A new study shows that a machine learning predictive model can identify which children with early-onset eczema are most likely to develop persistent asthma or allergic rhinitis by school age.
Atopic dermatitis, a common form of eczema, has long been recognized as an early stage in the “atopic march,” a progression that can lead to asthma and allergic rhinitis.
However, clinicians have historically struggled to determine which infants develop more severe and long-term respiratory disease. This uncertainty limits opportunities for early intervention and personalized care.
Machine learning asthma prediction model shows high accuracy
Researchers conducted a large retrospective birth cohort study using electronic health record data from 10,688 children diagnosed with atopic dermatitis before age 3.
Two machine learning asthma prediction models were developed. Comprehensive models using detailed clinical variables and simplified models based on routinely available clinical data.
Both models showed good performance in predicting moderate to severe persistent asthma from 5 to 11 years of age.
The comprehensive model achieved an area under the curve (AUC) of 0.893, while the simplified model showed nearly identical discrimination (AUC: 0.892). At a specificity of 95%, the sensitivity reached 40.4% and 36.2%, and the positive predictive value was 39.3% and 33.8%, respectively.
Prediction of allergic rhinitis and risk stratification
In addition to asthma, the model was also applied to predict allergic rhinitis.
Performance was more moderate, with AUC values of 0.793 and 0.773. However, the positive predictive value was significantly higher in high-risk groups, reaching over 70% in the comprehensive model.
Importantly, both the comprehensive and simplified models for asthma and allergic rhinitis showed good calibration, especially in children classified as highest risk.
This suggests that machine learning predictive tools may be effective in stratifying patients and identifying those who may benefit from closer surveillance or prevention strategies.
Implications for early individualized care
These findings highlight the potential for machine learning predictions to transform pediatric allergy care. By leveraging early clinical data, clinicians may be able to move beyond reactive treatments and adopt a proactive, individualized approach.
This study is limited by its retrospective design and reliance on data from a single health system, which may affect generalizability. Future studies are needed to validate these models across diverse populations and assess their impact in real-world clinical settings.
If confirmed, such tools could play an important role in early identification of high-risk children and enable targeted interventions that could change the trajectory of atopy progression.
reference
Chen W et al. Machine learning prediction of asthma and allergic rhinitis in children with early-onset atopic dermatitis. J Allergy Clinical Immunol. 2026;DOI:10.1016/j.jaci.2026.03.025.
Featured image: Evgeniia Primavera from Adobe Stock
