Non-attendance at diabetic eye screening and risk of sight-threatening diabetic retinopathy: a population-based cohort study.

AIMS/HYPOTHESIS: This study evaluated whether repeated non-attendance for diabetic eye screening is associated with the risk of sight-threatening diabetic retinopathy (STDR). METHODS: This was a cohort study of 6,556 residents with diabetes who were invited for screening between 2008 and 2011 in a population-based eye screening programme in inner London and who attended for their first-ever screen in 2008. The proportion of participants with STDR was evaluated in relation to the number of years in which screening was missed. RESULTS: The proportion of participants who did not attend screening decreased between 2009 and 2011 (annual reduction 1.6% [95% CI 0.9%, 2.3%]). The adjusted relative odds of STDR for 210 participants who did not attend two consecutive years of screening were 3.76 (95% CI 2.14, 6.61; p < 0.001), compared with participants who were screened annually. In 605 participants with mild non-proliferative retinopathy at the first screen, the adjusted relative odds of developing proliferative or moderate to severe non-proliferative retinopathy were 5.72 (95% CI 7.43, 22.83; p = 0.013) for 53 participants who missed two screens. CONCLUSIONS/INTERPRETATION: Patients who do not attend diabetic eye screening are at increased risk of developing STDR. Tracing of non-attenders with evidence of established retinopathy should be an important fail-safe procedure.

Prospective evaluation of an artificial intelligence-enabled algorithm for automated diabetic retinopathy screening of 30 000 patients.

BACKGROUND/AIMS: Human grading of digital images from diabetic retinopathy (DR) screening programmes represents a significant challenge, due to the increasing prevalence of diabetes. We evaluate the performance of an automated artificial intelligence (AI) algorithm to triage retinal images from the English Diabetic Eye Screening Programme (DESP) into test-positive/technical failure versus test-negative, using human grading following a standard national protocol as the reference standard. METHODS: Retinal images from 30 405 consecutive screening episodes from three English DESPs were manually graded following a standard national protocol and by an automated process with machine learning enabled software, EyeArt v2.1. Screening performance (sensitivity, specificity) and diagnostic accuracy (95% CIs) were determined using human grades as the reference standard. RESULTS: Sensitivity (95% CIs) of EyeArt was 95.7% (94.8% to 96.5%) for referable retinopathy (human graded ungradable, referable maculopathy, moderate-to-severe non-proliferative or proliferative). This comprises sensitivities of 98.3% (97.3% to 98.9%) for mild-to-moderate non-proliferative retinopathy with referable maculopathy, 100% (98.7%,100%) for moderate-to-severe non-proliferative retinopathy and 100% (97.9%,100%) for proliferative disease. EyeArt agreed with the human grade of no retinopathy (specificity) in 68% (67% to 69%), with a specificity of 54.0% (53.4% to 54.5%) when combined with non-referable retinopathy. CONCLUSION: The algorithm demonstrated safe levels of sensitivity for high-risk retinopathy in a real-world screening service, with specificity that could halve the workload for human graders. AI machine learning and deep learning algorithms such as this can provide clinically equivalent, rapid detection of retinopathy, particularly in settings where a trained workforce is unavailable or where large-scale and rapid results are needed.

Changes in detection of retinopathy in type 2 diabetes in the first 4 years of a population-based diabetic eye screening program: retrospective cohort study.

OBJECTIVE: Annual diabetic eye screening has been implemented in England since 2008. This study aimed to estimate changes in the detection of retinopathy in the first 4 years of the program. RESEARCH DESIGN AND METHODS: Participants included 32,340 patients with type 2 diabetes resident in three London boroughs with one or more screening records between 2008 and 2011. Data for 87,570 digital images from 2008 to 2011 were analyzed. Frequency of sight-threatening diabetic retinopathy (STDR) was estimated by year of screen for first screens and for subsequent screens according to retinopathy status at first screen. RESULTS: Among 16,621 first-ever screens, the frequency of STDR was 7.1% in 2008, declining to 6.4% in 2011 (P = 0.087). The proportion with a duration of diabetes of <1 year at first screen increased from 18.7% in 2008 to 48.6% in 2011. Second or later screens were received by 26,308 participants. In participants with mild nonproliferative retinopathy at first screen, the proportion with STDR at second or later screen declined from 21.6% in 2008 to 8.4% in 2011 (annual change -2.2% [95% CI -3.3 to -1.0], P < 0.001). In participants with no retinopathy at first screen, STDR declined from 9.2% in 2008 to 3.2% in 2011 (annual change -1.8% [-2.0 to -1.7], P < 0.001). Declining trends were similar in sociodemographic subgroups. CONCLUSIONS: After the inception of population-based diabetic eye screening, patients at lower risk of STDR contribute an increasing proportion to the eligible population, and the proportion detected with STDR at second or subsequent screening rounds declines rapidly.