Refractive Errors and Their Associated Factors in Schoolchildren: A Structural Equation Modeling

ABSTRACT Purpose To determine the prevalence of myopia and hyperopia in Shahroud schoolchildren and their risk factors Methods Optometric examinations including the measurement of uncorrected and corrected visual acuity as well as non-cycloplegic and cycloplegic refraction using retinoscopy were done for students. Generalized Structural Equation Modeling (GSEM) was used to determine direct and indirect effects of independent variables on myopia and hyperopia. Results The data of 5581 students with a mean age of 9.24 ± 1.7 years were used in this study. The prevalence of myopia was 5.0% (95%CI: 4.3–5.7) and the prevalence of hyperopia was 4.8% (95%CI: 4.0 − 5.5) in all schoolchildren. According to the GSEM results, the odds of myopia in rural areas were 0.55 compared to urban areas. A one-unit increase in the ocular AL increased the odds of myopia by 4.91 times. The interaction of sex and age on myopia was significant such that in girls, the odds of myopia increased by 20% for every one-year increase in age while no significant change was seen in boys. A one-unit increase in the ocular AL decreased the odds of hyperopia by 0.49 times. Moreover, the interaction of outdoor activity hours and sex on the prevalence of hyperopia was significant such that increased outdoor activity reduced the odds of hyperopia in girls while no significant correlation was found in boys. Conclusion Myopia and hyperopia had moderate prevalence. Axial Length had the largest direct association on myopia and hyperopia. Age and outdoor activity had weak associations on refractive errors.


Introduction
Refractive errors are the most common visual problem in all age groups that require billions of dollars for treatment and correction. They also have significant indirect effects on daily life, too. 1 According to the latest WHO report, uncorrected refractive error is responsible for moderateto-severe vision impairment in 123.7 million people worldwide. 2 Children are vulnerable group for refractive errors. These errors may result in ocular outcomes like amblyopia, educational outcomes like deterioration in school performance, and social outcomes such as decreased quality of life and isolation. 3 Due to the importance of this age group, Negrel proposed a protocol for evaluation of refractive errors in children in 2000. 4 After that, several studies investigated refractive errors in children worldwide. [4][5][6][7][8][9][10][11] Hyperopia is more common than myopia in children; however, recent studies suggest that the prevalence of these two refractive errors is very close in some ethnic groups and countries; moreover, due to the increasing trend of myopia, this refractive error seems to be more important than hyperopia in children. 12 According to recent reports, it is estimated that 3.36 billion people will have myopia in 2030, of whom 516.7 million will suffer from severe complications. 2 The difference in the prevalence of refractive errors from less than 1% to more than 50% suggests the diversity of the determinants of refractive errors in different parts of the world and their different effects. [4][5][6][7]10,11,[13][14][15][16][17][18][19][20][21][22][23][24] It is already clear that age, ocular biometric components, and lifestyle are associated with refractive errors. 25 This study was part of Shahroud Schoolchildren Eye Cohort Study with the aim of determining the prevalence of refractive errors and their associated factors in children.

Methods
This cross-sectional study was conducted in 2015. The target population was Shahroud schoolchildren aged 6-12 years. Shahroud is located in the northeast of Iran. Although the methodological details of this study were published elsewhere, 26 they are briefly presented in the following. All rural students (n = 1214) were invited to participate in the study, while cluster sampling was done in urban areas. In total 5410 urban students were invited in 473 clusters (classrooms). To achieve a sufficient sample size, 200 clusters were selected using systematic random sampling. Written informed consent was obtained from the students' parents. In the next step, the children's demographic data and past medical history were collected from their parents and recorded in predesigned questionnaires. Then, all students underwent optometric examinations on site. First, non-cycloplegic refraction autorefraction was done using the Nidek ARK-510A auto refractokeratometer. If a student wore spectacles, spectacle-corrected visual acuity was measured and lensometry was done. Then, uncorrected visual acuity was measured using the Nidek CP-770 chart projector at 3 m and the autorefraction data were refined using retinoscopy (Heine Beta, HEINE Optotechnic, Hersching, Germany). In each stage, first the right eye and then the left eye was examined and their results were recorded. Subjective refraction was done in students whose visual acuity was worse than 20/20. In the final stage of the optometric examinations, cycloplegic refraction was measured at least 30 minutes after instilling cyclopentolate 1% drops twice at a 5-minute interval using retinoscopy and auto refractometry. The Allegro Biograph (WaveLight AG, Erlangen, Germany) was used to measure biometric components in participants.

