Automated Machine Learning Analysis of Patients With Chronic Skin Disease Using a Medical Smartphone App: Retrospective Study
September 2023
in “
JMIR. Journal of medical internet research/Journal of medical internet research
”
TLDR Machine learning can predict symptoms and quality of life in chronic skin disease patients using smartphone app data, and shows that app use varies with patient characteristics.
In a retrospective study involving 368 patients with chronic hand and/or foot eczema or psoriasis vulgaris, automated machine learning (AutoML) was used to analyze data collected from a smartphone monitoring app. The study aimed to model the development of itching, pain, and Dermatology Life Quality Index (DLQI) over 6 months, as well as to assess app usage patterns. The light gradient boosted trees classifier model was the most accurate for predicting itching development, while the random forest classifier model was used for pain and DLQI development. App usage was analyzed with an elastic net blender model, revealing that higher BMI, higher disease activity, and higher anxiety levels were associated with increased app use, while older age was associated with decreased use. The study demonstrates the potential of smartphone apps and AutoML in managing chronic skin diseases and improving patient care.
