Prevalence of poor vision and its associated factors among middle school students: a cross-sectional study

Abstract

Background

Poor vision is a global public health challenge, with rising prevalence and earlier onset among Chinese adolescents. This study analyzed the prevalence and factors associated with poor vision among middle school students in Nantong.

Methods

Ophthalmological examinations data were collected from 196,969 students across 32 middle schools (2018 to 2023). The Cochran-Armitage trend test analyzed the temporal and grade-level trends. A multi-stage stratified cluster sampling survey collected psychological and behavioral questionnaire data (2018: 1,392 vs. 2023: 1,380) from 5 out of the 32 schools. Exploratory factor analysis (EFA) identified potential factor structure, multivariate logistic regression quantified adjusted odds ratios (OR), and gender-stratified structural equation model (SEM) examined association pathways.

Results

The prevalence of poor vision increased from 86.95% (2018) to 88.67% (2023), showed upward trends across grade-level (β = 0.027, P-trend < 0.001) and survey years (β = 0.007, P-trend < 0.001). Prevalence among females was higher (P < 0.001). In 2018 students, the eye behavior factor (OR = 1.35, 95%CI: 1.17 to 1.56) and the family heredity and environment factor (OR = 1.21, 95%CI: 1.06 to 1.40) were positively associated with poor vision. By 2023, the family heredity and environment factor (OR = 1.31, 95%CI: 1.14 to 1.50) remained significant, and the psychological factor was significantly associated factor (OR = 1.17, 95%CI: 1.03 to 1.34). Gender-stratified analysis revealed that the adaptive eye regulation factor was inversely associated with poor vision (β = -0.191, P = 0.049) in males, while in females, it showed a direct positive association (β = 0.178, P = 0.013).

Conclusion

The psychological factor is significantly associated with poor vision in students, with notable gender-specific pathways. These findings highlight the need for integrated, gender-tailored interventions to address this public health challenge.

Trial registration

This study was approved by the Ethics Committee of Affiliated Hospital of Nantong University (Approval No. 2021-K118-02).

Supplementary Information

The online version contains supplementary material available at 10.1186/s12889-026-26941-0.

Introduction

Poor vision, primarily driven by uncorrected refractive error, constitutes a leading public health challenge among children and adolescents globally [1]. The World Health Organization estimates that over 2.2 billion affected individuals worldwide [2]. Over the past three decades, the prevalence of poor vision among Chinese schoolchildren has increased rapidly, with a trend toward earlier onset. This condition profoundly impacts academic performance, psychological well-being, and long-term development [3].

The causes of poor vision among students are complex and primarily influenced by genetic and environmental factors [4]. For example, Wang et al. conducted a study from 2010 to 2015 on a prospective cohort of 19 primary schools and 22 middle schools in Guangzhou, China, and found that children with myopic parents had a significantly increased risk of developing myopia [5]. Additionally, a study by Wang from 2016 to 2021 surveyed middle school students in Shanghai’s Jing’an District through questionnaires, revealing that environmental factors such as academic pressure, eye usage habits, frequency of electronic product use, and outdoor activity duration significantly impacted visual health [6].

Emerging evidence suggests that the consequences of poor vision extend beyond visual acuity into the psychological domain. A large-scale cross-sectional study among 19,425 junior middle school students in rural China reported a significant positive association between visual impairment and poorer mental health, with eyeglasses usage associated with better psychological outcomes [7]. Similarly, a six-year multicenter study conducted in Jiangsu Province, China found a significant population level association between students with visual acuity below 4.0 and those reporting symptoms of depression compared to their peers with normal vision (OR = 1.90, 95%CI: 1.53 to 2.37) [8]. Although direct evidence in adolescents remains limited, studies in visually impaired populations suggest potential associations between visual problems and psychological well-being, warranting investigation in school-aged samples.

Gender represents another under-investigated dimension in this field. Epidemiological studies consistently report higher myopia prevalence among females than males during adolescence [9]. A 2025 study involving 2,595 Chinese children and adolescents demonstrated that while myopia prevalence was significantly higher in females (60.23% vs. 52.71%, χ² = 14.93, P < 0.05), no sex difference was observed in AL-related visual impairment, suggesting that gender disparities in poor vision may be driven by behavioral, psychosocial, or reporting differences rather than purely biological factors [10]. These patterns raise important questions about whether the pathways connecting psychological factors to poor vision are gender-specific, a hypothesis that has received minimal empirical testing in adolescent school-based populations.

Although a large amount of literature focuses on the clinical diagnosis of “myopia” and its long-term eye complications, this study focuses on the public health issue of “uncorrected visual impairment” detected through screening. Its primary impact is the immediate visual function of students in school, and the main intervention methods are visual health promotion and behavioral intervention, which are different in nature and intervention level from the clinical pathway management of “pathological myopia” or “visual impairment”.

