From the 2022 HVPA National Conference
Y. Alvin Liu M.D. (Wilmer Eye Institute, Johns Hopkins University School of Medicine)
Diabetic retinopathy (DR) is the leading cause of vision loss in the working age population around the world. Annual retinal exams allow at risk patients to be identified and treated promptly to prevent permanent blindness, leading the National Committee for Quality Assurance (NCQA) to designate DR screening adherence at the population level as a Healthcare Effectiveness Data and Information Set (HEDIS) quality measure. In 2020, this quality measure was updated to clarify that diabetic retinal exams performed by a rigorously-validated, FDA-cleared, autonomous artificial intelligence (AI) device will qualify. However, whether autonomous AI can make a difference in meeting this HEDIS metric has not been established before.
To investigate whether implementation of an autonomous AI system for DR screening in primary care clinics will improve DR screening compliance at an integrated healthcare system.
Year-on-year data was compared: August 2020 (prior to implementation) vs. August 2021 (1 year after implementation). Community based primary care clinics within Baltimore City and County, managed by Johns Hopkins Medicine, were included. 2 clinics with autonomous AI DR screening (“AI clinics”) were compared to 4 clinics without autonomous AI screening (“non-AI clinics”).
DR screening compliance rate in the AI clinics improved from 35.4% (1,127/3,182) to 58.8% (1,995/3,395). The compliance rate in the non-AI clinics improved from 30.0% (751/2,504) to 41.7% (1,080/2,593). The year-on-year improvement of 23.4% in AI clinics was significantly higher as compared to the improvement of 11.7% in the non-AI clinics (p<0.0002).
Conclusions and Clinical Implications
The implementation of autonomous AI DR screening in primary care clinics results in a significantly greater improvement in DR screening compliance. Thus, it has the potential to improve diabetic patients’ ophthalmic health and prevent irreversible blindness at a population level.