OBJECTIVE

To assess the diagnostic accuracy of point-of-care screening using Medios™ Artificial Intelligence (AI) in the diagnosis of diabetic retinopathy (DR).

METHODS

This is a multi-center, cross-sectional, instrument validation study among adult Filipinos with Type 2 diabetes seen at Endocrine specialty clinics from May to November 2021. Retinal images were captured by a minimally trained nurse using the Remedio Fundus on Phone (FOP). Images were interpreted separately by the Medios™ AI and three retina specialists. The primary outcome measure is the accuracy of Medios™™ AI in diagnosing DR compared to retina specialists’ findings using sensitivity and specificity, predictive values, and likelihood ratios.

RESULTS

A total of 182 subjects with Type 2 diabetes were included in the study. The sensitivity and specificity of the Medios™ AI in diagnosing any DR were 73.68% (95%CI, 57.99-85.03) and 83.74% (95%CI, 79.35-87.35), respectively, compared with the retinal specialists’ findings using the same images. The positive and negative predictive values were 34.57% (95%CI, 25.12-45.41) and 96.47% (95%CI, 93.62-98.07). The positive and negative likelihood ratios were 4.53 (95%CI, 4.26 4.82) and 0.31 (95%CI, 0.26-0.38). The overall diagnostic accuracy of Medios™ AI in detecting DR was 82.69% (95%CI, 78.47-86.23).

CONCLUSION

The Medios™ AI system showed acceptable diagnostic accuracy when used as a point-of-care screening tool in detecting DR among patients with Type 2 diabetes seen at Endocrine specialty clinics. This technology can be a useful screening tool for endocrinologists as it is relatively inexpensive, safe, and easily performed. It can also shorten the lead time from screening to referral and intervention.

Human ; Diabetes Mellitus ; Diabetic Retinopathy