1. Introduction

Myopia in children and adolescents seriously affects their quality of life and has become a significant global public health burden in the 21st century.[1] It is also a significant cause of the development of severe ocular diseases, such as retinal detachment, glaucoma, cataract, and myopic maculopathy, which greatly intensifies vision disability.[2] Aside from the high worldwide prevalence, East Asia and Southeast Asia have significantly higher rates than other regions, and the prevalence has been persistently increasing to alarming levels (>80%) in middle and high schools in East Asia.[3–5] By 2050, low vision ability will affect 5 billion people worldwide, and high-degree myopia will escalate the disease burden by 24% in the Asia-Pacific region.[6] Since 1985, the Ministry of Education and the National Health Commission of China (NHCC) have conducted a nationwide survey of student health and physical fitness every 5 years, including vision, height, weight, and exercise capacity. Since 2018, the NHCC has selected representative schools, including kindergartens, elementary schools, middle schools, and universities, across the country to conduct vision inspection and assessment to achieve coordinated surveillance of the demographic distribution of vision impairment levels and the evolution trend in a timely manner. In 2020, the NHCC required that myopia prevention and control would be a high-priority task for local governments.

As part of the national surveillance of common diseases and related social determinants, the ultimate goal is to understand how myopia physiologically develops and how major social determinants affect its development, as well as regulating personalized intervention. Currently, although some general suggestions, such as reducing near-work and academic pressure and increasing outdoor activities, are available, the establishment of standardized specialized intervention approaches based on age and vision-checkup information is still lacking, both in China and globally. This study leverages detailed vision-checkup information to initiate practical solutions to target certain groups for intervention. Specifically, in addition to establishing models to determine risk factors (e.g., age, gender, urban/rural residence, school type) and identifying high-risk students based on myopia prevalence and the prevalence of glasses wearing, we further rigorously identified intervention groups by combining 3 decisive variables (visual acuity [VA], spherical equivalence, and age) to establish alarm regions. By leveraging these detailed measures, we aimed to develop evidence-based strategies for targeted interventions aimed at reducing the burden of myopia in the identified at-risk populations.

2. Methods 2.1. Data collection and processing

The data were sourced from the Student Vision Surveillance Network in Yantai. This study was approved by the Research Ethics Committee of the Yantai Center for Disease Control and Prevention. Written informed consent was obtained from both participants and their parents or guardians. This cross-sectional study was conducted in September, 2021. Using a multistage sampling approach, a total of 35 schools (including elementary [grades 1–5], middle [grades 6–9], and high and vocational high [grades 10–12] schools) were randomly selected from 14 counties, and a specific number of classes were randomly chosen from each grade within each school, ensuring the participation of no fewer than 80 students from each grade. A total of 10,276 students participated in the study. Optometrists and/or ophthalmologists conducted eye examinations using a retro-illuminated logMAR chart for VA measurement (uncorrected and corrected by glasses). Uncorrected VA was tested for each eye in patients not wearing a thokeratology contact lens. For those wearing glasses, the naked-eye VA (without correction) measurement was followed by the corrected VA (wearing the prescription glasses) measurement 30 minutes later. Students with ocular diseases or ocular injuries were excluded. We further used a table-mounted noncycloplegic autorefractor (TOPCON: model KR-800, Japan) to measure the refractive error, where “spherical equivalent (SE)” (under nonciliary muscle palsy) for each eye was calculated as “sphere + half of the cylinder,” and the average of 3 readings was recorded. All autorefractors were calibrated daily to ensure measurement quality. We referred to myopia screening standards (National Health Commission of the People’s Republic of China, 2018) and defined myopia by combining uncorrected VA (abbreviated as “VA” in the sequel) and SE to avoid overestimation of myopic magnitude due to refractive errors.[7,8] Our threshold for myopia is VA <5.0 and SE <−0.50D. Students having myopia in any of 2 eyes are counted into the myopia group along with those wearing orthokeratology lenses. By referring to Chinese national standards (WS/T586-2018 and GB/T31178-2014), body mass index (BMI) (weight [kg]/height [m]2) is classified into “underweight,” “normal,” “overweight” or “obese” category (gender and age specific). The large sample size (n = 10,276) suffices for parameter estimation in the logistic regression.

