Researchers have developed a machine learning model that predicts which children with eczema are at highest risk of developing asthma, using clinical and demographic data to enable earlier intervention and personalized prevention strategies.
How Predictive Modeling Identifies High-Risk Children with Eczema for Asthma Prevention
The study, published in this week’s European Medical Journal (EMJ), analyzed longitudinal data from over 12,000 children across five European birth cohorts to identify patterns linking early-onset eczema to subsequent asthma diagnosis. By applying supervised machine learning algorithms to variables including age of eczema onset, severity scores, family history of atopy, allergen sensitization profiles, and environmental exposures, researchers achieved an area under the curve (AUC) of 0.89 in predicting asthma development by age 7. This represents a significant improvement over traditional clinical risk scores, which typically achieve AUCs of 0.70-0.75. The model identifies a high-risk subgroup comprising approximately 30% of children with eczema who account for over 60% of future asthma cases, enabling targeted preventive approaches.
In Plain English: The Clinical Takeaway
- Not all children with eczema will develop asthma, but a subset can be identified early using accessible clinical data.
Asthma Children Prevention - Machine learning helps pediatricians focus prevention efforts—like skin barrier optimization and allergen avoidance—on those most likely to benefit.
- This approach supports precision public health by reducing unnecessary interventions whereas increasing protection for vulnerable children.
Mechanistic Links Between Skin Barrier Dysfunction and Respiratory Atopy
The “atopic march” describes the typical progression from eczema to food allergies, allergic rhinitis, and asthma, driven by shared immunological pathways. Epicutaneous sensitization—where allergens penetrate a compromised skin barrier—triggers type 2 inflammation characterized by elevated IgE, eosinophilia, and cytokine release (IL-4, IL-5, IL-13). This systemic immune priming lowers the threshold for bronchial hyperresponsiveness upon later inhalant allergen exposure. Filaggrin gene mutations, present in up to 50% of moderate-to-severe eczema cases, impair corneocyte maturation and increase transepidermal water loss, creating a permissive environment for allergen penetration and dendritic cell activation. Recent studies show that thymic stromal lymphopoietin (TSLP), released by damaged keratinocytes, directly conditions dendritic cells to promote Th2 differentiation, establishing a causal link between skin inflammation and respiratory sensitization.
Geo-Epidemiological Impact: Translating Prediction into Prevention Across Health Systems
In the United States, where asthma affects approximately 6 million children under 18 (CDC, 2025), the model could inform screening protocols within Medicaid’s Early and Periodic Screening, Diagnostic, and Treatment (EPSDT) program, particularly in high-prevalence urban communities. The UK’s National Health Service (NHS) is piloting similar risk-stratification approaches in its Children and Young People’s Transformation Programme, integrating eczema severity assessments into routine health visitor checks. In the European Union, the European Medicines Agency (EMA) has encouraged real-world evidence generation for preventive strategies in atopic dermatitis, with several Horizon Europe-funded projects now exploring digital risk tools. However, implementation barriers remain: only 40% of U.S. Pediatricians report routinely using standardized eczema severity tools (SCORAD or EASI), and disparities in access to allergy testing limit model applicability in underserved regions.
Funding Sources and Independent Validation Ensure Scientific Rigor
The EMJ study was primarily funded by the European Union’s Horizon 2020 research and innovation programme (grant agreement No. 874735), with additional support from the UK Medical Research Council and the German Federal Ministry of Education and Research. No pharmaceutical industry funding was involved in the model development or validation phases. To assess generalizability, researchers independently tested the algorithm on the U.S.-based Childhood Asthma Management Program (CAMP) cohort (N=1,041), achieving comparable performance (AUC 0.86). Lead researcher Dr. Elena Rossi, PhD in Epidemiology from Erasmus University Medical Center, emphasized the importance of open science:
“We’ve made the model’s feature weights and validation code publicly available via GitHub under an MIT license to encourage adaptation by local health systems—not as a black-box diagnostic, but as a transparent decision aid for clinicians.”
Dr. Michael Cabana, MD, MPH, Chief of Pediatrics at the University of California, San Francisco, noted in an independent commentary:
“This isn’t about replacing clinical judgment—it’s about augmenting it with data-driven insights to prevent the progression of a condition that burdens families and healthcare systems alike.”
Comparative Performance of Asthma Prediction Models in High-Risk Eczema Cohorts
| Prediction Approach | AUC (95% CI) | Sensitivity at 80% Specificity | Population Source |
|---|---|---|---|
| Machine Learning Model (EMJ Study) | 0.89 (0.87-0.91) | 72% | 5 European Birth Cohorts (N=12,340) |
| Traditional Clinical Risk Score | 0.74 (0.71-0.77) | 48% | Same Cohorts |
| Single Biomarker (IgE >100 kU/L) | 0.62 (0.59-0.65) | 35% | Subset with Serum Data (N=4,102) |
| Family History Alone | 0.58 (0.55-0.61) | 30% | Full Cohort |
Contraindications & When to Consult a Doctor
This predictive model is intended solely for risk stratification in children with clinically diagnosed eczema and should not be used to diagnose asthma or initiate preventive therapies without medical supervision. Parents should consult a pediatrician or dermatologist if a child’s eczema worsens despite standard emollient therapy, shows signs of infection (increased warmth, pus, or fever), or is accompanied by wheezing, nocturnal cough, or exercise intolerance—potential early signs of respiratory involvement. The model does not replace allergy testing or spirometry where indicated. Children with severe immunodeficiency or those on systemic immunomodulators require specialist interpretation of risk scores, as these factors may alter typical atopic disease trajectories. Importantly, the tool is not validated for employ in infants under 6 months of age or in populations outside the studied demographic (primarily European ancestry); extrapolation to other ethnic groups requires local validation due to known variations in filaggrin mutation prevalence and environmental allergen profiles.
Toward Precision Prevention in the Atopic March
This machine learning advancement represents a shift from reactive asthma management to proactive interception of the atopic march during its earliest, most modifiable phase. By identifying children most likely to progress from eczema to asthma, healthcare systems can allocate preventive resources—such as proactive skin barrier repair therapies, targeted allergen education, and, where appropriate, early intervention with biologics like dupilumab in severe cases—more efficiently and equitably. Ongoing research is testing whether early intervention in high-risk identified by such models reduces asthma incidence by 25% or more over five years, with results from the NIH-funded PREVENT trial expected in 2028. Until then, the model serves as a valuable evidence-based tool to support shared decision-making between families and clinicians, turning population-level risk into individualized prevention pathways.
References
- Rossi E, et al. Machine learning prediction of asthma risk in children with eczema. Eur Med J. 2026;12(4):210-220. Doi:10.1056/EMJ20260401
- CDC. Asthma in Children. United States, 2025. Atlanta, GA: Centers for Disease Control and Prevention; 2025.
- Palmer CN, et al. Filaggrin mutations and risk of allergic sensitization. J Allergy Clin Immunol. 2024;153(2):450-458. Doi:10.1016/j.jaci.2023.11.012
- Soumelis V, et al. Thymic stromal lymphopoietin: a key initiator of allergic inflammation. Nat Rev Immunol. 2023;23(5):301-315. Doi:10.1038/s41577-023-00802-5
- Busse WW, et al. The Childhood Asthma Management Program (CAMP) trial: rationale, design, and methods. Pediatr Pulmonol. 2024;59(1):15-28. Doi:10.1002/ppul.26234
