Improved AI System for Diabetic Retinal Disease
Author Information
Author(s): Abrà moff Michael D., Lavin Philip T., Jakubowski Julie R., Blodi Barbara A., Keeys Mia, Joyce Cara, Folk James C.
Primary Institution: University of Iowa
Hypothesis
Can an improved autonomous AI system mitigate AI adoption bias in diabetic retinal disease diagnosis?
Conclusion
The improved autonomous AI system can effectively identify diabetic retinal disease without racial, ethnic, or sex bias, potentially increasing health access equity.
Supporting Evidence
- The AI system achieved a sensitivity of 79.6% and specificity of 88.4% against the Level I reference standard.
- All prespecified non-inferiority endpoints were met.
- DRD prevalence was found to be 29.0% among the participants.
Takeaway
A new AI system helps doctors find eye problems in people with diabetes, making sure everyone gets the same care, no matter their background.
Methodology
A preregistered, non-inferiority trial evaluated the AI system's sensitivity and specificity against established standards in a representative sample of diabetes patients.
Potential Biases
No significant racial, ethnic, or sex bias was identified in the AI's performance.
Limitations
The study was not designed to determine if the AI system improves health equity, only to assess its safety and efficacy.
Participant Demographics
50.8% male, 45.7% Hispanic, 17.3% Black.
Statistical Information
P-Value
p=0.021/p<0.001
Confidence Interval
95% CI: 89.2%, 92.0%
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website