Trends in Diabetic Retinopathy, Visual Acuity, and Treatment Outcomes for Patients Living With Diabetes in a Fundus Photograph–Based Diabetic Retinopathy Screening Program in Bangladesh

Key Points Question Is it feasible to develop a systematic model of fundus photograph–based eye screening for patients living with diabetes in Bangladesh? Findings In this cross-sectional study, 49 264 patients underwent fundus photograph–based eye screening during a 7-year period at 3 centers in Bangladesh. The primary prevalence rate of diabetic retinopathy was 33%. Meaning Given the very high prevalence of diabetes, there is an urgent need to scale up diabetic eye screening services in resource-poor countries to prevent retinopathy complications and vision loss.


Introduction
Type 1 and type 2 diabetes is rapidly increasing as a major cause of morbidity and mortality globally. 1,2 According to the World Health Organization and the International Diabetes Federation, the number of people living with diabetes was 425 million in 2017 and is projected to increase to 629 million in 2045. 3 The International Diabetes Federation estimated that 6.9 million people in Bangladesh were living with diabetes in 2017, increasing to more than 10 million in 2025. 3 VISION 2020: The Right to Sight initiative lists diabetic retinopathy (DR) as one of its priority eye diseases for Southeast Asia and other regions. 4 Diabetic retinopathy is a leading cause of earlyonset blindness among working-age people with diabetes worldwide. 5 It affects nearly all individuals with type 1 diabetes and approximately 77% of those with type 2 diabetes who have had diabetes for 20 years or longer. 6 The risk of DR is directly associated with the duration of individual diabetes cases and the level of glucose control maintained. 5 Bangladesh has among the world's largest number of people with diabetes, with 35% of the population older than 35 years receiving a diagnosis of diabetes or prediabetes. 7, 8 The public and private health care systems in Bangladesh lack effective models for implementing DR screening programs at the national level. The enormous screening burden cannot be effectively addressed by using DR screening clinics within eye hospitals alone. The current system fails to raise awareness and motivate patients to seek regular screening for DR, and almost no organizations are systematically engaged in screening for DR.
Diabetic retinopathy is asymptomatic until the later, more severe stages, but retinal screening can dramatically reduce the risk of blindness because treatment of retinopathy at an early stage allows for good vision to be retained. 9 With proper screening and treatment, almost all blindness from DR can be prevented. We present the findings from a DR screening program focused on identifying DR among individuals presenting for diabetes care in Bangladesh.

Population
The patients in this cross-sectional study were individuals with diabetes who were screened at 3 different hospitals in Bangladesh between June 1, 2010, and September 30, 2017. All data were collected at the 2 participating eye hospitals-Chittagong Eye Infirmary and Training Complex (CEITC) and the National Institute of Ophthalmology (NIO) and Hospital-as well as at Feni Diabetes Hospital (FDH). Ethical approval for the projects was granted by the local institutional review boards at each institution, with a memorandum of understanding established between Helen Keller International and each institution, and study ethical approval granted by John Hopkins Hospital Institutional Review Board. All eligible patients provided written informed consent to undergo DR screening as part of standard hospital policy. This study was performed according to Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines. 10 This project had 2 stages. In the first stage, a DR screening service was set up in 2 locations in 2009, a center based in an eye hospital (that included the reading center where the retinal images from the FDH and CEITC were graded by the team) at the CEITC and a screening center at the FDH. In both centers, people with diabetes were invited to undergo fundus photography-based DR screening. A 3-tier workforce was established that included primary and secondary nonmedical photographic graders with ophthalmologists (D.N. and J.A.) as arbitration graders. The education and counseling of the patients and the training of the screening team were coordinated by Helen Keller International with local health care professionals. Treatment pathways included laser, intravitreal injection therapy, and vitreoretinal surgical services. A licensed electronic medical record known as OptoMIZE (Emis Health) approved for DR screening was used to capture patient data and fundus photographs and to manage the patient workflow and triage for graders and ophthalmologists.
The second stage of the DR screening project involved a scaling up of the DR screening services in Dhaka, the capital city of Bangladesh. This second stage began in 2014 and was based in the Government of Bangladesh NIO and Hospital, which is the largest government tertiary eye care center in Bangladesh.
The screening locations all had a patient waiting area, a photography area equipped with a fundus camera, a grading area, and a dedicated space for counseling and health education. These clinics were set up within the retinal outpatient zone of each hospital.

