Threshold-aware diet-only hypertension triage using NHANES dietary recall

Ghanbari, Shahin; Homayounvala, Elaheh; Al-Sudani, Sahar

London Met Repository

Tools

Lists

Ghanbari, Shahin, Homayounvala, Elaheh and Al-Sudani, Sahar (2026) Threshold-aware diet-only hypertension triage using NHANES dietary recall. In: 14th International Conference on Frontiers of Intelligent Computing: Theory and Applications (FICTA-2026), 8-9 June 2026, London. (In Press)

Abstract
Documents
Details
Record

[+][-]

Abstract

Hypertension is a major public-health risk factor and a practical target for early risk stratification. In routine care, assessment depends on blood-pressure measurement, clinical history, medication information, anthropometric variables, and laboratory or examination data. These inputs may not always be available in low-burden digital or community screening workflows. This study examines whether 24-hour dietary recall alone can provide a useful prioritisation signal for hypertension follow-up. Adult participants from the National Health and Nutrition Examination Survey (NHANES) were analysed using diet-derived nutrient variables only, with clinical, laboratory, medication, and anthropometric predictors excluded from the input set. A calibrated machine-learning model selected on validation data was evaluated on a held-out test set using precision–recall discrimination and a pre-selected operating threshold. At this threshold, the model achieved a hypertension AUPRC of 0.248, positive predictive value of 0.327, recall of 0.249, specificity of 0.933, and approximately 87.8 alerts per 1,000 screened. These results suggest that dietary recall can enrich the group selected for follow-up, but the low recall means that many hypertension-positive participants would still be missed. The approach is therefore best interpreted as triage before confirmatory assessment, not as a diagnostic substitute.

Documents