Despite policy initiatives promoting vision health education, two research gaps persist. First, there is a lack of longitudinal data tracking evolving risk profiles, particularly psychological interplay. Second, we have an inadequate understanding of gender-specific potential pathways linking psychological factors to poor vision. This study analyzed the prevalence of poor vision (2018–2023) and its associated factors among middle school students in Nantong city, China. Using multivariable regression and structural equation model (SEM), we investigated potential structural differences in factors associated to poor vision among middle school students from 2018 to 2023 and explored potential gender-specific pathways. The findings can provide valuable insights for developing targeted interventions to address this escalating public health challenge.

Methods

Study design

This repeated cross-sectional study investigated the prevalence and structural differences associated with poor vision among middle school students in Nantong, China from 2018 to 2023. According to the provincial and municipal monitoring protocols [11, 12], we collected ophthalmological examinations data (n = 196,969) in 32 middle schools (26 junior middle schools, 5 senior middle schools, 1 vocational school). Annually, a multi-stage stratified cluster sampling survey was conducted and five schools (2 junior middle schools, 2 senior middle schools, 1 vocational school) were randomly selected based on the distribution of school types. Full grade coverage, randomly selecting 2 classes from each grade and conducting surveys on a whole class basis, inviting all students from the selected classes to participate. There are 59 variables included in the analysis of this study. According to the principle of events per variable (EPV) ≥ 20 (conservative estimation), a minimum of 1,180 students are required to participate in the questionnaire [13]. Questionnaires were distributed to participants in both survey years. In 2018, 1,414 questionnaires were distributed, with 1,392 valid questionnaires included in the final analysis after excluding 22 invalid responses (valid questionnaires rate: 98.4%). In 2023, 1,396 questionnaires were distributed, resulting in 1,380 valid responses following the exclusion of 16 invalid questionnaires (valid questionnaires rate: 98.9%) (Fig. 1). Descriptive statistics were used to examine the trend of prevalence, Exploratory factor analysis (EFA) was used to determine the potential structure of factors associated with poor vision during the survey period, and gender-stratified SEM was used to evaluate the direct and indirect associations between the psychological factor and poor vision (Fig. 2). All participants signed informed consent forms, and minors obtained the consent of their guardians. This study has been approved by the Ethics Committee of Affiliated Hospital of Nantong University (Approval No. 2021-K118-02). And the collected data are only for academic research purposes. All methods were performed in accordance with the relevant guidelines and the Declaration of Helsinki.

Fig. 1.

Study design flowchart

Fig. 2.

Structural equation model

Note: Factors from 2023 are denoted as Factor 1b-Factor 4b; Factor 1b: Psychological factor, Factor 2b: Adaptive eye regulation factor, Factor 3b: Active eye protection factor, Factor 4b: Family heredity and environment factor; (a) Structural Equation Model 1: initial model. (b) (Revised) Structural Equation Model 2: final gender-specific model; The coefficients marked on the path are standardized estimates; The solid regression line indicates statistical significance at a 95% confidence level; The dotted line indicates no statistical significance; *P < 0.05, **P < 0.01, ***P < 0.001

Inclusion and exclusion criteria

Inclusion criteria: (1) Full-time students officially registered at middle schools in Tongzhou District, Nantong City; (2) Aged 10 to 20; (3) Provide signed informed consent (from parents/guardians for minors). Exclusion criteria: (1) Without complete ophthalmological examination results; (2) Diagnosed with any ocular or systemic disease that may affect visual acuity or study outcomes; (3) Missing data (due to the COVID-19 or other reasons affecting data collection). Invalid questionnaire criteria: (1) Excessive missing data (> 20% of items); (2) Same value answer; (3) Logical contradictions in key responses; (4) Failure to follow filling instructions [14, 15].

Ophthalmological examinations data

Certified ophthalmic technicians conducted all vision assessments in the classroom under standardized lighting conditions of at least 300 lx. The bilateral uncorrected visual acuity was measured using the Standard For Logarithmic Visual Acuity Charts (GB11533-2011) [16]. Poor vision is defined as a monocular uncorrected visual acuity of less than 5.0 (worse-seeing eye) [17, 18], corresponding to approximately 0.00 logMAR or 20/20 Snellen. Poor vision is further categorized into mild (4.9 or approximately 0.10 logMAR or 20/25 Snellen), moderate (4.6 to 4.8, corresponding to approximately 0.22 to 0.40 logMAR or 20/50 to 20/32 Snellen), and severe (≤ 4.5, corresponding to approximately ≥ 0.52 logMAR or ≤ 20/63 Snellen) based on national standards [19]. We use worse-seeing eye for study, which is consistent with the World Health Organization’s recent emphasis on reporting uncorrected vision to better estimate the demand for refractive error services [20].

Questionnaires data

The study used a questionnaire titled “Middle School Student Health Status and Influencing Factors Questionnaire” to collect data (2018: Cronbach’s alpha:0.825; KMO:0.871, P < 0.001; 2023: Cronbach’s alpha:0.834; KMO:0.880). This tool is electronically managed through online platforms within the school environment. During the survey, students were asked to input their checklist barcode to facilitate automatic linking between the health factor questionnaire and physical examination records. The questionnaire utilized in this study was an official instrument from the Monitoring Project for Common Diseases and Health Influencing Factors among Students in Nantong City. Its validity is based on content validity, established through its development by expert panels and its adoption for standardized population-level health surveillance [12].

