Key Features.
The North East London Diabetes Study (NELDS) was established to investigate the epidemiology of diabetes-related outcomes and to support the development of equitable predictive models.
Based in North East London, England, the NELDS includes people with diabetes (92% type 2) registered with the Diabetic Eye Screening Programme (DESP) between January 2012 and December 2021. The cohort comprises 178 536 individuals aged ≥12 years at baseline (median age 57 years; interquartile range 47–68). Ethnic diversity is substantial, with 36% White, 34% South Asian, and 16% Black participants. By December 2021, 15% had died and 9.8% had moved out of the area.
The DESP dataset includes 5.5 million colour fundus photographs, standardized diabetic retinopathy grading, and socio-demographic information spanning 1 074 985 person-years of follow-up. These data are linked to primary and secondary care electronic health records (EHRs) containing clinical measurements, medication records, and comorbidity information. Prospective data collection is ongoing.
The scale, diversity, and unique integration of longitudinal ophthalmic imaging with linked EHR data of the NELDS make it well suited for developing accurate, equitable predictive models to inform personalized care and improve the prevention of diabetes complications.
Enquiries about collaboration and data access should be directed to the corresponding author.
Why was the cohort set up?
Diabetes is a growing public health problem affecting 1 in 10 adults globally [1]. The number of people with diabetes worldwide increased from 151 million in 2000 to 589 million in 2024 and is projected to rise to 853 million by 2050 [2]. About half remain undiagnosed [3]. Furthermore, an estimated 635 million adults have impaired fasting glucose or impaired glucose tolerance, which progresses to type 2 diabetes at a rate of 5%–10% per year [4–6]. In the UK, an estimated 5 million people have diabetes, costing the National Health Service (NHS) ∼6% of the health budget [7, 8]. In 2024 alone, there were an estimated 3.4 million diabetes-related deaths worldwide, with 37% occurring in the working-age population [2].
The early identification of individuals at risk of complications is crucial for targeted intervention. Studies have sought to estimate the risk of major complications, including all-cause mortality, myocardial infarction, stroke, renal disease, and sight-threatening diabetic retinopathy (STDR) [9–12]. Diabetic retinopathy (DR)—a major microvascular complication of diabetes and leading cause of blindness in the working-age population [13]—provides a window to systemic health, with studies demonstrating the utility of incorporating DR grading into predictive models for adverse health outcomes [14–17]. However, >80% of ophthalmic imaging datasets lack socio-demographic characteristics and population diversity [18]; ethnic minorities and people with type 1 diabetes are typically under-represented and model translation is lacking [19].
The North East London Diabetes Study (NELDS) is a large cohort study of people registered with the North East London Diabetic Eye Screening Programme (NEL DESP), designed to harness routinely collected electronic health record (EHR) data to improve prediction and early intervention for complications in people with diabetes. Data from the NEL DESP, funded through the NHS public health budget, are available from 2012 following the implementation of the region’s current database system. The study combines longitudinal socio-demographic data, colour fundus photographs, manually graded DR severities, and screening outcomes with linked primary and secondary care EHRs from an ethnically diverse cohort. This provides detailed phenotypic information across a wide range of biomarkers and health outcomes, facilitating the development of strategies to enhance care for people with diabetes.
Who is in the cohort?
The NELDS is a cohort study of 178 536 people diagnosed with diabetes aged ≥12 years who were registered for population diabetic eye screening in the NEL DESP between 3 January 2012 and 31 December 2021. Prospective data collection is ongoing. Ethics approval did not require individual consent, as the study relied solely on retrospective, anonymized data (further details in the ‘Ethics approval’ section).
The uptake of DR screening in the UK is high (consistently >80% annual uptake) [20], driven by incentives under the Quality and Outcomes Framework for GPs to refer those with newly diagnosed diabetes for eye screening. This ensures a comprehensive and systematically collected ophthalmic dataset for the NELDS. Further details for the national and NEL DESP are provided in the Supplementary methods. While data are not available for non-participants (eligible individuals with diabetes who did not attend) or the general population of individuals with diabetes, the strong representation of high-risk groups in the NEL DESP cohort provides valuable insights.
Table 1 summarizes key socio-demographic characteristics for the full cohort, as well as by diabetes subtype. The median age at baseline DESP appointment was 57 years (interquartile range 47–68) and 46% were female. The majority of individuals (92%) had type 2 diabetes, while 4.6% had type 1. The remainder (grouped together as ‘Other’ in Table 1) were recorded as Maturity Onset Diabetes of the Young (MODY) (n = 116), Other (n = 287), or Unknown (n = 6592). Individuals with type 1 diabetes were considerably younger at diagnosis (median age 21 years), with a longer median diabetes duration (8 years) compared with the overall cohort.
Table 1.
