Development and Validation of an Incident Hypertension Risk Prediction Model for Young Adults.

BACKGROUND

Identifying young adults at high risk of hypertension can improve blood pressure screening recommendations.

METHODS

We developed models to predict incident hypertension using diverse contemporary cohorts of young adults aged 18 to 39 years from Kaiser Permanente Southern California (derivation and internal validation) and 3 cohort studies (CARDIA [Coronary Artery Risk Development in Young Adults], FHS [Framingham Heart Study], HCHS/SOL [Hispanic Community Health Study/Study of Latinos]; external validation). Predictors included age, systolic and diastolic blood pressure, body mass index, smoking, social determinants of health, comorbidities, high- and low-density lipoprotein cholesterol, and pregnancy-related hypertensive disorders. We used Cox elastic net and random survival forests to develop sex-specific models and compared their performance to 2021 US Preventive Services Task Force blood pressure screening recommendations.

RESULTS

Among 355 524 adults from Kaiser Permanente Southern California (mean age, 29 years; mean systolic/diastolic blood pressure, 115/70 mm Hg), 11.7% developed hypertension in a median of 5.4 years. External validation showed good discrimination and calibration (Harrell's C-statistic, 0.76 and 0.82; Integrated Brier Score, 0.04 and 0.02 for men and women, respectively). Compared with US Preventive Services Task Force, a 10-year risk for hypertension of ≥15% in the external cohort using the new model showed similar sensitivity (men, 0.90 versus 0.89; women, 0.81 versus 0.81) and moderate improvement in specificity (men, 0.35 versus 0.25; women, 0.66 versus 0.44). Using the National Health and Nutrition Examination Survey, the prediction model estimated 28.5 million US young adults being at high risk of hypertension compared with 45.3 million by the US Preventive Services Task Force.

CONCLUSIONS

Compared with the US Preventive Services Task Force, the hypertension risk prediction model may be a more efficient tool to identify high-risk young adults for early intervention.