Automated Machine Learning Analysis of Patients With Chronic Skin Disease Using a Medical Smartphone App:...

Area of Science:

• Digital health
• Machine learning in dermatology
• Health informatics

Background:

• Digitalization in healthcare is increasing, but trust and technical expertise hinder mobile health app adoption.
• Automated machine learning (AutoML) offers a solution to empower healthcare professionals with mobile health technologies.

Purpose of the Study:

• To analyze clinical data from patients with chronic hand/foot eczema or psoriasis vulgaris using a smartphone monitoring app.
• To focus on predicting itching, pain, Dermatology Life Quality Index (DLQI), and app usage patterns.

Main Methods:

• Utilized AutoML to process a pseudonymized dataset of 368 patients, combining and preparing data from three primary sources.
• Developed and selected multiple machine learning classification models, including gradient boosted trees and random forest, for data analysis.

Main Results:

• Accurately modeled itching, pain, and DLQI development over six months using various AutoML classifiers.
• Identified key correlations influencing outcomes, such as BMI, age, disease activity, and anxiety scores.
• Found specific demographic and clinical factors associated with increased app usage.

Conclusions:

• Smartphone and AutoML techniques show significant potential for improving chronic disease management in dermatology.
• The study provides insights into data characteristics and outcomes for patients with chronic eczema or psoriasis.