Realistically, SB is an important factor in the formation of myopia, that is, b1 in Figure 1 is statistically positively significant. SB is conceptualized as any waking behavior characterized by an energy expenditure ≤1.5 metabolic equivalents (METs). For college students who were segregated during the COVID-19 epidemic, SB refers mainly to sedentary reading, and the relationship between sedentary reading and myopia has been consistently reported. The frequency and duration of sedentary reading were significantly and positively associated with the risk ratio of developing myopia in the United States (19), China (20), and Japan (21). Even in Sweden, where schoolwork is very stressful, this influence was positively significant (22). Therefore, we propose the first hypothesis of this study:

Myopia is a refractive error triggered by prolonged inappropriate indoor eye use, and PA apparently helps alleviate this disadvantage, namely b2 in Figure 1 is statistically negatively significant. During the COVID-19 epidemic, PA was one of the few ways to relieve eye stress among college students. Moderate PA reduced indoor eye time among college students, which was widely demonstrated. It is also worth noting that the potential of PA for myopia relief is enormous (23, 24). A World Health Organization report states that 70% of college students lack sufficient PA, and if college students were motivated to have more PA, it would go a long way toward alleviating the myopia epidemic. We, therefore, propose the second research hypothesis:

In the context of the COVID-19 epidemic, the relationship between ST and SB is significant, this phenomenon is even more pronounced in China. This is because during the COVID-19 epidemic, China required that almost every college student learn online. Chinese universities, unlike many European universities, require students to complete a vast amount of coursework, and online learning forces Chinese college students to spend more time completing coursework, significantly increasing sedentary time. Therefore, we propose the third research hypothesis:

From an academic perspective, ST and PA are mutually exclusive, and this relationship has been widely reported, that is, a1 in Figure 1 is statistically positively significant. This phenomenon is even more evident in college students, compared to their elementary and middle school student counterparts. This is because college students are unsupervised in their use of electronic devices, which creates two pathways for ST to influence PA. On the one hand, ST directly reduces the PA time, because for a given college individual, spare time is generally regarded as a constant (25). On the other hand, excessive ST would generate negative emotions and indirectly influence PA (26). Therefore, we propose the fourth research hypothesis:

Academically speaking, SB negatively influences PA, similar to the way ST influences PA, namely a2 in Figure 1 is statistically negatively significant. On the one hand, SB directly reduces the amount of time individual college students spend engaging in PA. On the other hand, excessive SB will cause irreversible physical issues such as obesity, varicose veins, and even lumbar disc herniation, influencing college students' physical functions and hindering them from physical activity. Therefore, we propose the fifth research hypothesis:

The relationship between ST and myopia has not been consistently reported. On the one hand, some scholars have suggested a positive relationship between ST and myopia (8, 27). On the other hand, some researchers have demonstrated no statistically significant relationship between ST and myopia within a large data sample (13, 14). Moreover, these arguments exist in many countries, such as Australia, the United States, China, and New Zealand (28).

Thus, we suggest that SB influences myopia through three pathways: (a) ST increases SB and leads to increased myopia, as in a1-b1, (b) ST decreases PA and leads to increased myopia, as in a2-b1, and (c) ST increases SB and leads to decreased PA, and ultimately increased myopia, as in a1-d-b2. Taken together, because the mediation consisting of SB and PA sequentially, completely mediated the influence of ST on myopia, the direct influence of ST on myopia was not significant. We therefore propose the sixth research hypothesis:

Experimental subjects were recruited for a nationwide experiment among Chinese college students during the COVID-19 epidemic, which aimed at estimating the potential psychological and physical changes of Chinese college students due to the COVID-19 epidemic. We strictly followed the multi-stage stratified random cluster sampling principles. First, we randomly selected seven universities in mainland China. Second, we randomly selected different classes in all universities. Third, we distributed paper questionnaires to selected students, and assigned special personnel to supervise the completion process of questionnaires. Data was collected from 13 Sep to 12 Dec, 2021. Finishing a questionnaire took about 20 min for college individuals. The inclusion standards of valid respondents were as follows: The information of respondents' demographic characteristics, myopia, ST, SB, and PA were validated, simultaneously. After excluding the invalid respondents, the final data size was 759. Our research was authorized by the Institutional Review Board of Xian Jiaotong University and obeyed the ethics code of the World Medical Association Declaration of Helsinki.