Summary of socio-demographic factors at baseline DESP visit. Results are presented for the full cohort (shaded in grey) and by diabetes subtype.
| Full cohort | Diabetes subtype |
|||
|---|---|---|---|---|
| Characteristic | N = 178 536 | Type 2 (N = 163 393) | Type 1 (N = 8148) | Other a (N = 6995) |
| Age (years) | ||||
| Median (Q1, Q3) | 57 (47, 68) | 57 (48, 68) | 32 (23, 46) | 63 (49, 76) |
| Sex [n (%)] | ||||
| Female | 81 714 (46) | 74 889 (46) | 3627 (45) | 3198 (46) |
| Male | 96 796 (54) | 88 483 (54) | 4520 (55) | 3793 (54) |
| Unknown | 26 (<0.1) | 21 (<0.1) | 1 (<0.1) | 4 (<0.1) |
| Diabetes duration (years) | ||||
| Median (Q1, Q3) | 2 (0, 8) | 2 (0, 8) | 8 (1, 17) | 1 (0, 5) |
| N missing (%) | 4259 (2.4) | 997 (0.6) | 79 (1.0) | 3183 (46) |
| Age at diabetes diagnosis (years) | ||||
| Median (Q1, Q3) | 52 (42, 62) | 52 (43, 62) | 21 (12, 33) | 54 (41, 66) |
| N missing (%) | 4259 (2.4) | 997 (0.6) | 79 (1.0) | 3183 (46) |
| Died [n (%)] | 26 547 (15) | 23 476 (14) | 666 (8.2) | 2405 (34) |
| Age at death (years) | ||||
| Median (Q1, Q3) | 79 (70, 86) | 79 (70, 86) | 69 (55, 80) | 80 (71, 86) |
| Ethnicity [n (%)] | ||||
| White | 64 592 (36) | 56 533 (35) | 5093 (63) | 2966 (42) |
| South Asian | 61 202 (34) | 58 880 (36) | 1109 (14) | 1213 (17) |
| Black | 29 103 (16) | 27 269 (17) | 960 (12) | 874 (12) |
| Any other Asian | 9838 (5.5) | 9457 (5.8) | 190 (2.3) | 191 (2.7) |
| Other | 5394 (3.0) | 4974 (3.0) | 255 (3.1) | 165 (2.4) |
| Unknown | 4833 (2.7) | 3073 (1.9) | 303 (3.7) | 1457 (21) |
| Mixed | 2384 (1.3) | 2093 (1.3) | 197 (2.4) | 94 (1.3) |
| Chinese | 1190 (0.7) | 1114 (0.7) | 41 (0.5) | 35 (0.5) |
| IMD quintile [n (%)] | ||||
| 1 | 61 097 (34) | 56 737 (35) | 2664 (33) | 1696 (24) |
| 2 | 58 135 (33) | 54 042 (33) | 2547 (31) | 1546 (22) |
| 3 | 30 395 (17) | 28 300 (17) | 1285 (16) | 810 (12) |
| 4 | 13 280 (7.4) | 12 163 (7.4) | 686 (8.4) | 431 (6.2) |
| 5 | 4281 (2.4) | 3801 (2.3) | 265 (3.3) | 215 (3.1) |
| Unknown | 11 348 (6.4) | 8350 (5.1) | 701 (8.6) | 2297 (33) |
| Smoking history [n (%)] | ||||
| Ever smoked | 52 758 (30) | 48 811 (30) | 2378 (29) | 1569 (22) |
| Never smoked | 100 000 (56) | 93 279 (57) | 4336 (53) | 2385 (34) |
| Unknown | 25 778 (14) | 21 303 (13) | 1434 (18) | 3041 (43) |
| Family history of CVD [n (%)] | 66 315 (37) | 61 794 (38) | 2228 (27) | 2293 (33) |
‘Other’ includes individuals for whom the type of diabetes was recorded as ‘Maturity Onset of the Young’, ‘Other’, or ‘Unknown’.
Findings are presented as median (interquartile range) for continuous variables and number (%) for categorical variables.
CVD, cardiovascular disease; IMD, Index of Multiple Deprivation.
The NELDS cohort demonstrates considerable ethnic diversity, with 36% of individuals self-reporting as being of White ethnicity, 34% as South Asian, and 16% as Black. The majority (67%) of individuals resided in the two most deprived Index of Multiple Deprivation (IMD) quintiles. Supplementary Table 1 further highlights ethnic disparities within deprivation, with Black (46% in Q1) and South Asian (36% in Q1) groups having a higher representation in the most deprived quintile of IMD compared with White individuals (29% in Q1). South Asian individuals were diagnosed with diabetes at a younger median age (47 years) than White individuals (56 years).
How often have they been followed up?