2.2. Logistic models for myopia and glass-wearing rates

In the logistic model

Logit(Pr(Yi=1))=β0+Xi,1β1+...+Xi,kβk,

Xi,j (i = 1,...,n, j = 1,...,k) represents the predictor j for student i. In a previous study,[9] 3 factors (age [6–15 years old], sex, and school type) were considered to establish a logistic model for predicting the prevalence of refractive errors among suburban children (Yongchuan, southwestern China). We considered 4 predictors (gender [male = 1, female = 0], district [urban = 1, rural = 0], age [continuous], and school type [nominal: elementary = 1, middle = 2, high = 3, vocational high = 4]). For the myopia population, the distributional mode (age = 15.5) in the urban area was approximately 2 years greater than that in the rural area (age = 13.5), and the overall population had a bimodal distribution. No apparent distributional differences existed between males and females. The gender and district baseline effects amount to intercept adjustments in Eq. (1). In Figure 1, log(rate/(1 − rate)) roughly linearly increases with age (from 6 to 14 years) and then levels off (from 15 to 18 years). This suggests segmented logistic regression (segment 1 [age ≤14.5], segment 2 [age ≥14.5]). In segment 1, school type (1–4) had a sample size (5287, 1844, 9, and 3). In segment 2, school type (2–4) has a sample size of (941, 1842, 349) with a breakdown of (541, 1394, 349) (urban) and (400, 448, 0) (rural).

Figure 1.:

Observed logit (myopia prevalence rate) vs. age.

2.3. Statistical methods

Statistical analysis was performed using the R statistical software package. Continuous variables were presented as mean ± standard deviation, while categorical data were expressed as percentages. The comparison of categorical data was conducted using the Chi-square test or Fisher exact probability test. A P < .05 was considered statistically significant.

3. Results 3.1. Marginal myopia rates

A total of 10,320 participants were included in the survey, excluding individuals who did not undergo both vision and refraction examinations, 10,276 school children (5157 [50%] male students) were included in the study. Sixty percent of them were urban students and 63% of them had myopia (65% of them wore glasses) (Table 1). Females had a higher myopia rate than males (66% vs. 60%, P < .001), the urban district had a higher myopia rate than the rural (67% vs. 58%, P < .0001), and high schools had higher myopia rates than vocational high schools (92% vs. 81%, P < .0001). BMI slightly increased with age (ρ = 0.42, 95% confidence interval [CI] = 0.40–0.43). The sample sizes among school types (elementary, middle, high, vocational high) are (2793, 1609, 1403, 352) (urban) and (2494, 1176, 448, 0) (rural), where (vocational) high schools tend to be located in the urban district.

Table 1 - Two-way distributions of study participants (n = 10,276). Category Boys (n, %) Girls (n, %) Overall (n, %) Myopia Wearing glasses Myopia Wearing glasses Myopia Wearing glasses School type  Elementary 941 (36) 355 (38) 1109 (42) 453 (41) 2050 (39) 808 (39)  Middle 1168 (84) 765 (65) 1271 (91) 937 (74) 2439 (88) 1702 (70)  High 828 (90) 714 (86) 879 (95) 774 (88) 1707 (92) 1488 (87)  Vocational high 168 (81) 135 (80) 117 (81) 107 (91) 285 (81) 242 (85) District  Urban 1985 (64) 1324 (67) 2120 (69) 1515 (71) 4105 (67) 2839 (69)  Rural 1120 (55) 645 (58) 1256 (61) 756 (60) 2376 (58) 1401 (59) BMI category  Underweight 122 (61) 82 (67) 136 (55) 73 (54) 258 (58) 155 (60)  Normal 1663 (63) 1085 (65) 2121 (68) 1470 (69) 3784 (66) 2555 (68)  Overweight 537 (60) 343 (64) 581 (66) 362 (62) 1118 (63) 705 (63)  Obese 783 (55) 459 (59) 538 (60) 366 (68) 1321 (57) 825 (62) Age (years)  6 28 (11) 11 (39) 44 (15) 12 (27) 72 (13) 23 (32)  7 72 (14) 17 (24) 101 (20) 24 (24) 173 (17) 41 (24)  8 132 (28) 32 (24) 134 (30) 29 (22) 266 (29) 61 (23)  9 180 (38) 60 (33) 206 (44) 87 (42) 386 (41) 147 (38)  10 216 (53) 76 (35) 258 (61) 112 (43) 474 (57) 188 (40)  11 224 (59) 115 (51) 252 (69) 127 (50) 476 (64) 242 (51)  12 322 (72) 179 (56) 413 (87) 275 (67) 735 (80) 454 (62)  13 316 (82) 188 (59) 366 (90) 255 (70) 682 (86) 443 (65)  14 434 (88) 316 (73) 460 (90) 354 (77) 894 (89) 670 (75)  15 446 (89) 343 (77) 417 (95) 341 (82) 863 (92) 684 (79)  16 267 (87) 231 (87) 289 (92) 261 (90) 556 (90) 492 (88)  17 338 (89) 288 (85) 323 (93) 296 (92) 661 (91) 584 (88)  18 130 (88) 113 (87) 113 (88) 98 (87) 243 (88) 211 (87) Overall 3105 (60) 1969 (63) 3376 (66) 2271 (67) 6481 (63) 4240 (65)

