Transferrin Receptor and Sociodemographic Factors: A Multimodal Model for Heavy Drinking Risk Stratification in Women

Objective This study aimed to develop and validate a clinically actionable nomogram-based prediction model integrating sociodemographic factors and biological biomarkers to identify heavy alcohol consumption among adult women in the United States. Methods Data were extracted from the 2021 to 2023 National Health and Nutrition Examination Survey (NHANES), including 994 eligible adult women (18 + years, past 12-month alcohol use, and complete data on key variables). The Boruta algorithm (random forest-based) was used for feature selection to identify critical predictors of heavy drinking. A logistic regression model was constructed, and performance was evaluated via discrimination(Receiver operating characteristic, ROC), calibration (calibration curve, Hosmer-Lemeshow test), and clinical utility (decision curve analysis, DCA). Internal validation was performed using 1000 bootstrap resamples to ensure robustness. Results Eight key predictors were selected: transferrin receptor, ferritin, high-sensitivity C-reactive protein (HS-CRP), alpha-1-acid glycoprotein, income-to-poverty ratio, education level, country of birth, and age. Multivariate logistic regression revealed significant associations with heavy drinking: transferrin receptor (OR = 1.19, 95% CI = 1.10–1.29, p < 0.001), ferritin (OR = 1.00, 95% CI = 1.00–1.01, p = 0.003), alpha-1-acid glycoprotein (OR = 2.98, 95% CI = 1.36–6.51, p = 0.006), lower education levels (e.g., high school graduates: OR = 3.51, p < 0.001 vs. college graduates), non-U.S. birth (OR = 2.44, p < 0.001), and younger age (OR = 0.96, p < 0.001). The model showed good discrimination with a corrected AUC of 0.710, excellent calibration (alignment between predicted and observed probabilities), and significant net clinical benefit via DCA. Conclusion The nomogram-based model effectively identifies adult women at risk of heavy alcohol consumption by combining biological markers of iron dysregulation with key sociodemographic risk factors. This model provides a powerful tool for enhancing early detection and facilitating personalized interventions.