Abstract 4359477: Evaluation of PREVENT Equations for Estimating Cardiovascular Disease Risk in Individual Asian American and Native Hawaiian and Pacific Islander Groups

Abstract

Background: The race-free Predicting Risk of cardiovascular disease (CVD) EVENTS (PREVENT) equations estimate 10-year risk for total CVD. However, their accuracy in individual Asian or Native Hawaiian and Pacific Islander (NHPI) ethnic groups remain unknown. Research Question: What is the risk prediction accuracy of the PREVENT Base and Full Equations in Asian and NHPI ethnic groups? Methods: A retrospective cohort study was conducted among adults aged 30-79 years without a history of CVD who self-reported as non-Hispanic (NH) White, NH Asian, or NH NHPI and were actively enrolled as members of Kaiser Permanente Southern California on September 30 th , 2009, with follow-up through 2019. Asian participants were further disaggregated into ethnic groups. Ten-year risk for total CVD was estimated using the PREVENT Base and Full (Base plus hemoglobin A1c, urine albumin-creatinine ratio, and social deprivation index) equations. Model discrimination was evaluated using Harrell’s C-index, and calibration assessed using predicted-to-observed event ratios. Results: Among 542,848 adults (424,277 NH White; 110,855 NH Asian; 7,716 NH NHPI), 31,556 CVD events occurred over 10 years. Harrell’s C-index for the PREVENT Base equation was 0.764 (95% CI = 0.761–0.767) in NH White, 0.773 (95% CI = 0.765–0.779) in NH Asian, and 0.757 (95% CI = 0.733–0.780) in NH NHPI adults ( Table ). PREVENT Full equation showed similar C-indexes. Among Asian ethnic groups, PREVENT Base and Full equation discrimination was highest in Chinese (0.810) adults and lowest in Vietnamese (0.735) adults. The PREVENT Base and Full equations underestimated total CVD risk in NH White (mean calibration 0.63–0.84) and NH NHPI (mean calibration 0.74–0.86) adults. In contrast, the PREVENT Base equation overestimated total CVD risk in overall NH Asian adults (mean calibration 1.18) and most disaggregated Asian ethnic groups (mean calibration 1.04-1.39), while the PREVENT Full equation underestimated total CVD risk in overall NH Asian (mean calibration 0.96) and generally well calibrated or underestimated total CVD risk in disaggregated Asian ethnic groups (mean calibration 0.83–1.07). Conclusions: Despite the overall strong discrimination of the PREVENT Base and Full equations in predicting total CVD over 10 years, calibration varied across disaggregated Asian and NHPI ethnic groups, highlighting the importance of accounting for population heterogeneity in CVD risk assessment.