DOI: 10.1097/js9.0000000000005550 ISSN: 1743-9159

External validation and incremental value of adding the neutrophil-to-lymphocyte ratio to a pneumonia prediction nomogram in burn patients with inhalation injury: a single-center retrospective cohort study

Xiulan Huang, Shuhan Yue, Qinghua Wu, Ziyu Zhou, Caoyi Liu, Qiao Lei, Jingwei Zhang
Show PDF Cite

Background:

Burn patients with inhalation injury face a significantly elevated risk of pneumonia. Although Li et al developed the first nomogram for pneumonia prediction in this population, external validation remains limited, and the model lacks systemic inflammation markers or comprehensive burn severity scores.

Methods:

This single-center retrospective cohort study included 218 patients with full-thickness burn and inhalation injury (January 2020 to December 2025). External validation of Li’s model was performed. The incremental value of adding the neutrophil-to-lymphocyte ratio (NLR) was evaluated using area under the curve (ΔAUC), net reclassification improvement (NRI), and integrated discrimination improvement (IDI). Model performance was assessed by discrimination, calibration, and decision curve analysis. A sensitivity analysis was conducted in the cohort with complete bronchoscopy findings.

Results:

The original model (tracheal intubation/tracheostomy, albumin, and glucose) achieved an AUC of 0.889 [95% confidence interval (CI): 0.842–0.936] with good calibration (Hosmer–Lemeshow P = 0.891). Adding NLR increased the AUC to 0.913 (95% CI: 0.869–0.956, P = 0.106), with significant improvements in NRI (0.589, 95% CI: 0.302–0.876, P < 0.001) and IDI (0.079, 95% CI: 0.032–0.125, P < 0.001). In contrast, adding the Abbreviated Burn Severity Index did not significantly improve overall discrimination (ΔAUC = 0.001, P = 0.794; IDI = 0.007, P = 0.272), although the NRI showed modest improvement (0.321, P = 0.031). The modified nomogram incorporating tracheal intubation/tracheostomy, albumin, glucose, and NLR was constructed and showed improved calibration. Decision curve analysis confirmed higher clinical net benefit for the NLR-incorporated model. Sensitivity analysis showed that the predictive value of NLR remained robust after adjusting for inhalation injury severity.

Conclusions:

Li’s nomogram demonstrated acceptable performance in this independent cohort. Adding NLR significantly improved risk reclassification and overall discrimination, providing a simple and practical tool for early pneumonia risk stratification in burn patients with inhalation injury.

More from our Archive