Article Text

Download PDFPDF

Systematic screening for diabetic retinopathy (DR) in Hong Kong: prevalence of DR and visual impairment among diabetic population
  1. Jin Xiao Lian1,2,
  2. Rita A Gangwani1,
  3. Sarah M McGhee3,
  4. Christina K W Chan1,
  5. Cindy Lo Kuen Lam2,
  6. Primary Health Care Group,4,
  7. David Sai Hung Wong1
    1. 1Department of Ophthalmology, The University of Hong Kong, Hong Kong, Hong Kong
    2. 2Department of Family Medicine and Primary Care, The University of Hong Kong, Hong Kong, Hong Kong
    3. 3School of Public Health, The University of Hong Kong, Hong Kong, Hong Kong
    4. 4Department of Family Medicine and Primary Health Care, Hospital Authority, Hong Kong, Hong Kong
    1. Correspondence to Dr Rita A Gangwani, Department of Ophthalmology, The University of Hong Kong, Room 301, Block B, Cyberport 4, 100 Cyberport Road, Hong Kong, Hong Kong; gangwani{at}, dr.rita_gangwani{at}


    Purpose To determine the prevalence of diabetic retinopathy (DR), sight threatening DR (STDR), visual impairment and other eye diseases in a systematic DR screening programme among primary care Chinese patients with diabetes mellitus (DM) in Hong Kong.

    Methods Screening for DR was provided to all subjects with DM in public primary care using digital fundus photography according to the English national screening programme. STDR was defined as preproliferative DR (R2), proliferative DR (R3) and/or maculopathy (M1). The presence of other eye diseases was noted. Visual impairment was classified as none (visual acuity in the better eye of 6/18 or better), mild (6/18 to >6/60) and severe (6/60 or worse).

    Results Of 174 532 subjects screened, most had never been screened before. The prevalence of DR was 39.0% (95% CI 38.8% to 39.2%) and STDR 9.8% (95% CI 9.7% to 9.9%). The most common DR status was R1 (35.7%), followed by M1 (8.6%), R2 (3.0%) and R3 (0.3%). The prevalence of mild and severe visual impairment was 4.2% and 1.3%, respectively. Subjects with STDR had a higher prevalence (9.8%) of visual impairment than those without (3.5%).

    Conclusions DR was prevalent in this population and one in 10 had STDR. This suggests the need for systematic screening to ensure timely referral to an ophthalmologist for monitoring and/or treatment.

    • Retina
    • Public health
    • Epidemiology
    • Vision

    Statistics from

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


    Diabetic retinopathy (DR) is one of the commonest microvascular complications of diabetes mellitus (DM) that can lead to severe visual impairment. It is a major cause of blindness among adults aged <70 years in developed countries.1 ,2 It results in productivity loss, reduced quality of life and a significant healthcare burden to individuals, healthcare systems and society.

    Many studies have found a high prevalence of DR among subjects with DM.3 ,4 Screening for DR is accepted as an important means of detecting sight threatening DR (STDR) and preventing blindness.2 ,5 Systematic screening for DR started in the UK in 20036 ,7 and the prevalence of any DR was 19% to 30% among those with type 2 DM.7–10 Most data on screening outcomes are from predominantly Caucasian populations with only limited data on DR prevalence among Chinese.4 ,11

    In Hong Kong (HK), Chinese are the majority of the population and face an increasing burden of DM with prevalence of 10% in adults aged 20–79 years and 20% among those aged ≥65 years.12 ,13 There has been no systematic screening and the prevalence of DR was unknown prior to this study. After piloting,14 systematic screening commenced in 2010 using the grading and quality standards of the English National Screening Programme (UK guidelines).15 Systematic screening means having a system of identifying all diabetic patients (via a HK-wide public patients’ database), inviting them and screening using digital fundus photography. This screening programme was included as a component of a chronic disease programme launched by the hospital authority, the major public healthcare provider. The programme, known as the Risk Assessment and Management Programme,16 is provided at primary care clinics to all those with DM who attend public general outpatient clinics (GOPC). In this study, we measured the prevalence of DR and other eye diseases and the extent of visual impairment among these subjects using the data generated from the screening programme. This is the first population-based study on DR in HK and probably the first standardised data collection in a Chinese population. It provides information on the burden of disease to assist planning of screening services for Chinese populations in mainland China, as well as HK.


