AI screening tool increases eye exam referrals for African American patients

Per News-Medical, investigators at the Wilmer Eye Institute, Johns Hopkins Medicine published exploratory, peer-reviewed findings in npj Digital Medicine (April 13) showing that an FDA-approved AI-assisted diagnostic screening program was associated with higher diabetic eye exam referrals for African American patients with diabetes. The News-Medical article reports the researchers conducted a retrospective analysis of 3,745 adult patients seen between August 2020 and September 2022, and that 3,352 patients received referrals from primary care providers while 393 patients received recommendations involving the AI-assisted screen. The study explicitly focused on two historically disadvantaged groups: African American patients and patients covered under Medicaid, per News-Medical.

Editorial analysis: The News-Medical summary does not publish full methodological tables, so readers should consult the original npj Digital Medicine paper for model performance metrics, false positive/negative rates, and how the AI outputs were integrated into primary care workflows. The article names the program as an FDA-approved AI-assisted screening tool, but does not provide the algorithm name, version, or deployment architecture in the summary.

AI-based screening embedded at the point of primary care can change referral patterns even when diagnostic workflows remain clinician-led. For practitioners, the key operational questions are how the screening output was presented to PCPs, what thresholds triggered a referral, and whether the AI screening altered clinician behavior versus simply documenting additional findings. The quoted caution from T.Y. Alvin Liu, M.D. underscores a common gap between referral generation and patient adherence: "A referral [from a primary care provider] doesn't guarantee people will attend a diabetic eye exam, even if it's needed," News-Medical reports.

For practitioners: obtain the full npj Digital Medicine article to inspect sensitivity, specificity, and subgroup performance for the AI tool; verify whether performance held across age and socioeconomic strata. Observers should also track subsequent implementation reports that document follow-up rates, appointment completion, and whether additional care-navigation interventions were paired with AI screening to convert referrals into completed exams.