Definitions
In this study, spherical equivalent was determined based on cycloplegic refraction using retinoscopy. For proper comparison of the results with other refractive errors studies in children, 4 myopia was defined as a spherical equivalent of equal to or worse than −0.5D and hyperopia was defined as a spherical equivalent of equal to or worse than +2D. Severe myopia and hyperopia was defined as a spherical equivalent worse than −6D and worse than +4D, respectively.
To calculate the mean daily hours of outdoor activity, physical and sport activities of the students as well as hours spent daily on watching TV or playing computer games on weekdays and holidays were inquired from parents. Outdoor activity was defined as playing, walking, running, jumping, riding a bicycle, gardening or any work in the daylight. The total hours of daily outdoor activity and playing outdoor sports on weekdays was calculated and multiplied by six, and the result was added to the hours of outdoor activities on holidays. The total hours of outdoor activity in a week was then divided by seven to calculate the mean daily hours of outdoor activity. The same method was applied to calculate the mean time of watching TV and playing computer games.

Statistical analysis
The prevalence of different refractive errors is reported as percentage with 95% confidence interval. The design effect and sampling weight were considered for calculating 95% confidence intervals using post standardization.
Using generalized structural equation models (GSEM), we fit regression models to estimate direct and indirect associations of variables on outcomes of interest including binary ones. In these models, all parameters are estimated simultaneously and the association of each variable on other variables is calculated considering all other correlations. Therefore, using this model and considering myopia and hyperopia as the response variables, the relationships between the studied variables were investigated and odds ratios and pathway diagrams were determined. All analyses were performed at a significance level of 0.05.

Ethical consideration
The study was conducted in accordance with the Helsinki Declaration. All procedures involving children were approved by the Ethics Committee of Shahroud University of Medical Sciences. We obtained written informed consent from all parents or students' legal guardians and also oral consent from all the students.

Results
Of 6624 randomly selected students, 5620 participated in the study. The data of 5586 students were available for analysis of this study. The mean age of the participants was 9.24 ± 1.7 years, 4443 students lived in urban areas and 53.9% of them were boys. Table 1 presents the distribution of uncorrected, spectacle-corrected, presenting, and best corrected visual acuity according to the visual acuity group.

Refractive error
Cycloplegic refraction was not measured in 192 students and 5394 students underwent cycloplegic retinoscopy refraction. To present the spherical equivalent mean and distribution, one case with a spherical equivalent of −18.25 D was excluded from the study as outlier data.
The mean spherical equivalent was 0.86 diopter (D) (95% CI: 0.82-0.90) in 5393 children aged 6-12 years. Supplementary Figure S1 shows the different category of spherical equivalent by age groups of participants. The mean spherical equivalent was 0.86D (0.81-0.90) in boys and 0.86D (0.80-0.92) in girls (p = .885). A linear myopic shift was seen with increasing in age, as the highest spherical equivalent was seen in 6-year-old children (1.05D) and the lowest was seen in children aged 12 years (0.53D). The mean spherical equivalent was lower in urban children (0.85D, 95% CI: 0.81-0.89) compared to rural children (0.91D, 95% CI: 0.86-0.97), but the difference was not significant (p = .096).

