Development and validation of a nomogram for differentiating bacterial from non-bacterial necrotizing pneumonia in children: a retrospective cohort study

Vollständiger Abstract

Worum geht es in dieser Arbeit?

Background Necrotizing pneumonia (NP) in children requires etiology-dependent treatment, yet early etiological differentiation remains challenging due to the low sensitivity and slow turnaround of conventional microbiological methods. This study aimed to develop and validate a nomogram for predicting non-bacterial NP using routinely available clinical parameters. Methods This single-center retrospective study enrolled children (<18 years) with contrast-enhanced computed tomography (CT)–confirmed NP at Tianjin Children’s Hospital (January 2020–January 2025). Patients were classified as bacterial NP (BNP) or non-bacterial NP (NBNP) based on pathogen identification via culture, polymerase chain reaction (PCR), or targeted next-generation sequencing, and randomly split into training and validation sets (7:3). Predictors were selected by univariable screening followed by stepwise multivariable logistic regression, and a nomogram was constructed. Performance was assessed using receiver operating characteristic (ROC) analysis, calibration plots, and decision curve analysis. Results Of 149 children (76 BNP, 73 NBNP; training n = 104, validation n = 45), three independent predictors of NBNP were identified: fever duration (odds ratio [OR] = 4.580, p = 0.004), CRP (OR = 0.272, p = 0.009), and white blood cell count (WBC) (OR = 0.257, p = 0.025). The nomogram achieved an AUC of 0.816 (95% CI: 0.735–0.896) in the training cohort. Bootstrap internal validation (1,000 resamples) of the validation cohort yielded a bias-corrected AUC of 0.846 (95% CI, 0.732–0.959), confirming stable discriminative performance. Calibration was satisfactory (Hosmer–Lemeshow p = 0.857 and 0.738), and decision curve analysis demonstrated positive net benefit. Conclusion A parsimonious nomogram based on fever duration, WBC count, and CRP level offers a practical bedside tool for early etiological differentiation of NP in children, potentially guiding timely pathogen-directed therapy before microbiological confirmation. External multicenter prospective validation is warranted.