Introduction

The vision loss expert group (VLEG) reported that diabetic retinopathy (DR) accounted for 1.07% of blindness and 1.25% of moderate to severe visual impairment (MSVI) in 2015 [1]. Despite global efforts to reducing visual impairment, the prevalence of DR is increasing (Crude prevalence 1990: 0.03%; 2015: 0.04%) while the prevalence of all other causes of visual impairment is decreasing consequent to concerted global efforts [1]. This disparity is explained by the estimated global demographic changes between 2019 and 2030–24.8% increase in people with diabetes (International Diabetes Federation, IDF, 2019 estimation: 463 million; 2030 projection: 578 million) [2], 10.7% increase in population (2019 estimate: 7.71 billion; 2030 projection: 8.54 billion) [3], 13.1% longer longevity of people (life expectancy: 2019 estimation: 64.2 years; 2030 projection: 72.6 years) [3], and the increasing aging population of 65+ years (2019 estimate: 9.1%; 2030 projection: 11.7%) [3]. Table 1 lists the key elements of diabetes care in the world [2].

Table 1 Key elements in global status of diabetes [2].

There is a concurrent rise in the worldwide cost of diabetes care; in 2030, the direct cost could be USD 825 billion (2019 estimate: USD 760 billion) [2] and total cost, both direct and indirect, could be USD 2.1–2.5 trillion (2015 cost: USD 1.3 trillion) [4]. However, there is gross inequity on planned expenditure on diabetes and its complications globally (the region with higher prevalence and the low-middle income countries, LMIC, are spending less). As one of the six building blocks of the WHO health system, health finance is an important consideration. In less developed countries, the consideration of the cost of DR care assumes more significance for three reasons: (a) higher resource allocation for cataract surgery and uncorrected refractive error; (b) a significant amount of out-of-pocket spending, and occasionally, catastrophic health expenditure; (c) late presentation of patients with advanced diabetic retinopathy and vision loss.

Progression of DR follows a particular pattern from non-proliferative DR (NPDR) to proliferative DR (PDR). Typically, it takes years for NPDR to convert to sight-threatening diabetic retinopathy (STDR). The STDR includes macular edema and proliferative diabetic retinopathy (PDR). Using retinal photography, a recent review estimated the global burden of DR in people with diabetes mellitus (DM) could be as high as 27% [5]. The estimated quality-adjusted life years (QALY) loss due to visual impairment is also high (−74.93 years per 100,000 person-years in a study from Korea) [6].

The key to reducing visual impairment and blindness for a chronic disease like DR is early detection and treatment of STDR. Screening is a useful tool, and DR meets nearly all criteria laid down by Wilson and Jungner [7]. Screening for DR has also added benefit to identifying people at risk of other diabetes-related complications such as neuropathy, nephropathy, cardiovascular disease (CVD), stroke, and peripheral vascular disease [8].

The regional and national perspectives have been evolving globally to incorporate newer advances and innovations in DR screening. However, there are variations as per the existing public health requirements and available resources. The objective of this review was to identify potential areas for improvement to enable global coverage with DR screening. We reviewed the currently available and recently updated DR screening guidelines across the globe, one from each region of the World Health Organization (WHO)/International Agency for the Prevention of Blindness (IAPB). We compared them with the International Council of Ophthalmology (ICO) guidelines for evaluating the similarities and variations in screening for STDR.

Methods

A search of available electronic databases, including the WHO, ICO, IAPB, VISION 2020 Right to Sight, American Academy of Ophthalmology (AAO) sources, was completed to identify the existing country or professional ophthalmological society- approved DR guidelines for people with type 1 (T1) DM and type 2 (T2) DM. The reference terms were “diabetic retinopathy,” “screening,” “guidelines,” and “practice pattern.” We reviewed 12 guidelines available in the English-language published or updated in the last 5 years and selected one from each region of the IAPB. The publication period was chosen as the previous 5 years to include only those guidelines that have possibly incorporated the recent DR screening updates (Table 2a, b).

