*2.3. Statistical Analysis*

In the first analytic step, we conducted univariate non-parametric analysis and reported chi-square test, as well as reporting parametric paired sample t-test results, in order to assess to what extent self-reported progression of myopic symptoms (dependent variable) differs by individual characteristics. A *p*-value of <0.05 was considered to be statistically significant. In the second analytic step, we fit multivariate logistic regression models to examine the association between digital screen use, outdoor activity, lighting condition, and self-reported myopia progression, after controlling for individual traits and pre-pandemic vision condition. The analyses were performed using STATA version 15.0 (Stata, StataCorp LLC, College Station, TX, USA) software.

#### **3. Results**

In Table 1, descriptive statistics of subject background information and results from the univariate nonparametric analysis are presented. Of the 3405 subjects that satisfied the inclusion criteria, 1358 (39.9%) reported myopic symptoms, 1647 (48.4%) were female, 2234 (65.6%) were in primary, 269 (7.9%) were in lower-secondary, and 902 (26.5%) were in upper-secondary schools, 540 (15.6%) were in rural areas, 248 (7.3%) were in urban– rural transitional areas, and 2627 (77.1%) were in urban cities. Among them, 1374 (40.4%) reported suffering from myopia prior to the COVID-19 pandemic. In terms of daily digital screen use, we first calculated daily digital screen use in diopter hours (dh), for which the sample mean is 10.0 diopter hours (SD = 6.3). We also reported daily digital screen use in unadjusted hours, which has a sample mean of 3.9 h (SD = 2.3). As for outdoor exercise, 620 (18.2%), 398 (11.7%), 1583 (46.5%), and 804 (23.6%) subjects reported as frequent,


**Table 1.** Univariate parametric and nonparametric analysis.

as "feels okay."

somewhat frequent, somewhat infrequent, and infrequent, respectively. For indoor lighting condition, 208 (6.1%) reported it being "too dim," 96 (2.8%) "too bright," and 3101 (91.1%)

Notes: <sup>a</sup> *p*-value based on χ<sup>2</sup> test, <sup>b</sup> *p*-value based on *t*-test.

Under univariate nonparametric analysis, we examined the association between subject background characteristics, daily digital screen use, outdoor exercise, indoor lighting condition, and progression of myopic symptoms during the COVID-19 pandemic. First, myopic symptoms do not differ by subjects' sex (χ<sup>2</sup> = 1.124, *p* = 0.289) or location of residence (χ<sup>2</sup> = 3.452, *p* = 0.178). Second, subjects who reported myopic symptoms are more likely to be in lower-secondary and upper-secondary but less so in primary schools (χ<sup>2</sup> = 179.580, *p* < 0.001). Third, subjects who suffer from a pre-pandemic myopia condition are also more likely to report symptomatic myopia progression during the COVID-19 pandemic (χ<sup>2</sup> = 338.785, *p* < 0.001). Fourth, subjects who reported myopic symptoms on average engage in 3.7 more diopter hours than subjects who did not (*p* < 0.001). The same finding holds true without diopter adjustment, which is 1.4 unadjusted more so for subjects who reported myopic symptoms (*p* < 0.001). Fifth, subjects' frequency of participation in outdoor exercise is associated with the likelihood of reporting myopic symptoms (χ<sup>2</sup> = 36.015, *p* < 0.001). Sixth, indoor lighting condition, particularly lighting

condition that is too dim or too bright, is associated with progression of myopic symptoms (χ<sup>2</sup> = 68.347, *p* < 0.001).

The distribution of the computed diopter hours result is displayed by enrolment level and by e-learning device type in Figure 1, from which two observations can be made. First, higher grade levels are associated with more intensive daily digital screen use in diopter hours. Second, smartphone is the most commonly used e-learning device reported across all grade levels.

**Figure 1.** Daily use of e-learning devices among school-age children.

The multivariate logistic regression analysis was used to examine the association between e-learning screen use, outdoor activity, lighting condition, and myopia development, after adjusting for subjects' sex, grade, and location. Most strikingly, findings in Table 2 indicate that every diopter hour increase in daily e-learning screen use is significantly associated with progression of myopia symptoms (OR: 1.074, 95% CI: 1.058–1.089; *p* < 0.001). Since the sample average of daily digital screen use is 10.0 diopter hours, this result would imply substantial risks of myopic progression for the typical subject. In addition, subjects who engage in four to six times (OR: 0.745, 95% CI: 0.568–0.977; *p* = 0.034) and one to three times (OR: 0.829, 95% CI: 0.686–0.991; *p* = 0.048) of outdoor exercise per week are significantly less likely to report myopia symptoms than subjects who have no outdoor exercise each week. Finally, indoor lighting that is "too dim" (OR: 1.686, 95% CI: 1.226–2.319; *p* = 0.001) or "too bright" (OR: 1.529, 95% CI: 1.007–2.366; *p* = 0.036) is significantly associated with a higher likelihood of myopia symptoms, relative to the comfortable indoor lighting condition. The association between pre-pandemic myopia condition and progression of myopia symptoms is also statistically significant (OR: 2.814, 95% CI: 2.376–3.334; *p* < 0.001).


