Development and validation of a prediction model for in-hospital mortality in patients with severe pneumonia: A retrospective cohort study

Abstract

Background: Severe pneumonia is a major cause of mortality worldwide, and accurate early risk assessment is clinically important. This study aimed to develop and validate a mortality prediction model based on routinely available clinical variables. Methods: In this retrospective cohort study, we enrolled 300 patients with severe pneumonia: 200 in the training cohort and 100 in the validation cohort. Univariate analysis and least absolute shrinkage and selection operator (LASSO) regression were used to identify nine key predictors for model development. Model performance was evaluated using receiver operating characteristic (ROC) curves, calibration curves, and decision curve analysis (DCA) in both cohorts. Multivariable logistic regression was further performed, and a nomogram was constructed for mortality prediction. Results: Baseline characteristics were comparable between the training and validation cohorts. The prediction model demonstrated good discrimination in the training cohort, with an area under the ROC curve (AUC) of 0.855 (95% CI: 0.801–0.908), and acceptable discrimination in the validation cohort, with an AUC of 0.750 (95% CI: 0.640– 0.861). Calibration and decision curve analyses indicated favourable clinical utility. Multivariable logistic regression identified age, coronary heart disease, pulmonary hypertension, and mechanical ventilation as independent predictors of in-hospital mortality in severe pneumonia. Conclusion: The proposed nomogram may serve as a preliminary tool for estimating in-hospital mortality risk in patients with severe pneumonia. External validation in independent populations is required before clinical application.