Prevalence
The prevalence of myopia was 5.0% (95% CI: 4.3-5.7) in all children. Table 2 shows the prevalence of myopia according to age, sex, and living place. The prevalence of myopia was 5.6% in girls and 4.5% in boys. The changes of myopia prevalence with age were not linear, but the prevalence of myopia increased from 2.88% at 6 years to 9.99% at 12 years (p < .001).
The prevalence of hyperopia was 4.75% (95% CI: 4.01-5.49) in all students with no inter-gender difference (p = .090). The chi-square test for trend showed a significant decrease in the prevalence of hyperopia with age; the highest and lowest prevalence of hyperopia were seen in children aged 7 years and 12-year-old children, respectively (p < .001). There was no significant difference in the prevalence of hyperopia between urban and rural children (p = .806). Considering severe hyperopia as a spherical equivalent worse than +4 D, 0.56% (95% CI: 0.36-0.76) of the students had severe hyperopia. The prevalence of anisometropia was 1.18% (95% CI: 0.87-1.5) in this study.
In the GSEM model, considering myopia as the main response variable, the odds ratio of myopia in rural areas were 0.55 compared to urban areas, therefore rural children had almost half of the odds of developing myopia as the urban children (Table 3). A one-unit increase in the ocular AL increased the odds of myopia by 4.91 times. The interaction of sex and age on myopia was significant such that in girls, the odds of myopia increased by 20% for every one-year increase in age (OR = 1.20, 95% CI: 1.04-1.38; P value = .011) while no significant change was seen in boys (OR = 0.98, 95% CI: 0.87-1.10, P value = .711) in a multiple logistic regression model.
Supplementary Figures S2 and S3 show the fitted models for the association of different independent variables with myopia and hyperopia respectively. Any direct path between independent variables and outcomes was  considered as direct association. Indirect association was defined as those paths that influence outcomes through affecting other variables.
In investigating associated factors with myopia, sex and age had a significant effect on ocular AL; sex and living place had a significant effect on the duration (hours) of outdoor activity; and sex, age, and living place had a significant association on hours spent on computer games and watching TV (Table 3, Supplementary figure S2). Table 4 presents the total associations of variables on myopia by direct and indirect ways. According to Table 4, age, sex, ocular AL, and living place correlated with myopia; among these variables, only age and sex had direct and indirect associations and only their indirect effects correlated with myopia.
In another GSEM considering hyperopia as the main response variable, a 1-mm increase in ocular AL decreased the odds of hyperopia by 0.48 times. Moreover, the interaction of the duration (hours) of outdoor activity and sex on hyperopia was significant; in other words, increased outdoor activity decreased the odds of hyperopia in girls (OR = 0.995, 95%CI: 0.992-0.999, P value = .009) while no significant correlation was seen in boys (OR = 1.000, 95% CI: 0.998-1.003, P value = .942) in a multiple logistic regression model. Simultaneously, sex and age had a significant association on ocular AL; sex and living place had a significant association on the duration (hours) of outdoor activity; and sex, age, and living place had a significant association on hours spent on computer games and watching TV (     Table 4 also shows the direct and indirect associations of variables on hyperopia. According to results, sex, age, and AL had a significant relationship with hyperopia, and direct effects of AL correlated with hyperopia.

