Autonomous Artificial Intelligence Increases Access and Health Equity in Underserved Populations with Diabetes

Area of Science:

• Ophthalmology and Public Health
• Artificial Intelligence in Healthcare
• Health Equity Research

Background:

• Diabetic eye disease (DED) is a leading cause of preventable blindness, necessitating annual screening for diabetic patients.
• Historically low adherence to DED testing guidelines and disparities in access disproportionately affect minority populations.
• The FDA has cleared autonomous artificial intelligence (AI) for DED diagnosis in primary care settings.

Approach:

• A retrospective study compared DED testing adherence rates between primary care sites with and without autonomous AI implementation from 2019 to 2021.
• A difference-in-difference analysis using logistic regression was performed on data from 17,674 (2019) and 17,590 (2021) adult patients with diabetes.
• Sites were categorized as 'non-AI' (no AI deployment) or 'AI-switched' (AI deployed between 2019 and 2021).

Key Points:

• Autonomous AI implementation was associated with a 36% higher odds ratio of adherence to DED testing compared to non-AI sites.
• AI improved access for Black/African American patients, with an 11.9% adherence increase in AI-switched sites versus a 1.2% decrease in non-AI sites.
• AI helped close care gaps, significantly reducing the adherence disparity between Asian Americans and Black/African Americans from 15.6% to 3.5%.

Conclusions:

• Real-world deployment of autonomous AI in a large healthcare system demonstrably improved adherence to DED screening, a key Healthcare Effectiveness Data and Information Set (HEDIS) measure.
• Autonomous AI enhanced patient access to DED testing and promoted health equity, particularly benefiting historically disadvantaged populations.
• Findings suggest autonomous AI is a valuable tool for improving diabetic eye care outcomes and reducing blindness, though further prospective validation is needed.