AI models predict glaucoma risk using self-reported health data

A new study demonstrates that artificial intelligence (AI) models can effectively predict glaucoma risk using self-reported health data, potentially improving early detection in large-scale, low-resource settings.

Researchers analyzed data from 8,205 participants in the All of Us Research Program, identifying glaucoma diagnoses based on electronic health records. They trained 3 machine learning models, including penalized logistic regression, XGBoost, and a neural network, to assess glaucoma risk using demographic, lifestyle, and medical history data from surveys.

XGBoost performed best, achieving an area under the receiver operating characteristic curve (AUROC) of 0.890, followed by logistic regression (AUROC 0.772). Key predictive factors included age, type 2 diabetes, and family history of glaucoma.

Reference
Ravindranath R, Naor J, Wang SY. Artificial Intelligence Models to Identify Patients at High Risk for Glaucoma Using Self-reported Health Data in a United States National Cohort. Ophthalmol Sci. 2024;5(3):100685. doi: 10.1016/j.xops.2024.100685. PMID: 40151359; PMCID: PMC11946806.