Discussion
This study reported the prevalence of refractive errors and their associated factors in an Iranian population aged 6-12 years. Since 2000 when Negrel et al. 3 proposed refractive error studies in children, many studies have been conducted in this regard worldwide using this protocol. Most of these studies adopted the cycloplegic method and similar definitions of refractive errors. 4,7,9,11,[14][15][16][27][28][29][30][31][32][33] However, lifestyle changes in recent years and increased computerization of daily activities require further research in this regard. 34 The results of the present study were compared with studies that used cycloplegic refraction and similar definitions. One of the important aspects of the present study was using SEM analysis for determining the risk factors associated with refractive errors. This analysis provides a better scheme of the direct and indirect association of independent variables with refractive errors.
In this study, a visual acuity worse than 20/40 was seen in 0.8% and 0.1% of the children according to the presenting and best-corrected visual acuity, respectively.
A look at previous studies shows that the prevalence of visual impairment based on presenting visual acuity varies from about 1% in Bojnourd (Iran) 22 to 11.26% in Shanghai. 32 Based on the corrected visual acuity, the prevalence has been reported from 0.03% in Dezful (Iran) 10 to 7.4% in Chile. 35 Comparison of the results of the present study with other studies indicates that children in this study had a low prevalence of visual impairment.
Myopia was seen in 5.0% of the students in this study, although the prevalence of myopia changed from 2.88% in children aged 6 years to about 10% in 12-year-olds. There are various reports of the prevalence of myopia in similar studies, from 1.2% in Nepal 5 to more than 50% in Indian 36 and Pakistani 37 children. A meta-analysis of children aged 1-18 years showed myopia in 69% of East Asian children and 5.5% of Indian children 38 (the main limitation of this study was using different refractions for cut points). A recent meta-analysis by Hashemi et al. 12 found a total prevalence of 11.7% for myopia in children worldwide, ranging from 4.9% in South-East Asia to 18.2% in Western Pacific region. A review of the literature shows a lower prevalence of myopia in the present study compared to African and some European countries. Considering the increasing trend of myopia in previous years, 12 it is expected that recent studies report a higher prevalence, which should be considered when comparing the results of this study with previous studies. Comparison of the results of this study with previous studies also suggests that myopia has had an increasing trend in Iranian children. However, a comparison of myopia prevalence with previous studies conducted in Iran or countries such as India 7 or Pakistan, 39 which are not economically very different from Iran, shows that the prevalence of myopia in students of this study is not low and even higher than in countries such as India and Pakistan ( Table 6). The prevalence of hyperopia was 4.75% in this study. According to Table 6, the prevalence of hyperopia ranges from 1% to about 20% in children across the world. A review study by Castagno et al. 60 found that the prevalence of hyperopia ranged from 5% in 7-year-old children to 2-3% in children aged 9-14 years. Hashemi et al. conducted a review study 12 and found a total prevalence of 4.6% for hyperopia across the world according to a spherical equivalent of equal to or worse than +2.00 D. In that review study, the lowest and highest prevalence of hyperopia was seen in South-East Asian (2.2%) and American countries (14.3%), respectively. Previous studies conducted in Iran showed a hyperopia prevalence of 16% in Dezful 10 and 19% in Mashhad. 20 However, the prevalence of hyperopia was lower in the present study. Some other recent studies also reported a lower prevalence. 33,55 It could be that the increasing trend of myopia prevalence is one of the reasons for lower prevalence of hyperopia in current study. 12 Another reason for this may be the use of cycloplegic retinoscopy refraction, because most of the previous studies reported refractive errors based on cycloplegic auto-refraction. It has been already reported that autorefraction overestimates hyperopia in cycloplegic conditions. 61 Epidemiologic, clinical, laboratory, and even genetic studies have shown different aspects of the risk factors of refractive errors. [62][63][64][65] The existing knowledge suggests that one factor alone cannot be held responsible as the main reason for myopia and this error is caused by a complex interplay of different causes. One of the features of SEM analysis used in the present study is to investigate both the direct and indirect associations of different variables on refractive errors. Therefore, variable associations were defined better by investigating a predefined theory and explicit assessment of measurement errors in this analysis. As mentioned earlier, according to the SEM, living in urban areas, age, sex, and AL were associated with myopia. Among these variables, the increase in AL had the most direct association on myopia. The association of age with myopia was mainly indirect and through influencing AL. Sex was also associated with myopia, both directly and indirectly. Each of these risk factors has been investigated in several studies. As reported earlier, the odds of myopia increased with age in girls. Previous studies reported contradictory and inconsistent results regarding the effect of age and gender on myopia in children. [4][5][6]23,33 The majority of the published results suggest a higher prevalence of myopia in male participants. [4][5][6]23,33,[66][67][68] Regarding age, it is already established that children are born with hyperopia and undergo a myopic shift early in life to achieve a mean refractive error of +1.25 D by age two, 69 and then slowly during schoolyears they have a slow myopic shift such that most residual hyperopia is lost in the population. As more children become myopic in some environments, the prevalence of myopia becomes higher in adolescents and young adults. 