BMI = body mass index.

3.2. Myopia rate model

From segment 1 (Table 2), gender, district, and age have significant effects, where females (urban) have a higher baseline rate (at the beginning of schooling) than males (rural). There was no significant school-type effect (elementary vs. middle school) or BMI effect after controlling for the other predictors. Educational stress intensity was stationary during this period. No gender or district-age interactions exist. The observed and fitted myopia rates roughly agreed with each other (Fig. 2). From segment 2, females have a significant disadvantage (vs. males) at the age of 14.5 years old, and vocational high school has improved vision (vs. middle school) (Table 2). No age effect, gender, age, or district-age interaction exists. Segment 1 shows a significant district effect at baseline (grade 1), while this effect disappears at baseline (14.5 years old) in segment 2. Compared to their rural peers, the urban students (in segment 2) may not have experienced the early growth period (before age = 14.5 years old) with intensive use of electronic devices, for example, smartphones are not popular during childhood. Since vocational high school students (career-oriented) generally bear less academic burden than their high school peers do, their vision tends to be better. A previous study[9] reported that children in academically challenging schools were at a higher risk than those in regular schools.

Table 2 - Parameter estimation for the logistic regression model (myopia). Predictor Estimate Std. error z value Pr(> z ) Odds ratio (95% CI) Segment 1 (age ≤14.5)  Intercept −5.57 0.19 −28.76 <2e−16  Gender‡ −0.39 0.06 −6.89 6e−12 0.68 (0.61–0.76)§  District‡ 0.33 0.06 5.73 1e−08 1.39 (1.24–1.55)§  Age (continuous) 0.55 0.02 28.27 <2e−16 1.73 (1.66–1.79)§  Middle school*,‡ 0.17 0.10 1.69 0.09 1.19 (0.97–1.45)  High school*,‡ 11.60 175.18 0.07 0.95 1.09 (0–∞)  Vocational high school*,‡ −1.50 1.23 −1.22 0.22 0.22 (0.02–2.48) Segment 2 (age ≥14.5)  Intercept 2.44 1.06 2.31 0.02  Gender‡ −0.46 0.13 −3.62 0.0003 0.63 (0.49–0.81)§  District‡ −0.10 0.15 −0.67 0.50 0.90 (0.67–1.22)  Age (continuous) 0.01 0.07 0.12 0.90 1.01 (0.88–1.16)  High school†,‡ 0.21 0.19 1.12 0.26 1.23 (0.85–1.79)  Vocational high school†,‡ −0.74 0.23 −3.23 0.001 0.48 (0.30–0.75)§

CI = confidence interval.

*In segment 1, the middle, high, and vocational high school effects are relative to the elementary school.

†In segment 2, the high and vocational high school effects are relative to the middle school.

‡Predictors are coded as: gender (male = 1, female = 0), district (urban = 1, rural = 0). School type is a nominal variable (elementary = 1, middle = 2, high = 3, vocational high = 4).

§The bold predictors are significant.

Figure 2.:

Myopia rate (observed and fitted) vs. age and BMI. BMI = body mass index.

