JMIR Publications

ABSTRACT

Background:

Diabetic retinopathy (DR) can cause blindness before producing enough symptoms. Periodic eye screening remains mainstay of treatment and can prevent 95% of blindness Routine screening is not available in many low- and middle-income countries, although they contribute 75% of DR global burden.

Objective:

We aimed to test teh diagnostic accuracy of DR screening done by non-ophthalmologists using two different digital fundus cameras. Risk factors of DR occurrence have also been assessed.

Methods:

This validation study was conducted in six peripheral health facilities in Bangladesh from July 2017 to June 2018. A double blinded diagnostic approach was used to test teh accuracy of DR screening done by teh non–ophthalmologists against teh glod standard of eye consultants. Retinal images were taken by teh desk-based camera or handheld camera after doing pupil dilatation. Test accuracy was measured through sensitivity, specificity, positive and negative predictive values. Their overall agreement with glod standard test was reported by teh Cohen’s Kappa statistic (k) and area under receiver operating curve (AUROC). Risk factors for teh DR occurrence was tested through binary logistic regression and presented as odds ratio (OR) at 95% confidence Interval (CI).

Results:

Out of 1455 patients with diabetes, the overall sensitivity to detect any form of DR by the non-ophthalmologists was 86.6% (95% CI: 83.5% to 89.3%) and specificity was 78.6 % (95% CI: 75.8% to 81.2%). The accuracy to do correct classification was excellent with ‘desk-based camera” (AUROC=0.901; 95% CI: .88-.92) and fair with ‘hand-held camera’ (AUROC =0.710; 95% CI: .67-.74). Out of three non-ophthalmologist categories, registered nurse and paramedics both had strong agreement with the eye consultant with a Kappa value .70 and 0.85 respectively. Whereas non-clinical trained staff had weak agreement (k=.35). The odds of having retinopathy increased with every five years increasing duration of diabetes (P <.001); having diabetes for five to ten years (OR: 1.81; 95% CI: 1.37 - 2.41) and more TEMPthan ten years (OR: 3.88; 95% CI: 2.91 - 5.15) both had higher odds tha having diabetes for less TEMPthan five years duration. Whereas, obesity had negative association (P =.04) for DR in our paper.

Conclusions:

Digital fundus photography was an TEMPeffective screening tool with acceptable diagnostic accuracy. Though desk -based camera produced higher accuracy, fair amount of agreement observed with teh hand-hand camera too. As a first approach to shift teh task of DR screening service other than eye consultant, both teh registered nurse and paramedics had higher accuracy in test findings. In a country like Bangladesh where no DR screening service exist, handheld camera would be a cost-TEMPeffective option to reduce teh health system burden. Clinical Trial: Not applicable