    Subjects and screening procedure

    Subjects with DM who were screened from August 2010 to March 2014 were included. Screening consisted of measuring habitual and pin hole visual acuity (VA) using the Early Treatment Diabetic Retinopathy Study (ETDRS) chart, slit lamp biomicroscopy to check the anterior chamber and dilatation of pupils with Mydriacyl (Tropicamide) 1% eye drops (SA Alcon-Couvreur NV, Belgium). Digital, colour, non-stereoscopic, retinal photographs were captured using a non-mydriatic auto fundus camera (AFC-230, NIDEK, Japan and Canon CR-DGi retinal camera, Canon, Japan) including two photographic fields (45°) for each eye: one centred at the optic disc and the other at the macula. Digital images and patient details were securely transferred from clinics to the grading centre using OptoMize software (V.1.1; digital healthcare, Cambridge, UK) and Health Level 7 standard protocols (Ann Arbor, Michigan, USA).

    Grading procedure

    Photographs were systematically graded by trained optometrists and ophthalmologists using a spatial resolution of 1024×768 pixels and Optomize software. They examined images for the presence and severity of DR according to the UK guidelines (table 1).6 Each photograph underwent primary grading (by an optometrist). All the photographs with any DR or any other eye disease and 15% of the photographs without any DR were further graded by a secondary grader (another optometrist). If there was any discrepancy in the grades allocated, the photograph was assessed by an arbitration grader (an ophthalmologist) who made the final decision. The final grade could therefore be allocated by the primary grader if negative, by agreement between the primary and secondary grader or by the arbitration grader. Quality assurance was ensured by recording sensitivity and specificity for each grader where sensitivity ≥95% and specificity ≥85% were considered acceptable. The graders underwent periodic tests to ensure the accuracy of grading. Additionally, external validation procedures compared the grading results of these graders against international grading experts (Liverpool grading centre, UK) and showed good agreement. STDR was said to be present if any features of maculopathy (M1), preproliferative DR (R2) or proliferative DR (R3) were found. Photographs that were not assessable or could not be graded were considered ungradable (U).

    Table 1

    Diabetic retinopathy (DR) classification used in our grading for DR screening based on UK guidelines6

    The DR grade for each subject was based on the worse DR grade of the two eyes. If one eye was non-STDR (ie, R0 or R1) and the other eye was graded as U, the subject would be classified as U and referred to the specialist clinic for further management. If one eye was STDR (ie, R2, R3 or M1) and the other eye U, the subject would be classified as STDR and referred. Those graded as R0 or R1 in both eyes were graded as non-STDR and were not referred. The better VA of the two eyes represented the subject's VA. The graders also noted the presence of other eye diseases which included but were not limited to epiretinal membrane, age-related macular degeneration (AMD), increased cup-to-disc ratio, retinal artery occlusion, retinal vein occlusion, macular hole, optic nerve disorders including anterior ischaemic optic neuropathy and arterial emboli. These subjects were referred to the specialist ophthalmology clinic as deemed appropriate by the ophthalmologist (arbitration grader).

    Data collection and statistical analysis

    OptoMize software stores all the screening data including VA, date of birth and sex. Duplicate visits due to rebooking and ‘test’ visits for initial testing of the software amounted to 1190 visits (1007 duplicates, 183 tests) and were deleted; then the first screening record was selected for each subject.

    The ETDRS VA was converted into Snellen VA as used in the clinical setting and accepted in the DR grading software. We classified the VA into three groups: no visual impairment (VA in the better eye ≥6/18), mild visual impairment (VA in the better eye between 6/18 and 6/60) and severe visual impairment (VA in the better eye ≤6/60).17

    We exported these data from OptoMize to Microsoft Excel and STATA V.13.1. Descriptive analyses summarised the prevalence of DR, visual impairment and other eye diseases with 95% CIs. Level of visual impairment was compared between the groups with and without STDR using the χ2 test. All subjects classified as U were excluded in this step. A p value of 5% was used as the significance level.


    From August 2010 to March 2014, screening was performed at 28 public GOPCs across all seven hospital clusters in HK with 29 trained and accredited optometrists and two ophthalmologists performing the grading. A total of 174 532 diabetic subjects were screened with information on age available for 99.9% (mean 64 years, SD 11) and on sex for 86.1% subjects (52.7%; 79 198/150 351 female).

