Development and Validation of a Diagnostic Model - The Hypertension Population Risk Tool (HTNPoRT) - to Predict Hypertension and Describe Risk Profiles: A Population-Based Cross-Sectional Study of Canadians

Abstract

Background: Hypertension is preventable, and existing models predicting hypertension lack use for individual and population health planning.

Methods: The Hypertension Population Risk Tool (HTNPoRT) was derived from 19,643 adult respondents in the Canadian Health Measures Survey. Sex-specific logistic regression models predicting hypertension presence were developed using 16 predictors: 4 sociodemographics, 3 psychosocial measures, 2 health status indicators, 5 health behaviours, and 2 chronic conditions.

Results: The final HTNPoRT models were discriminating (c-statistic, men: 0.86; women: 0.88), and well-calibrated in the overall population (observed v. predicted relative difference, men: 1.02%; women: 1.41%) and nearly all equity-relevant subgroups (179 out of 181). Age, diabetes, and body mass index were most influential predictors of hypertension seen on SHAP-derived risk profiles, while predictability of adiposity measures differed across sex.

Conclusions: The public and health policymakers can use the models and risk profiles of HTNPoRT to support planning and decision-making on addressing the hypertension burden.