Toward a Deeper Understanding of Predicting Risk of Cardiovascular Disease Events for 10-Year Atherosclerotic Cardiovascular Risk: Subgroup Fairness and Predictive Value of Social Determinants of Health.

BACKGROUND

The American Heart Association's Predicting Risk of Cardiovascular Disease Events model offers a modern, race-free approach to risk prediction, but its subgroup fairness and the added value of social determinants of health (SDOH) remain underexplored.

METHODS

To evaluate Predicting Risk of Cardiovascular Disease Events in 10-year atherosclerotic cardiovascular disease prediction regarding fairness and value of SDOH predictors, we conducted a retrospective cohort study of 554 675 adults aged 30 to 79 years, using deidentified electronic health records from Truveta, a multisystem US data platform. Subgroup fairness was assessed using percentile calibration plots and Cross Concordance Index metric. The incremental value of SDOH was evaluated by comparing discrimination, calibration, and fairness across models.

RESULTS

The 10-year atherosclerotic cardiovascular disease event rate was 1.8%. Most subgroups exhibited consistent calibration and the Cross Concordance Index values. The most pronounced disparities were observed between White and Asian participants (event rate: 10.3% versus 6.6% at the 95th percentile; Cross Concordance Index=0.849 versus 0.679), and private and public insurance groups (event rate: 0.7% versus 1.5% at the 25th percentile; Cross Concordance Index=0.578 versus 0.859). Adding SDOH predictors had minimal effects on model performance.

CONCLUSIONS

Predicting Risk of Cardiovascular Disease Events showed fairness across most demographic and SDOH subgroups, supporting its practical use to predict atherosclerotic cardiovascular disease risk. Adding SDOH predictors offered minimal incremental benefit, reinforcing the original equations' utility as a reliable and fair tool for general populations.