Prediction of oral food challenge outcomes via ensemble learning

Oral Food Challenges (OFCs) are essential to accurately diagnosing food allergy due the limitations of existing clinical testing. However, some patients hesitant undergo OFCs, while those willing suffer from limited access allergists in rural/community healthcare settings. Despite its success predicting patient outcomes other settings, few applications machine learning have been developed. Thus, this study, we seek leverage methodologies for OFC outcome prediction. Retrospective data was gathered 1,112 who collectively underwent a total 1,284 and consisted factors including serum-specific Immunoglobulin E (IgE), IgE, skin prick tests (SPTs), comorbidities, sex, age. Using these features, multiple models were constructed predict three common allergens: peanut, egg, milk. The best performing model each allergen an ensemble random forest (egg) or Learning Concave Convex Kernels (LUCCK) (peanut, milk) models, which achieved Area under Curve (AUC) 0.91, 0.96, 0.94, milk, respectively. Moreover, all such had sensitivity specificity values ≥89%. Model interpretation via SHapley Additive exPlanations (SHAP) indicates that specific along with wheal flare SPTs, highly predictive outcomes. results analysis suggest has potential reveal relevant further study.