Procedures
The DR screening clinics offered a "walk-in" system whereby patients with diabetes were referred from the eye clinics on the same day within the FDH, NIO, and CEITC, and patients were directed to the DR screening clinics. Two standard fundus photographs with dilated pupils-fovea centered and disc centered-were taken of each eye using a table-mounted fundus camera (Topcon TRC-NW8; Topcon Corp). These photographs were graded by primary and secondary graders according to a standard retinopathy (R) and maculopathy (M) grading classification. 11 The images were transmitted electronically to a grading area according to the primary screening location. The outcomes of DR

Training of Graders
We developed a 3-tier grading team that involved a nonmedical workforce of nurses, optometrists, and paramedics for primary and secondary fundus photograph grading. Only medical retina specialist ophthalmologists performed arbitration grading. This grading model was based on the National Health Service Diabetic Eye Screening Programme (NHS DESP) used in England. 11 Graders were trained through workshops and on-site practical sessions using standardized materials from the NHS DESP and Moorfields Eye Hospital. The staff responsible for undertaking DR screening and grading included ophthalmic nurses, paramedics, and nurses and ophthalmologists. All staff underwent appropriate training and local accreditation by a retina specialist from Moorfields Eye Hospital that included the following core and mandatory training modules: (1) national screening program: principles, processes, and protocols; (2) diabetes and its relevance to retinopathy screening including the anatomy, physiology, and pathologic characteristics of the eye and its clinical relevance; and (3) detecting retinal disease and classifying DR.

Quality Assurance Standards and Fail-safe
During photographic grading by primary graders, there was an automated internal quality assurance system whereby 10% of normal R0M0 photographs graded by the primary grader were automatically sent for secondary grading. A monthly online tool was developed for all DR screening graders to be assessed using an international version of the test and training module, as is conducted in the UK NHS DESP. A retina specialist from Moorfields Eye Hospital conducted external quality assurance annually during each project. Internal quality assurance was conducted by the medical retina specialist at each center, and individualized training was provided to improve education and training for DR grading. As part of the program, the medical retina specialists received training on the latest laser photocoagulation guidelines by a retina specialist from Moorfields Eye Hospital. During the program, a fail-safe officer from Helen Keller International monitored the technical, screening, and counseling activities to ensure optimal delivery. To reduce patient risk, all patients with positive screening results who were logged as requiring eye treatment were monitored as part of internal quality assurance. This method also served to identify any barriers or delays in treatment.

Outcomes
The primary objective of this study is to describe the grades of DR detected from fundus photograph-based eye screening for patients with diabetes in Bangladesh. The secondary objectives were to explore the presenting visual acuity (VA) levels in the population with DR across the 3 regions in Bangladesh and their association with vision loss. Patient data were evaluated at quarterly intervals for the duration of the project to ensure that only new cases of DR were captured for this study. Treatment outcomes for laser photocoagulation, intravitreal anti-vascular endothelial growth factor injections (intravitreal injections), and vitrectomy were validated against hospital logbooks to ensure that the first treatment outcome was captured. The World Health Organization classification for visual impairment comprises categories 0 to 5; blindness comprises categories 3 to 5. Mild or no visual impairment (level 0) is indicated by a VA 6/6 to 6/18, moderate visual impairment (level 1) is indicated by a VA greater than 6/18 to 6/60, severe visual impairment (level 2) is indicated by a VA greater than 6/60 to 3/60, and blindness (levels 3, 4, and 5) is indicated by a VA of greater than 3/60 to no light perception. 12

Statistical Analysis
The data set was analyzed from  Figure 1 for examples of DR grades). The Table presents the age, sex, and worst grade of retinopathy and maculopathy for the screened population. The combined prevalence of pre-PDR was 3.9%, the combined prevalence of PDR across all 3 sites was 7.8%, and the combined prevalence of maculopathy was 19.2% (Figure 2). The (Figure 4).
There was a statistically significant association between VA and DR grade as determined by

