The purpose of this study is to evaluate the performance of a deep learning algorithm for retinopathy of prematurity (ROP) screening in Nepal and Mongolia.
This was a retrospective analysis of prospectively collected clinical data.
Clinical information and fundus images were obtained from infants in two ROP screening programs in Nepal and Mongolia.
Fundus images were obtained using the Forus 3nethra neo in Nepal and RetCam® Portable in Mongolia. The overall severity of ROP was determined from the medical record using the International Classification of ROP (ICROP). The presence of plus disease was independently determined in each image using a reference standard diagnosis. The Imaging and Informatics for ROP (i-ROP) deep learning (DL) algorithm, which was trained on images from the RetCam® was used to classify plus disease, as well as assign a vascular severity score (VSS) from 1-9.
Main outcome measures
The main outcome measures were area under the receiver operating characteristic (AUC-ROC) and area under the precision recall curve (AUC-PR) for the presence of plus disease or type 1 ROP, and association between VSS and ICROP disease category.
The prevalence of type 1 ROP was found to be higher in Mongolia (14.0%) than in Nepal (2.2%, p < 0.001) in these data sets. In Mongolia (Retcam images), the AUC-ROC for exam-level plus disease detection was 0.968 and AUC-PR was 0.823. In Nepal (Forus images), the AUC-ROC for exam-level plus disease detection was 0.999 and AUC-PR was 0.993. The ROP vascular severity score was associated with ICROP classification in both datasets (p < 0.001). At the population level, the median [interquartile range] VSS was found to be higher in Mongolia (2.7 [1.3–5.4]) as compared to Nepal (1.9 [1.2–3.4], p < 0.001).
These data provide preliminary evidence of the effectiveness of the i-ROP DL algorithm for ROP screening in neonatal populations in Nepal and Mongolia, using multiple camera systems, and provide useful data for consideration in future clinical implementation of AI-based ROP screening in low- and middle-income countries.
More than 2 billion people are thought to be living with some form of vision impairment worldwide. Yet relatively little is known about the wider impacts of vision loss on individual health and well-being, particularly in low- and middle-income countries (LMICs). This study estimated the associations between all-cause vision impairment and self-reported symptoms of anxiety and depression among older adults in Kogi State, Nigeria.
Individual eyes were examined according to the standard Rapid Assessment of Avoidable Blindness methodology, and anxiety and depression were assessed using the Washington Group Short Set on Functioning–Enhanced. The associations were estimated using multivariable logistic regression models, adding two- and three-way interaction terms to test whether these differed for gender subgroups and with age.
Overall, symptoms of either anxiety or depression, or both, were worse among people with severe visual impairment or blindness compared with those with no impairment (OR=2.72, 95% CI 1.86 to 3.99). Higher levels of anxiety and/or depression were observed among men with severe visual impairment and blindness compared with women, and this gender gap appeared to widen as people got older.
These findings suggest a substantial mental health burden among people with vision impairment in LMICs, particularly older men, underscoring the importance of targeted policies and programmes addressing the preventable causes of vision impairment and blindness.
The number of patients with visual impairment and blindness from glaucoma is rapidly increasing with wide-ranging impacts for individuals and societies. However, the disease often goes undiagnosed for a long time, especially in low- and middle-income countries where healthcare services are limited. This paper presents the results of a pilot programme, which integrated targeted glaucoma screenings of people aged ≥40 y in community-based eye care services in the Ganjam district of Odisha state, India.
Using routine programme data, descriptive statistics were produced for the characteristics of patients participating in the screening programme and the rate and uptake of glaucoma referrals. Bivariate analysis was used to examine associations between patient characteristics, clinical risk factors and glaucoma diagnosis.
Out of 23 356 individuals aged ≥40 y screened for glaucoma over a period of 18 mo, 2219 (9.5%) were referred and 2031 presented for further examination. Among them, almost half (n=968, 48%) were diagnosed with glaucoma, representing a screening to diagnosis conversion rate of 4.14% (95% CI 3.9 to 4.4%). A positive diagnosis of glaucoma among suspects was associated with female sex, age >60 y, visual impairment, vertical cap-to-disc ratio ≥0.6:1, intraocular pressure ≥30 mmHg and shallow anterior chamber (p<0.001).
