Development and validation of a nomogram integrating inflammatory-nutritional-metabolic composite indices for predicting short-term adverse outcomes in hospitalized patients with pneumonia

Objective To develop and validate a prediction model integrating multidimensional inflammatory-nutritional-metabolic composite indices for early risk stratification of short-term adverse outcomes in hospitalized pneumonia patients. Methods A total of 839 hospitalized pneumonia patients admitted between January 2020 and December 2023 were retrospectively enrolled and randomly divided into a training cohort ( n = 587) and a validation cohort ( n = 252) at a 7:3 ratio. The primary endpoint was in-hospital intensive care unit (ICU) admission and/or invasive mechanical ventilation (IMV). Independent predictors were selected using least absolute shrinkage and selection operator (LASSO) regression combined with multivariable logistic regression, and a nomogram was constructed accordingly. Random Forest and XGBoost machine learning models were additionally built to corroborate variable importance. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC), calibration curves, decision curve analysis (DCA), and net reclassification improvement (NRI)/integrated discrimination improvement (IDI). A simplified integer-based THM score was subsequently derived from the three highest-weighted inflammatory-nutritional-metabolic indices. Results The overall incidence of short-term adverse outcomes was 17.6% (148/839). Following LASSO selection and multivariable analysis, nine independent predictors were identified: monocyte-to-high-density lipoprotein cholesterol ratio (MHR), triglyceride-glucose (TyG) index, neutrophil-to-lymphocyte ratio (NLR), hemoglobin-albumin-lymphocyte-platelet (HALP) index, peripheral oxygen saturation (SpO₂), partial pressure of arterial oxygen to fraction of inspired oxygen ratio (PaO₂/FiO₂), CURB-65 score, procalcitonin (PCT), and age. The nomogram achieved AUCs of 0.871 and 0.848 in the training and validation cohorts, respectively, significantly outperforming the Pneumonia Severity Index (PSI; AUC = 0.752) and CURB-65 (AUC = 0.728) (both p < 0.05). The Hosmer-Lemeshow (H-L) test confirmed satisfactory calibration (validation cohort p = 0.621), and DCA demonstrated superior net clinical benefit over traditional scores across a wide range of threshold probabilities. NRI and IDI analyses confirmed significant incremental predictive value over both reference tools (all p < 0.001). The THM simplified score (AUC = 0.804) stratified patients into low-risk (4.7%), intermediate-risk (16.6%), and high-risk (25.2%) groups, with a significant gradient trend ( p < 0.001). Conclusion The nomogram and THM simplified score integrating inflammatory-nutritional-metabolic composite indices significantly outperform conventional PSI and CURB-65 in predicting short-term adverse outcomes among hospitalized pneumonia patients, providing practical quantitative tools for early precise triage and individualized clinical decision-making.