Prevalence and risk factors associated with diabetic retinopathy: a cross-sectional study in northern Xinjiang

Abstract

To assess the prevalence and risk factors of diabetic retinopathy (DR) in northern Xinjiang. A hospital-based cross-sectional study was performed on patients above 20 years old who were enrolled in chronic diabetes management in six areas of northern Xinjiang. All subjects received a standardized questionnaire interview of baseline characteristics. Fundus photographs were captured and reviewed by optometrists and ophthalmologists using DR grading software to determine the diagnosis and severity of DR according to the UK guidelines. A total of 3253 patients with diabetes were surveyed, among whom 2972 subjects with graded fundus photographs. The age-standardized prevalence of DR was 17.9% and vision-threatening diabetic retinopathy (VTDR) was 8.7%. In the binary logistic regression model, younger age, male gender, longer duration of diabetes, higher fasting blood glucose, hypertension, and treatment methods (insulin use compared to diet only) were independent risk factors for DR. These risk factors were associated with VTDR except the male gender. The prevalence of DR and VTDR in diabetic patients in northern Xinjiang was 17.9% and 8.7%, respectively. The younger age, male gender, higher fasting blood glucose, hypertension, and insulin use were the main risk factors for DR.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1038/s41598-026-55556-z.

Introduction

Diabetes mellitus (DM) is a series of diseases characterized by chronic hyperglycemia. The prevalence of DM in males (4.3-9.0%) and females (5.0-7.9%) has been increasing worldwide from 1980 to 2014 and is one of the greatest public healthcare challenges^1,2. Diabetic retinopathy (DR) is a common complication of diabetes and the main cause of blindness among the working population in many countries. Retinopathy advances to a severe stage (preproliferative DR or proliferative DR) when visual impairment occurs but it is usually asymptomatic at the initial stage (background DR)³. Over one-third of the 285 million patients with diabetes around the world have signs of DR and a third of them have visual impairment symptoms in different degrees. This affects the quality of life of patients and imposes a burden on society and families^4,5.

A meta-analysis of data from 35 population-based studies around the world in 2008 showed that the prevalence of any DR, PDR, and VTDR, (severe retinopathy and macular edema) in diabetic patients were estimated to be 34.6%, 7.0%, and 10.2%, respectively⁶. So far, there is no national population data on the prevalence of DR in China. A meta-analysis conducted in 2019 estimated that in some parts of China, the prevalence of any DR, nonproliferative DR (NPDR) and proliferative DR (PDR) was 18.45%, 15.06%, and 0.99%, respectively⁵. Even though many studies have examined the epidemiological aspects of DR, the rising incidence of diabetes and regional differences mean that there is still a need to better understand the risk factors of DR.

The UK National Diabetic Retinopathy Screening Program was launched in 2003 and covered the country in 2008. The aim was to detect and reduce diabetes-related vision loss as early as possible. However, a national survey has not been carried out in China. Xinjiang province, located in the northwest of China, has been particularly short of epidemiological characteristics of diabetic ocular complications data. Currently, the Lifeline Express Foundation has implemented DR screening program in several general hospitals on the Chinese mainland in order to achieve early detection and treatment. When compared with the slit lamp, the screening efficiency is improved using an easy-to-operate non-mydriatic fundus camera, and the image recording ensures accuracy, repeatability and objectivity of the results. The DR automatic screening model proposed in recent years has higher accuracy, sensitivity and specificity, which will play a more important role if extensively adopted in DR screening⁷.

In this study, we used data from the Lifeline Express Foundation DR screening program to investigate the prevalence and risk factors for DR in diabetic patients in northern Xinjiang. The findings provide scientific data for the development of government strategies for blindness prevention and treatment.

Methods

The current study used data from the Lifeline Express Foundation DR screening program in 6 hospitals (including Changji People’s Hospital, Qitai County People’s Hospital, Manas County People’s Hospital, Hutubi County People’s Hospital, Huocheng county People’s Hospital, and Shawan County People’s Hospital) in northern Xinjiang from January 2018 to January 2019. Before the screening, patients were recruited through community visits, text messages, phone calls, television stations, and posters. Community workers gathered subject information and conducted preliminary screening. The study included patients who were registered in the local diabetic chronic disease management system or previously diagnosed with diabetes by a physician. Diabetes mellitus was defined as fasting venous blood glucose level ≥ 7.0 mmol/L after an overnight fast of 8–12 h and 2-hour postprandial venous blood glucose leve ≥ 11.1 mmol/L, whereas the glycaemic variable we collected was fasting fingertip capillary blood glucose (FBG) from routine records. All subjects participated in the program for free and were fully informed about the purpose and process of the study. The study followed the Declaration of Helsinki and was approved by the Ethics Committee of the Guangzhou Aier eye hospital. Each subject provided written informed consent.

