Researchers at the Mayo Clinic have developed an artificial intelligence (AI)-powered digital biomarker that can identify a high-risk subgroup of children with asthma who are more susceptible to serious respiratory infections and asthma exacerbations. Their study, “Artificial Intelligence Biomarker Detects High-Risk Childhood Asthma Subgroup for Respiratory Infections and Exacerbations,” was recently published in the Journal of Allergy and Clinical Immunology.
Researchers analyzed data from more than 22,000 children born between 1997 and 2016, using natural language processing (NLP) algorithms to evaluate electronic health records. By applying two validated tools — the NLP-based predetermined asthma criteria (NLP-PAC) and asthma predictive index (NLP-API) — researchers categorized children into four distinct asthma phenotypes.
The most vulnerable group, identified as NLP−PAC+/NLP−API+, showed a significantly higher risk of developing pneumonia or influenza A/B and experiencing asthma exacerbations by age three. This subgroup also had elevated rates of respiratory syncytial virus (RSV) infections in early childhood.
Researchers noted that his is the first large-scale study to demonstrate that an AI-driven digital biomarker can detect a high-risk asthma phenotype for acute respiratory infections as early as infancy.
The findings suggest that the AI biomarker could be used to guide early intervention strategies, potentially improving outcomes for children with asthma who are at greater risk of complications from respiratory infections.
In addition to clinical implications, the study also confirmed previous observations that the PAC/API-positive subgroup exhibits distinct immunologic characteristics, such as eosinophilic inflammation and a higher likelihood of allergic rhinitis.
The research underscores the growing role of AI in precision medicine, particularly in pediatric care, where early identification of risk can lead to more targeted and effective treatment plans.