The questionnaire covers multiple aspects and aims to comprehensively evaluate student health. Including demographic information, lifestyle behaviors (dietary practices, and exercise routines), indoor visual habits, outdoor visual habits and screen time and outdoor activities. In this study, the grade-level analysis adopts China’s official terminology from the Guidelines for the use of English in public service areas (GB/T 30240.6–2017): junior middle school (7-9th grade), senior middle school (10-12th grade) and vocational school (10-12th grade) [21] (Additional File A. Questionnaire).

Quality control measures

This study follows the national and provincial quality control standards for monitoring common diseases among students, ensuring standardized operations through unified training and assessment, using calibrated and qualified instruments, and strictly following physical examination standards. The city adopts an electronic monitoring system for questionnaire filling and data entry, and 5% of the samples are selected by the Center for Disease Control and Prevention for on-site review. The health and education departments jointly review qualifications and carry out full supervision to comprehensively ensure data quality [12].

Statistical Analysis

Descriptive statistics were used to describe the characteristics of the study population, representing categorical variables as frequencies (%) and continuous variables as mean (SD). The Cochran-Armitage trend test was used to evaluate the temporal and grade-level trends of the prevalence of poor vision, with the β coefficient representing the magnitude of the trend.

To identify factors associated with poor vision, univariate logistic regression was performed (P < 0.05 significance threshold). We used the variance inflation factor (VIF < 5) to evaluate multicollinearity [22]. To investigate the potential structure of associated factors, we conducted EFA using principal component analysis with varimax-rotation method. The analysis depends on the variables included and the resulting EFA structure. After conducting Kaiser-Meyer-Olkin (KMO) test (> 0.5) and Bartlett sphericity test (P < 0.001) to evaluate the applicability of EFA, this dimensionality reduction technique was applied to significant variables in univariate analysis (P < 0.05). Factors with eigenvalues > 1 were retained, and variables with factor loadings ≥ 0.40 were included, while considering parallel analysis and theoretical interpretability [23].

We constructed a multivariate logistic regression model using EFA derived factor scores to estimate the adjusted odds ratio (OR) and 95% confidence interval (95%CI) for each factor. A gender-stratified SEM was implemented using the Mplus software to examine psychological-behavioral pathways and gender-specific potential pathways. First, a measurement model was established with four latent variables. All observed variables were treated as ordered categorical. The measurement model was validated through internal consistency (CR > 0.7; rho_A > 0.6), convergent validity (AVE > 0.5), and discriminant validity (HTMT < 0.85). Subsequently, a structural model was built to test the pathways from the latent variables to the observed outcome. The model was estimated separately for male and female groups using the weighted least squares mean and variance adjusted (WLSMV) estimator, which is robust for categorical data. The initial model included all possible paths. The model was evaluated by comparing fit index (CFI > 0.9), root mean square error of approximation (RMSEA < 0.100), χ², degrees of freedom (df), tucker-lewis index (TLI > 0.9) and weighted root mean square residual (WRMR < 1.0). Standardized coefficients were reported for all paths. Based on the statistical significance of the paths (p < 0.05), the model was simplified by removing non-significant pathways for each gender group, resulting in a revised SEM model. Finally, considering the nested structure of the data (students within the school), we implemented a comprehensive robustness analysis comparing two approaches: the cluster robust standard errors (CRSE) approach and the bootstrap analysis with 5,000 iterations without clustering adjustment [24].

All analyses were performed using STATA 17.0 (StataCorp LLC, College Station, TX, USA) and Mplus 7.4 (Muthén & Muthén, Los Angeles, CA, USA), with statistical significance set at P < 0.05 (two-tailed test).

Results

Demographic characteristics

The study monitored 196,969 middle school students in Nantong from 2018 to 2023. Among these, 172,147 exhibited poor vision, yielding an overall prevalence rate of 87.40% (Table S1). Annual rates of poor vision were reported as 86.95%, 86.09%, 86.57%, 85.73%, 89.08%, and 88.67% for the years 2018 to 2023, respectively. The Cochran-Armitage trend test demonstrated a significant increase in poor vision prevalence with advancing grade-level (β = 0.027, z = 55.55, P-trend < 0.001). Additionally, the prevalence exhibited an upward trend over the years (β = 0.007, z = 15.64, P-trend < 0.001). Gender-specific analysis revealed that female students across all grades consistently had a higher prevalence of poor vision compared to male students (P < 0.001) (Table 1 and Table S2).

Table 1.