70 City dwelling is an environmental risk factor investigated in many studies, 10,17,21,71 and several studies reported its relationship with myopia. [72][73][74][75][76] Living in the city may be associated with several risk factors such as increased use of computer systems and increased nearwork activities. Anthropometric indexes like height and BMI are also different in people living in urban and rural areas.
One of the new hypotheses regarding myopization is decreased outdoor activity and increased indoor activity. 77 It seems that outdoor activity is decreasing in the urban lifestyle while it is much more common in many rural areas due to lack of modernization. Several studies investigated the hypothesis of the effect of outdoor activity on myopia. 62,[78][79][80][81][82][83][84][85][86][87][88][89][90] In a clinical trial, the incidence of myopia was 10% lower in children who had more outdoor activity. 91 The factor that may play a role in lack of myopization with more outdoor activity is light. 77 Several studies showed the role of strong ambient light in lack of myopization. 77 Its mechanism of action is probably that light stimulates dopamine secretion in the retina; in turn, this neurotransmitter prevents ocular elongation during development and myopization. 77 However, although the prevalence of myopia was higher in urban areas, no significant correlation was found between myopia and outdoor activity. The reasons for the difference between the results of this study and previous studies may be the lower prevalence of myopia in the present study or the effect of living in the city on myopia through mechanisms other than outdoor activity. Due to the cross-sectional nature of the present study, longitudinal studies are recommended to better identify risk factors of myopia.
Regarding the interaction of the female sex and outdoor activity and its protecting role in reducing the prevalence of hyperopia, the possible effect of estrogen should be mentioned. The results of both observational and in vivo studies suggest that estrogen can induce structural and biomechanical changes in the cornea that affect the corneal curvature and its refractive power, leading to a relative myopic shift and reduction of hyperopia. 92 The topical estrogen therapy has been proposed as a method of correcting hyperopia. 92 On the other hand, outdoor activity and exposure to light can enhance the activity of estrogen. 93 Increased ocular AL was one of the most important variables that correlated with increased prevalence of myopia in the SEM. Similar to the role of AL in myopia development, it seems that there are sufficient clinical, epidemiologic, and genetic investigations to prove this hypothesis, although some studies [94][95][96] emphasized that the axial length to corneal curvature had the main role in refractive errors. Age, sex, and AL were associated with hyperopia; age had an indirect relationship with hyperopia through its effect on AL. Although the indirect relationship between age and hyperopia was not significant, ageing can reduce the odds of hyperopia through increasing myopia. Comparison of the results of myopia and hyperopia suggests that living in urban areas not only does not affect hyperopia, but also increases myopization. The urban lifestyle characteristics, including more indoor activity and increased use of electronic devices like the computer and tablets increase the odds of myopia. 97 One of the most important strengths of this study was the use of SEM to determine the associated factors of refractive errors and presented a new feature for these risk factors. However, it should be noted that our study was limited to one of the Iranian cities and the results cannot be generalized to the Iranian population. Moreover, the data were obtained from a crosssectional study and the observed correlations do not indicate causality.
According to the results of this study, the prevalence of hyperopia was lower only in girls with more outdoor activity (mean outdoor activity was 61 and 74 minutes in hyperopic and non-hyperopic girls and 122 and 121 minutes in hyperopic and non-hyperopic boys, respectively), which confirmed the results of previous studies 62 and the hypotheses we put forward. These results are important for policymaking and developing correct interventions for lowering incidence of refractive errors. More attention also should be paid to urban girls who had higher odds of developing Myopia.

Data Availability
The data that support the findings of this study are available from the corresponding author, [MHE], upon reasonable request.

Disclosure statement
No potential conflict of interest was reported by the author(s).