    Among the 174 532 subjects, 5.6% (9777/174 532) had ungradable fundus photographs for one or both eyes (table 2) which is in line with the UK national minimum standard of <10%.15 The prevalence of any DR at screening was 39.0% (68 058/174 532) and of STDR 9.8% (17 116/174 532). The most common DR status was R1 with a prevalence of 35.7%, followed by M1 with 8.6%, R2 with 3.0% and R3 with 0.3%.

    Table 2

    Prevalence of diabetic retinopathy (DR; N=174 532)

    In the screened population, 99.7% (173 973/174 532) subjects had information on VA. Among them, 94.5% (164 456/173 973) had no visual impairment, with VA ≥6/18 (table 3). Mild and severe visual impairments were found in 4.2% (7263/173 973) and 1.3% (2254/173 973), respectively. After excluding 9777 subjects with ungradable photographs, subjects with STDR had a significantly higher risk of visual impairment than those without (p<0.001; table 4).

    Table 3

    Visual impairment level in the screened population (N=173 973)

    Table 4

    Visual impairment level by groups with and without STDR

    At screening, 13.4% (23 312/174 532) of subjects were found to have other eye diseases (table 5). Among the ungradable photographs, cataract was the commonest diagnosis and other cases included asteroid hyalosis and vitreous haemorrhage.

    Table 5

    Prevalence of other eye diseases detected at screening (N=174 532)


    This study estimated the prevalence of any DR and STDR when subjects with DM first received systematic screening in a naïve population (who had not been subjected to systematic screening in the past). By systematic screening we mean a system of identifying all diabetic patients (via a HK-wide public patients database), inviting them for DR screening using digital fundus photography, grading the photographs, acting on the results (ie, referral to ophthalmologists within set time targets) and rescheduling subjects for rescreening based on a protocol, that is, those not referred are usually scheduled for rescreening after ≥12 months unless they are considered at high risk of DR in which case they will be rescreened sooner. We found a high prevalence of any DR (39.0%) and STDR (9.8%) among diabetic subjects in the public primary care setting. Nearly one in three had some level of DR which confers a higher risk of progression to STDR and nearly one in 10 had STDR already which requires referral to a specialist ophthalmology clinic for follow-up and treatment. Very few studies have previously identified the prevalence of DR in HK18 ,19 and none have included, like our study, all those who attend public primary care clinics. In one study on a selected population with DM in HK, the prevalence of DR was 28.4%.18 In another community-based screening programme in a selected population from an integrative community health centre, the prevalence of DR was 30.7%.19 It is difficult to compare these estimates with our study due to the selection of populations and different definitions of DR.

    The prevalence of DR or STDR has been reported in other naïve populations but comparisons are also difficult due to different definitions of DR. Since the programme in HK is run according to the UK guidelines we might compare our findings with those at the establishment of systematic screening in Liverpool8 when the prevalence of DR and STDR was 25.3% and 6.0%, respectively. At the start of national screening in Wales, the initial prevalence was 30.9% and 4.41%, respectively, among those with type 2 DM.9 Both these results are lower than the findings from our study. One possible reason might be the shorter duration of DM among the subjects in the two studies when compared with ours. The median duration was 3.2 years in Liverpool and the mean duration was 5.3 years in Wales. In the population studied in HK, the mean duration of DM was 7.7 years.20 A meta-analysis of population-based studies in mainland China reported the prevalence of any DR as 23% (95% CI 17.8% to 29.2%), non-proliferative DR (NPDR) as 19.1% (13.6% to 26.3%) and PDR as 2.8% (1.9% to 4.2%) among those with DM but the included studies varied by examination method and definitions for DR.21 The prevalence of STDR and maculopathy was not considered. Two population-based studies reported the prevalence of vision-threatening DR, defined as severe non-proliferative DR, PDR or with clinically significant macular oedema as 6.3% and 6.4%, respectively, in mainland China.22 ,23 The prevalence of STDR found in our study may be higher than these Chinese reports because we included moderate NPDR and maculopathy in our definition of STDR.

    The most common type of STDR found at screening in our study was maculopathy which can lead to central vision loss. The definition of maculopathy varies between different classification systems.4 We included subjects with mild maculopathy, that is, those with dot haemorrhage and/or microaneurysm within one disc diameter of fovea with VA ≤6/12 as per the DR classification in the UK guidelines. We also found 0.3% of our subjects had R3 and might need immediate laser treatment.