The epidemiological characteristics of poor vision

Grade	Gender	Prevalence of poor vision (%)						Total (%)	Grade average (%)	Gender Chi-square test
		2018	2019	2020	2021	2022	2023			χ2	P
1st of Junior Middle School	Male	76.86	76.44	76.56	77.12	79.01	77.96	77.53	81.86	689.72	P < 0.001
	Female	87.24	86.02	85.92	85.89	87.26	85.56	86.42
2nd of Junior Middle School	Male	80.58	81.96	82.20	80.66	84.99	83.23	82.58	86.12	557.72	P < 0.001
	Female	89.06	90.29	90.48	88.00	91.15	89.20	89.86
3rd of Junior Middle School	Male	84.06	84.39	84.58	85.13	87.08	86.67	85.41	88.51	455.92	P < 0.001
	Female	91.77	91.48	91.57	90.89	92.38	92.26	91.77
1st of Senior Middle School	Male	--	--	--	--	92.96	93.33	93.13	94.48	30.94	P < 0.001
	Female	--	--	--	--	95.21	96.64	95.82
2nd of Senior Middle School	Male	90.59	89.62	90.65	90.10	91.20	91.66	90.74	92.90	222.03	P < 0.001
	Female	95.47	94.55	94.97	95.11	95.53	94.22	95.05
3rd of Senior Middle School	Male	--	--	--	--	92.35	94.10	93.13	94.98	58.86	P < 0.001
	Female	--	--	--	--	96.16	97.46	96.78
Overall (%)		86.95	86.09	86.57	85.73	89.08	88.67	87.40

Note: The Cochran-Armitage trend test demonstrated a significant increase in poor vision prevalence with upward trends across grade-level (β = 0.027, z = 55.55, P-trend < 0.001) and survey years (β = 0.007, z = 15.64, P-trend < 0.001); Data for 1^st and 3^rd senior middle school grades were only available from 2022 onward, as these grades were not included in the surveillance program prior to that year; Missing data are denoted by “--”

No significant differences existed between the 2018 and 2023 questionnaire samples regarding demographic characteristics (grade, gender) or visual acuity distribution (P > 0.05) (Table S3).

Univariate analysis

Univariate logistic regression analysis found that several covariates were significantly associated with poor vision in both study years. Sociodemographic factors, particularly gender and poor parental vision, remain important predictors. In addition, modifiable lifestyle factors such as eye care, outdoor activities, and sleep remain important in these two years (Table S4).

Diagnoses of multicollinearity

In the univariate analysis of poor vision, we conducted multicollinearity diagnostics for variables demonstrating statistically significant associations. Variance inflation factor (VIF) was used to assess potential multicollinearity. The research results indicate that there is no multicollinearity (2018 max VIF: 1.40; 2023 max VIF: 3.12) (Table S5). These findings emphasize the robustness of our model specifications over the past two years.

Exploratory factor analysis

EFA was conducted on variables significant in univariable regression (P < 0.05). For 2018, the KMO was 0.600, and the Bartlett sphericity test confirmed the existence of a association between the variables and the adequacy of the factorial reduction (χ² = 1,138.745, P < 0.001). Parallel analysis (Horn’s method, 300 random simulations, 95%) suggests retaining 3 factors (adjusted eigenvalues > 1) (Table S6, Figure S1). However, considering that the eigenvalue of the fourth factor is close to 1 and it contains a variable of significant theoretical importance in the study of poor vision, after taking into account the scree plot and cumulative variance contribution rate, we ultimately extracted four factors and explained 56.60% of the variance. Factor 1a (School lifestyle factor): This factor was comprised of average daily time spent watching TV and number of eye exercises done daily. Factor 2a (Eye behavior factor): This factor was defined by behaviors such as reading or looking at screens while walking or riding and whether lights are turned off when looking at night. Factor 3a (Family heredity and environment factor): This factor included parents’ visual condition and number of physical education classes per week. Factor 4a (Classroom eye policy factor): This factor captured variables associated to regularly exchange class seats and lighting conditions when reading or writing at night.

For 2023, factorability was similarly confirmed (KMO = 0.668; Bartlett sphericity test: χ² = 3,418.204, P < 0.001). Taking the same consideration, we extracted four factors to explain 66.74% of the variance. Factor 1b (Psychological factor): This factor was comprised of items associating to feeling hopeful, feeling happiness and sense of happy in life. Factor 2b (Adaptive eye regulation factor): This factor included gender, the number of physical education classes per week and average daily sleep duration. Factor 3b (Active eye protection factor): This factor was defined by number of eye exercises done daily and average daily outdoor activity time during daylight hours. Factor 4b (Family heredity and environment factor): This factor, consistent with 2018, contained participation in tutorial classes and parents’ visual condition. (Table 2, Table S4).

Table 2.

The factor matrix of varimax-rotation

Variable	Factor 1a	Factor 2a	Factor 3a	Factor 4a	Communalities
Average daily time spent watching TV	0.7357	0.0962	0.032	0.0308	0.5525
Average daily time spent using the computer	0.7109	0.1844	-0.0414	-0.0869	0.5486
Number of eye exercises done daily	0.5803	-0.1893	-0.3087	-0.1077	0.4795
Average daily sleep duration	0.5067	-0.4183	0.2081	0.2562	0.5407
Reading or looking at screens while walking or riding	0.0154	0.8197	-0.0244	-0.0277	0.6735
Whether lights are turned off when looking at night	0.0756	0.8018	0.0476	0.0629	0.6549
Parents’ visual condition	-0.0687	0.0837	0.7402	-0.0554	0.5627
Number of physical education classes per week	0.3154	-0.2349	0.4663	0.3796	0.5162
Regularly exchange class seats	-0.0835	0.0094	-0.014	0.7841	0.6221
Lighting conditions when reading or writing at night	-0.0964	-0.1243	0.4905	-0.4945	0.5099
Variance contribution rate (%)	18.75	16.07	11.32	10.45	--
Cumulative variance contribution rate (%)	18.75	34.82	46.15	56.60	--

Variable	Factor 1b	Factor 2b	Factor 3b	Factor 4b
Sense of happy in life	0.9147	0.0359	0.0191	-0.0011	0.8383
Feeling happiness	0.9109	0.0555	0.0036	0.0170	0.8331
Feeling hopeful about the future	0.8122	0.0415	-0.0354	0.0147	0.6629
Grade	0.0935	0.8446	-0.1592	0.0724	0.7526
Number of physical education classes per week	0.0133	0.8020	0.1297	-0.0241	0.6608
Average daily sleep duration	0.0533	0.7234	0.2477	-0.1273	0.6037
Number of eye exercises done daily	-0.1283	0.0188	0.7532	-0.1446	0.6050
Average daily outdoor activity time during daylight hours	0.1408	0.0840	0.7242	0.1192	0.5655
Parents’ visual condition	0.0145	0.2099	-0.1296	0.7198	0.5791
Participation in tutorial classes	0.0121	-0.2786	0.0935	0.6973	0.5727
Variance contribution rate (%)	23.75	20.15	12.22	10.62	--
Cumulative variance contribution rate (%)	23.75	43.89	56.12	66.74	--

Note: Factors from 2018 are denoted as Factor 1a-Factor 4a; 2023 factors as Factor 1b-Factor 4b; Factor 1a: School lifestyle factor, Factor 2a: Eye behavior factor, Factor 3a: Family heredity and environment factor, Factor 4a: Classroom eye policy factor, Factor 1b: Psychological factor, Factor 2b: Adaptive eye regulation factor, Factor 3b: Active eye protection factor, Factor 4b: Family heredity and environment factor

Multivariate logistic regression analysis

Adjusted models identified several factors significantly associated with poor vision. For 2018, the school lifestyle factor (OR = 0.67, 95%CI: 0.59 to 0.76, P < 0.001) was inversely associated factor with poor vision among middle school students. In contrast, the eye behavior factor (OR = 1.35, 95%CI: 1.17 to 1.56, P < 0.001), the family heredity and environment factor (OR = 1.23, 95%CI: 1.07 to 1.41, P = 0.003) showed positive associations. In 2023, the active eye protection factor (OR = 0.76, 95%CI: 0.67 to 0.88, P < 0.001) and the adaptive eye regulation factor (OR = 0.81, 95%CI: 0.71 to 0.93, P = 0.003) were inversely associated factors with poor vision among middle school students. Conversely, the psychological factor (OR = 1.17, 95%CI: 1.03 to 1.34, P = 0.016) and the family heredity and environment factor (OR = 1.31, 95%CI: 1.14 to 1.50, P < 0.001) showed positive associations (Table 3).

Table 3.

Multivariable logistic regression with factor scores

Variable	OR	95%CI	P
2018
Factor 1a (School lifestyle factor)	0.67	0.59 to 0.76	< 0.001
Factor 2a (Eye behavior factor)	1.35	1.17 to 1.56	< 0.001
Factor 3a (Family heredity and environment factor)	1.23	1.07 to 1.41	0.003
Factor 4a (Classroom eye policy factor)	0.94	0.83 to 1.08	0.394
2023
Factor 1b (Psychological factor)	1.17	1.03 to 1.34	0.016
Factor 2b (Adaptive eye regulation factor)	0.81	0.71 to 0.93	0.003
Factor 3b (Active eye protection factor)	0.76	0.67 to 0.88	< 0.001
Factor 4b (Family heredity and environment factor)	1.31	1.14 to 1.50	< 0.001

Abbreviations: OR, odds ratio; CI, confidence interval

Structural Equation Model

Based on the 2023 EFA and multivariable regression results, we constructed a SEM to examine the associations among four latent factors (Fig. 2a). Continuous and ordered categorical variables were treated as sequential numeric scales. All measurement models demonstrated robust psychometric properties. Factor loadings exceeded 0.60 (P < 0.001), confirming indicator reliability (P < 0.001) [25] (Table S7).

The measurement model was evaluated for internal consistency reliability, convergent validity and discriminant validity, compound reliability (CR) and rho_A were used to assess internal consistency reliability. The average variance extracted (AVE) was used for testing convergent validity. The heterotrait-monotrait (HTMT) ratio was used to evaluate discriminant validity. CR values were higher than 0.7. The rho_A was higher than 0.6, and AVE was higher than 0.5 indicating better internal consistency reliability and convergent validity. The HTMT result was lower than 0.85, thus verifying the discriminant validity of the measurement model (Table S8).

Model potential path analysis

We used Mplus for analysis, conducted robustness checks considering school-level clustering, and performed Bootstrap tests (95%CI). After removing paths that were not statistically significant, the final model is shown in Fig. 2 (b). The fit indices for the revised models were good (males: χ² = 24.659, df = 13, TLI = 0.986, CFI = 0.991, RMSEA = 0.036, RMSEA 95% CI = [0.012 to 0.058], WRMR = 0.598; females: χ² = 13.725, df = 12, TLI = 0.997, CFI = 0.998, RMSEA = 0.014, RMSEA 95% CI = [0.000 to 0.043], WRMR = 0.014) (Table S9).

For males, the psychological factor was positively associated with the adaptive eye regulation factor (CRSE: β = 0.081, CRSE = 0.06, 95% CI = [-0.035 to 0.198], P = 0.172; Bootstrap: β = 0.110, SE = 0.05, Bootstrap 95% CI = [0.010 to 0.206], P = 0.029). Conversely, the adaptive eye regulation factor showed inversely associated with poor vision (CRSE: β = -0.191, CRSE = 0.10, 95% CI = [-0.382 to -0.001], P = 0.049; Bootstrap: β = -0.245, SE = 0.06, Bootstrap 95% CI = [-0.362 to -0.126], P < 0.001). The robustness analysis revealed that the association between the psychological factor and the adaptive eye regulation factor became non-significant after accounting for school-level clustering, whereas the associated with the adaptive eye regulation factor and the vision remained stable (Table 4). Multicollinearity diagnostics indicated no concerning levels of collinearity between the latent variables (VIF = 1.003).

Table 4.

Results of standardization robustness analysis

Path	CRSE				Bootstrap
	Coefficients	CRSE	95%CI	P	Coefficients	SE	Bootstrap(95%CI)	P
Male
Factor 1b → Factor 2b	0.081	0.06	[-0.035 to 0.198]	0.172	0.110	0.05	[0.010 to 0.206]	0.029
Factor 2b → poor vision	-0.191	0.10	[-0.382 to -0.001]	0.049	-0.245	0.06	[-0.362 to -0.126]	< 0.001
Female
Factor 1b → Factor 2b	0.204	0.05	[0.100 to 0.308]	< 0.001	0.219	0.05	[0.122 to 0.313]	< 0.001
Factor 1b → poor vision	0.178	0.07	[0.035 to 0.295]	0.013	0.165	0.07	[0.027 to 0.300]	0.016

Abbreviations: CRSE Cluster robust standard errors

For females, a positive association was found between the psychological factor and the adaptive eye regulation factor (CRSE: β = 0.204, CRSE = 0.05, 95% CI = [0.100 to 0.308], P < 0.001; Bootstrap: β = 0.219, SE = 0.05, Bootstrap 95% CI = [0.122 to 0.313], P < 0.001). Additionally, the psychological factor was positive direct associated with poor vision (CRSE: β = 0.178, CRSE = 0.07, 95% CI = [0.035 to 0.295], P = 0.013; Bootstrap: β = 0.165, SE = 0.07, Bootstrap 95% CI = [[0.027 to 0.300], P = 0.016). These results indicate that these associations remain stable even after considering clustering at the school level (Table 4). The VIF confirmed the absence of multicollinearity concerns in the latent variables (VIF = 1.021).

Discussion

Primary findings

This large-scale study reveals a persistently high and increasing burden of poor vision among middle school students in Nantong from 2018 to 2023. Prevalence increasing significantly across grades and survey years. Female students consistently showed higher prevalence. Our comprehensive analysis reveals differences in the structure of factors associated with poor vision among the students between 2018 and 2023. In the students of 2023, the psychological factor showed a statistically significant association with poor vision, while the eye behavior factors did not show a significant association. The factors showed inversely associated with poor vision in this group mainly include “adaptive eye regulation” and “active eye protection”. Furthermore, no significant association was observed between the screen exposure and the poor vision in this group. In contrast, among the student in 2018, the main factor inversely associated with poor vision was “school lifestyle,” and its association pattern was different from that of the 2023 students. Further gender-stratified analysis show that for males, the adaptive eye regulation factor was inversely associated with poor vision, while in females, it had a direct positive association with poor vision. The results indicate that there are differences in the structure of factors associated with poor vision among student surveyed at different time points. Specifically, the factors associated with poor vision in the student in 2018 focused on environmental and behavioral dimensions, while for the student in 2023, it was a multidimensional pattern that included psychological, behavioral, and adaptive regulation. The differences at the group level suggest that when carrying out visual health interventions, attention should be paid to the composition characteristics of associated factors in different groups, and psychological support and gender differentiation strategies should be considered.

Comparison with other studies

The prevalence of poor vision among middle school students in Nantong, China, has shown an overall upward trend from 2018 to 2023, with an average of 87.40% and temporary fluctuations. This trend is consistent with the epidemic pattern in Jiangsu Province (86.63% in 2019) and exceeds the national average (78.72% in 2019), highlighting regional severity [26]. There was a significant temporary decline in 2021 (2020: 86.57%; 2021:85.73%, χ² = 7.88, P = 0.005), which is temporarily consistent with the implementation of classroom lighting intervention measures in the entire district [27]. This observation is consistent with evidence that environmental changes may affect visual outcomes, although the causal relationship needs further verification [28, 29]. In addition, in the following years, this upward trend resumed and reached 89.08% in 2022 (χ² = 188.20, P < 0.001). As grades increase, the observed upward trend in poor vision is consistent with broader patterns reported nationally and internationally [30], indicating that academic pressure and environmental factors may associated with shaping visual health across different age groups [31]. The significant differences between the second and first grades of senior middle school (χ² = 27.57, P < 0.001) and between the second and third grades (χ² = 46.37, P < 0.001), deserve further investigation, which may be influenced by sampling heterogeneity or learning adaptation.

Gender differences in poor vision prevalence were evident, with female students having a higher rate than males. This is consistent with international research trends, such as studies on Polish adolescents [9], where female adolescents exhibited larger equivalent refractive error values than males. Despite conflicting data on the prevalence of poor vision associated with gender, increasing evidence suggests that female students are more susceptible to its impact [32]. Biological factors, including early puberty and accelerated eye development, may associated with these differences, while social behavioral factors may amplify the differences [33].

EFA has revealed differences in the structure of factors associated with poor vision among middle school students between 2018 and 2023. For the student in 2023, it was a comprehensive association model of psychology-behavior-physiology. In the students of 2023, the psychological factor showed a statistically significant association with poor vision, which was not significant among students in 2018, indicating that this was a potential pathway that may be associated with poor vision [34]. Meanwhile, in both surveys, family genetics and environmental factors have been significant, confirming their association with visual development [35]. It is worth noting that among students in 2023, the eye behavior factor was no longer significant, reflecting differences in major environmental such as increased academic stress [36].

Multivariate logistic regression showed the structure of factors associated with poor vision (2018 vs. 2023). Consistent with previous studies in China and the United States [37, 38], gender, parental myopia status, and daily outdoor activity time were strong association factors. Genetic factors continue to significantly positively associated with poor vision [35], while the psychological factor showed a statistically significant association with poor vision in 2023 students, and the eye behavior factor was no longer significant. Meanwhile, there are differences in the structure of factors inversely associated with poor vision. They had been divided into two different parts: the active eye protection factor and the adaptive eye regulation factor [39]. In addition, screen exposure factor was no longer significant, which means that its association with poor vision may have been redistributed in the factor structure. These findings challenge traditional environmental models and emphasize the associated with psychological factors, and suggest that the prevention of poor vision needs to enter a new stage, focusing on a comprehensive psychological-behavioral model [40].

SEM revealed different gender-specific potential pathways between the psychological factor and the poor vision, and the measurement model confirmed its robustness. For males, the adaptive eye regulation factor was direct associated with poor vision. In contrast, for females, the psychological factor had a direct and significant positive association with poor vision. These findings highlight that female students show a stronger association with psychological. These findings are supported by previous research [41, 42].

The potential association pathways indicate that there are differences in the pathways associated with poor vision among different gender groups. In the male students, the adaptive eye regulation factor was inversely associated with poor vision, while the psychological factor did not show a robust direct association. In the female, the psychological factor was not only associated with the behavioral factor, but also directly positively associated with poor vision. Cui et al. found that there was no gender difference in the association between axial length and visual impairment, indicating that gender differences in visual outcomes may depend on other factors, which is consistent with the psychological pathway we studied [10]. The interaction between learning stress, reduced outdoor activity, and psychological distress may activate cortisol-mediated pathways, leading to dysregulation of retinal dopamine signaling [43], as demonstrated by our gender-stratified logistic regression results (Table S10).

Strength and limitations

This study has several significant advantages. Firstly, the repeated independent cross-sectional surveys conducted in 2018 and 2023 can effectively describe the structural differences in factors associated with poor vision at the population level. This design is particularly suitable for differences in factor association patterns during a certain period of time. Secondly, constructing a structural equation model based on gender, this study revealed the differences in the potential pathways of psychological factors and poor vision between males and females, providing important evidence for developing gender specific intervention strategies. Lastly, the study integrated objective vision screening data and questionnaire data, and systematically analyzed multi-level associated factors from environment to psychology through various statistical methods such as exploratory factor analysis, multivariate regression, and structural equation modeling, enhancing the comprehensiveness and reliability of the results.

