Machine learning-based identification of inflammatory biomarkers for predicting pulmonary consolidation in children with Chlamydia pneumoniae infection.

Objective: This study aimed to identify core inflammatory biomarkers through machine learning approaches and develop an accessible online risk calculator to predict pulmonary consolidation in children with Chlamydia pneumoniae infection, addressing the current lack of effective early warning tools. Methods: This retrospective case-control study enrolled 42 children with C. pneumoniae infection (consolidation group: 26 cases; non-consolidation group: 16 cases) between January 2020 and December 2024. Five machine learning algorithms, including least absolute shrinkage and selection operator (LASSO) regression, support vector machine-recursive feature elimination (SVM-RFE), Random Forest, XGBoost, and LightGBM, were employed for feature selection, and core predictive factors were identified through consensus validation across these algorithms. K-means clustering analysis was performed on the key inflammatory markers, and an online risk assessment system based on HTML5 technology was developed. Results: The five machine learning algorithms consistently identified lactate dehydrogenase (LDH), C-reactive protein (CRP), and erythrocyte sedimentation rate (ESR) as core inflammatory markers for predicting pulmonary consolidation. All three indicators were significantly higher in the consolidation group compared with the non-consolidation group ( Conclusion: LDH, CRP, and ESR are key indicators for predicting pulmonary consolidation in children with C. pneumoniae infection. The online risk assessment system developed based on these three routine laboratory parameters demonstrates good clinical usability and practicality, enabling early identification of high-risk patients to guide individualized treatment decisions.