Effectiveness of task-shifting for the detection of diabetic retinopathy in low- and middle-income countries: a rapid review protocol

Diabetic retinopathy (DR) is the most frequent microvascular complication of diabetes and a significant cause of vision loss in adults [1]. Vision loss due to DR can largely be prevented through regular retinal screening for timely identification of DR, followed by referral and treatment [2‐5].

DR screening has been shown to be cost-effective when compared to no screening or opportunistic screening [3, 6, 7]. However, cost-effectiveness requires high coverage and this is challenging in low- and middle-income (LMICs) countries [8] where the available eye health services are clustered around the cities and access to them from the remote and rural populations is often difficult [9]. Few countries worldwide have been successful in establishing population-based screening programmes that reach most of the patients with diabetes [3]. The uptake of retinal screening depends on context-specific factors related to the service provider, the wider organization and structure of the health system and the patient [10‐12].

A crucial limitation to deliver DR screening in LMICs is the low availability of ophthalmologists, particularly in sub-Saharan Africa (SSA), where the mean ophthalmologist density per million population is 3.7 [13]. In this context, it is not realistic to rely on ophthalmologists to examine the retinas of the growing population of people living with diabetes (PLWD) unless it is at the expense of the delivery of other critical eye care services. This is not only a challenge for LMICs, and programmes in high-income countries (HICs) have overcome this issue by successfully adopting task-shifting in DR screening by training retinal graders, thus freeing ophthalmologists time to perform other tasks [14]. The WHO defines task-shifting as the transfer of tasks to existing cadres of healthcare workers with shorter training and fewer qualifications or to newly created cadres who receive a competency-based training for the specific task [15]. There is some evidence of the successful use of task-shifting in SSA in other health fields including diabetes [16, 17]. However, it is not clear whether task-shifting interventions for the detection of DR in LMICs results in an increase in the uptake of screening and higher DR detection rate.

There are several factors that can potentially affect the effectiveness of task-shifting interventions for DR screening. Firstly, the choice, level and training of the cadre performing the screening is a key consideration and is likely to be different depending on the composition and competencies of the health workforce in a particular country. The quality of the screening is measured by the diagnostic test accuracy (DTA) of the method of choice in the hands of non-ophthalmologists [18]. It is recommended that population-based screening programmes should reach over 80% sensitivity and 95% specificity [6, 19]. Screening performed by photographers has been shown to be more accurate when done by photographers with specialist medical or eye care qualifications [20]. There is also evidence that non-medical retinal image graders can achieve the threshold level of sensitivity and specificity in both mydriatic and non-mydriatic retinal assessment [18].

Secondly, the screening technique used can also affect the DTA. The recommended technique is photography-based screening which adds the possibility of quality assurance [14]. DR screening programs in LMICs are gradually adopting retinal photography. However, in many clinics without cameras, other methods of retinal examination that are not necessarily effective as screening techniques are still used [18]. The sensitivity of direct ophthalmoscopy in detecting DR is low (65%) and indirect ophthalmoscopy requires extensive training and experience [21]. Slit lamp bio-microscopy has a high sensitivity and specificity but also requires significant training and the availability of a slit lamp. Optical coherence tomography (OCT) for the detection of diabetic maculopathy is expensive and rarely available outside tertiary level eye departments in LMICs.

Thirdly, the use of mydriasis can improve the effectiveness of task-shifting by increasing the number of gradable images [18]. However, non-mydriatic imaging presents several advantages for screening programmes, like reduced screening time, less inconvenience for patients and importantly no limitations for using cadres that may not be allowed to instil eye drops.

Finally, an important consideration in LMICS is the need to reach rural and remote populations with difficult access to eye health services. Outreach services delivered by non-ophthalmologists can be an answer to this problem, but must ensure that DTA is maintained [18, 20]. Evidence suggests that DTA is higher in hospital-based screening services compared with outreach services [15]. Several approaches for DR screening in outreach services that can lend themselves to task-shifting have shown promising accuracy levels. These include telemedicine using digital imaging, the use of artificial intelligence-supported cameras or smart phone-based technology for image capture [22‐26].

This rapid review aims to identify and synthesise the peer reviewed literature on the effectiveness of task-shifting for the detection of DR in LMICs. It is an important topic given the predicted epidemic of diabetes mellitus which will disproportionally affect LMICs which will require major prevention strategies to prevent the corresponding increase in blindness caused by diabetic retinopathy. There is a body of evidence to support the effectiveness of task-shifting in DR screening in high-income settings which suggests that this may be a viable strategy in LMICs. However, given the differences in structure, organization and resourcing in health systems in LMICs, evidence from HICs may not be generalizable to these settings. The aim of this rapid review is to summarise the published literature reporting the effectiveness of task-shifting interventions for the detection of diabetic retinopathy by non-ophthalmologists low- and middle-income countries.

Differences in health systems organization, structure and resources will determine the need and success of task-shifting interventions for DR screening. The review will examine how these interventions have been used and/or tested in LMICs. This rapid review will provide context-specific evidence to inform policy makers and programme managers tasked with designing and implementing services to prevent and manage diabetes and its complications in low- and middle- income settings.

In balancing the time demands of a rapid review methodology, we have prioritized the peer-reviewed literature as we considered the more robust evidence for effectiveness will be found in published scientific journals. We acknowledge that the limits in searches and time may potentially miss some evidence which could have been published before the period included and in the grey literature.