Development and validation of a risk prediction model for severe community-acquired pneumonia associated pulmonary embolism

Pulmonary embolism (PE) may be difficult to detect in patients with severe community-acquired pneumonia (SCAP) because symptoms and physiological abnormalities overlap with those of severe infection and respiratory failure. Early identification of patients at high risk of PE may facilitate timely diagnostic evaluation. This study aimed to identify factors associated with PE among patients with SCAP admitted to the intensive care unit (ICU) and to establish and validate a nomogram for identifying those at high risk of PE. Clinical data from 3148 ICU patients with SCAP were retrospectively analyzed using the Medical Information Mart for Intensive Care IV (MIMIC-IV, version 2.2; 2008–2019) database. A random 7:3 split was used to form the training cohort ( n = 2204) and validation cohort ( n = 944). Independent predictors were determined by least absolute shrinkage and selection operator followed by multivariable logistic regression, and these predictors were used to build a nomogram. The model was assessed in terms of discrimination, calibration, and clinical net benefit using the area under the receiver operating characteristic curve (AUROC), calibration curves and Brier scores, and decision curve analysis (DCA), respectively. Among the 3148 ICU patients with SCAP, PE was identified in 152 patients (4.83%). Age, sex, heart rate, creatinine, prothrombin time, activated partial thromboplastin time, and partial pressure of oxygen were independently associated with PE in patients with SCAP. The model achieved AUROC values of 0.683 (95% CI, 0.635–0.731) in the training cohort and 0.658 (95% CI, 0.567–0.749) in the validation cohort. The Brier scores were 0.0464 and 0.0416 in the training and validation cohorts, respectively, and the calibration curves showed good consistency between predicted and observed risks. Bootstrap internal validation with 1000 resamples yielded an optimism-corrected AUROC of 0.661. DCA indicated a clinical net benefit for threshold probabilities of approximately 0.03–0.15. This study constructed the first model for estimating PE risk among ICU patients with SCAP and validated its performance. The model demonstrated modest discrimination and good calibration and may facilitate early PE risk assessment and stratification in this population. However, its predictive performance and clinical applicability require further validation in independent external cohorts. Not applicable. This study is a retrospective observational analysis based on the MIMIC-IV database.