Leveraging Artificial Intelligence to Prevent Vision Loss From Diabetes

This study is a parallel cluster randomized clinical trial to determine whether AI-BRIDGE, an autonomous AI-based protocol of screening for DR, improves DR screening and follow-up rates compared to usual care. In this study, DR screening refers to the detection of DR either via referral of patients for an eye-care appointment with an eye-care provider in usual-care or via point-of-care AI-based DR screening for patients in AI-BRIDGE. The trial will be conducted across 8 clinics from four health systems that are Federally Qualified Health Centers. Each clinic will be randomly assigned to either usual care or the AI-BRIDGE intervention.

Randomization is done at the clinic level in this study. One of the two clinics within each of the four health systems will be randomly selected to be in the AI-BRIDGE arm (i.e., a total of four clinics will use AI-BRIDGE during the intervention period) and the other clinics will be in the control arm (i.e., a total of four clinics will be in usual care).

Primary Objective:

Compare the proportion of patients who follow-up with recommended eye care in the AI-BRIDGE and usual-care arms within 6 months of the recommendation.

Secondary Objective:

Compare the proportion of patients who complete DR screening, defined as: (AI-BRIDGE) an AI-interpretable image obtained at the primary care visit; or, for patients who did not complete AI screening or did not start AI-screening, an in-person screening eye care visit completed within 6 months of the primary care visit; (Usual-care) an in-person screening eye care visit completed within 6 months of the primary care visit.

Clinics Inclusion Criteria:

• Clinic must be from a health system with at least two clinics that meet eligibility criteria and will agree to join.
• Number of unique patients with diabetes should be >550 over any continuous 15-month period prior to the clinic providing a letter of support to participate in the trial.
• No point-of-care screening system in use for DR.
• Have an electronic health record to track patient data and agree to share limited data set as requested.
• Have a member of the clinical team who can commit to serving as the clinic champion for this study.
• Have the IT infrastructure to deploy an artificial-intelligence-based eye camera in the clinic.

Patient Inclusion Criteria:

• Diagnosis of type 1 or 2 diabetes
• No known diabetic retinopathy
• No diabetic retinopathy screening in the past 12 months

Patient Exclusion Criteria:

• Have a documented DR screening in the electronic health record within 12 months of the date of the primary care visit
• Contraindication includes diagnosed with macular edema, severe non-proliferative retinopathy, proliferative retinopathy, radiation retinopathy, or retinal vein occlusion
• Pregnant