Derivation and External Validation of a Risk Prediction Model for Pulmonary Embolism in Patients With Lung Cancer: A Large Retrospective Cohort Study

ObjectiveTo investigate the risk factors of pulmonary embolism in patients with lung cancer and develop and validate a novel nomogram scoring system-based prediction model. MethodWe retrospectively analyzed the clinical data and laboratory characteristics of 900 patients with lung cancer who were treated, including patients with lung cancer without pulmonary embolism (LC) and patients with lung cancer with pulmonary embolism (LC + PE). The patients were randomly divided into derivation and internal validation groups in a 7:3 ratio. Using logistic regression analysis, a diagnostic model of the nomogram scoring system was developed by incorporating selected variables in the derivation group and validated in the internal and external validation groups (n = 108). ResultSeven variables (adenocarcinoma, stage III-IV LC, indwelling central venous catheter, chemotherapy, and the levels of serum albumin, hemoglobin, and D-dimer) were identified as valuable parameters for developing the novel nomogram diagnostic model for differentiating patients with LC and LC + PE. The scoring system demonstrated good diagnostic performance in the derivation (area under the curve [AUC]; 95% confidence interval [CI], 0.918; 0.893, 0.943; sensitivity, 88.5%; specificity, 80.5%), internal validation (AUC; 95% CI, 0.921; 0.884, 0.958; sensitivity, 90.5%; specificity, 80.4%), and external validation (AUC; 95% CI, 0.929; 0.875, 0.983; sensitivity; 85.0%; specificity; 87.5%) groups. ConclusionIn this study, we constructed and validated a nomogram scoring system based on 7 clinical parameters. The scoring system exhibits good accuracy and discrimination between patients with LC and LC + PE and can effectively predict the risk of PE in patients with LC.