Qdata Geographic Atrophy

Qdata® Geographic Atrophy (GA) is an exclusive, real-world dataset of 800,000+ de-identified patients with GA in the U.S. This dataset is created by curating and linking electronic health records and ophthalmic images from the American Academy of Ophthalmology IRIS® Registry (Intelligent Research in Sight). Qdata Geographic Atrophy curated variables include diagnosis by source, visual acuity and lesion location (subfoveal vs. nonsubfoveal), lesion size and lesion growth rate. Specific tables include patient demographics, procedure, observation, provider, payer information and more.

For more information on Qdata Geographic Atrophy, visit veranahealth.com  or complete this form  to request a sample dataset or speak with a team member.

GA diagnosis

Until Feb. 17, 2023, no FDA-approved treatment for GA existed, resulting in under-coded GA ICD-10 diagnoses. Qdata Geographic Atrophy provides GA confirmation based on structured ICD-10 codes, as well as from notes and images.

Image derived GA diagnosis is generated from 30 degree field of view (FoV) fundus autofluorescence (FAF) images, volumetric optical coherence tomography (OCT), and 30 degree FOV infrared reflectance (IR) images using deep learning (DL) models. All images have gone through image quality assessment and location assessment to filter out fovea centered gradable images. In the ground truth dataset, the presence of GA is defined following criteria provided in the CAM3 report.

GA lesion location (subfoveal vs. nonsubfoveal)

Qdata Geographic Atrophy uses a combination of ICD-10 codes, notes, and images to identify and confirm GA lesion location. Lesion location provides valuable information about the severity of GA, with patients with subfoveal GA being more likely to experience central vision loss and legal blindness, and a worse prognosis than patients with nonsubfoveal involvement only.

Image derived GA lesion location is generated from volumetric OCT images using DL models, all OCT images have gone through image quality assessment and location assessment to filter out foveal centered gradable images. In the GA with subfoveal involvement is defined as GA presence at the central fovea pixel, the same definition as GA clinical trials.

GA lesion size

GA lesion size is calculated from 30 degree FoV FAF images using Verana Health developed machine learning (ML) models. GA lesion size is defined as the total area of all GA lesions in an FAF image.

Number of GA lesions

GA lesion number is calculated from 30 degree FoV FAF images using Verana Health developed ML models. GA lesion number is defined as the total number of all GA lesions in an FAF image.

GA lesion permiter

GA lesion perimeter is calculated from 30 degree FoV FAF images using Verana Health developed ML models. GA lesion perimeter is defined as the total perimeter of all GA lesions in an FAF image.

GA lesion growth rate

GA lesion growth rate is calculated from 30 degree FoV FAF images using Verana Health developed ML models. GA lesion growth rate is defined as:(GA lesion size t2 - GA lesion size t1)/(t2 - t1) where t2 and t1 are at least 6 months apart.

Use Cases

• Calculate average VA by time since first GA diagnosis
• Calculate average number of visits by year since first GA diagnosis
• Calculate average VA (ETDRS letters) and number of measurements per patient eye within a year of first GA diagnosis, stratified by age group