Participants undergo annual or more frequent DR screening as part of routine NHS care, with recent policy changes allowing biennial screening for low-risk individuals [21]. During the study period (2012–21), the NEL DESP dataset includes a total of 1 130 390 appointments and >5.5 million retinal images spanning >1 million person-years. The median follow-up duration was 6.4 years (interquartile range [IQR] 3.2, 9.2), with 61.1% of patients having ≥5 years of follow-up from their baseline DESP appointment (Fig. 1).
Figure 1.
Distribution of patients in the DESP cohort. (a) Number of patients by year of baseline DESP appointment, showing recruitment patterns over time. (b) Distribution of follow-up duration in years, indicating the range of longitudinal data available for analysis.
By the end of 2021, 15% of individuals in the cohort had died, with a median age at death of 79 years (IQR 70, 86). 9.8% had moved out of the area. Supplementary Table 2 compares the baseline characteristics of the 43 972 individuals permanently lost to follow-up (due to death or moving out of area) with the continuing cohort. These individuals were older with longer diabetes duration and more likely to have missing data.
What has been measured?
The NEL DESP dataset includes colour fundus photographs, standardized grading for DR (Supplementary methods), and socio-demographic information. Linked EHR data were available for 99.5% of individuals, capturing all historic and contemporary primary and secondary care records, providing clinical measurements, medication details, and comorbidity data. The methodology for defining clinical measurements and diabetes-related comorbidities from EHR data are detailed in the Supplementary methods—‘Characterization at baseline’. The complete final list of included SNOMED CT codes and numbers of EHR entries is available in the Supplementary Excel file.
What has it found?
Clinical measurements
High completeness of linked primary care data for key clinical measurements was achieved, with >95% of individuals having at least one measurement for haemoglobin A1C (HbA1c), total cholesterol, systolic blood pressure, estimated glomerular filtration rate (eGFR), and body mass index (BMI) (Supplementary Table 3). Detailed clinical characteristics are presented in Table 2 and Supplementary Table 4. Individuals with type 1 diabetes had a higher median HbA1c value (69 mmol/mol) compared with those with type 2 (53 mmol/mol). Notable ethnic differences were observed in systolic blood pressure, BMI, and eGFR, with higher cardiovascular risk profiles in White and Black individuals compared with South Asian and Chinese groups (Supplementary Table 4).
Table 2.
Summary of clinical measurements and medication history at baseline DESP visit. Results are presented for the full cohort (shaded in grey) and by diabetes subtype.
| Full cohort | Diabetes subtypea |
|||
|---|---|---|---|---|
| Characteristic | N = 178 536 | Type 2 (N = 163 393) | Type 1 (N = 8148) | Other (N = 6995) |
| Systolic blood pressure (mmHg) | ||||
| Median (Q1, Q3) | 133 (124, 141) | 133 (125, 141) | 123 (115, 132) | 134 (126, 142) |
| N missing (%) | 4283 (2.4) | 3244 (2.0) | 712 (8.7) | 327 (4.7) |
| BMI (kg/m²) | ||||
| Median (Q1, Q3) | 29.2 (25.9, 33.5) | 29.4 (26.2, 33.6) | 24.6 (21.9, 28.3) | 29.0 (25.5, 33.3) |
| N missing (%) | 8999 (5.0) | 7356 (4.5) | 1072 (13) | 571 (8.2) |
| HbA1c (mmol/mol) | ||||
| Median (Q1, Q3) | 54 (48, 67) | 53 (48, 66) | 69 (56, 85) | 50 (45, 64) |
| N missing (%) | 30 437 (17) | 26 224 (16) | 2687 (33) | 1526 (22) |
| eGFR (ml/min/1.73 m²) | ||||
| Median (Q1, Q3) | 74 (60, 87) | 74 (60, 87) | 82 (60, 90) | 60 (60, 80) |
| N missing (%) | 37 733 (21) | 32 524 (20) | 3288 (40) | 1921 (27) |
| Total cholesterol (mmol/l) | ||||
| Median (Q1, Q3) | 4.59 (3.97, 5.27) | 4.60 (3.97, 5.28) | 4.50 (3.95, 5.13) | 4.55 (3.93, 5.23) |
| N missing (%) | 7455 (4.2) | 5067 (3.1) | 1756 (22) | 632 (9.0) |
| Retinopathy grade, worst [n (%)] | ||||
| R0 | 127 312 (71) | 118 477 (73) | 4157 (51) | 4678 (67) |
| R1 | 41 759 (23) | 37 079 (23) | 3081 (38) | 1599 (23) |
| R2 | 3914 (2.2) | 3195 (2.0) | 442 (5.4) | 277 (4.0) |
| R3 | 2270 (1.3) | 1735 (1.1) | 364 (4.5) | 171 (2.4) |
| Unknown | 3281 (1.8) | 2907 (1.8) | 104 (1.3) | 270 (3.9) |
| Maculopathy grade, worst [n (%)] | ||||
| M0 | 163 274 (91) | 150 222 (92) | 6901 (85) | 6151 (88) |
| M1 | 11 981 (6.7) | 10 264 (6.3) | 1143 (14) | 574 (8.2) |
| Unknown | 3281 (1.8) | 2907 (1.8) | 104 (1.3) | 270 (3.9) |
| Best VA (LogMAR) | ||||
| Median (Q1, Q3) | 0.00 (−0.10, 0.20) | 0.00 (−0.10, 0.20) | 0.00 (−0.10, 0.00) | 0.00 (0.00, 0.20) |
| N missing (%) | 1570 (0.9) | 1352 (0.8) | 58 (0.7) | 160 (2.3) |
| Anti-hypertension medication [n (%)] | 112 376 (63) | 105 606 (65) | 2291 (28) | 4479 (64) |
| Lipid-lowering medication [n (%)] | 109 465 (61) | 103 208 (63) | 2345 (29) | 3912 (56) |
| Non-insulin hypoglycaemic medication [n (%)] | 132 967 (74) | 127 055 (78) | 1981 (24) | 3931 (56) |
| On insulin [n (%)] | 27 973 (16) | 19 393 (12) | 7296 (90) | 1284 (18) |
‘Other’ includes individuals for whom the type of diabetes was recorded as ‘Maturity Onset of the Young’, ‘Other’, or ‘Unknown’.
