When developing the NHS Diabetic Eye Screening Programme in England, we needed to consider 11 different stages that are listed in Table 1.
It was critical for diabetologists, ophthalmologists, public health doctors and optometrists to speak with one voice; otherwise, we would never have established the programme. Assessment and treatment facilities are available in England as part of our National Health Service, but this question becomes much more relevant in developing countries where treatment facilities may not be so readily available. There is no point in screening for sight-threatening diabetic retinopathy if treatment facilities are either not available or inadequate.
In England, everyone has a Primary Care Physician (GP) and so we have to obtain details from Primary Care on those diagnosed with diabetes and everyone has an NHS identifier number. A letter is sent out to everyone with diabetes aged 12 years and over to invite them for a diabetic eye screening appointment once a year. National leaflets have been produced to explain about diabetic retinopathy and about the screening test and what happens if screen-positive diabetic retinopathy is found. We have included information in the leaflet that the screening is not a diagnostic test and hence will only detect at best 90% of sight-threatening diabetic retinopathy and not other eye conditions. There has also been active engagement with patient organisations. There are appropriate exclusion criteria for those who do not need to be invited for screening, e.g. those already under ophthalmology and terminally ill.
Software has been developed to provide a single collated list of people with diabetes and for call recall, screening, grading and audit. In the database, the images are attached to patient details and confidentiality of patient data is a priority. The database is regularly backed up and an IT infrastructure has been established for capture and transmission of images to and from the cameras.
In England, we use non-mydriatic cameras and undertake mydriatic photography on all people with diabetes aged 12 years and older. The two 45° fields captured by the English Screening Programme are shown in Fig. 1 together with the one 45° field used by Scotland and the seven 30° stereo fields that are used as a reference standard against which screening tests are judged. The English Screening Programme sets a minimum camera specification and tests all prospective cameras that meet this minimum specification on patients who are known to have specific features of diabetic retinopathy. The specification document is a fairly lengthy document which includes the following statements: The unit must be capable of providing a minimum field of view of 45° horizontally and 40° vertically at the specified resolution (at least 30 pixels/degree). The unit must be capable of accommodating refractive errors of ±15 D as detailed in EN ISO 10940. The internal fixation aid should be capable of positioning the eye to capture the fields of regard specified below. The ‘field of regard’ of the fundus camera must make it relatively straightforward for an appropriately trained and competent retinal screener to capture images centred on (1) the foveal area and (2) the optic disc. In addition, the ‘field of regard’ of the fundus camera must be able to capture images as defined by the area covered by fields 3–7 of the seven-field protocol used in the Early Treatment Diabetic Retinopathy Study . The list of cameras that are currently approved to be used in the English Screening Programme is published by Public Health England on their webpage .
There has been a progressive increase in the size of uncompressed images from the modern camera backs which are now over 20 MB—the English NHS Diabetic Eye Screening Programme recommends capture of images that are compressed to a size of 1–2 MB. This level of compression has not been shown to lose any clinically significant information [13,14,15].
When considering whether to routinely dilate the pupil of people with diabetes attending for screening the study that was influential in the decision-making process was a population-based screening study  of 1549 people with diabetes who had received non-mydriatic one-field digital photography followed by mydriatic two-field digital photography and a reference standard examination by an experienced ophthalmologist whose examination was tested separately against seven-field stereo-photography . The sensitivity for one-field non-mydriatic photography was 86.0% (95% CI, 80.9–91.1%), the specificity was 76.7% (95% CI, 74.5–78.9%) and the poor-quality image rate was 19.7% (95% CI, 18.4–21.0%). The sensitivity for two-field mydriatic photography was 87.8% (95% CI, 83.0–92.6%), the specificity was 86.1% (95% CI, 84.2–87.8%) and a poor-quality image rate was 3.7% (95% CI, 3.1–4.3%). This study led to the approach used in England of two-field mydriatic photography and the approach in Scotland of staged mydriasis with one-field non-mydriatic photography and with dilation only if poor-quality images were obtained. The correlation  with age led Northern Ireland to only routinely dilate those aged 50 years and over.
It has been demonstrated that there is a strong correlation  between age and poor-quality image rates in diabetic retinopathy screening, for both non-mydriatic and mydriatic photography. Hence, publications [19,20,21,22] with small numbers in a young age range are not relevant to population-based screening programmes where many of the individuals to be screened are over 60 years.
In any population-based screening programme, it is necessary to balance acceptability to the individuals being screened and cost-effectiveness of the screening method with detection rates of sight-threatening diabetic retinopathy. Population-based screening programmes that use non-mydriatic photography like the Scottish Screening Programme  usually capture one field centred on the fovea and those that use mydriatic photography like the English Screening Programme usually capture a second field that is centred on the disc, which also give a second view of the macular area. In 1989, Moss  demonstrated that for eight retinopathy levels, the rate of agreement with seven stereoscopic fields ranges from 80% for two 30° stereo fields to 91% for four 30° stereo fields. In 2003, Scanlon  reported that two-field mydriatic digital photography gave a sensitivity of 80.2% (75.2–85.2) and specificity of 96.2% (93.2–99.2) in comparison with seven-field stereo-photography. In the latter study, 15.3% of seven-field sets were ungradable compared with 1.5% of the two-field digital photographs.
