Predictors of Flares and Disease Severity in Patients With Atopic Dermatitis Using Machine Learning

Area of Science:

• Dermatology
• Epidemiology
• Biostatistics

Background:

• Atopic dermatitis (AD) is a chronic condition with unpredictable flares impacting quality of life.
• Current severity classifications and treatment decisions often overlook the impact of flares.

Purpose of the Study:

• To validate flare predictability for atopic dermatitis (AD) severity.
• To quantify the predictive importance of flares on AD severity and vice versa.

Main Methods:

• Utilized the Danish Skin Cohort, analyzing data from 878 patients with AD.
• Employed quantile regression to associate 2022 flare frequency with 2023 patient-reported severity.
• Applied boosted random forests to identify predictors of flares and AD severity.

Main Results:

• A higher number of annual flares in 2022 significantly correlated with patient-reported AD severity measures in 2023.
• Flare frequency predicted Patient-Oriented Eczema Measure and Dermatology Life Quality Index scores, even after adjusting for baseline severity.
• Flare characteristics (severity, duration, number) were key predictors of AD severity, while AD severity predicted flare frequency.

Conclusions:

• Increased flare frequency in atopic dermatitis (AD) is linked to poorer disease prognosis and reduced quality of life.
• Flares are crucial indicators for assessing AD severity and predicting future disease course.
• Establishing flare thresholds in treatment decisions is recommended for enhanced disease control and patient well-being.