3.3. Glass-wearing rate model

As shown in Table 3, the logistic model had 3 significant predictors: age, BMI, and VA. Age had a significant positive effect (P < .0001; odds ratio = 1.25; 95% CI = 1.21–1.29). Thus, students of younger ages tend not to wear glasses promptly for correction. For the glass-wearing myopic group, the correlation between VA and VA improvement (due to glass correction) was −0.21 (95% CI = −0.24, −0.18, P < .0001). The correlation between age and VA improvement (due to correction) was 0.09 (95% CI = 0.06–0.12, P < .0001), and vision worsened as age increased.

Table 3 - Parameter estimation for the logistic regression model (glass-wearing). Predictor Estimate Std. error z value Pr(> z ) Odds ratio (95% CI) Intercept 25.71 0.80 32.16 <2e−16 Age 0.22 0.01 15.45 <2e−16 1.25 (1.21–1.29) BMI 0.02 0.01 2.21 0.03 1.02 (1.00–1.04) Vision acuity −6.33 0.17 −37.44 <2e−16 0.0018 (0.0013–0.0025)

BMI = body mass index, CI = confidence interval.

3.4. Form the alarm regions by combining age, VA, and SE

The prevalence and degree of myopia showed an age-dependent increase. SE distributions were positively skewed for 6 to 8-year-olds and negatively skewed for 9 to 15-year-olds.[9] The shift from positive to negative skew occurs during the critical period of myopia onset, typically around years age of 8 to 9. It is important to note that the diagnosis of myopia using cycloplegic auto-refraction may lead to overestimation.[10] Furthermore, defining myopia based on a combination of VA and SE measurements may introduce bias. To examine the temporal relationship between these factors, we calculated and plotted the correlations (with 95% CIs) in Figure 3 (left panel), where a significant negative correlation between VA and SE was observed in grade 1 (early myopia development), and the correlation profile gradually increased and reached a stable level from grade 7 onward. Thus, the VA and SE did not deteriorate synchronously during the early stages. Between ages 8 and 9, the period was characterized by a shift in correlation from negative to positive. The coincidence of the SE distribution shape shift[9] and the change in the correlation sign between VA and SE (in our study) during the same period intensifies the previous speculation that this period is critical for developing myopia and requires prompt intervention. In the left panel of Figure 3, the difference in the correlation between VA and SE among high and vocational high schools (grade 10–12) was not statistically significant, as indicated by the overlapping confidence intervals. The scatter plots of VA and SE in the middle and right panels (grades 1 and 13) show noticeable differences. In particular, the grade 13 scatter plot displays a pear-shaped region, and we define the complementary area as the alarming region(for the grade 1 student intervention). We hypothesize that VA and SE worsening at early ages (e.g.,<8 years) are unstable and lack synchronization. Consequently, a substantial proportion of students classified as “myopic” based on diagnosis may be falsely identified and may not conform to the well-defined joint distribution observed in grade 13, where myopia development becomes solidified and irreversible.

Figure 3.:

Correlation evolution trend (VA, SE) (grades: 1–12), scatter plots (VA, SE) (grades: 1–13), and a proposed alarm region (VA, SE) for grade 1. SE = spherical equivalence, VA = vision acuity.