Patient demographic and clinical characteristics

All participants were interviewed using a uniform questionnaire. Patient’s data, including gender, age, nationality, education background (< 9 or ≥ 9 years), family history of diabetes (yes or no), type of diabetes, duration of DM, treatment method (diet only, oral antidiabetic drugs, insulin or insulin combined with oral antidiabetic drugs), hypertension (informed by the doctor or suggested for anti-hypertension drugs), self-monitoring of blood glucose (SMBG) and smoking status (the duration of smoking status was not recorded) were obtained. SMBG was defined as daily, weekly or irregular based on the monitoring frequency in the previous year; smoking status was classified as current smokers or current non-smokers. Trained nurses measured the height (m) and weight (kg) of subjects and their body mass index (BMI) was calculated as follows: weight divided by height squared(Kg/m²).

Blood glucose data were obtained from the most recent records of fasting fingertip blood glucose(FBG) documented in community diabetes management files or health examination records, which obtained from a fresh capillary whole blood sample via fingerstick puncture after at least 8 h of overnight fasting, with no food or water intake in the morning prior to measurement. Color non-stereo retinal photographs were captured for each eye under non-mydriatic conditions by the same professional photographer using a digital non-mydriatic fundus camera (FundusVue, Mingda Medical, Taiwan), comprising two 45° fields of view (macular centered and disc centered). The fundus images were transmitted via Eye Grader software (V1.0, Telemedicine Cloud Platform, Guangzhou, China) to the Shantou International Ophthalmic Center for diagnostic confirmation and grading. Further examination was recommended for patients if retinopathy was found in the screening.

Grading procedure

All fundus photographs were sent to the reading centre of the Joint International Eye Center and reviewed and graded using a spatial resolution of 1024 × 768pixel. The customized DR grading software VisDR has magnifying, brightness, contrast adjustment, red free, and statistical analysis functions. Diagnosis and severity were classified by professional and qualified optometrists and ophthalmologists according to the UK guidelines (Table 1)⁸. Each photograph was graded primarily by artificial intelligence. Photographs with DR or other eye diseases, as well as 15% of the negative photographs randomly selected were submitted to two professional optometrists for second grading. Any variance between the two gradings was resolved by an arbiter (an ophthalmologist) for the final decision. Quality control was accomplished by recording sensitivity and specificity for each grader where sensitivity ≥ 95% and specificity ≥ 85% were considered satisfactory. Additionally, periodical tests were performed to ensure the accuracy of the results.

Table 1.

Disease grading protocol in UK guideline for Diabetic Retinopathy (DR)⁹

Retinopathy (R)	R0	None
	R1	Background	Microaneurysm(s), Retinal haemorrhage(s)+any exudate
	R2	Preproliferative	Venous beadingVenous loop or reduplicationIntraretinal microvascular abnormality (IRMA)Multiple deep, round, or blot haemorrhagesCotton wool spots in addition to the above features
	R3	Proliferative	New vessels on disc (NVD)New vessels elsewhere (NVE)Pre-retinal or vitreous haemorrhagePreretinal fibrosis tractional retinal detachment
Maculopathy (M)	M0		No maculopathy
	M1		Exudate within 1 disc diameter (DD) of the centre of the foveaCircinate or group of exudates within the maculaRetinal thickening within 1 DD of the centre of the fovea (if stereo available)Any microaneurysm or haemorrhage within 1 DD of the centre of fovea only
Unclassifiable (U)			If opacity, poor view on photographs

DR was classified as background DR (R1), preproliferative DR (R2), proliferative DR (R3) based on the severity of retinopathy. The presence of any type of R2, R3 or maculopathy (M1) was defined as vision-threatening diabetic retinopathy (VTDR). Photographs that could not be evaluated or graded were considered ungradable (U). The grading depended on the worse eye if there was any discrepancy between the two eyes. The grading followed the definition of the other eye if one eye was ungradable.

