI 0.89, 95% CI 0.52–1.24). In addition, depressive symptom score showed better discrimination and NRI in both female and male validation sets. Conclusions: Our findings suggest that including depressive symptom score in PREVENT could improve the prediction of CVD with low cost and without invasive procedures.