Table 2 Literature review of DR and DM screening.

The final section of the guideline and reason for such selection is as follows: Canada (2018; North America-Canada practices systematic tele-screening), England (2017; Europe-England practices national screening), India (2019; South-East Asia-India is home to the second largest population of DM), Kenya (2017; Africa- Kenya has developed a detailed DM/DR protocol), New Zealand (2016; Western Pacific- New Zealand practices systematic national DR screening), and AAO Preferred Practice Pattern (PPP) (2019; guidelines used in Latin America and Eastern Mediterranean countries) [9,10,11,12,13,14]. All these guidelines were compared between them and against the ICO (2017) guidelines (Table 2a, b) [15].

Questions

The ICO guidelines on diabetic eye care provide a framework for developing regionally applicable DR screening guidelines. We reviewed the ICO guidelines and identified questions relevant to DR screening. The answers to these questions are central components for any DR screening program.

  1. 1.

    Classification of DR: Which classification is easy to use and reliable that could be applied with optimal training of human resources?

  2. 2.

    Systemic factors in DR: What are the target parameters associated with DM care that impacts DR outcome?

  3. 3.

    Screening and referral for DR: Why, Who, How, and What?

Results

The following answers to the questions derived from the review of the guidelines are included in this communication.

Definition and classification of DR: what is easy to use and yet reliable?

The earliest classification of DR is the Airlie House classification [16]. The proposed classification, with little modification, was, used in the Diabetic Retinopathy Study (DRS) and, for the first time, the standards of photo-documentation using stereo photographs of 7 standard fields (around the optic disc and macula) were laid [17]. Later, this classification was further modified and used in the Early Treatment of Diabetic Retinopathy Study (ETDRS) [18]. The ETDRS introduced a new term- the clinically significant macular edema (CSMO) [18]. The ETDRS also measured the DR severity scale into 13 levels.

The ETDRS classification became the new gold standard of the DR severity scale; it was suitable for research but suffered from its complexity. In 2003 the Clinical Disease Severity Scale for DR proposed a new classification, the International Classification of Diabetic Retinopathy (ICDR) [19]. There are five categories in DR; diabetic macular edema (DMO), when present, was classified into three categories. (Table 3). Using the optical coherence tomography heatmap, the DMO is also classified into “center involving” and “non-center involving” DMO [20]. All examined/selected guidelines currently follow the ICDR classification, and color fundus photography is the recommended standard for DR screening.

Table 3 International classification of diabetic retinopathy (ICDR) classification [19].

Systemic factors in DR: what are the target parameters?

Control of diabetes mellitus and many associated co-morbidities is necessary for maximum treatment benefit to people with STDR [21,22,23,24,25]. This consists of a variable combination of retinal laser and intravitreal anti-vascular endothelial growth factor (VEGF) injection and vitreoretinal surgery [21].

There are several modifiable systemic factors of DM, but the two risk factors with the most convincing evidence and affordable treatment are hyperglycemia and hypertension. The ACCORD and its follow-up studies provided recent evidence that intensive glycaemic control remains beneficial for reducing DR progression. The legacy effect is evident in people with type 2 DM [22, 24]. The Diabetes Control and Complications Trial (DCCT) evaluated intensive control of hyperglycemia in Type 1 DM, and long term results showed definite benefits in risk reduction of DR [25]. The United Kingdom Prospective Diabetes Study (UKPDS) has shown a decreased incidence of DR with tight control of blood pressure and glucose in patients with Type 2 DM [23]. In the Wisconsin Epidemiologic Study of Diabetic Retinopathy (WESDR), diastolic blood pressure was a significant predictor for DR progression to PDR over 14-year follow-up in people with T1DM [26]. The DM guidelines from different countries of the region [27,28,29,30,31,32,33] have set targets for diabetes, blood pressure, and cholesterol control (Table 4). Other important systemic factors are kidney disease (greater association with T1DM) [34], microalbuminuria, anemia (for retinopathy progression and DMO) [35], and obesity (strong relationship with insulin resistance) [36]. With the recent surge of novel anti-diabetes strategies, the initial worsening of DR should be monitored before the retina begins to stabilize, as observed with insulin therapy initiation [37]. A multi-disciplinary approach and close interaction between the diabetologist and ophthalmologist helps, and housing them together is beneficial.

