A simplified composite risk score integrating inflammation, infection, nutrition, and oxygenation to predict adverse outcomes in bacterial pneumonia: a retrospective study

To develop and internally validate a prognostic risk score integrating age, infection severity, systemic bacterial response, and physiological reserve for adverse outcomes in patients with bacterial pneumonia. This single-center retrospective observational study included 270 consecutive patients hospitalized with bacterial pneumonia at Wuxi No.2 People’s Hospital between January 2021 and December 2023, comprising 214 survivors and 56 non-survivors. Baseline clinical characteristics, laboratory parameters, and oxygenation indices at admission were retrospectively collected from medical records. The Clinical Pulmonary Infection Score (CPIS) was calculated from documented clinical, laboratory, radiological, and microbiological information. Twenty-one baseline candidate predictors were evaluated using LASSO Cox regression. To assess the robustness of variable selection, the LASSO Cox analysis was repeated in 1,000 bootstrap resamples, and the selection frequency of each candidate predictor was calculated. Based on the LASSO regularization path, bootstrap selection frequencies, model parsimony, and clinical interpretability, age, CPIS, PCT, and albumin were retained and subsequently entered into a multivariable Cox proportional hazards model. Time-dependent receiver operating characteristic (ROC) curve analysis was performed to evaluate discriminative performance at 15 and 20 days. The final Cox model was separately subjected to bootstrap internal validation using 1,000 bootstrap resamples to estimate optimism-corrected model performance. Calibration and decision curve analyses were also performed. Patients in the non-survival group were significantly older than those in the survival group (69.61 ± 8.19 vs. 62.33 ± 4.65 years, P < 0.001), had a higher prevalence of chronic obstructive pulmonary disease (44.64% vs. 27.10%, P = 0.011), a higher proportion of intensive care unit admission (39.29% vs. 21.50%, P = 0.006), and higher CPIS values ( P < 0.001). At admission, NLR and PCT levels were significantly elevated in the non-survival group, whereas albumin and PaO₂/FiO₂ were significantly reduced. In the 1,000 bootstrap LASSO analyses, age, CPIS, PCT and albumin showed selection frequencies of 92.0%, 86.8%, 86.5%, and 66.8%, respectively, whereas the highest selection frequency among the remaining candidate predictors was 28.4% for CRP. Based on the LASSO regularization path and bootstrap selection stability, these four variables were retained for multivariable Cox regression. Age, CPIS, PCT, and albumin were each independently associated with in-hospital mortality and were incorporated into the prognostic risk score. The risk score yielded time-dependent AUCs of 0.728 (95% CI: 0.624–0.832) at 15 days and 0.773 (95% CI: 0.627–0.919) at 20 days, compared with 0.706 (95% CI: 0.613–0.800) and 0.674 (95% CI: 0.511–0.837), respectively, for CURB-65. The final model had an apparent C-index of 0.779, with an optimism-corrected C-index of 0.762 after bootstrap validation. Decision curve analysis suggested a positive net benefit across selected threshold probabilities at the 15-day prediction horizon. A four-predictor risk score comprising age, CPIS, PCT, and albumin showed moderate discrimination for in-hospital mortality in patients hospitalized with bacterial pneumonia and may serve as an adjunctive tool for early risk stratification. Given the retrospective single-center design and the absence of independent external validation, these findings should be considered preliminary. Further prospective multicenter studies and external validation are required before routine clinical application. Not applicable.