Leveraging machine learning models to forecast pediatric allergic rhinitis exacerbation risk based on environmental...

Area of Science:

• Environmental Health
• Pediatric Allergy
• Machine Learning in Medicine

Background:

• Acute exacerbations of pediatric allergic rhinitis (AR) lack reliable prediction tools.
• Individualized risk assessment integrating clinical and environmental data is needed.

Purpose of the Study:

• To develop an interpretable machine learning model for predicting 90-day AR exacerbation risk in children.
• To conduct a preliminary assessment of the model's transportability to an independent cohort.

Main Methods:

• A development cohort (n=2,000) linked NHANES, EPA Air Quality, and NOAA data.
• Clinical predictors (TNSS, IgE, eosinophils) and environmental exposures (PM2.5, PM10, O3, NO2, temp, humidity) were analyzed.
• L2-regularized logistic regression (LR) was the primary model, benchmarked against other ML algorithms using cross-validation and multiple imputation.

Main Results:

• The LR model achieved an internal AUC of 0.81 and external AUC of 0.71.
• Key predictors included Total Nasal Symptom Score (TNSS), same-day PM2.5, and total IgE.
• Children exposed to high PM2.5 (>75 μg/m3) had ~5x higher exacerbation odds (OR=5.08).

Conclusions:

• An interpretable LR model combining symptom burden and PM2.5 exposure shows potential for predicting pediatric AR exacerbations.
• Preliminary findings suggest model transportability, but a small external cohort limits definitive validation.
• Larger prospective studies are needed for clinical implementation.