Table 4 Recommended systemic and lifestyle targets for non-pregnant adults with DM [26,27,28,29,30,31,32].

Screening and referral for DR: Why Who, How, and When

Why should DR screening be established?

The vast majority of patients who develop DR have no symptoms until the late stages due to DMO and PDR complications, such as vitreous hemorrhage or tractional retinal detachment. The late presentation of people with an advanced disease state is a worldwide phenomenon related to a lack of public knowledge, awareness, and social deprivation [38,39,40,41]. Direct medical costs for DR care are substantial, so also the indirect costs of visual impairment with respect to loss of productivity, increasing hospital admissions, and decreased quality of life [42]. From a public health perspective, blindness, and its treatment cost results in poverty at the individual level and retards economic development at the national level [43]. Good health and wellbeing (Sustainable Development Goal, SDG 3) is intimately connected with SDG 1 and 2 (No poverty and Zero hunger) [44]. All guidelines agree to screen for diabetic retinopathy, and DR screening fits all chronic disease screening criteria [45].

Who should perform DR screening?

In the past, countries have relied on ophthalmologists and physicians to screen all people with diabetes. With the shift from ophthalmoscopy to digital retinal photography, technically trained and certified screeners such as optometrists and allied ophthalmic personnel, including trained retinal photographers, are more cost-effective [46]. The reliability of screening by optometrists and/or retinal photographers has reached 91% sensitivity, and 78% specificity [47] against the National Institute for Clinical Excellence (NICE, UK) recommended acceptance level for DR screening at 80% sensitivity, 95% specificity (and clinical failure rate <5%) [48]. In many countries, the existing law does not allow optometrists to dilate pupils without ophthalmologist supervision. This is a barrier, but reading and grading fundus photos obtained in a non-mydriatic camera could overcome this barrier with the current technology.

How often and where should DR screening be done

Most guidelines suggest that people with T2DM are screened first at the time of diagnosis, and people with T1 DM are screened first at puberty or five years after the diagnosis.

The follow-up care depends on the disease severity; it could vary from 1 to 2 years when there is no retinopathy to half-yearly review once the retinopathy is stabilized after adequate treatment. Longer intervals of follow-up have also been suggested after cohort studies on T2 DM [49]. There are different opinions on the time of retinal examination during pregnancy (Table 5). Examining all people with DM at fixed facilities (eye hospital/ophthalmology services) may not always be feasible in many low-middle income countries (LMICs) and in countries/regions that do not have enough workforce and technology resources. In most countries, there is no standardized, systematic nation-wide screening for DR. In the studied guidelines, only England and New Zealand practice a national DR screening policy. Maintaining a national registry of patients with DM will help direct patients with DM for periodic screening efficiently. Annual eye evaluation may not be cost-effective for low-risk patients [50].

Table 5 DR Screening strategy [8,9,10,11,12,13,14,15].

Studies from India, Malawi, and the USA have shown that screening closer to people and/or coupled with the application of laser to eyes with STDR detected in screening improves compliance and is cost-effective, including the gain in quality-adjusted life years (QALY) [51,52,53]. Based on the current resources, we suspect reaching people with mobile services or opportunistic screening in mass medical congregations would continue for quite some time in all LMICs.

What should be included in DR screening?

The ICO screening guideline suggests a record of disease and treatment history (duration, status, and medication). The basic eye examinations include presenting (and spectacles corrected, if any) vision and fundus photography [15]. A comprehensive eye examination would also include slit-lamp examination, intraocular pressure measurement (gonioscopy when intraocular pressure is high), and dilated eye examination. However, essentially, a screening examination should be short enough for people to adopt it and yet informative enough for intelligent referrals.