Undeniably, this study has certain limitations. Firstly, although repeated cross-sectional designs can describe differences at the population level, due to the independent sample size of the two surveys, it is not possible to track individual level changes or infer causal relationships between factors and outcomes. Meanwhile the comparability of factor structures may be influenced by subtle differences in variable selection and measurement tools. The factor structures for 2018 and 2023 are independently explored each year, so structural changes cannot be directly compared. Secondly, structural equation modeling has specific requirements for data distribution and sample size. The handling of categorical variables in this study may have a certain impact on model fitting. In addition, SEM results represent statistical association structures rather than causal mechanisms. Thirdly, although multiple confounding factors have been controlled for, there may be unmeasured confounding variables (such as genetic details, subtle differences in family socioeconomic status) that affect the results. Moreover, questionnaire data relies on self-report, which may associate with recall bias or social expectation bias. Additionally, the sample of this study was sourced from Tongzhou District, Nantong City, and the results have reference value when reflecting regions with similar socio-economic backgrounds. However, caution should be exercised when generalizing to the whole country or other regions with significant differences. Finally, in exploratory factor analysis, while it is reasonable to retain the fourth factor based on theoretical importance, it may introduce some subjectivity. In the future, the stability of factor structure can be further tested through confirmatory factor analysis or in independent samples.

Policy implications

Based on the core findings of this study, namely that psychological factors have a significant associated factor for poor vision among middle school students in 2023, and there are significant gender differences in their impact pathways, we suggest constructing a precise intervention system that integrates psychological support and vision health management. Firstly, a brief psychological health assessment should be embedded in the existing student vision screening, and a “vision-psychology” dual track intervention channel should be established to provide psychological support such as cognitive-behavioral therapy and guidance on eye use behavior to high-risk students simultaneously; Secondly, implement gender-specific strategies, focus on designing stress management and psychological support courses for female, and strengthen gamified sports activities for male to consolidate their adaptive eye regulation behavior; Furthermore, through digital platforms linking home and school doctors, students’ visual, behavioral, and psychological data can be dynamically monitored and early warning linkage can be achieved; Finally, based on optimizing the hardware such as classroom lighting environment, we will combine the seat rotation and classroom rest system to deeply integrate healthy eye habits into daily teaching management. This system integrates multidimensional linkage and precise policy implementation, offering an alternative to traditional one-way behavioral intervention by incorporating both psychological-behavioral dimensions into a comprehensive governance approach.

Conclusion

This study reveals that the problem of poor vision among middle school students remains severe in 2023, and its associated factors are composed of a comprehensive psychological-behavioral model. It is worth noting that psychological factor was significantly associated with poor vision in 2023, and showed gender-specific pathways: in male, the association is mainly achieved through the adaptive eye regulation factor, while in female, it is reflected as a direct positive association between the psychological factor and the poor vision. Based on evidence-informed hypotheses and risk profiling, it is necessary to construct integrated environment optimization, school intervention, and targeted psychological support, especially comprehensive prevention and control strategies that combine gender-specific pathways, to more effectively address this increasingly prominent public health challenge.

Abbreviations

AVE

Average variance extracted

CFA

Confirmatory factor analysis

CFI

Comparative fit index

CI

Confidence interval

CR

Compound reliability

CRSE

Cluster robust standard errors

df

Degree of freedom

EFA

Exploratory factor analysis

EPV

Events per variable

GB

National standards of the people’s republic of China

HTMT

Heterotrait-monotrait

KMO

Kaiser-Meyer-Olkin

OR

Odds ratio

RMSEA

Root mean square error of approximation

SCL-90

Symptom checklist-90

SEM

Structural equation model

TLI

Tucker-lewis index

VIF

Variance inflation factor

WRMR

Weighted root mean square residual

Authors’ contributions

GQ, Y-QZ and XZ conceived and designed this study. L-JD collected the data for analysis. L-JD and B-SN carried out the statistical analysis. B-SN, L-JD and J-TS wrote the first draft of the manuscript. GQ, Y-QZ, XZ, Y-HJ, YY and J-TS have made key revisions. All authors have contributed to the scientific discussion and approved the final draft.

Funding

B-SN acknowledges support from the Postgraduate Research & Practice Innovation Program of Jiangsu Province (grant no. KYCX24_3542). GQ acknowledges support from the Jiangsu Provincial Research Hospital (grant no. YJXYY202204-YSA03).

Data availability

The data from the current study are not publicly available due to the privacy of the minor participants but are available from the corresponding author, Prof. Gang Qin, upon reasonable request (tonygqin@ntu.edu.cn).

Declarations

Ethics approval and consent to participate

This study was approved by the Ethics Committee of Affiliated Hospital of Nantong University (Approval No. 2021-K118-02), and all methods were performed in accordance with the relevant guidelines and regulations. Participants and their legal guardians (when participants were minors) gave informed consent to participate in the study before taking part.

Consent for publication

Not applicable.

Competing interests

Gang Qin is an editorial board member of BMC Public Health. The remaining authors have no conflicts of interest to declare.