TOPLINE:

A machine learning (ML) model (XGBoost) showed a modest but statistically significant improvement over a conventional regression model in predicting 180‑day readmissions among children hospitalized for asthma.

METHODOLOGY:

• Researchers developed and validated an ML model (XGBoost) to predict readmissions among children hospitalized for asthma and compared it with a conventional regression model.
• The study included 173,730 encounters of children aged 4-18 years hospitalized for asthma exacerbations between January 2016 and December 2024, using an administrative database from 47 tertiary children’s hospitals in the US.
• Predictors included sociodemographic factors, asthma severity, clinical outcomes, chronic conditions, prior healthcare utilization, hospital characteristics, and timing-related factors.
• Models were trained on data from 36 hospitals (137,854 encounters) and externally validated on a test set of 11 hospitals (35,876 encounters). Performance was assessed using the area under the receiver operating characteristic curve (ROC AUC) and the area under the precision-recall curve (PR AUC).
• The primary outcome was inpatient readmission for asthma exacerbation within 180 days of discharge.

TAKEAWAY:

• Overall, readmission within 180 days occurred in 10.1% of the training set and 9.8% of the test set.
• The ML model showed a modest but statistically significant improvement in predicting pediatric asthma readmissions compared with the conventional regression model, with a higher ROC AUC (0.718; 95% CI, 0.709-0.727 vs 0.702; 95% CI, 0.692-0.712) and PR AUC (0.271; 95% CI, 0.256-0.286 vs 0.250; 95% CI, 0.236-0.264; P < .001 for both).

IN PRACTICE:

“Despite the moderate performance of an ML model, this study represents an important benchmark in the use of administrative data for pediatric readmission prediction, demonstrating that even limited data may support early identification of high-risk children that is generalizable across children’s tertiary hospitals in the United States,” the authors wrote.

SOURCE:

The study was led by Jonathan M. Gabbay, Albert Einstein College of Medicine, Bronx, New York. It was published online on April 15, 2026, in Pediatric Pulmonology.

LIMITATIONS:

Patients could not be tracked across hospitals outside the database network, which may have led to incomplete follow-up for readmissions. The database did not capture outpatient visits and prescription medications. Case identification relied on diagnostic codes, which could have introduced misclassification bias.

DISCLOSURES:

This study was supported by a Career Development Award provided by the Institute of Clinical and Translational Research at Albert Einstein College of Medicine. The authors reported having no conflicts of interest.

This article was created using several editorial tools, including AI, as part of the process. Human editors reviewed this content before publication.