Related Experiment Video
Updated: Jun 20, 2026
Symptom Assessment of Patients with Allergic Rhinitis Using an Allergen Exposure Chamber
Published on: March 3, 2023
Multimodal EHR-Based Prediction of Pediatric Asthma Exacerbations
Zhengkang Fan1, Jinqian Pan1, Mengxian Lyu1
1Department of Health Outcomes and Biomedical Informatics, University of Florida, Gainesville, Florida, USA.
Insights
Predicting pediatric asthma exacerbations is challenging. Machine learning models using electronic health records show promise, identifying symptom notes and medication use as key risk factors for future events.
Area of Science:
- Pediatric Pulmonology
- Clinical Informatics
- Computational Health
Background:
- Pediatric asthma exacerbations lead to frequent emergency department visits and hospitalizations.
- Current risk prediction for pediatric asthma exacerbations is limited, with no established consensus risk scores.
Purpose of the Study:
- To evaluate computable phenotypes (CAPriCORN and COMPAC) for predicting pediatric asthma exacerbations.
- To assess the performance of machine learning models in predicting exacerbations over 6-, 12-, and 24-month periods.
Main Methods:
- Utilized UF Health electronic health records (EHRs) from 2011-2023.
- Trained machine learning models including XGBoost with cross-validation and hyper-parameter optimization.
- Defined exacerbations using diagnosis codes and systemic corticosteroid prescriptions.
Main Results:
- XGBoost demonstrated the best predictive performance.
- SHapley Additive exPlanations (SHAP) identified note-derived symptom terms and rescue medication use as significant predictors.
- The developed framework is interpretable and integrates text data.
Conclusions:
- An interpretable, text-integrated framework using EHR data can predict pediatric asthma exacerbations.
- This approach may enhance child-specific risk stratification.
- Potential to inform EHR-based decision support for proactive asthma management.
Abstract:
Pediatric asthma exacerbations are a frequent cause of emergency department (ED) visits and hospitalizations, yet accurate risk prediction remains limited and no consensus risk scores exist. Using UF Health electronic health records (EHRs) from 2011-2023, we evaluated two computable phenotypes (i.e., CAPriCORN and COMPAC) to predict exacerbations over 6-, 12-, and 24-month horizons. Exacerbations were defined using a validated composite of diagnosis codes from ED, inpatient, or outpatient encounters combined with systemic corticosteroids prescriptions. Several commonly used machine learning (ML) models were trained with stratified five-fold cross-validation, Bayesian hyper-parameter optimization, and Youden's J thresholding. XGBoost achieved the best performance, with SHapley Additive exPlanations (SHAP) highlighting note-derived symptom terms and rescue-medication use as dominant predictors. Future work will focus on external validation and assessment of generalizability. This interpretable, text-integrated framework may support child-specific risk stratification and inform EHR-based decision support for timely pediatric asthma management.
Related Concept Videos
01:30Asthma-IV: Diagnostic and Management
Clinical Assessment for Asthma:
This is the first step in diagnosing and managing asthma. It includes:
01:26Asthma-II: Pathophysiology and Classification
Additionally, environmental and genetic factors play crucial roles in determining an individual's susceptibility to asthma and the severity of their condition.
Critical processes in asthma pathophysiology include:
01:13Asthma III: Clinical Manifestations
01:24Asthma-III: Symptoms and Complications
Classification of Asthma
01:20Asthma: Pathogenesis and Management
Asthma is classified as allergic and non-allergic. Allergens such as dust mites, pollen, and pet dander trigger allergic asthma, while factors like cold air, intense emotions, or exercise can induce non-allergic asthma.
01:28Asthma I: Introduction