**Table 2.** Multivariate logistic regression analysis.

Notes: OR = odds ratio; CI = confidence interval.

#### **4. Discussion**

A substantial portion of primary and secondary schools in China were closed between January and May of 2020, and a majority of school-age children had to resort to online e-learning using computers, smartphones, or TVs. Importantly, these remote learning arrangements may present new risk factors for youth vision development as a consequence of changes in daily physical and eye-use behavior among children. To the best of our knowledge, this study is one of the first to examine the association between digital screen use, outdoor activity, lighting condition, and myopia development among school-age children in the context of a nationwide remote learning experiment during the COVID-19 outbreak in China.

Using a large-scale national survey, we presented three main findings. First, we found that the duration of daily digital screen use among school-age children during the COVID-19 outbreak in China is substantial, measuring at 3.9 h daily on average. Once weighting by viewing-distance is considered, near-vision exposure to digital devices rises to 10.0 diopter hours daily. Using multivariate logistic regression analysis, we found that each additional diopter hour of digital screen use is associated with a higher likelihood of symptomatic myopia development, which can translate into significant risks considering the extended periods of time young children spend in front of digital screens daily. Second, approximately one in four school-age children in our sample do not perform any outdoor exercise during the COVID-19 school closures, and more than 70 percent engage in outdoor exercise less than three times per week. Prior studies have highlighted the positive influence outdoor playtime can have on visual acuity as well as the associated health risks the lackthereof [7]. In our analysis, we found that more frequent outdoor exercise is generally associated with a lower likelihood of myopia development. Third, approximately one in

ten school-age children in our sample report that indoor lighting condition is either too dim or too bright at home. As the main venue for learning activities during the COVID-19 pandemic, indoor lighting conditions critically affect children's vision development, and poor lighting conditions are associated with a higher likelihood of worsening vision status among children in the study sample.

We contribute to the current literature on youth public health and build on prior studies on adolescent health that examine the association between near-vision electronic use [24], outdoor activity [7], lighting condition [21], and myopic vision progression by assessing these risk factors in conjunction and leveraging an extended period of COVID-19-pandemic-induced remote learning. While this study does not directly assess how student learning is affected, prior studies have underscored the critical negative impact myopic vision can have on student learning, particularly if corrective vision interventions are not afforded [25]. In this regard, our findings tend to confirm speculative predictions of a myopia boom during the COVID-19 pandemic [3] and are consistent with a recent longitudinal study on Chinese youth that indicated the positive association between digital screen exposure and prevalence of myopia [26] while complementing a recent cross-sectional study that examined how home confinement can have adverse effects on youth vision health [27], with richer information on digital screen use and indoor lighting condition. Based on our findings, we speculate that extended periods of school closure due to public health crises, such as the COVID-19 pandemic, and consequent alternative learning arrangements at home can increase the risks of inducing a myopia boom among school-age children. However, mitigation strategies such as limiting nearvision digital screen use duration, increasing outdoor exercise frequency, and improving indoor lighting conditions may prove to be effective in reducing, or delaying, such a youth visual health crisis [28]. While these findings are context-specific to school-age youth in China, the broader behavioral and policy implications are broadly relevant for a wide range of countries whose education system is making or is expecting to implement similar e-learning accommodations during the COVID-19 pandemic.

Finally, a limitation in this study worth mentioning is relying on subjects' self-reports rather than specialist eye examinations. While professional ophthalmic evaluations would be ideal to obtain detailed information on refractive error, prior studies have suggested that subjects' self-assessment of vision status does not differ systemically from professional ophthalmic evaluations [29]. Additionally, another reason for adopting self-reported measures is to allow for rapid and large-scale survey rollout [30], which would not have been feasible given the logistical and social-distancing requirements due to the COVID-19 pandemic. Nonetheless, it has been increasingly common in optometry and visual science studies to leverage questionnaire survey designs, considering the relatively high costeffectiveness [24]. Future studies may find it useful to conduct professional ophthalmic evaluations in lieu of collecting subject self-reports.
