Machine learning–based identification of moderate-to-severe atopic dermatitis in US patients using the TARGET-DERM AD registry.

Challenge

Identifying patients with moderate-to-severe AD in large real-world datasets typically requires manual chart review or IGA-based classification, which is resource-intensive and limits scalability for population-level analyses. There was no validated, automated approach to distinguish moderate-to-severe AD patients from milder cases using structured EHR data in the TARGET-DERM registry.

Solution

A machine learning algorithm was developed and validated using TARGET-DERM AD registry data to identify patients with moderate-to-severe AD based on structured EHR variables, creating a scalable, automated classification tool applicable to large real-world populations.

Impact

A validated ML-based patient identification algorithm enables rapid, reproducible cohort construction for population-level AD analyses without manual IGA review, directly supporting Amgen/Kyowa Kirin's development programs by enabling scalable real-world evidence generation from the TARGET-DERM infrastructure.