Findings are presented as median (interquartile range) for continuous variables and number (%) for categorical variables. DR outcomes are recorded as per the UK National Screening Committee classification system (see Supplementary methods).
logMAR, logarithm of the minimum angle of resolution; VA, visual acuity.
Individuals with type 2 diabetes presented with higher median systolic blood pressure and BMI and lower median eGFR compared with those with type 1 diabetes. Correspondingly, a higher proportion of people with type 2 diabetes had commenced antihypertensive (65% vs 28% type 1), lipid-lowering (63% vs 29% type 1), and non-insulin hypoglycaemic (78% vs 24% type 1) medications (Table 2).
Diabetic retinopathy
At baseline, 71% of the overall cohort had no detectable retinopathy (R0). However, as shown in Table 2, individuals with type 1 diabetes had a notably higher baseline prevalence of DR (48% R1 or worse) and diabetic maculopathy (14% M1) compared with those with type 2 diabetes (26% R1 or worse, 6.3% M1). A clear trend for worsening cardiometabolic risk factors, longer diabetes duration, and younger age at diagnosis was observed with increasing DR severity (Supplementary Tables 5 and 6). A total of 30 766 individuals developed referable DR (M1, R2, or R3) at some stage during the study period.
Comorbidities and incident events
As shown in Table 3, hypertension was the most common prevalent comorbidity (90 456 cases), followed by renal failure (23 289 cases). Hypertension and renal failure also exhibited the highest incidence rates (41.5 and 20.8 per 1000 person-years, respectively). The incidence rates for coronary heart disease and stroke were 10.6 and 5.1 per 1000 person-years, respectively. As further illustrated in Supplementary Table 7, the incident rates for all comorbidities were observed to be higher in individuals with more severe baseline retinopathy. The majority of cases for all these conditions were ascertained from primary care records (Supplementary Figure 1).
Table 3.
Prevalent cases of major comorbidities at baseline DESP visit and subsequent incident events. Incidence rates expressed per 1000 person-years.
| Comorbidity | Prevalent cases | Crude prevalence (%) | Age-standardized prevalence a [% (95% confidence interval)] | N at risk | Incident cases | Crude incidence rate/1000 person-years | Age-standardized incidence rate/1000 person-years a [n (95% confidence interval)] |
|---|---|---|---|---|---|---|---|
| Coronary heart disease | 21 350 | 12.0 | 3.8 (3.8, 3.9) | 157 186 | 9680 | 10.6 | 3.8 (3.7, 3.9) |
| Foot ulcer | 2625 | 1.5 | 1.0 (0.9, 1.2) | 175 911 | 3463 | 3.3 | 2.1 (1.8, 2.4) |
| Hypertension | 90 456 | 50.7 | 20.4 (20.2, 20.6) | 88 080 | 17 911 | 41.5 | 20.9 (20.5, 21.4) |
| Peripheral artery disease | 3289 | 1.8 | 0.6 (0.5, 0.6) | 175 247 | 3580 | 3.4 | 1.4 (1.3, 1.5) |
| Renal failure | 23 289 | 13.0 | 4.5 (4.4, 4.6) | 155 247 | 18 168 | 20.8 | 8.2 (7.9, 8.5) |
| Stroke | 5593 | 3.1 | 1.0 (1.0, 1.0) | 172 943 | 5231 | 5.1 | 1.4 (1.4, 1.5) |
Age-standardized prevalence and incidence rates were calculated based on the World Health Organization 2000–2025 World Standard Population.