The importance of targeted screening for glaucoma using simple referral criteria to identify patients at high risk of vision loss who can benefit from treatment is critical to slow the progression of the disease and the prevention of blindness. Further studies assessing costs of the targeted screening, the role of technology in improving programme effectiveness and efficiency and the longer term compliance with treatment are needed to support glaucoma policy frameworks, guidelines and clinical practice.
Background/aims Quantity of cataract surgery has long been an important public health indicator to assess health accessibility, however the quality of care has been less investigated. We aimed to summarise the up-to-date evidences to assess the real-world visual outcomes after cataract surgery in different settings.
Methods A systematic review was undertaken in October 2021. Population-based cross-sectional and longitudinal studies reporting vision-related outcomes after cataract surgery published from 2006 onward were included. A meta-analysis was not planned.
Results Twenty-six cross-sectional studies from low-income and middle-income countries (LMICs) and five cross-sectional studies from high-income countries (HICs) were included. The proportions of participants with postoperative presenting visual acuity (VA) ≥0.32 (20/60) were all over 70% in all HICS studies, but mostly below 70% in LMICS studies, ranging from 29.9% to 80.5%. Significant difference in postoperative VA was also observed within countries. The leading causes for postoperative visual impairment (defined mostly as presenting VA <20/60) mainly included refractive error, ocular comorbidities and surgical complications including posterior capsule opacification, except for one study in Nigeria wherein the leading cause was aphakia. Only four population-based cohort studies were included with 5–20 years of follow-up time, generally demonstrating no significant changes in postoperative visual outcomes during the follow-up.
Conclusions We observed large inequality in the visual outcomes and principal causes of visual impairment after cataract surgery among different countries and regions. Structured quality control and enhancement programmes are needed to improve the outcomes of cataract surgery and reduce inequality.
In the past 3 y, considerable attention has been paid to the importance of eye health as a global development priority, first by the World Report on Vision launched by the World Health Organization (WHO) in October 2019,1 then by the Lancet Global Health Commission on Global Eye Health, published in February 20212 and most recently by the Vision for Everyone Resolution adopted by the United Nations (UN) General Assembly in July 2021.3 All three documents provide compelling evidence on the magnitude, impact and costs of poor vision and call for coordinated global efforts to improve access to and the quality and equity of eye care services worldwide. The documents argue that the progress towards the Sustainable Development Goals (SDGs)4 will not be sufficient without integrating eye health within universal health coverage (UHC)5 and highlight the role of quality evidence and research in achieving this ambitious global agenda. The Lancet Global Health Commission gives particular attention to contextually relevant solution-focused implementation research and to translating research findings into policy and practice.2
International non-governmental organisations (iNGOs) have been at the forefront of the global efforts to eliminate avoidable vision impairment and improve access to eye care services in low- and middle-income countries (LMICs) for several decades. Their role in mobilising resources, service delivery, building local capacity and advocacy has been well recognised. In recent years, a number of iNGOs have made considerable investments in generating high-quality evidence and have become important players in eye health implementation research.
There are a number of critical strengths that iNGOs bring to international research. First, iNGOs have a good understanding of service delivery realities and frontline knowledge gaps and can help identify research questions that fit local priorities and meet the needs of local governments, healthcare providers and communities. Second, integrating implementation research into ongoing programmatic activities is both practical and cost-effective. Third, iNGOs are closely engaged with national and local decision-making processes and are well positioned to facilitate the uptake and use of research findings.
This special issue clearly illustrates how iNGOs can contribute to generating crucial evidence to answer operational questions they face and to advance global knowledge on eye health. We are delighted to introduce a collection of articles that present evidence from secondary data analysis and implementation research integrated into eye health programmes supported by the iNGO Sightsavers in sub-Saharan Africa and Asia. This is the first of three supplements to be led by Sightsavers in 2022 and 2023. The two later supplements will focus specifically on evidence from our neglected tropical diseases (NTDs) programmes. In this supplement, we present research findings on a range of eye conditions, including cataracts, unaddressed refractive error (URE), glaucoma, trachoma and onchocerciasis. The articles draw on evidence from 12 countries and address a number of topics pertinent to eye health policies and services in LMICs. We are particularly pleased that several articles are based on a secondary analysis of our programmatic data and have been authored by the technical experts and managers supporting our programmes in the field.
