[Construction and validation of a risk prediction model for secondary sepsis in pneumonia]

This bidirectional cohort study included 452 patients with pneumonia (modeling cohort) from the Department of Respiratory and Critical Care Medicine at Jinan People's Hospital retrospectively enrolled between January 2021 and December 2023, and 300 patients with newly diagnosed pneumonia (validation cohort) prospectively recruited from January to December 2024. The least absolute shrinkage and selection operator regression (10-fold cross-validation, λ=0.023) was initially applied for dimensionality reduction. Subsequently, Firth-penalized logistic regression was used to construct the predictive model, followed by the development of a P-Sep scoring system. The system was validated using the bootstrap method (500 resamples) and an independent prospective cohort. Multivariate analysis identified four independent predictors of interest: invasive mechanical ventilation (OR=5.12, 95%CI 3.05-8.61); positive blood culture (OR=4.23, 95%CI 2.38-7.51); lactate ≥ 2 mmol/L (OR=3.15, 95%CI 1.92-5.18); and serum amyloid A ≥100 mg/L (OR=2.58, 95%CI 1.52-4.39). The investigators established that the P-Sep score (0-12 points) was an independent predictor denoting low risk (0-5 points), intermediate risk (6-8 points), and high risk (≥9 points). In the modeling cohort, the area under the curve was 0.87 (95%CI 0.83-0.91; sensitivity: 85.2%; specificity: 89.3%). In the validation cohort, the area under the curve was 0.86 (95%CI 0.82-0.89; sensitivity: 84.8%; specificity: 89.1%). In addition, based on data from both cohorts, the predictive performance of the P-Sep score was compared with that of the Quick Sequential Organ Failure Assessment and CURB-65. The results demonstrated that the P-Sep score exhibited favorable predictive efficacy in the warning of secondary sepsis in patients with pneumonia.