Attendance patterns and socio-demographic disparities in screening uptake
Data from the NELDS have provided critical insights into the impact of socio-demographic factors on diabetic eye care, with a core focus on addressing health inequalities. Contrary to some prior assumptions, our analyses revealed higher odds ratios (ORs) of attending screening for South Asian [OR 1.16, 95% confidence interval (CI) 1.09–1.23], Chinese (OR 1.91, 95% CI 1.39–2.71), and Any other Asian (OR 1.30, 95% CI 1.17–1.45) individuals, with no significant difference observed for Black individuals compared with White [22]. Attendance was positively associated with socio-economic status (OR 1.06 per higher IMD quintile, 95% CI 1.03–1.08) [22] while younger age was associated with lower attendance (OR per decade increase 1.17, 95% CI 1.15–1.19) [22].
Ethnic and age disparities in progression to STDR
Beyond the initial uptake, our research established contemporary incidence rates for progression to STDR and highlighted significant ethnic and age disparities [23]. Despite similar or better attendance at initial screening, Black individuals experienced a 57% increase in the hazards of STDR (HR 1.57, 95% CI 1.50–1.64) and South Asian individuals showed a 36% increase (HR 1.36, 95% CI 1.31–1.42) when compared with White participants. Moreover, younger age groups consistently showed higher hazards of STDR [23]. These findings collectively point to deeper systemic and/or biological factors contributing to health inequalities in disease progression, even when the initial screening engagement is comparable.
Policy implications and equitable risk stratification
These disparities have direct policy implications. Following the introduction of biennial screening in England for low-risk individuals in October 2023 [21], simulations using NELDS data revealed that a blanket biennial approach would delay diagnosis for 56.3% of STDR and 43.6% of proliferative DR cases, disproportionately affecting Black, South Asian, and young individuals [24]. To counteract this, we developed a simplified risk score using age, diabetes duration, and ethnicity, which reduced delayed STDR diagnoses from 39% to 25% while achieving a 43% reduction in screening appointments compared with annual screening [25]. Extensive Patient and Public Involvement and Engagement activities [26] explored perceptions of artificial intelligence (AI)-assisted screening. While 24% of people living with diabetes and 37% of healthcare practitioners provided detailed comments, both groups largely viewed the introduction of AI as inevitable and beneficial for efficiency and cost savings [27]. However, concerns were raised regarding job losses, data security, dehumanization of the patient–practitioner relationship, and AI accuracy across different ethnic groups, highlighting the necessity for targeted education and transparent communication to build trust for successful AI integration [28].
What are the main strengths and weaknesses?
A key strength of the NELDS cohort is its rich ethnic diversity. Only 36% self-reported as having White ethnicity, with substantial proportions from South Asian (34%) and Black (16%) communities. This broad ethnic representation is particularly pertinent given that the rates of diabetes are notably higher among Black and South Asian groups in England compared with White groups [29].
This contrasts sharply with other major cohorts, including the Scottish Diabetes Research Network National Diabetes Cohort (73.1% White) [30], SCONe (77.5% White) [31], and the Danish DD2 and Swedish ANDIS cohorts, which are presumably ethnically homogeneous [32, 33]. While the Singapore SEED study offers Asian-population insights, it lacks White and Black representation, and relies on repeat examinations rather than longitudinal EHR data linkage [34].
Furthermore, more than two-thirds of the individuals reside in the two most deprived IMD quintiles, consistently with national trends showing that ethnic-minority groups are disproportionately affected by socio-economic deprivation [35]. This enables the in-depth investigation of socio-economic inequalities in diabetes outcomes and access to care.
The integration of longitudinal ophthalmic imaging data alongside expert DR grading and linked EHRs is a distinguishing feature of the NELDS, positioning the cohort to contribute to the emerging field of oculomics [36]. While large biobank cohorts such as UK Biobank contain some retinal images [37], they lack systematic DR screening data and detailed ophthalmic EHR data, limiting their utility for developing predictive models that incorporate ophthalmic biomarkers. The NELDS also provides a multi-ethnic external validation set for models developed in other cohorts.
Despite its strengths, the NELDS has some limitations. Images and DR gradings from predecessor screening programmes prior to 2012 are not included. As with all large-scale observational studies, there may be missing or conflicting data within routinely collected EHRs, e.g. regarding diabetes type or date of diagnosis. The retrospective capture of Hospital Episode Statistics is limited to records from April 2015 onwards, potentially underestimating the baseline comorbidity prevalence, though this does not affect the prospective capture of incident events. There is also the potential to miss some mortality data due to the timing of data extraction. Finally, detailed outcomes from specialist ophthalmic services may have been incompletely fed back into the DESP. Future work will address this through linkage to tertiary ophthalmic EHR data [38].
Can I get hold of the data? Where can I find out more?
The NELDS data are not publicly available due to restrictions on data sharing and the highly sensitive nature of the linked, longitudinal records, which include clinical images. The programme is open to considering collaborative work on a case-by-case basis. Approved collaborations requiring access to the data for analysis must be conducted within the secure, approved Trusted Research Environment (TRE) at Homerton Healthcare NHS Foundation Trust, London, UK. Enquiries regarding collaboration opportunities should be directed to the corresponding author.
Ethics approval
Ethical approval was obtained from the NHS Health Research Authority and Health and Care Research Wales (IRAS project ID 265637). Approval from the head of research governance at the lead clinical centre was obtained. Individual consent forms were not deemed necessary for ethics approval, as the study was confined to the utilization of retrospective anonymized data, stored and analysed within a secure TRE (see Supplementary methods). The study was conducted in accordance with the Declaration of Helsinki.
Supplementary Material
Acknowledgements
The authors thank Simon Lam and Alex Figaro from the North East London ICB System Delivery & Data Management team for extracting primary and secondary EHR data for the NELDS cohort.
Contributor Information
Abraham Olvera-Barrios, NIHR Biomedical Research Centre, Moorfields Eye Hospital NHS Foundation Trust & University College London Institute of Ophthalmology, London, United Kingdom.
Alasdair N Warwick, NIHR Biomedical Research Centre, Moorfields Eye Hospital NHS Foundation Trust & University College London Institute of Ophthalmology, London, United Kingdom; Institute of Cardiovascular Science, University College London, London, United Kingdom.
Matilda Pitt, Department of Population Health and Policy, City St George’s, University of London, London, United Kingdom.
Zeinab Schaefer, Department of Population Health and Policy, City St George’s, University of London, London, United Kingdom.
Schanhave Santhirasekaran, Department of Population Health and Policy, City St George’s, University of London, London, United Kingdom.
Ryan Chambers, Diabetes, Homerton Healthcare NHS Foundation Trust, London, United Kingdom.
Louis Bolter, Diabetes, Homerton Healthcare NHS Foundation Trust, London, United Kingdom.
John Anderson, Diabetes, Homerton Healthcare NHS Foundation Trust, London, United Kingdom.
Jiri Fajtl, School of Computer Science and Mathematics, Kingston University, London, United Kingdom.
Sarah Barman, School of Computer Science and Mathematics, Kingston University, London, United Kingdom.
Reecha Sofat, Department of Pharmacology and Therapeutics, University of Liverpool, Liverpool, United Kingdom.
Aroon Hingorani, Institute of Cardiovascular Science, University College London, London, United Kingdom.
Catherine Egan, NIHR Biomedical Research Centre, Moorfields Eye Hospital NHS Foundation Trust & University College London Institute of Ophthalmology, London, United Kingdom.
Adnan Tufail, NIHR Biomedical Research Centre, Moorfields Eye Hospital NHS Foundation Trust & University College London Institute of Ophthalmology, London, United Kingdom.
Alicja R Rudnicka, Department of Population Health and Policy, City St George’s, University of London, London, United Kingdom.
Christopher G Owen, Department of Population Health and Policy, City St George’s, University of London, London, United Kingdom.
the ARIAS Research Group:
Nayani G Adhikari, John Anderson, Sarah Barman, Louis Bolter, Ryan Chambers, Lakshmi Chandrasekaran, Umar Chaudhry, Catherine Egan, Jiri Fajtl, Aaron Lee, Abraham Olvera-Barrios, Christopher G Owen, Paolo Remagnino, Alicja R Rudnicka, Royce D Shakespeare, Adnan Tufail, Charlotte Wahlich, Alasdair N Warwick, and Roshan Welikala
Author contributions
A.O.-B. and A.N.W. designed the study with input from C.G.O. and A.R.R. A.O.-B. and A.N.W. analysed and interpreted the data. The initial manuscript version was written by A.O.-B. and A.N.W. All authors critically reviewed, edited, and approved the final manuscript. The Artificial Intelligence & Automated Retinal Image Analysis Systems (ARIAS) Research Group: John Anderson, Sarah Barman, Louis Bolter, Ryan Chambers, Lakshmi Chandrasekaran, Umar Chaudhry, Emily Y. Chew, Catherine Egan, Jiri Fajtl, Frederick L. Ferris, Aroon D. Hingorani, Aaron Y. Lee, Abraham Olvera-Barrios, Christopher G. Owen, Paolo Remagnino, Alicja R. Rudnicka, Royce Shakespeare, Reecha Sofat, Adnan Tufail, Alasdair N. Warwick, Charlotte Wahlich, Roshan Welikala, Kathryn Willis, Yue Wu.
Supplementary material
Supplementary material is available at IJE online.
Conflicts of interest
None declared.
Funding
This work is funded by the Wellcome Collaborative Award (2022, 224390/Z/21/Z). The research was supported by the National Institute for Health and Care Research (NIHR) Biomedical Research Centre based at Moorfields Eye Hospital NHS Foundation Trust and UCL Institute of Ophthalmology. The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR, or the Department of Health and Social Care.
Use of artificial intelligence (AI) tools
AI-assisted tools were used in a limited capacity during the preparation of this manuscript, restricted to minor rephrasing of the text for clarity and generating short code snippets to support data processing. AI tools did not contribute to the study design, data collection, analysis, interpretation of results, or the intellectual content of this work. All AI-generated outputs were reviewed, verified, and edited by the authors, who take full responsibility for the accuracy and integrity of this manuscript.
References
- 1. World Health Organization. Global report on diabetes. Global report on diabetes. 2016. https://www.who.int/publications/i/item/9789241565257.
- 2. International Diabetes Federation. IDF Diabetes Atlas 11th Edition. 2025. https://diabetesatlas.org/atlas/.
- 3. Ogurtsova K, Guariguata L, Barengo NC et al. IDF diabetes Atlas: Global estimates of undiagnosed diabetes in adults for 2021. Diabetes Res Clin Pract 2022;183:109118. [DOI] [PubMed] [Google Scholar]
- 4. Ong KL, Stafford LK, McLaughlin SA et al. Global, regional, and national burden of diabetes from 1990 to 2021, with projections of prevalence to 2050: a systematic analysis for the Global Burden of Disease Study 2021. Lancet 2023;402:203–34. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Hostalek U. Global epidemiology of prediabetes - present and future perspectives. Clin Diabetes Endocrinol 2019;5:5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Tabák AG, Herder C, Rathmann W, Brunner EJ, Kivimäki M. Prediabetes: a high-risk state for diabetes development. Lancet 2012;379:2279–90. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.10 Year Vision: For diabetes prevention, care and treatment. Diabetes UK. https://www.diabetes.org.uk/about-us/news-and-views/we-present-10-year-vision-diabetes-care-house-commons.
- 8.Cost of devastating complications highlights urgent need to transform diabetes care in the UK. Diabetes UK. https://www.diabetes.org.uk/about-us/news-and-views/cost-complications-highlights-urgent-need-transform-diabetes.
- 9. Fan R, Zhang N, Yang L et al. AI-based prediction for the risk of coronary heart disease among patients with type 2 diabetes mellitus. Sci Rep 2020;10:14457. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. van Dieren S, Beulens JWJ, Kengne AP et al. Prediction models for the risk of cardiovascular disease in patients with type 2 diabetes: a systematic review. Heart 2012;98:360–9. [DOI] [PubMed] [Google Scholar]
- 11. van der Leeuw J, van Dieren S, Beulens JWJ et al. The validation of cardiovascular risk scores for patients with type 2 diabetes mellitus. Heart 2015;101:222–9. [DOI] [PubMed] [Google Scholar]
- 12. van der Heijden AA, Nijpels G, Badloe F et al. Prediction models for development of retinopathy in people with type 2 diabetes: systematic review and external validation in a Dutch primary care setting. Diabetologia 2020;63:1110–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Olvera-Barrios A, Mishra AV, Schwartz R et al. Formal registration of visual impairment in people with diabetic retinopathy significantly underestimates the scale of the problem: a retrospective cohort study at a tertiary care eye hospital service in the UK. Br J Ophthalmol 2023;107:1846–51. [DOI] [PubMed] [Google Scholar]
- 14. Orsi E, Solini A, Bonora E et al. ; Renal Insufficiency And Cardiovascular Events (RIACE) Study Group. Retinopathy as an independent predictor of all-cause mortality in individuals with type 2 diabetes. Diabetes Metab 2023;49:101413. [DOI] [PubMed] [Google Scholar]
- 15. Barlovic D et al. The association of severe diabetic retinopathy with cardiovascular outcomes in long-standing Type 1 diabetes: a longitudinal follow-up. Diabetes Care 2018;41:2487–94. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Vy G, B C, X W, Jjw L, Bc Z. Prospective association between diabetic retinopathy and cardiovascular disease-a systematic review and meta-analysis of cohort studies. J Stroke Cerebrovasc Dis 2016;25:1688–95. [DOI] [PubMed] [Google Scholar]
- 17. Ck K, Tc R, Lh C, Jl G, Mj A. Diabetic retinopathy predicts all-cause mortality and cardiovascular events in both type 1 and 2 diabetes: meta-analysis of observational studies. Diabetes Care 2011;34:1238–44. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Silva PS, Zhang D, Jacoba CMP et al. Automated machine learning for predicting diabetic retinopathy progression from ultra-widefield retinal images. JAMA Ophthalmol 2024;142:171–7. 10.1001/jamaophthalmol.2023.6318 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Ndjaboue R, Ngueta G, Rochefort-Brihay C et al. Prediction models of diabetes complications: a scoping review. J Epidemiol Community Health 2022;76:896–904. [DOI] [PubMed] [Google Scholar]
- 20.NHS diabetic eye screening (DES) programme - GOV.UK. 2025. https://www.gov.uk/health-and-social-care/population-screening-programmes-diabetic-eye.
- 21.Diabetic eye screening intervals extended for people at lowest risk. GOV.UK https://www.gov.uk/government/publications/diabetic-eye-screening-longer-intervals-for-people-at-lowest-risk/diabetic-eye-screening-intervals-extended-for-people-at-lowest-risk.
- 22. Olvera-Barrios A, Seltene M, Heeren TFC et al. Effect of ethnicity and other sociodemographic factors on attendance at diabetic eye screening: a 12-month retrospective cohort study. BMJ Open 2021;11:e046264. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Olvera-Barrios A, Owen CG, Anderson J et al. ; ARIAS Research Group. Ethnic disparities in progression rates for sight-threatening diabetic retinopathy in diabetic eye screening: a population-based retrospective cohort study. BMJ Open Diabetes Res Care 2023;11:e003683. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Olvera-Barrios A, Rudnicka AR, Anderson J et al. ; ARIAS Research Group. Two-year recall for people with no diabetic retinopathy: a multi-ethnic population-based retrospective cohort study using real-world data to quantify the effect. Br J Ophthalmol 2023;107:1839–45. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Pitt M, Olvera-Barrios A, Anderson J et al. ; ARIAS Research Group. A simple score-based strategy to improve equity of the UK biennial diabetic eye screening protocol among people deemed as low risk. Diabetologia 2025;68:1157–68. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Willis K, Chaudhry UAR, Chandrasekaran L et al. ; Study Advisory Group members. What are the perceptions and concerns of people living with diabetes and National Health Service staff around the potential implementation of AI-assisted screening for diabetic eye disease? Development and validation of a survey for use in a secondary care screening setting. BMJ Open 2023;13:e075558. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Wahlich C, Chandrasekaran L, Chaudhry UAR et al. Patient and practitioner perceptions around use of artificial intelligence within the English NHS diabetic eye screening programme. Diabetes Res Clin Pract 2025;219:111964. [DOI] [PubMed] [Google Scholar]
- 28. Willis K, Shakespeare R, Chandrasekaran L et al. ; ARIAS Research Group. What are the perceptions and concerns of people living with diabetes and National Health Service staff around the potential implementation of AI-assisted screening for diabetic eye disease? Diabetic Medicine 2026;43:e70165. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Goff LM. Ethnicity and Type 2 diabetes in the UK. Diabet Med 2019;36:927–38. [DOI] [PubMed] [Google Scholar]
- 30. McGurnaghan SJ, Blackbourn LAK, Caparrotta TM et al. Cohort profile: the Scottish Diabetes Research Network national diabetes cohort - a population-based cohort of people with diabetes in Scotland. BMJ Open 2022;12:e063046. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Tochel C, Bernabeu MO, McTrusty A et al. SCONe: a community-acquired retinal image repository enabling ocular, cardiovascular and neurodegenerative disease prediction. BMJ Health Care Inform 2025;32:e101236. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Christensen DH, Nicolaisen SK, Berencsi K et al. Danish centre for strategic research in Type 2 Diabetes (DD2) project cohort of newly diagnosed patients with type 2 diabetes: a cohort profile. BMJ Open 2018;8:e017273. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Ahlqvist E, Prasad RB, Groop L. Subtypes of Type 2 diabetes determined from clinical parameters. Diabetes 2020;69:2086–93. [DOI] [PubMed] [Google Scholar]
- 34. Majithia S, Tham Y-C, Chee M-L et al. Cohort profile: the Singapore Epidemiology of Eye Diseases study (SEED). Int J Epidemiol 2021;50:41–52. [DOI] [PubMed] [Google Scholar]
- 35. Charlton J, Rudisill C, Bhattarai N, Gulliford M. Impact of deprivation on occurrence, outcomes and health care costs of people with multiple morbidity. J Health Serv Res Policy 2013;18:215–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Zhu Z, Wang Y, Qi Z et al. Oculomics: Current concepts and evidence. Prog Retin Eye Res 2025;106:101350. [DOI] [PubMed] [Google Scholar]
- 37. Warwick AN, Curran K, Hamill B et al. ; UKBB Eye and Vision Consortium. UK Biobank retinal imaging grading: methodology, baseline characteristics and findings for common ocular diseases. Eye (Lond) 2023;37:2109–16. 10.1038/s41433-022-02298-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.BMA Foundation - Research funding. https://x.com/BMAFoundationMR/status/1851346515719311751
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.