Developing and Testing a Low Cost Opportunistic Glaucoma Screening Model by Non-physician Graders in Vietnam

Background: According to the International Diabetes Federation (IDF), the prevalence of diabetes will increase by 51% from 463 million in 2019 to 743 million in 2045 among adults between the ages of 20-79 years. This increase is projected to be 15% in Europe, 33% in North America, 74% in South East Asia and 143% in Africa from 2019 to 2045. About 3 in 5 participants who have diabetes in Africa remain undiagnosed. Whereas the total diabetes related health expenditure is 494 billion USD in high income countries, and 264 billion USD in middle income countries, low income countries expend only 1 billion USD. Therefore, diabetes is a huge problem worldwide especially in low income countries. Diabetic retinopathy is the most common microvascular complication of diabetes which leads to blindness if left untreated. Diabetic retinopathy is one of the leading causes of blindness especially among the working group when vision is very important for daily living. Diabetic retinopathy is identified in a third of diabetic patients. Diabetic retinopathy (DR) however can be treated effectively but early detection and timely referrals are essential for successful outcomes. Early diagnosis, timely referrals and access to treatment through a comprehensive and systematic surveillance of the diabetic population has been shown to be clinically effective and cost effective. The United Kingdom (UK) has a national population based diabetic retinopathy screening programme which has been very successful. These screening programmes are non-existent or inadequately developed in many resource constrained settings. A study was recently concluded in Vietnam in which non-physician graders were trained to screen for diabetic retinopathy among the population of diabetic patients. The study also assessed the accuracy of DR grading among medical graders versus non-physician graders in Vietnam. The aim of this current study is to develop and test a low cost opportunistic glaucoma screening model during diabetic retinopathy screening using the same non physician graders currently used in the recently concluded diabetic retinopathy (DR) study. Introduction Glaucoma is a progressive optic neuropathy with characteristic optic nerve changes and functional visual field defects which in part is related to a raised intraocular pressure (IOP). The disease is characterized by typical pattern of damage to the optic nerve, loss of retinal ganglion cells, and thinning of the retinal nerve fiber layer. Glaucoma can be divided into two main types based on the morphology of the anterior chamber angle: there is the open angle and the angle closure glaucoma. These two types of glaucoma have different risk factors, different natural histories and different management. There are other forms of classification which include primary and secondary glaucomas depending on the etiology of the glaucoma. Glaucoma is the leading cause of irreversible blindness globally. An estimated 60.5 million persons were said to be affected by glaucoma in 2010. Glaucoma is a huge problem worldwide especially in developing countries. The disease however can be managed with medical, laser, and surgical forms of treatment. Reduction of intraocular pressure by 30-50% from baseline usually stops progression of the disease. It has been reported that the increasing life expectancy especially of Asians and Africans may lead to an increase in the prevalence of glaucoma by 74% from 2013 to 2040. It is projected that the prevalence of glaucoma will increase from 64.3 million to 111.8 million in 2040, and that glaucoma will disproportionately affect people residing in Africa and Asia. Glaucoma often progresses unnoticed by the patient until central visual acuity is affected, therefore early detection of the disease is important before symptoms develop. The rate of undiagnosed disease however is quite high in developing countries with a large proportion not being aware of the disease. It is estimated that about 95% are not aware of their glaucoma in developing countries while 50% are unaware of the disease in developed countries. Early detection of glaucoma requires some form of screening. Screening of any disease is a form of secondary prevention and the goal is often to prevent undesired outcomes. In the context of glaucoma, screening helps to diagnose it in the asymptomatic stage so that treatment can be started earlier to slow down or prevent progression of the disease to blindness. In a recent study on the burden of undetected untreated glaucoma in the United States, it was reported that the odds of having undiagnosed glaucoma was highest among blacks with a 4.4 times greater odds compared with Caucasians. This disparity was attributed to uneven access to health care by blacks. One way to alleviate this is to conduct population based screening for glaucoma. But this has not been found to be cost effective in developed countries. In a very recent study published in the Lancet Global Health, however, population screening was reported to be cost effective in China. In a study on the cost effectiveness of community screening for glaucoma in rural India, it was reported that community screening would prevent 2,190 person years blindness over a 10 year period. It was concluded that community population screening may be cost effective if targeted at 40-69 years' age group and if implemented in urban areas. A study in China recently demonstrated using a decision analytic Markov model that population screening for both primary open angle glaucoma (POAG) and primary angle closure glaucoma (PACG) was cost effective in both rural and urban China. It is possible that screening of two major blinding eye diseases (glaucoma and diabetic retinopathy) using the same facilities is likely to be more cost effective. Opportunistic glaucoma screening can be more cost effective especially when it is piggy bagged into another screening programme. Diabetic retinopathy screening has been ongoing with the use of optic nerve photos. It is possible to train the same non- physician graders who are currently screening