A simplified nomogram for predicting deep vein thrombosis: development in NHANES and retrospective single-center external validation

Abstract

Abstract Background Deep vein thrombosis (DVT) is a major cardiovascular burden, and existing risk-stratification tools often rely on laboratory biomarkers, limiting their utility in primary care. We developed and externally validated a simplified, laboratory-free nomogram to predict DVT in community settings. Methods The model was developed using 39,692 participants from NHANES(2005–2018; self-reported physician-diagnosed DVT) and externally validated in 3,136 hospitalized Chinese patients(2016–2023; ultrasound-confirmed DVT). Predictors were selected by LASSO and fitted using multivariable logistic regression incorporating the NHANES complex survey design. Performance was assessed by AUC, calibration (MAE), and decision curve analysis (DCA). Results Four predictors were retained: age, smoking, physical activity, and diabetes. In development, the nomogram showed good discrimination (AUC, 0.816), comparable to XGBoost (0.820). Diabetes (OR, 2.84; 95% CI, 2.36–3.43) and ever-smoking (OR, 2.25; 95% CI, 1.92–2.65) were the strongest predictors. On external validation, discrimination was maintained (AUC, 0.792; 95% CI, 0.768–0.816) with good calibration (MAE, 0.022) and clinical net benefit across thresholds probabilities. Conclusions This four-variable nomogram may help identify adults at higher risk of DVT in primary-care settings. It is intended as an exploratory screening aid, not a standalone diagnostic test or for suspected acute DVT. Prospective validation is required before clinical implementation.