The supplement opens with a guest editorial by Keel and Cieza of the WHO, who present the recent UN developments in support of eye health and highlight global priorities for action to facilitate the journey of member states towards universal eye health. The key messages of the guest editorial are echoed in the commentary by Jones, who discusses the state of eye health funding in LMICs and calls for a step change in delivering more and better financing for eye health within UHC.
The substantial impact of vision on other global health priorities is highlighted in the article by Gascoyne et al., who included disability metrics in a visual impairment survey in Kogi State in Nigeria and present new evidence on the intersection of vision impairment and mental health.
In recognition of the multiple health priorities facing populations in LMICs and the finite resources to address them, a number of articles explore questions on how to make the right policy choices and maximise the effectiveness and efficiency of eye care delivery. Hamill et al. report on the results of a pilot study assessing the feasibility and affordability of mapping areas with medium to low transmission of onchocerciasis in Ghana and Nigeria. They argue that as more and more countries approach onchocerciasis elimination thresholds, questions regarding the cost-effectiveness of treatment in hypo-endemic areas become more and more pressing and the role of implementation research generating such evidence should not be underestimated.
Another article in the series addresses questions of efficiency and equity of services in the context of trachoma elimination. Ul Hassan et al. use programmatic data from seven trachoma-endemic countries (Ethiopia, Kenya, Mozambique, Zambia, Tanzania, Uganda and Nigeria) and compare the effectiveness of different community mobilisation and case-finding approaches with a specific focus on gender. The article stresses the importance of routine multicountry service data in developing recommendations and guidance relevant to a variety of settings.
In India, Buttan et al. also analysed programmatic data from a pilot study that integrated targeted glaucoma screening into existing eye care services and assessed the feasibility and benefits of this approach. Glaucoma is a rapidly emerging global eye health priority that is very challenging for resource-poor settings and operational data on how to address it in a pragmatic and cost-effective way are lacking. Integrating implementation research into ongoing eye care programmes provides an excellent opportunity to build the evidence base so urgently needed by the sector.
Another global priority addressed in this supplement is developing an eye health workforce that is able to meet the growing needs for eye care globally. Two articles focus on task shifting, a strategy frequently used in eye health to address the constant shortages of eye care specialists in LMICs. In a commentary on primary eye care, Yasmin and Schmidt review lessons learned from eye care programmes in Tanzania, Sierra Leone and Pakistan. They argue that while the integration of certain eye care tasks into the roles of primary health and community workers is feasible, a system thinking approach is critical to ensure it is done in a coherent and sustainable way. Further, in a study in Liberia, Tobi et al. show that teachers can be trained to conduct vision screening of schoolchildren to an acceptable level of accuracy. However, based on the data generated by the programme, the authors warn that careful considerations should be given to teachers’ supervision and quality assurance systems.
Two further articles focus on patient-centred care, the theme that features centrally in the World Report on Vision1 and the new eye health agenda. Shrestha et al. explore characteristics and perspectives of patients with postoperative trichiasis in the Hadiya Zone of Ethiopia, while Bechange et al. report results of a multicountry qualitative study considering patients’ perspectives on cataract surgery in rural areas of Kenya, Zambia and Uganda. Both articles argue that patients’ perceptions of quality of care are important determinants of their health-seeking behaviour and need to be better understood to maximise the effectiveness and efficiency of surgical outreach.
Two remaining articles address the role of evidence and innovations in eye health. A commentary by Bartlett et al. describes Sighsavers’ experience of working with the Federal Ministry of Health of Nigeria on strengthening national electronic medical records systems and specifically the adaptation of the electronic data collection tool known as the Trachomatis Trichiasis (TT) Tracker commonly used in trachoma programmes to the needs of paediatric cataract services. Finally, Jolley et al. use the evidence gap map approach and present the current state of evidence from cataract-related systematic reviews relevant to LMICs. The authors point to the shortage of evidence on cataract-related health systems, equity and impact—the gaps also highlighted by the World Report on Vision and the Lancet Commission.
