Predicting pediatric severe asthma exacerbations: an administrative claims-based predictive model.

Area of Science:

• Pediatric pulmonology
• Health informatics
• Machine learning in healthcare

Background:

• Childhood asthma exacerbations pose a significant health burden.
• Existing predictive models often overlook crucial social determinants of health (SDOH) and demographic factors.
• There is a need for models that incorporate race, ethnicity, and SDOH to predict asthma outcomes.

Purpose of the Study:

• To develop and evaluate a predictive model for childhood asthma exacerbations.
• To explore the predictive value of race, ethnicity, rural-urban commuting area (RUCA) codes, and Child Opportunity Index (COI) for asthma outcomes.
• To assess the importance of various SDOH in predicting asthma-related hospitalizations and emergency department (ED) visits.

Main Methods:

• Utilized insurance claims data from the Arkansas All-Payer Claims Database.
• Identified a cohort of 22,631 children aged 5-18 with asthma and continuous Medicaid enrollment.
• Employed conditional random forest models to predict asthma-related hospitalizations and ED visits.

Main Results:

• The predictive model achieved an area under the curve (AUC) of approximately 72-73%.
• Previous asthma-related hospitalizations or ED visits were the strongest predictors of future events.
• Medication use (reliever and controller) also contributed to predictive accuracy.

Conclusions:

• The developed model demonstrated moderate accuracy in predicting childhood asthma exacerbations.
• Race, ethnicity, RUCA codes, COI, and ICD-10 SDOH measures did not significantly enhance the model's predictive power.
• Future research may need to explore alternative or refined SDOH metrics for improved prediction.