4. Conclusion

Some research investigations have indicated that the annual incidence of myopia (aged between 7 and 15 years) remains roughly unchanged across urban and rural settings, and the prevalence rate in urban schools has reached 80%.[6,10] From this cross-sectional study with both vision and glass-wearing data collected, the myopia prevalence (63%) in the elementary and secondary schools in Yantai City is close to that of developed countries in East Asian nations such as Singapore, Hong Kong, and Taiwan,[11] and higher than some previous studies in China.[12] Myopia becomes common as early as 6 years of age (rate = 13%), and the prevalence rate increases throughout childhood to reach 80% by 12 years of age. The rates are (39%, 88%, 81%, and 92%) for (elementary, middle, vocational high, high) schools, respectively. The findings from this study further reveal a significant association between myopia prevalence and factors such as sex, geographical district, and age up to 14.5 years, with this association subsequently shifting to encompass sex and school type once beyond the threshold of 14.5 years. Gender disparity was significant for the entire adolescent period, with females having a higher rate than males. Although urban areas have a disadvantage (vs. rural) at the beginning of schooling, our study implies that the district effect likely disappears for the 15–18 age group. Recent reports from Japan further accentuate elevated prevalence rates, notably reaching 77% in elementary schools and 95% in middle schools.[10–12] Similar results have been reported in urban children who aged 7 to 12 years old are reported and the progression of myopia decelerates with age.[13] More stressful education forces children to spend more time on nearby things and reduces the time spent on outdoor activities. This is regarded as an important risk factor.[14,15] Our study implies that urban students are likely to bear more vision burden and suffer higher myopia prevalence rates at the start of schooling. This likely arises from the cumulative and intensive usage of smartphones and/or other electronic devices during the preschool growth period. This observation aligns harmoniously with established precedent findings.[9,14,15] Results on rural school students with low educational pressure have been reported,[1] where the myopia prevalence rate was 2.4% (29.4%) in grade 1 (7), and higher grades had higher rates, a notably lower incidence in contrast to our study. The low prevalence in rural and vocational high schools in our study suggests that Chinese students may not be highly genetically susceptible, and environmental factors likely play a major role in myopia development. The increasing prevalence rate, corresponding to advancing age, can be largely attributed to the heavy educational burden from elementary school.[16] We assume that genetic factors, exemplified by parental VA, are inherently randomized among all predictor levels. Environmental factors (e.g., levels of near-work and/or outdoor activity) are clearly classified by school type (academically challenging and regular).[9] Consequently, their incorporation as supplementary predictors within the logistic model may not be essential.

A study focusing on a cohort of individuals aged 5 to 16 years in rural Yunan Province, China, indicated that only 18.9% of myopic students wear glasses.[17] A recent survey in Shanghai spanning from 2017 to 2018 reported that some parents were unwilling to make their children wear glasses by mistakenly believing that wearing glasses would become irreversible and negatively affect their physical appearance.[18] Forty-eight percent of the parents believed that wearing glasses would worsen vision, and 39% would not wear glasses in time for children with myopia. The reasons include, but are not limited to, lack of myopia screening, lack of eye-care awareness, and the misconception that wearing glasses negatively affects vision or recovery.[19,20] As high-degree myopia increases the risk of vision loss and ocular diseases, prompt prevention and/or intervention strategies are desirable to alleviate myopia progression. Health education is in demand to raise public awareness of wearing glasses on time to improve eyesight. In our study, more than one-third of myopic students did not wear glasses. The reasons may include: parents did not pay attention to it, the student’s physical examination results were not sent to parents, and some students may not have informed their parents that they require glasses due to a low degree of myopia. Our results showed that age and VA had significant effects on glass wearing, possibly in conjunction with BMI. No glass correction challenges effective learning in classrooms, especially for younger students.

Our temporal (VA, SE) correlation profile reveals the critical myopia development period (age 8–9) which has already been proposed from another viewpoint.[9] The age-specific joint (VA, SE) distributions further help us to propose vision-checkup-based alarm regions for early intervention (e.g., grade 1). Due to the strong eye accommodation ability of young children, the “Standards for Screening Refractive Errors for Primary and Secondary School Students”[21] recommends the standardized use of vision combined with nonmydriatic refraction as a national myopia survey method. However, the results exhibited certain deviations. This causes false positives and an overestimation of the myopia rate. Therefore, we need to perform more rigorous mydriatic optometry in medical institutions to correct preliminary screening results and improve the accuracy of the survey results.

To extensively conduct comparative and confirmatory studies among many published studies, the major obstacle is that the predictor joint distributions (Pr(X1,...)) are often not available, and only marginal prevalence rates (stratified by single predictors) are presented. To overcome this, we propose obtaining access to Pr(X1,...) by referring to independent databases and refining the logistic model (using the set of predictors considered in each study) until model-based marginal rates highly agree with the reported rates.

Our study has certain limitations. First, this cross-sectional study only reported the status quo with regard to myopia and glass wearing. Although our cross-sectional study involved more grades than some other cross-sectional studies,[22] a large-scale longitudinal study[23] would help investigate myopia progression in more detail. Second, this study focused on visual inspection without involving other factors (e.g., genetic, environmental, and behavioral habits). Data collection for follow-up (5–10 years) studies is currently underway to develop better interventions to prevent or slow down myopia development.

Acknowledgments

The authors would like to thank all study participants