Statistical analysis

Age-standardized prevalence rates were calculated using the 2010 population census data of Xinjiang⁹. SPSS software version 25.0 (SPSS, Chicago, Illinois, USA) was used to perform data analysis. Subjects with incomplete data or binocular U grading were excluded from the analysis. To assess potential risk factors for DR and VTDR, a t-test was used for continuous variables and a chi-square test was used for categorized variables in univariate analysis. Age, gender, diabetes duration, family history of diabetes, FBG, Treatment method and hypertension and SMBG were included as covariates based on clinical relevance. All covariates were simultaneously entered into the model using the ‘Enter’ method for multivariable analysis. A binary logistic regression model was used to evaluate the odds ratio (OR) and 95% confidence interval (CI) of the potential risk factors for DR and VTDR. A P-value < 0.05 was considered statistically significant.

Results

Patient demographic and clinical characteristics

A total of 3253 diabetic patients were investigated, 2972 having graded fundus photographs as well as complete demographic and clinical data. 259 (8.0%) subjects were excluded because of type 1 diabetes (22), incomplete data (57) or ungradable fundus photographs (202). Demographic and clinical characteristics of patients are shown in Table 2.

Table 2.

Demographic and clinical characteristics of patients with diabetes.

		Subjects with graded pictures (n = 2972)
		MEAN ± SD (range)
Age (year)		61.52 ± 10.07(31–88)
Height (m)		1.63 ± 0.08(1.24–1.87)
Weight (Kg)		69.85 ± 11.61(33.0-139.0)
BMI (Kg/m2)		26.19 ± 3.45(14.67–48.1)
Duration of diabetes (y)		8.53 ± 6.24(1–35)
FBG (mmol/L)		8.78 ± 2.42(2.1–24.0)
		N(percent)
Gender	Male	1451(48.8)
	Female	1521(51.2)
Education background	≤ 9 years	2597(87.4)
	> 9 years	375(12.6)
Hypertension	Yes	1380(46.4)
	No	1592(53.6)
Smoking status	Current smokers	623(21.0)
	Current no-smokers	2349(79.0)
Family history of diabetes	Yes	671(22.6)
	No	2301(77.4)
Treatment method	Diet	230(7.7)
	Oral drugs	1491(50.2)
	Insulin	398(13.4)
	Both	853(28.7)
SMBG	Daily	95(3.2)
	Weekly	825(27.8)
	Irregular	2052(69.0)

BMI, body mass index; FBG, fasting blood glucose; SMBG, self-monitoring of blood glucose.

Prevalence of DR and VTDR

DR and VTDR had age-standardized prevalence of 17.9% (95% CI, 0.167–0.192) and 8.7% (95% CI, 0.078–0.096), respectively. According to the standardized calculation of age, the prevalence of R1, R2, R3 and M1 were 12.7% (95% CI, 0.117–0.138), 2.5% (95% CI,0.020–0.031), 2.6% (95% CI, 0.021–0.031) and 5.3% (95% CI, 0.045–0.061), respectively (Table 3).

Table 3.

Prevalence and grading of DR.

	Number of cases in subjects	Crude proportion of subjects (%)	Age-standardized proportion of subjects (95%CI)
DR	673	22.6	17.9(0.167-0.192)
R1M0	346	11.6	9.2(0.083-0.101)
VTDR	327	11.0	8.7(0.078-0.096)
M1	199	6.7	5.3(0.046-0.060)
R1M1	133	4.5	3.5(0.030-0.041)
R2M0	51	1.7	1.3(0.010-0.017)
R2M1	44	1.5	1.1(0.008-0.015)
R3M0	76	2.6	2.0(0.016-0.025)
R3M1	22	0.7	0.6(0.003-0.009)

DR, diabetic retinopathy; VTDR, vision-threatening diabetic retinopathy; R1, background diabetic retinopathy; R2, pre-proliferative diabetic retinopathy; R3, proliferative diabetic retinopathy; M0 no maculopathy; M1, maculopathy

Assessment of associated risk factors

The characteristics of patients were classified based on the presence or absence of DR and VTDR (Table 4). In univariate analysis, DR was significantly associated with a family history of diabetes, duration of DM, latest FBG, hypertension, treatment methods and SMBG (all P ≤ 0.001), the results in VTDR were similar, except for the family history of diabetes in VTDR(P = 0.016). Age (P = 0.081) and gender were marginally linked to DR (P = 0.06). In multivariate-adjusted analysis, we discovered that younger age (OR, 0.0951; 95% CI, 0.940–0.963), male gender (OR, 0.812; 95% CI, 0.664–0.993), longer duration of diabetes (OR, 1.142; 95% CI, 1.121–1.163), higher FBG (OR, 1.132; 95% CI, 1.090–1.176), hypertension (OR, 1.456; 95% CI, 1.182–1.794) and treatment method (insulin and insulin combined with oral antidiabetic drugs) were independent risk factors for DR (Table 5). The risk factors were comparable to those of VTDR, except for the male gender (P = 0.664). Smoking and educational background had no statistical effect on DR or VTDR.

