AI Support Tools Recommended for Pediatric Atopic Dermatitis Management

Michael Brandwein, PhD

Investigators led by Michael Brandwein, PhD, MYOR Diagnostics in Yerushalayim, Israel, suggested that machine learning predictive modeling using non-invasive and accessible inputs could be useful in stratifying an infant’s risk of developing atopic dermatitis from birth.

Data from their study was presented at the American Academy of Pediatrics (AAP) 2021 Virtual Conference during the session “Stratifying Risk for Early Onset Atopic Dermatitis from Birth: An AI-Empowered Clinical Decision Support Tool”.

Previous studies reported on the additional risk factors caused by a parental history of atopic dermatitis as well as other atopic conditions.

Prior to the current study, it was unclear as to whether the combined analysis of familial history and other elements allowed for an understanding of the interaction between the 2 variables, as well as the factors behind an infant’s risk of developing atopic dermatitis.

As such, Brandwein and colleagues employed various precision models in their study to assess an infant’s risk of developing atopic dermatitis from birth.

The Study

Brandwein and colleagues utilized a geographically diverse cohort, as well as a myriad of risk factors.

The investigators combined study data from 4 separate pediatric cohorts, with each one being assessed for the development of atopic dermatitis within 15 months of birth.

A total of 2469 infants were evaluated in the study based off available data from the cohorts.

Data on an infant’s paternal and maternal history of atopic dermatitis, asthma, food allergies, birthweight, gender, maternal smoking, and ethnicity were also obtained during the study.

Predictive models were trained and validated on the combined dataset.

The Findings

Investigators concluded that infants with a paternal history of atopic dermatitis, food allergies, or asthma had a higher odds ratio of developing atopic dermatitis than those with maternal atopic dermatitis, food allergies, and asthma.

Additionally, receiver operating characteristic curve analyses showed an area under the curve of 0.83 for the random forest model, which outperformed a logistic regression and XGBoost model.

They added that the model a maximum accuracy of 78%, with corresponding sensitivity of 63%, specificity of 87%, and a positive predictive value of 75%.

Brandwein and colleagues espoused the act of using these models for the education of pediatric risk of atopic dermatitis for clinicians and caregivers.

“Knowledge of an infant's risk can inform both caregivers and medical professionals as to timely interventions to mitigate the discomfort associated with atopic dermatitis,” the team wrote.