Biostatistical Associations and Machine-Learning Classification of Measured Blood Pressure Threshold Status Among U.S. Adults: NHANES August 2021–August 2023

This cross-sectional analysis utilized publicly available NHANES data collected from August 2021-2023. The study included 6,123 adults for whom at least one valid systolic and one valid diastolic blood pressure reading was available.We examined the association of demographic and lifestyle characteristics with blood pressure status, and evaluated three machine-learning classifiers. Participants were classified according to their mean measurements for the objective of this analysis: systolic blood pressure ≥130 mmHg or diastolic blood pressure ≥80 mmHg. Survey-adjusted logistic regression was based on 5,985 complete cases. We evaluated logistic regression, random forest, and support vector machine models using a stratified training–test split and five-fold cross-validation on the training data. Age, male sex, body mass index, and lifetime cigarette-smoking history were associated with the outcome. Test AUCs ranged from 0.654 to 0.669. Lowering the logistic-regression threshold from 0.50 to 0.31 increased recall from 0.343 to 0.856, while specificity fell from 0.783 to 0.432. The level of discrimination was moderate, and an external validation process is required before any clinical application.