Table 4.

Related risk factors for DR and VTDR in these subjects.

		Non-DR(n=2299)	DR(n=673)	P-Value	Non-VTDR(n=2645)	VTDR(n=327)	P-Value
Age (year)		61.70±10.13	60.93±9.86	0.081	61.56±10.05	61.20±10.27	0.538
Gender (Male)%		1112(47.9)	350(52.0)	0.060	1295(49.0)	156(47.7)	0.669
BMI (Kg/m2)		26.22±3.51	26.10±3.26	0.389	26.22±3.47	25.98±3.32	0.221
Education background (≤9 y)%		2020(87.9)	577(85.7)	0.144	2313(87.4)	284(86.9)	0.759
Smoking status (%)		485(21.1)	138(20.5)	0.740	563(21.3)	60(18.3)	0.218
Family history of diabetes(%)		483(21.0)	188(27.9)	<0.001	580(21.9)	91(27.8)	0.016
Duration of diabetes (y)		7.27±5.46	18.82±6.80	<0.001	7.88±5.78	13.77±7.26	<0.001
FBG (mmol/L)		8.53±2.23	9.62±2.83	<0.001	8.53±2.23	9.62±2.83	<0.001
Hypertension (%)		1197(52.1)	395(58.7)	0.002	1381(52.2)	211(64.5)	<0.001
Treatment method(%)				<0.001			<0.001
	Diet	211(9.2)	19(2.8)		224(8.5)	6(1.8)
	Drug	1338(58,2)	153(22.7)		1427(54.0)	64(19.6)
	Insulin	238(10.4)	160(23.8)		311(11.8)	87(26.6)
	Both	512(22.3)	341(50.7)		683(25,8)	170(52.0)
SMBG%				0.001			<0.001
	Daily	64(2.8)	31(4.6)		75(2.8)	20(6.1)
	Weekly	613 (26.7)	212(31,5)		721(27.3)	104(31,8)
	Irregular	1622(70.6)	430(63.9)		1849(69.9)	203(69.0)

DR, diabetic retinopathy; VTDR, vision-threatening diabetic retinopathy; BMI, body mass index; FBG, fasting blood glucose; SMBG, self-monitoring of blood glucose.

Table 5.

Binary logistic regression model analysis ofDR and VTDR in these subjects.

Risk factors			Multivariate adjusted DR				Multivariate adjusted VTDR
			B	P	OR	95%CI	B	P	OR	95%CI
Age			-0.050	< 0.001	0.951	0.940–0.963	-0.051	< 0.001	0.951	0.935–0.966
Gender			-0.208	0.042	0.812	0.664–0.993	0/057	0.664	1.059	0.819–1.368
Duration of diabetes			0.133	< 0.001	1.142	1.121–1.163	0.122	< 0.001	1.130	1.105–1.155
Family history of diabetes			0.061	0.617	1.063	0.836–1.351	0.016	0.917	1.016	0.751–1.375
FBG			0.124	< 0.001	1.132	1.090–1.176	0.095	< 0.001	1.100	1.048–1.154
Hypertension			0.376	< 0.001	1.456	1.182–1.794	0.581	< 0.001	1.788	1.364–2.342
Treatment method	Diet			< 0.001				< 0.001
		Oral drugs	-0.194	0.467	0.823	0.488–1.390	-0.298	0.422	0.742	0.359–1.535
		Insulin	1.274	< 0.001	3.575	2.068–6.179	0.987	0.010	2.682	1.271–5.659
		Both	0.914	< 0.001	2.494	1.461–4.257	0.420	0.266	1.522	0.725–3.195
SMBG		Daily		0.581				0.162
		Weekly	-0.123	0.648	0.884	0.521–1.499	-0.305	0.325	0.737	0.402–1.352
		Irregular	-0.139	0.596	0.870	0.521–1.454	-0.483	0.108	0.617	0.342–1.113

