Answers are grounded in this paper and cite their sources.
Obesity develops through interacting familial, dietary, and lifestyle factors, highlighting the need for BMI-free screening tools to identify susceptibility before excess weight develops. This study developed an Obesity Risk Score (ORS) using 13 questionnaire predictors and compared three weighting strategies: consensus weights from five nutritionists (Expert ORS), Gradient Boosting-derived weights (GB ORS), and weights from three machine learning models (ML-3 ORS). Data-driven weights were derived from the current weight status of 2,084 records using nested cross-validation. All scores increased progressively across healthy weight, overweight, and obesity. The Expert ORS showed the strongest association with weight status (Spearman ρ = 0.458; AUC = 0.740 for obesity and 0.759 for overweight/obesity), classified 98.5% of individuals with obesity as High risk, and flagged 47.8% of non-overweight individuals as High risk. Although the scores ranked individuals similarly (ρ = 0.874–0.973), categorical agreement was limited (κ = 0.094–0.422). Twenty machine learning models reproduced the scores with high accuracy (XGBoost R² ≥ 0.997; best macro-F1 = 0.924–0.966). Family history was the most influential predictor. Expert-informed BMI-free scoring offers a transparent approach to assessing obesity susceptibility, requiring prospective and external validation before predictive use.
No takes yet. Share an insight, caveat, or question.
Dakdouk et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: