Development, internal validation, and independent external evaluation of a nomogram for pulmonary embolism risk in hospitalized patients

To develop and internally validate a nomogram for pulmonary embolism (PE) risk in hospitalized patients and to evaluate its performance in an independent hospital cohort. We retrospectively analyzed 4,380 eligible inpatients from Dongyang Hospital Affiliated to Wenzhou Medical University. Least absolute shrinkage and selection operator regression was followed by multivariable logistic regression with backward Akaike information criterion selection, and the cohort was randomly divided 70:30 for development and internal validation. Independent evaluation used 1,376 CTPA-evaluable first admissions from the First Affiliated Hospital of Wenzhou Medical University. The locked original intercept and unrounded coefficients were applied without refitting, and 95% confidence intervals were obtained using 2,000 stratified bootstrap resamples. The development cohort included 233 patients with PE and 2,833 without PE, and the internal validation cohort included 104 patients with PE and 1,210 without PE. Thirteen predictors were retained in the nomogram. The AUC was 0.767 (95% CI, 0.735–0.799) in development and 0.808 (95% CI, 0.766–0.850) in internal validation. The independent cohort comprised 519 patients with PE and 857 without PE; the locked model yielded an AUC of 0.720 (95% bootstrap CI, 0.691–0.746). The Brier score was 0.222 (95% bootstrap CI, 0.211–0.232), with a calibration intercept of 0.339 and slope of 0.606. The nomogram retained moderate discrimination in an independent hospital cohort. Application of the locked original equation quantified systematic risk underestimation and identified higher net benefit than the treat-all and treat-none strategies across threshold probabilities of approximately 25%-65%. These findings support transportability while indicating the need for local recalibration and prospective comparison with established diagnostic pathways.