DR, diabetic retinopathy; VTDR, vision-threatening diabetic retinopathy; OR, odd ratio; CI, confidence interval; FBG, fasting blood glucose

Discussion

Epidemiological studies on DR in the Xinjiang region remain scarce.To address this gap, a cross-sectional survey was conducted in community hospitals across Northern Xinjiang. However, differences in study design and diagnostic procedures have made comparisons across studies more challenging. Several grading systems and early disease management protocols for DR screening have been published previously^10,11. Some of these protocols, evertheless, propose unduly complex grading systems, which may increase the burden on ophthalmic services in hospitals. Our working group adopted the grading strategy from the UK guidelines, simplifies existing screening protocols and has shown higher sensitivity and specificity than the Exeter criteria¹². This strategy helps implement quality-controlled screening services for diabetic patients.

Our study found that the prevalence of DR and VTDR was 17.9% and 8.7%, respectively, after standardizing their ages.The primary contributing factors of retinopathy were younger age, male gender, higher FGB, hypertension and treatment methods (insulin or insulin combined with oral anti-diabetic drugs).

In recent years, numerous studies have reported on the prevalence of DR. A survey in Hong Kong¹³revealed a 39.0%prevalence of DR and a 9.8% prevalence of VTDR among diabetic patients.The SEED found 26.2% and 6.93% prevalence of DR and VTDR, respectively, among Chinese¹⁴. The 2023 national DR prevalence stood at 16.3%¹⁵, a meta-analysis revealed that the overall prevalence of DR was 23.0% in 19 regions in China⁵, Liu¹⁶ and Zhang et al.¹⁷ conducted additional research in the northern (37.07% vs. 28.7%) and southern (27.55% vs. 26.9%) regions of China. These studies reveals substantial variations in DR prevalence across Chinese regions and international (DR 16.3%-39.0%, VTDR 6.93%-9.8%). Notably, none of these studies presented findings from the Xinjiang province. while our DR prevalence falls within the moderate range, VTDR prevalence exceeds comparative studies. As no prior studies on DR prevalence in Xinjiang region were identified, longitudinal comparisons remain unfeasible.

Disparities in prevalence may stem from multiple contributing factors. First, significant variations exist across studies in protocol design, participant selection criteria, and diagnostic classification systems. Secondly, Xinjiang’s unique geographical position-characterized by a temperate continental climate, mountainous terrain, aridity, and significant temperature variations-has shaped a local diet predominantly centered on high-carbohydrate foods like rice and flour-based staples, meat, and even animal fats. This dietary pattern provides ample calories to combat cold but offers minimal intake of dietary fiber from vegetables. Such nutritional imbalances severely undermine blood glucose control and weight management, All of these factors may be potential factors that acceleate the progression of the diabete. Finally, disparities in primary healthcare resources across regions present critical challenges. Located in northwest China, Xinjiang contends with limited medical resources, relatively lower economic development, and vast distances that impede transportation. Consequently, health management education remains markedly inadequate¹⁸. These factors collectively diminish diabetic patients’ willingness to seek medical care, with some even refusing comprehensive eye examinations. This significantly obstructs timely diagnosis and disease management. Many patients seek medical attention only after experiencing significant vision decline, leading to treatment delays. Collectively, this may explain why DR prevalence remains moderate while VTDR prevalence is notably elevated.

Our results indicate that diabetes course, hypertension, and insulin use as primary risk factors influencing the progression of both DR and VTDR, Similar findings have been documented in prior studies. Longer diabetes duration^13,]19 correlates with higher risks of both DR and VTDR. Prolonged hyperglycemia induces persistent damage to microvessels, causing leakage of intravascular fluid components into interstitial spaces. This results in retinal edema, ischemia, and neovascularization.Clinically, diabetes and hypertension commonly coexist^20–22. Comorbid hypertension increases DR risk by 39.7% and VTDR risk by 77.3% in our study, concurrent hypertension exacerbates microvascular injury, leading to further clinical deterioration. Our findings demonstrate that insulin therapy increases the risk of DR and VTDR compared to dietary management alone, aligning with Tomic et al.‘s research^23–25. However, these findings do not necessarily indicate that insulin itself accelerates DR progression. Increased insulin therapy typically implies the development of diabetes. Consequently, insulin therapy serves as an indicator for predicting DR and can function as an early warning signal. This alerts clinicians to closely monitor the advancement of retinopathy.

