Exploring Machine Learning Applications in Pediatric Asthma Management: Scoping Review.

Area of Science:

• Pediatric Pulmonology
• Computational Health
• Biomedical Informatics

Background:

• Machine learning (ML) offers a novel approach to predicting asthma-related outcomes in children.
• This integration is crucial for advancing pediatric healthcare and asthma management strategies.

Purpose of the Study:

• To conduct a scoping review of studies published since 2019 on ML algorithms for pediatric asthma.
• To analyze the applications and predictive performance of these ML models.
• To identify trends, common algorithms, and key outcomes in ML-driven pediatric asthma research.

Main Methods:

• A comprehensive literature search was conducted across major databases (Ovid MEDLINE, Embase, Cochrane Library, CINAHL, Web of Science) from January 2019 to July 2023.
• Studies involving ML models for predicting asthma outcomes in children (<18 years) were included.
• Risk of bias was assessed using the Prediction Model Risk of Bias Assessment Tool.

Main Results:

• Fifteen studies met the inclusion criteria, utilizing ML techniques like logistic regression (47%) and random forests (40%).
• Key applications included predicting exacerbations (XGBoost, AUROC 0.76), classifying phenotypes (SVM, AUROC 0.79), diagnosis (ANN, AUROC 0.63), and identifying risk factors (Random Forests, AUROC 0.88).
• Most studies had low to moderate risk of bias, but limitations included data quality, sample size, and interpretability.

Conclusions:

• ML demonstrates diverse applications in predicting pediatric asthma outcomes, with varying model strengths.
• XGBoost, SVM, ANN, and Random Forests show significant predictive capabilities for different asthma-related outcomes.
• Future research must prioritize data quality, larger sample sizes, and enhanced model interpretability for clinical translation.