    We found that 5.5% subjects had already experienced visual impairment with 1.3% of those screened meeting the criteria for severe visual impairment (≤6/60 in the better-seeing eye). A community-based study on diabetic subjects in HK showed a higher prevalence of any visual impairment (11.3%) but slightly lower prevalence of severe visual impairment (0.7%) than our study.17 Other studies have shown a higher prevalence of visual impairment in diabetic populations than non-diabetic populations.24 Although we do not know how much of the current visual impairment can be attributed to STDR, we did find that STDR was associated with visual impairment. We did not determine the cause of the visual impairment or whether it was correctable, but all patients with STDR or other eye diseases who required treatment were referred to ophthalmologists for further assessment.

    We also identified other eye diseases increasing the value of the DR screening since these subjects can also be referred for further management. The prevalence of other eye diseases should be interpreted with caution because this is a screening programme for DR and the quality assurance was ensured for the grades of DR, not other eye diseases. Our previous studies on detection of increased cup-to-disc ratio and detection of AMD in a diabetic screening programme have highlighted the detection of these conditions on a pilot cohort (n=2182).25 ,26

    A strength of this study is that it is the first and largest population-based study on DR in HK with around 170 000 diabetic subjects and these should be representative of the diabetic population in public primary care. Second, the screening was carried out using a standard protocol with quality assurance. To the best of our knowledge, this is the first study using standard protocols to estimate the prevalence of DR in a Chinese diabetic population. Good-quality data are imperative to identify disease burden and support planning and implementation of screening services. However, there were several limitations. First, prevalence was based on grading of the photographs without confirmation in the specialist clinic. This may overestimate the prevalence of DR and STDR. However, internal and external validation of our screening and grading showed good sensitivity and specificity. Internal validation results, which consisted of sensitivity and specificity results of grading by optometrists, were 95.2% (range 85.9–97.3, SD 5.26) and 92.2% (range 84.6–97, SD 4.68), respectively. External validation results showed a high degree of agreement (κ value=0.7) between our graders and international graders for the grades of STDR. We used non-stereoscopic digital retinal images for the grading of DR, which can pose a challenge in grading certain lesions like retinal thickening in maculopathy, because we adhered to the method of the English National Screening Program as in Liverpool and other UK regional screening programmes.

    The findings from this study have important implications for decision makers in HK as well as in mainland China. China, in 2013, had the largest number of diabetic subjects (98.4 million) in the world but systematic screening for DR is not currently available.12 We found a high proportion of subjects with R1 who are at risk of progression to STDR.27 This highlights the need for screening in China, as in the West where population-based screening for DR has already been prioritised.28 ,29

    In conclusion, we found a high prevalence of DR and STDR among the Chinese diabetic population in a public primary care setting in HK. This emphasises the need for systematic DR screening.


    The authors would like to thank Dr W Lai, Ms R Sum, all the primary care clinics’ teams and all the graders of the DR screening programme.



    • Collaborators Primary Health Care group of Hong Kong Hospital Authority includes Dr MY Wong, Dr YF Wong, Dr WS Chu, Dr WS Tsui, Dr SK Kwong, Dr YC Li, Dr VK Chao, Dr PF Chan, Dr KP Lai, Dr YK Yiu, Dr W Luk, Dr SN Fu, Dr MT Hui, Dr J Liang and Dr ML Chan.

    • Contributors RAG: Contribution to the design of the work and acquisition of data; analysis and interpretation of the data; drafting the work and revising it critically for important intellectual content. JXL, CKWC: Contribution to the analysis and interpretation of data; drafting the work and revising it critically for important intellectual content. SMMG: Made substantial contribution to the conception and design of the work; analysis and interpretation of the data, drafting the work and revising it critically for important intellectual content. CLKL and Primary Care Group: Contributions to acquisition of data; revising critically for important intellectual content. DSHW: Made substantial contribution to the conception and design of the work; drafting the work and revising it critically for important intellectual content.

    • Funding Azalea Foundation and Lee Hysan Foundation have provided funding support for diabetic retinopathy screening. The Diabetic Retinopathy screening programme was initially conducted by the Eye Institute, The University of Hong Kong, now known as ‘Department of Ophthalmology,’ The University of Hong Kong and is being currently performed under the ‘Risk Assessment and Management Programme’ of ‘Hospital Authority,’ Hong Kong.

    • Competing interests None declared.

    • Ethics approval Institutional Review Board, Hospital Authority.

    • Provenance and peer review Not commissioned; externally peer reviewed.

    • Data sharing statement The unpublished data are still being acquired for continuous screening programme which is the study project.