Our study also considered independent risk factors such as younger age, FBG and male gender. The likelihood of developing retinopathy decreases with age. A similar conclusion was reached in population-based research in mainland China²⁶, which revealed that the threat to vision increased even if the incidence decreased with age. The lower prevalence of DR in old diabetic patients appears to be associated with an increase in mortality, the risk of death from medium-and high-risk conditions increases with age²⁷. After adjusting for factors such as diabetes duration, our results showed that each additional year of age was associated with an approximately 4% reduction in the risk of DR (OR = 0.96 per year). This suggests that, given the same disease duration, advanced age itself may be a slight protective factor. This could be because the subtype of diabetes in younger-onset patients exerts a greater systemic impact and more severe cumulative damage. Simultaneously, survival biases cannot be ignored, as multiple factors may collectively contribute to the observed reduction in DR prevalence with increasing age. FBG was found to be an independent risk factor in both DR and VTDR groups. Previous research has shown that the effect of blood glucose on retinopathy is persistent^28,29, which is consistent with our findings. It should be noted that while diabetes was diagnosed using venous blood glucose criteria, the regression analysis used fasting capillary glucose, which is widely accepted for routine diabetes monitoring due to its convenience and good concordance with venous glucose under fasting conditions. The male gender in our study had a 23.6% higher risk of developing DR than the female gender, but had no significant effect on VTDR. This is consistent with the results of LALES³⁰. Nevertheless, other pieces of research³¹ indicate that gender has no significant effect on DR. It remains to be determined whether the higher risk of DR in men is related to factors such as lifestyle and dietary habits.

In univariate analysis, family history of diabetes and SMBG were linked to the occurrence of DR and VTDR but did not demonstrate a statistically significant difference after adjusting for all other confounders. Uhlmann et al.³² reported that the occurrence of DR is influenced by genetic factors (determined gene expression), multifactorial variability and environmental factors. Our findings further suggest that the occurrence of DR or VTDR is the result of a combination of environment and heredity. More than two-thirds of patients in our study had irregular SMBG, which was similar to the baseline data of the COMPASS study (SMBG in Patients with Diabetes on Insulin Study)³³. A survey conducted in Northeast China revealed that 77.9% of participants considered relevant examinations necessary only when visual symptoms occurred³⁴. These findings indicate that in some regions of China, screening for DR remains inadequate, and health management education is severely lacking. SMBG, as part of the health management strategy of diabetics, has far-reaching significance in hyperglycemia control. Our screening found that most community institutions in China lack appropriate equipment for capturing fundus images, resulting in undetected conditions and delayed treatment. Eun Young Choi and colleagues developed a ChatGPT-4 automated risk calculator³⁵, enabling healthcare workers to efficiently identify patients requiring retinal examinations using solely medical history and laboratory data. This approach would benefit numerous clinicians and patients with limited access to healthcare services in the future.

There are some drawbacks to this study. First, The absence of stratified sampling for screening locations and populations reduced representativeness and introduced selection bias. The exclusion of severe, hospitalized, and deceased cases led to underestimation of DR prevalence (survivor bias). Second, This study did not employ ordinal or partial proportional odds models for DR severity, nor were model assumptions formally tested. Although clinical considerations informed this choice, the absence of such models represents an analytical limitation. Finally, no more recent data from the same program or cohort were generated during the study period, meaning that updated information on the study population or outcomes after 2019 could not be incorporated.

Conclusions

The prevalence of DR and VTDR in diabetes patients in northern Xinjiang differ from those in other areas. The prevalence of VTDR is markedly higher, implying that more diabetics in Xinjiang are facing the risk of vision loss and blindness. As such, local health care and medical institutions should develop an appropriate DR screening program and conduct health education to lower the risk of blindness and the burden on families and society.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Data curation: Xiaoye Gao, Yinu Ma, Xueyi Chen.Formal analysis: Xiaoye Gao, Yinu Ma.Validation: Xueyi Chen, Jinglin Zhang.Writing – original draft: Xiaoye Gao, Yinu Ma, Jinglin Zhang, Xueyi Chen.Writing – review & editing: Xiaoye Gao, Yinu Ma, Xueyi Chen.

Data availability

All data supporting the findings of this study are available within the paper and its Supplementary Information.

Declarations

Competing interests

The authors declare no competing interests.