We would like to thank all the authors for the time they spent preparing the manuscripts and for sharing their valuable experiences and findings in an open and transparent way. Many thanks to our reviewers for their time and constructive comments and advice.
We are preparing this supplement at the time of the unfolding coronavirus disease 2019 pandemic. While the scale and impact of this unprecedented global crisis is yet to be fully understood, the increasing importance of evidence in decision-making is indisputable and the role of implementation research is becoming more prominent than ever before. We hope you enjoy reading these articles and find them useful in your own programmatic decisions and policy choices.
Evidence indicates that school-based vision screening by trained teachers is an effective way of identifying and addressing potential vision problems in schoolchildren. However, inconsistencies have been reported in both the testing methods and accuracy of the screeners. This study assessed the prevalence of refractive errors and accuracy of screening by teachers in Grand Kru County, Liberia.
We conducted a retrospective analysis of data from four schools where, in February 2019, children were screened for refractive errors by trained teachers and then re-examined by ophthalmic technicians. One row of five optotypes of the Snellen 6/9 (0.2 logMar) scale (tumbling E chart) was used at a distance of 3 m. The prevalence of visual impairment and associations with sex, age and school were explored. Sensitivity, specificity and predictive values were calculated.
Data were available for 823 of 1095 eligible children with a mean age of 13.7 y (range 5–18) and male:female ratio of 1:0.8. Poor vision was identified in 24 (2.9%) children with no differences by either sex or age but small differences by school. Screening by teachers had a sensitivity of 0.25 (95% confidence interval [CI] 0.077 to 0.423) and a specificity of 0.996 (95% CI 0.992 to 1.000). Positive and negative predictive values were 0.667 (95% CI 0.359 to 0.975) and 0.978 (95% CI 0.968 to 0.988), respectively. The results were influenced by a high number of misclassifications in one of the four schools.
Teachers can be trained to conduct vision screening tests on schoolchildren to an acceptable level of accuracy, but strong monitoring and quality assurance systems should be built into screening programmes from the onset. In settings like Liberia, where many children do not attend school regularly, screening programmes should extend to community platforms to reach children out of school.
In 2014, Sightsavers developed the first evidence gap map (EGM) to assess the extent and quality of review-level evidence on cataract relevant to low-and middle-income countries. The EGM identified 52 studies across five broad themes. This paper reports the update of the EGM conducted in 2021 and changes to the extent and quality of the evidence base. We updated the EGM using the exact process conducted to develop the original. Searches were run to 14 September 2021, and two independent reviewers selected eligible studies, critically appraised them and extracted data using the Supporting the Use of Research Evidence checklist. A summary quality assessment was shared with the authors for comments. Forty-six new reviews were identified, and the EGM now includes 98 reviews. The new reviews predominantly focus on treatment and risk factors. The overall methodological quality was found to be improved, with 13/46 reporting high confidence in findings. EGMs remain a useful tool for policy-makers to make informed decisions and periodic updates are important to assess changes and to refine the focus for future research. The EGM highlights significant disparity in the topics addressed by reviews, with health system interventions particularly neglected.
Background. The majority of patients with retinoblastoma, the most common intraocular cancer of childhood, are found in low-and middle-income countries (LMICs), with leukocoria being the most common initial presenting sign and indication for referral. Findings from the current study serve to augment earlier findings on the clinical presentation and outcomes of children with retinoblastoma in Uganda. Methods. This was a retrospective study in which we reviewed records of children admitted with a diagnosis of retinoblastoma at the Uganda Cancer Institute from January 2009 to February 2020. From the electronic database, using admission numbers, files were retrieved. Patient information was recorded in a data extraction tool. Results. A total of 90 retinoblastoma patients were studied, with a mean age at the first Uganda Cancer Institute (UCI) presentation of 36.7 months. There were more males (57.8%) than females, with a male to female ratio of 1.37 : 1. The majority (54.4%) had retinoblastoma treatment prior to UCI admission. The most common presenting symptoms were leukocoria (85.6%), eye reddening (64.4%), and eye swelling (63.3%). At 3 years of follow-up after index admission at UCI, 36.7% of the patients had died, 41.1% were alive, and 22.2% had been lost to follow-up. The median 3-year survival for children with retinoblastoma in our study was 2.18 years. Significant predictors of survival in the multivariate analysis were follow-up duration (), features of metastatic spread (), history of eye swelling (), and bilateral enucleation (). Conclusions. The majority of children who presented to the Uganda Cancer Institute were referred with advanced retinoblastoma, and there was a high mortality rate. Retinoblastoma management requires a multidisciplinary team that should include paediatric ophthalmologists, paediatric oncologists, ocular oncologists, radiation oncologists, and nurses.
