Acta Ophthalmologica

Handheld cameras have been shown to increase access to diabetic retinopathy (DR) screening in a cost-effective manner, particularly in low-resource settings and hard to reach populations. We believe such devices should also be compared against modalities used in the current practice, in a way to assess their performance in a real-world scenario. In some regions of Brazil, a country with an enormous population of individuals with diabetes and devoid of a national strategy for DR screening, binocular indirect ophthalmoscopy (BIO) corresponds to the current practice.

We, herein, present the comparison of images obtained with a native portable retinal camera (Eyer, Phelcom Technologies, São Carlos, Brazil) against BIO, performed by retinal specialists. Such images were obtained during a volunteer-based DR screening effort performed in 2019, in an underserved area of Bahia State, Northeastern Brazil, which offered retinal examination for 791 individuals with diabetes mellitus aged over 18 years old (mean age $60.9 \pm 11.1$ years, 61.6% females). Image acquisition was performed along a 6-h period by a team of nine examiners with heterogenous experience, the ambient lighting conditions of the healthcare facility, and the high-burden setting were compatible with the real-world scenario. Patients with a previous history of laser treatment were excluded, resulting in a total of 700 individuals (1400 eyes).

The ungradable rate was 14.9% of eyes. Among the remaining 1191 gradable eyes, kappa statistics—simple ($K_s$) and weighted ($K_w$)—revealed an almost perfect agreement between image reading and BIO for any DR, more than mild DR, and vision-threatening (severe NPDR or PDR, and/or DME) DR ($K_s=0.8615, 0.8799$ and $0.8631$, respectively). Almost perfect agreement was also found among different levels of DR and for DME presence ($K_w=0.8369$ and $K_s=0.8753$, respectively).

The already-mentioned advantages of handheld cameras for DR screening associated with telemedicine may bridge the gaps of insufficient eyecare personnel and infrastructure, while decreasing the burden on busy eye clinics or on the scarce specialists. The almost perfect agreement reported, herein, in terms of all the proposed outcomes makes the case for handheld cameras even more compelling for the real world of low-to middle-income countries. By bringing real-world validation to a DR screening strategy based on mobile devices in a high-burden setting, we hope to reinforce the important role of portable cameras in the prevention of avoidable blindness.