Clear protocols need to be in place for management of people with poor-quality images. In the English Screening Programme, all people with poor-quality images are referred for examination by slit lamp biomicroscopy.
The English NHS Diabetic Eye Screening Programme routinely measures Visual Acuity at screening, but it is recognised that it is not sufficiently sensitive on its own to be a screening tool [24, 25]. Hence, it needs to be used in conjunction with other features that are detected at grading.
The diabetic retinopathy grading classification that has the best evidence base is the Early Treatment Diabetic Retinopathy Study (ETDRS) final Retinopathy Severity Scale  because it provided the first detailed classification system for retinopathy severity based on a natural history study of untreated eyes. However, this relies on detailed grading of stereo-photographs of seven fields of each eye. This scale did not grade lesions in the macular area. The ETDRS study did define ‘clinically significant macular oedema’ which was a level at which laser treatment was advised, but this was based on stereo-photography and the study did not recommend a referral level for closer observation before laser treatment was recommended.
Table 2 shows the International Classification  which was developed by the American Academy of Ophthalmology in 2002 and recommends that any level of retinopathy more severe than mild retinopathy (defined as the presence of microaneurysms only) warrants examination by an ophthalmologist. However, this is too early a referral level for use in the English Screening Programme and the referral level for the English Screening Programme is also listed in Table 2.
Table 3 shows the risks of progression to proliferative diabetic retinopathy as recorded in the Early Treatment Diabetic Retinopathy Study . Screening Programmes need to accept a certain level of risk. In the English programme, we needed to decide whether we were prepared to accept a 6.2% risk or an 11.3% risk that a patient who has been screened and given a 1-year appointment develops proliferative DR before their next screen. We opted for the 11.3% risk which is the equivalent to moderate non-proliferative diabetic retinopathy on the ETDRS final Retinopathy Severity Scale . We also had to develop a definition for maculopathy referral based on two-dimensional markers. The ETDRS study did not classify maculopathy, but it did make recommendations on what constituted clinically significant macular oedema requiring laser treatment as shown in Table 4. We opted for three referral criteria, based on two-dimensional photographic markers and measurement of Visual Acuity:
Exudate within 1 disc diameter (DD) of the centre of the fovea (Fig. 2).
Circinate or group of exudates within the macula (Fig. 3).
Any microaneurysm or haemorrhage within 1DD of the centre of the fovea only if associated with a best VA of ≤ 6/12 (if no stereo) (Fig. 4).
A minimum screen resolution is recommended  when viewing the images for grading which has progressed as the technology of screens has advanced. The current minimum acceptable standard for screen resolution is a vertical resolution of 1080 (1920 × 1080) with an achievable and recommended standard of a minimum of 1200 (1920 × 1200 or higher). It is recommended that a minimum of 60% of the image should be viewable on the grading screen to avoid too much scrolling to see the full image.
To ensure that whole screening programme is provided by a trained and competent workforce a minimum qualification  is required for screeners and graders in the English programme. Evidence of ongoing continuous professional development and taking the monthly External Quality Assurance Test sets [29, 30] is also required—all 1500 graders in the English Screening Programme are required to take a monthly test set of 20 image sets and their grading of these images is compared against a guide grade. An International Version of the qualification  and the monthly test and training set  for screeners working outside the UK is available.
An important part of any screening programme is the introduction of Quality Assurance. The purpose of introducing Quality Assurance is to reduce the probability of error and risk, ensure that errors are dealt with competently and sensitively, help professionals and organisations improve year on year, and set and keep under review national standards.
The NHS Diabetic Eye Screening Programme has developed three Key Performance Indicators and nine other Quality Standards . These are given in Table 5 with the three Key Performance Indicators being shown in the right-hand column. A programme board which includes local health service representatives and national Quality Assurance team representatives oversees the results of a programmes performance against the standards four times a year, and if a programme is performing poorly, they are expected to improve or the service may be recommissioned to be provided by a different provider. Graders who perform poorly on test sets undergo extra training and have all of their work second graded until their performance improves.
In addition, an External Quality Assurance visit to all regional programmes who undertake Diabetic Eye Screening as part of the NHS Diabetic Eye Screening Programme is undertaken every 3 years. EQA visits are an integral part of Diabetic Eye Screening Quality Assurance. Formal EQA visits to a screening programme provide the forum for a review of the whole multidisciplinary screening pathway and an assessment of the effectiveness of team working within the screening centre and associated referral sites.