Predicting macrolide resistance in pediatric Mycoplasma pneumoniae pneumonia: A machine learning modeling study.

PURPOSE: To develop a machine learning-based clinical prediction model for macrolide-resistant Mycoplasma pneumoniae pneumonia (MRMPP) in children, facilitating early identification of resistant cases and guiding targeted therapeutic interventions.

METHODS

In this retrospective, single-center study, we developed a stacking ensemble prediction model using demographic, laboratory, and inflammatory data from pediatric patients with MPP. A feature selection protocol was implemented to identify key predictors. The final model was validated using both internal cross-validation and an independent external temporal cohort. Model interpretability was assessed using SHapley Additive exPlanations (SHAP).

RESULTS

The stacking ensemble model achieved an area under the curve (AUC) of 0.857 during internal validation, with a sensitivity of 0.769 and specificity of 0.841; the AUC during external validation was 0.812. Key predictive factors included interleukin-17 A (IL-17 A), interferon-gamma (IFN-γ), C-reactive protein (CRP), albumin-to-globulin ratio (A/G), History of pre-hospital macrolide use, and Pre-hospital course. The model is implemented as a web tool, facilitating rapid assessment of resistance risk.

CONCLUSION

The machine learning model developed in this study can initially identify children at high risk for MRMPP, serving as a data-driven decision-making tool for the rational use of antibiotics in clinical practice and demonstrating significant clinical translational value.