What are the follow-up and referral criteria

The key indicators of DR screening’s success are the robustness of the referral system and compliance with these referrals. Three outcomes could emerge from a DR screening episode (a) a routine re-examination when there is no or mild retinopathy, (b) non-urgent referral for moderate non-sight-threatening DR and (c) urgent referral for sight-threatening DR. Most guidelines agree that annual eye check-up is necessary when retinopathy is not detected (some countries recommend two years) and this interval is reduced depending on the degree of retinopathy. The ICO has suggested different guidelines for LMIC- it is twice longer for mild retinopathy (1–2 years in LMIC and 6–12 months in high-income countries, HIC). England and New Zealand guidelines have a systematic referral pattern: England (R0–R3) and New Zealand (R0–R5), depending on the disease severity. All guidelines are unanimous on the referral of all people with STDR and those with reduced visual symptoms. However, crucial to the right referral is the quality of the retinal image and grading of these images. All guidelines suggest referrals of all people with ungradable fundus photographs. Table 5 lists the DR screening strategy [8,9,10,11,12,13,14,15,16].

Discussion

All the included guidelines uphold the fundamental aim of DR screening to identify STDR. There are many similarities and a few variations in the screening guidelines (Table 5). It appears that two regions, the Eastern Mediterranean and Latin America region, follow the AAO PPP. We also believe the ICO guideline is used in many LMICs that have not yet developed their DR guidelines. Important areas of dissimilarity are the technical details, screening personnel, the need for mydriasis, and the choice of retinal camera.

Screening personnel

A 2015 ICO survey estimated 232,866 ophthalmologists globally, and they were unequally distributed, much less in low-income countries (3.7/million population) than in high-income countries (76.2/million population), and within the LMICs, they were located more in urban than in rural areas [54]. In the same year, 2015, there were 415 million people with diabetes globally, and 75% of them lived in LMICs [55]. Given that the global annual growth of the number of ophthalmologists is 2.9% and the yearly global increase in people with diabetes is 4.47% [2], it would be impossible for ophthalmologists alone to screen all people with diabetes all the time. Therefore, we advocate that non-ophthalmologists be trained to capture retinal images and use computer-aided grading of DR images, where available.

Need for mydriasis

Mydriatic examination generally allows screening of a greater retinal area compared to an undilated fundus view. Pupil dilatation enables the ophthalmologists to have a good view of the retina and a photographer to obtain artifact-free retinal photographs. However, dilatation for all patients and a routine mydriatic screening test that makes the person wait for at least 30 min and incapacitates the individual for near work for the next couple of hours may not always be necessary. With newer technology of non-mydriatic devices (such as non-mydriatic retinal camera and optical coherence tomography angiography), one could reserve pupillary dilatation to people where a readable/gradable image could not be obtained and obviously for those who need treatment. To achieve universal coverage in countries with steep increases in numbers of people with diabetes, innovative approaches in using non-mydriatic cameras should be encouraged. Policies that prevent mydriasis by non-clinician or non-clinical environment should be reviewed, risk-assessed, and situation-specific strategies tailored to facilitate mydriasis.

Creating referral system

One of the prime motivating factors of a screening program for DR is the evidence-based standardized care for all levels of STDR that includes laser, intravitreal anti-VEGF injection, and vitreoretinal surgery [56,57,58,59]. The detailed referral system practiced in England and New Zealand national screening is not practically possible everywhere, certainly not in LMICs. Ideally, countries must develop/revise their country-specific DR screening and treatment guidelines based on the local resources and trained human resources for health. The global benchmark, such as the ICO guideline that has taken different resources into account, is a good option as a national guideline or a template to develop/revise national policies.

Future of DR screening

Borne out of technological advances and policy planning, there are emerging game-changers. These should be considered in new or revised DR screening guidelines. These include camera technology, artificial intelligence, e-health, and universal health coverage.