for diabetic retinopathy to also screen for glaucoma using the same facilities and images. The use of non-physician graders to conduct screening in low-resource settings is particularly important as these are the group of health workers who have been trained to screen for diabetes. In addition, it reduces the burden on scarce, highly-trained health care providers in developing countries. Study Aims: * To develop a training course for screening glaucoma using fundus images obtained during diabetic retinopathy screening by non-physician graders. * To test this training course among non-physician graders in Vietnam Main research question Can trained non-physician graders accurately screen for glaucoma during diabetic retinopathy screening? Methods Study design: This is an uncontrolled interventional experimental before and after study in which a minimum of 42 non-physician graders shall be trained to screen for glaucoma using optic nerve photos obtained for diabetic retinopathy (DR) screening in Vietnam. The study flow will be as follows: * Non-physician graders are consented, enrolled and undergo testing on the standard image set at baseline * The non-physician graders complete the training course. There are questions at the end of each module of the course and participants must pass the revision questions. Those failing to pass any aspects of the revision questions must review those parts of the course. * Non-physician graders will undergo a post course test on the standard image set. Their performance will be compared to baseline and also to that of ophthalmologists who are taking part in the DR grading programme, but who have not taken the optic nerve grading course • Non-physician graders will undergo a post course test on the standard image set. Their performance will be compared to baseline and also to that of ophthalmologists who are taking part in the DR grading programme, but who have not taken the optic nerve grading course Standard Image training set and Test Set: These will each consist of about 50 normal optic nerve images of people without glaucoma and about 50 images of people with glaucoma, obtained from the on-going ORBIS CAFÉ DR screening programme (in which the graders are working), population-based eye studies in the UK (NICOLA) and the standard Glaucomatous Optic Neuropathy Evaluation (GONE) set of images. (https://gone-project.com/newgone/) Training and evaluating competencies: Training shall be done online, using the Cyber-sight website of Orbis, The course has been created specifically for the study by the investigators, modified from materials used to teach physicians on the World Glaucoma Association and GONE websites and training courses for ophthalmology residents. A minimum threshold of 70% accuracy on a previously-designed test set will be considered as adequate competency. Participants who score less than 70% on the test will have a teacher led training course after which the participants will repeat the test again. Recruitment: Vietnam non-physician and ophthalmologist DR image graders: These persons are already trained to screen and grade diabetic retinopathy and will be recruited by asking them about their willingness to take part in the study. Informed consent will be obtained, and participants will be allowed to participate voluntarily .and if they told that if they choose not to participate in the study, their jobs will not be affected in anyway. A minimum of 42 non-physician and 12 ophthalmologist graders shall be included in the study Planned statistical analysis of results: Statistical Sample Size \& Methods: The minimum sample size of 42 participants was calculated using the paired sample size formula for quantitative outcome data with confidence level of 95% and statistical power of 80%. Using the statistics from the Glaucoma Optic Neuropathy Evaluation study (GONE), we assumed that the mean baseline score would be the same as the 5th percentile score on the GONE evaluation course which is 29%. At the end of the course, we assumed that the mean score would be the same as the 45th -50th percentile score on the GONE evaluation course which is 60%. Briefly the GONE online course was designed to evaluate participant's skills in evaluating the optic nerve head for glaucoma. It was started in 2015 and as of October 2021, there had been 45,619 attempts by 11,287 individuals. However, when only first attempts and those who had completed at least 80% of the course were considered, this number reduced to 3666 participants. This course consists of 15 disc images in all. Sensitivity and specificity (with 95% confidence Intervals (CI's)), positive and negative predictive values, area under the receiver operating characteristic (ROC) curve and kappa statistic comparing non-physician and ophthalmologist graders in Vietnam with the grades on the standard image sets shall be performed. The performance of non-physician graders after training will be compared to that before training, and also to that of ophthalmologists who have not taken the optic nerve course but are working as DR graders in the CAFÉ programme, with the main outcome for comparison being the area under the receiver-operator curve. Data Protection issues Participants from whom optic nerve images were obtained shall not be formally enrolled into this study. Only their de-identified optic nerve images will be used in this study and analysis of images shall be done in a de-identified fashion. Although optic nerve images are considered personal data, all images are fully anonymized and de-identified. Blanket permission for research use of images was obtained in the NICOLA, HANDAN and the GONE studies. These studies have full ethical approvals and obtained consent from participants to use the images in future studies. All electronic copies of data shall be stored on password-protected computers with written log books and password protected access to buildings. All paper copies of data will be locked in a specific filing cabinets in QUB. All optic nerve photos will be anonymized, completely de-identified and kept in a safe, password-protected server at QUB. All data produced from this study shall be completely anonymized.