Discussion
We screened for DR using fundus photography in 2 eye centers (CEITC and NIO) and 1 diabetic hospital (FDH) and found a high prevalence of DR, with substantially higher rates in the eye centers.
Male patients were at higher risk of prevalent DR than female patients across all age groups in the 3 centers, and DR was associated with older age, consistent with previous reports. [13][14][15] Considering the large gap in care in Bangladesh, we focused primarily on the government system, which provides the most effective avenue to reach low-income patients in Bangladesh because government facilities are most widely used by low-income people. Bangladesh has one of the highest incidence rates of diabetes globally. 6,9,10 We report a higher DR prevalence in the eye hospital clinics than in diabetic hospital clinics in Bangladesh. This finding may be associated with the higher affluence of urbanization in Bangladesh and the associated change in dietary habits that has increased the risk of diabetes, or it could be associated with the fact that those who present directly to eye hospitals have a higher likelihood of having eye problems. The prevalence of DR in the Bangladesh populations studied is similar to rates of DR in upper-middle-income countries and highincome countries such as the United Kingdom, the United States, and Singapore. [16][17][18] The rates of maculopathy in Bangladesh were similar to the rates of diabetic macular edema in the United States. 19 The similarity in high prevalence rates of DR and maculopathy between Bangladesh and upper-middle-income countries and high-income countries may be associated with the high incidence of diabetes and changing lifestyles in a transforming context, the higher incidence of childhood obesity, physical inactivity, and urbanization. 20,21 The prevalence of blindness in Bangladesh was estimated in 2003 to be 1.53% (95% CI, 1.31%-1.75%) in individuals 30 years or older. 22 The main cause of blindness was cataracts (80%), with a blindness rate of 1.4% in the CEITC and 1.1% in the NIO. The report estimated that 25% of patients with diabetes had DR, although DR was not associated with blindness or low vision. 22 Nearly one-third of older adults have diabetes in Bangladesh. Our DR screening centers screened male and female patients in considerable numbers from both urban and rural areas, and the age groups were distributed across a broader range from 10 to 116 years. The rates of blindness in our study were more than twice those estimated in 2003, but the exclusive focus on those with diabetes likely explains this difference.
For DR screening to be sustainable, either government funding or cost recovery is needed. In our programs, an important reason for engagement of patients in the DR screening centers was the use of a partial cost recovery system. The model piloted in this program-namely, housing the DR screening in eye hospitals-reduces costs and increases the rate of identification of DR. The Bangladesh DR screening program uses an international-standard grading center with DR software, and the grading center reviews fundus photographs electronically transmitted from 3 screening centers. While patients wait, photographs are remotely reviewed by a team of graders, and recommendations for follow-up are relayed, enabling screening center staff to immediately advise on treatment plans, which reduces the risk that patients will not receive the information or treatment

Limitations
We report a number of study limitations. Snellen VA was recorded and was converted into logMAR VA for analysis. The accuracy of an ungradable image diagnosis was not complete, and the documentation of cataracts as ocular comorbidity and underlying DR cause was suboptimal.
However, missing DR grading outcomes were minimized because the screening DR grades were extracted from the electronic patient record and externally validated by a second data spreadsheet.
In this project, we measured the key performance indicator of a first treatment episode (laser, intravitreal injection, or vitrectomy) after a positive retinopathy and maculopathy screening episode.
Our screening programs did not follow up with the longer-term outcomes for patients after they received the primary treatment because this clinical treatment audit would remain under the auspices of the treating hospital eye center.

Conclusions
There are several policy implications of our study for Bangladesh with respect to current practice for DR screening and prevalence of diabetes in the country. Diabetic retinopathy is recognized as a health priority in the government of Bangladesh's National Eye Care Plan, overseen by the NIO. 23 Diabetic retinopathy screening can be implemented successfully in Bangladesh; expanding such programs could prevent many cases of unnecessary blindness. The Bangladeshi government should identify and support centers that carry out this work. Our DR screening programs have provided strong evidence for future consideration of a national model for DR screening, grading, treatment, and prevention integrated into the government of Bangladesh's National Eye Care Plan on a larger scale across the whole country.
The merits of scaling up the DR screening model include the fact that grading can be performed remotely and 1 site with expertise can be used by multiple locations. The costs of retinal photography have decreased, with many lower-cost devices now available. The major challenge is identifying committed organizations to take ownership of the expanded program and ensuring that those who require care will have coverage to allow for treatment. Identifying the best economic model to ensure success remains a challenge. The government of Bangladesh's National Eye Care Plan is currently under review by a government-appointed committee, with Helen Keller International as a member of the task force.