Diabetic retinopathy is a leading cause of preventable blindness, especially in low-income and middle-income countries (LMICs). Deep-learning systems have the potential to enhance diabetic retinopathy screenings in these settings, yet prospective studies assessing their usability and performance are scarce.
We did a prospective interventional cohort study to evaluate the real-world performance and feasibility of deploying a deep-learning system into the health-care system of Thailand. Patients with diabetes and listed on the national diabetes registry, aged 18 years or older, able to have their fundus photograph taken for at least one eye, and due for screening as per the Thai Ministry of Public Health guidelines were eligible for inclusion. Eligible patients were screened with the deep-learning system at nine primary care sites under Thailand’s national diabetic retinopathy screening programme. Patients with a previous diagnosis of diabetic macular oedema, severe non-proliferative diabetic retinopathy, or proliferative diabetic retinopathy; previous laser treatment of the retina or retinal surgery; other non-diabetic retinopathy eye disease requiring referral to an ophthalmologist; or inability to have fundus photograph taken of both eyes for any reason were excluded. Deep-learning system-based interpretations of patient fundus images and referral recommendations were provided in real time. As a safety mechanism, regional retina specialists over-read each image. Performance of the deep-learning system (accuracy, sensitivity, specificity, positive predictive value [PPV], and negative predictive value [NPV]) were measured against an adjudicated reference standard, provided by fellowship-trained retina specialists. This study is registered with the Thai national clinical trials registry, TCRT20190902002.
Between Dec 12, 2018, and March 29, 2020, 7940 patients were screened for inclusion. 7651 (96·3%) patients were eligible for study analysis, and 2412 (31·5%) patients were referred for diabetic retinopathy, diabetic macular oedema, ungradable images, or low visual acuity. For vision-threatening diabetic retinopathy, the deep-learning system had an accuracy of 94·7% (95% CI 93·0–96·2), sensitivity of 91·4% (87·1–95·0), and specificity of 95·4% (94·1–96·7). The retina specialist over-readers had an accuracy of 93·5 (91·7–95·0; p=0·17), a sensitivity of 84·8% (79·4–90·0; p=0·024), and specificity of 95·5% (94·1–96·7; p=0·98). The PPV for the deep-learning system was 79·2 (95% CI 73·8–84·3) compared with 75·6 (69·8–81·1) for the over-readers. The NPV for the deep-learning system was 95·5 (92·8–97·9) compared with 92·4 (89·3–95·5) for the over-readers.
A deep-learning system can deliver real-time diabetic retinopathy detection capability similar to retina specialists in community-based screening settings. Socioenvironmental factors and workflows must be taken into consideration when implementing a deep-learning system within a large-scale screening programme in LMICs.
This paper is based on qualitative research carried out in a diabetic retinopathy (DR) programme in three districts of Pakistan. It analyses the organisation and delivery of DR services and the extent to which the interventions resulted in a fully functioning integrated approach to DR care and treatment. Between January and April 2019, we conducted 14 focus group discussions and 37 in-depth interviews with 144 purposively selected participants: patients, lady health workers (LHWs) and health professionals. Findings suggest that integration of services was helpful in the prevention and management of DR. Through the efforts of LHWs and general practitioners, diabetic patients in the community became aware of the eye health issues related to uncontrolled diabetes. However, a number of systemic pressure points in the continuum of care seem to have limited the impact of the integration. Some components of the intervention, such as a patient tracking system and reinforced interdepartmental links, show great promise and need to be sustained. The results of this study point to the need for action to ensure inclusion of DR on the list of local health departments’ priority conditions, greater provision of closer-to-community services, such as mobile clinics. Future interventions will need to consider the complexity of adding diabetic retinopathy to an already heavy workload for the LHWs.