Development and validation of an immune-based nomogram model for predicting severe adenovirus pneumonia in hospitalized children

TL;DR

This study developed a nomogram model using immune markers to predict severe adenovirus pneumonia in hospitalized children, improving early risk assessment.

Contribution

The study introduces a novel immune-based nomogram model for predicting severe adenovirus pneumonia in children.

Findings

Seven immune-related factors were identified as significant predictors of severe adenovirus pneumonia.

The nomogram model showed strong discrimination with an AUC of 0.836 in the training set and 0.913 in the cross-validation set.

The model demonstrated good calibration and potential for clinical use in risk stratification.

Abstract

Human adenovirus (HAdV) is a significant cause of severe pneumonia in children that often causes sequelae. Although immune disorders are known to be associated with disease progression, comprehensive immunological predictors have not been identified. The purpose of this study was to explore the ability of multiple immunological indicators to predict severe adenovirus pneumonia (SAP) and to develop an immune-based nomogram for the early prediction of SAP in children.

This study involved a retrospective analysis of children with adenovirus pneumonia who were hospitalized and received treatment at the Department of Respiratory Medicine, Capital Center for Children's Health, Capital Medical University between January 2017 and June 2025. Patients were stratified into mild and severe groups on the basis of clinical manifestations. They were subsequently randomly allocated at an 80:20 ratio into a training set and a cross-validation set for nomogram development and validation. R software (version 4.4.3) was used for statistical analysis, and effect sizes are expressed as odds ratios (ORs) and 95% confidence intervals (CIs).

Among the 1220 cases included, 357 (29.3%) were classified as severe. In both the training and cross-validation sets, patients with SAP were younger and had longer hospital stays (all P < 0.001). After the adjustments for age and sex, logistic regression in the training set revealed seven significant factors associated with SAP occurrence in children: Mycoplasma pneumoniae infection (OR = 1.372, 95% CI: 1.04–1.809, P < 0.001); complement component 3 (C3) (OR = 0.234, 95% CI: 0.132–0.417, P < 0.001) and 4 (C4) (OR = 0.075, 95% CI: 0.018–0.31, P < 0.001); immunoglobulin G (IgG) (OR = 1.049, 95% CI: 1.028–1.071, P < 0.001); and the percentages of CD3+ [CD3+ (%)] (OR = 0.965, 95% CI: 0.952–0.979, P < 0.001), CD4+ [CD4+ (%)] (OR = 0.950, 95% CI: 0.934–0.967, P < 0.001), and CD19+ cells [CD19+ (%)] (OR = 1.042, 95% CI: 1.028–1.057, P < 0.001). Furthermore, logistic regression of the validation set revealed C3, C4, IgG, CD3+(%), CD4+ (%) and CD19+(%) as consistent predictors of SAP across datasets. Incorporating the significant factors improved model discrimination, increasing the area under the curve (AUC) from 0.627 to 0.836 in the training set and from 0.678 to 0.913 in the cross-validation set. The final nomogram model based on the significant factors demonstrated strong calibration and discrimination (C-index: 0.731 in the training set, 0.812 in the cross-validation set), supporting its potential for clinical risk stratification.

This study identified and validated seven independent factors significantly associated with SAP in children. A nomogram incorporating these factors was developed and it demonstrated favourable discriminative performance and good calibration. The model exhibited high sensitivity and maintained high predictive accuracy across both datasets, indicating its potential clinical utility for individualized risk stratification.

Figures