Camera technology

The concept of 7-field fundus photography (300 views) originated with the DRS in the early 1970s. The 7-field fundus photos were arranged around the optic disc in a particular sequence [17, 18]. Since then, camera technology and techniques have undergone sweeping changes. In brief, these changes are black and white to color photography, film-based to digital photography, narrow (300) field to wide (450–500, and ultra-wide (2000) field imaging. Recent studies have demonstrated that digital photography is as good as film-based photography [60]; monoscopic digital fundus photographs also match the rigor of stereo photos [61]; and a mosaic of four- or five- 450 fields could create a field of view reaching the area covered by classic 7-field fundus photography [62]. Single ultra-widefield color fundus photo is also reported good for DR grading [63]. One drawback of digital retinal imaging is its inability to identify and grade macular edema accurately. Over the years, the camera hardware technology has advanced to produce less bulky and hand-held cameras, smartphone-based cameras, and non-mydriatic fundus cameras. This has immensely increased the opportunity to reach people closer to their home and transmit their images to a remotely located reading site for grading and management planning. In the last two decades, the quality of handheld cameras has considerably improved with better prediction capability [64, 65]. A good-quality fundus image is one where both disc and macula are captured in the same frame with good focus and illumination; a poor-quality fundus image captures either disc or macula and has poor illumination (Fig. 1).

Fig. 1: Example of fundus photo obtained in DR screening using non-mydriatic camera.
figure 1

Upper panel: 3 Nethra™ (Forus, India)- good (left) and poor (right) quality fundus image. Lower panel: Visuscout™ (Zeiss, Germany)- good (right) and poor (left) quality fundus image.

A technical comparison of the currently available non-mydriatic cameras is listed in Table 6. Technology has improved over the years; the questions at present are (1) what is the minimum field required for DR screening in most instances; (2) what are financial resources, as the camera cost steeply increases with an increase in the field of view; (3) what is the choice between mydriatic and non-mydriatic camera for retinal photography and its applicability in DR screening. A systematic review on the use of non-mydriatic cameras reported that two-field retinal photographs were predictable (sensitivity 91%, specificity 94%). It matched the predictability of images obtained by the mydriatic fundus camera once the ungradable images (18.4%) are excluded [66].

Table 6 Technical comparison of commercially available non-mydriatic cameras.

Artificial intelligence

Artificial intelligence (AI) in health care uses complex algorithms to emulate human cognition to analyze complex medical data without direct human input. It provides a well-defined output to the end-user. In several large studies, deep learning (one of the tools of AI) in DR has shown good sensitivity (87% to 100%), specificity (87% to 98%), and receiver operating characteristic curve (AUC; 0.93 to 0.99) for referable DR and/or STDR [67,68,69]. The current technology cannot capture macular edema (would improve once OCT images are included). AI has not addressed the issue of images captured in different cameras, the image quality, and the field of view. Despite technological advances, deep learning is still not tried in a real-world screening of DR. The final barriers are the health policy of different countries and the trust of both clinicians and patients in the machine verdicts [70].

E-healthcare-ophthalmology

E-health refers to the intersection of medical informatics, public health, and business, referring to health services and information delivered or enhanced through the Internet and related technologies [71]. The World Health Assembly (WHA) 2018 acknowledged digital technologies’ potential to play a major role in improving public health worldwide (WHA 71.7) [72]. The most integral part of e-health and telemedicine is robust information and communication technology (ICT). Broadly, telemedicine applications are divided into (1) “asynchronous” (Store-and-forward) that involves an exchange of pre-recorded data between two or more individuals; and (2) “synchronous” (Real-time) that involves the simultaneous presence of individuals for an immediate exchange of information. Both are possible, though; the asynchronous method is more suitable for DR screening. The patient’s management information is shared at the soonest possible time, and appropriate care is offered after that.

In India, remote screening has been used in DR successfully using a low-cost portable retinal camera [73, 74]. Realizing the importance of tele-screening, the Canadian Retina Research Network (CR2N) has recently laid guidelines to standardize tele-screening methodology in DR. The important elements of these guidelines are DR severity classification and recommendation for retinal photography (two-450 field fundus photo, one centered on the disc and the other centered on the macula; 600 horizontal and 450 vertical), and optical coherence tomography (imaging macula) standards [75]. Some of the important barriers of e-health that applies equally to tele-screening of DR include the availability of reliable ICTs in regions where it matters the most (such as in LMICs), the language of communication across different parts of the world, human, culture, and behavior of shifting from face-to-face encounters to virtual ones and the local legal requirements.

Universal health care

Universal health coverage (UHC) (equitable access, quality care, no catastrophic financial hardship) is the world’s aspiration [76], and the continuum of health intervention by integrated people-centered eye care (IPCEC) is WHO’s recommendation [77]. The IPCEC addresses the full spectrum of eye conditions, according to people’s needs and throughout their life course [77]. It has four layers of care from community to tertiary care. Adhering to both the principles of UHC and IPCEC, different components of DR screening could be conveniently distributed from community level to tertiary level care [78] (Table 7).

Table 7 DR screening activities at different levels of health care [78] along the WHO guidelines of integrated people-centered eye care.

Systemic care

Monitoring for systemic factors forms an essential aspect of the management of DM and DR. The targets for systemic care are generally uniform across the geographic territories, barring minor differences (Table 4). There is consensus that hyperglycemia and hypertension could be managed as per the available resources at primary through tertiary centers (Table 7). This would result in a decrease in microvascular (including DR) and macrovascular complications of DM. In line with the UHC, the closer this intervention is taken to the community, the higher likelihood of compliance.

Task sharing

Screening of DR in a cost-effective way is important and necessary when there is a huge burden of DM and a shortage of human resources, particularly in regions with a high prevalence of diabetic eye disease [79]. Effective and efficient use of the available trained human non-ophthalmologist workforce would free the ophthalmologists to perform a technically more difficult task in DR care, such as delivering retinal laser, performing intravitreal injections/vitreoretinal surgery, and follow up care of the treated patients.

Screening is not without limitations. These include: (a) screening can reduce the risk of developing a disease or its complications but cannot offer a guarantee of protection; (b) there is a minimum risk of false-positive and false-negative results; (c) false-positive results could lead to distress and possibly unnecessary treatment; (d) false-negative results could lead to false reassurance to patients and doctors. Screening is effective only when it is combined with proper referral and timely treatment. Therefore, it is imperative that screening is not established without creating suitable referral pathways, appropriate treatment, and follow-up care.

Weakness and strengths

The main weakness of this analysis is the limited number of guidelines examined and confined to the English language only. Incidentally, more countries have guidelines for DM (including one from the IDF for T2DM) than DR. The only global guideline for DR is the one recommended by the ICO. Also, we selected English-language DR guidelines, one from each IAPB region and the ones recently published.

The strength of this analysis was the evaluation of the most recent (2016–2019) published DR guidelines. Given the technological advances of devices used in DR detection, increased advocacy, and friendly eye health policy, the inclusion of the DR guidelines formulated in the recent five years was important. This report also compares the current commercially available non-mydriatic fundus camera, which is soon likely to be the standard of DR screening. The analysis also provided evidence on gaps in guidelines, and recommendations are made based on potential solutions from practices in different health systems.

In conclusion, a uniform protocol for DR screening in each country would help improve case detection. National guidelines on timely and evidence-based treatments should be put in place to complement a good screening program. Using newer technology of the camera, e-health, artificial intelligence, and the use of available health care personnel beyond ophthalmologists such as the allied eye health personnel will improve universal coverage of screening. International and national policies need to prioritize DR screening and treatment to align with universal health coverage to improve the efficiency of the screening programs in diabetic retinopathy.