The main independent variable in our empirical research was ST. We designed a fill-in-the-blank question to ask respondents how many minutes they used their phones last week as follows:

In order to ensure the authenticity of the data, individuals were asked to look for the “healthy use of mobile phone” option (Android system) or the “screen use time” option (Apple System) in the “Settings” option of the mobile phone and report it in the questionnaire after converting it into minutes.

The information of SB of respondents was collected by the following question in our research:

PA was obtained by the International Physical Activity Questionnaire (IPAQ), which has proven to be a standard questionnaire that effectively displayed the intensity of physical activity of the respondents. The days, hours, and minutes of vigorous, moderate, and mild exercise every day last week were systematically investigated, respectively. The index obtained from the three activities above were multiplied by 8.0, 4.0, and 3.3, respectively, and finally generated the metabolic equivalent (MET) as follows:

Minutes_vigorous, Minutes_moderate, Minutes_mild denoted the minutes of vigorous, moderate, and mild physical activity performed in the last week. After taking the required number of days as the constraint, A three category variable was obtained, 1, 2, and 3 denote low, mid, and high PA, respectively.

The dependent variable in our research was myopia, a self-administered question was used to investigate the myopia of respondents as follows:

Covariates were set to control the statistical bias in the estimation, contained respondents' gender (1, 2 denoted male or female, respectively), education attainment (1, 2, and 3 represented Freshman or Sophomore, Junior or Senior, Master or Doctor, respectively), self-perceived family wealth (1, 2, and 3 denoted low, medium, high, respectively), residence (1, 2 denoted urban and rural, respectively), subjective health status (1, 2, and 3 denoted low, medium, high, respectively). The above factors have been proven to have a significant influence on myopia (29, 30).

We chose to use MPLUS 8.2 to analyze the potential relationship between ST and myopia in Chinese college students during the COVID-19 epidemic. A series of SEM models were built for this study based on the framework of a serial multiple mediator models that includes multiple mediators that are sequentially linked within a single model (31). This method was appropriate for our research because two moderations were sequentially placed in the association between ST and myopia. Heterogeneous information of respondents was shown in descriptive statistics by the myopia group, to reveal the potential relationship between ST, SB, PA, and myopia preliminarily.

Following the extant literature, we hypothesized that SB and PA indirectly participated in the relationship between ST and myopia. In order to prove the hypothesis, we first verified the relationship between the above four factors through a series of regression models, to provide empirical support for subsequent SEM analysis.

And then, an SEM was designed to describe the potential relationship between ST and myopia, with a serial multiple mediator of SB and PA, sequentially. A modification index (MI) was used to optimize our model through the specific code of modindices in MPLUS. Direct and indirect effects of the correlation between ST and myopia were estimated simultaneously. As for the multivariate non-normality, bootstraps (5,000 times) was the repeated sampling technology used in our research to reduce bias. Meanwhile, we took all covariates into consideration.

Table 1 briefly showed the results of the descriptive statistics. The percentage of college students with myopia accounted for 82.87% of the total students, and mild, moderate, and severe myopia in students consisted of 17.13, 27.01, and 11.33%, respectively. Meanwhile, the proportion of students without myopia in the full sample was only 17.13%, which indicated that myopia was extremely prevalent among Chinese college students during COVID-19.

Additionally, Table 1 revealed the statistically significant differences in ST, SB, and PA among individuals with different levels of myopia. To further reveal the association between SB, ST, PA, and myopia, we graphically depicted the relationships between SB and myopia, ST and myopia, and PA and myopia as shown in Figure 2 in (A), (B), and (C), respectively. In the context of descriptive statistics, Figure 2A verifies that high SB predicts high myopia, Figure 2B verifies that nearly no significant relationship between ST and myopia, and Figure 2C verifies that high PA predicts low myopia. In addition, we found some outliers in Figures 2A,B, that support our use of the “Winsor” command for removing the outliers.

As shown in the correlation results, SB, PA, and several covariates were tightly associated with myopia, while there was no statistically significant correlation between ST and myopia. Warranting particular attention, there was a statistically positive significant relationship between ST and SB, and a statistically negative relationship between ST and PA. The results of the correlation analysis preliminarily indicated that a serial multiple mediator that encompassed ST and PA sequentially, might mediate the relationship between ST and myopia.

We set up three regression models to further verify the relationship between ST, SB, PA, and myopia. It was worth noting the several outliers in the sample (as shown in Figure 2), therefore we used the “winsor 2” command in stata to shrink the tails of all regressions by 10%. Model one focused on the relationship between ST and SB, holding all else equal, an increase of 10% in ST was associated with a 1.2% increase in SB, statistically significant (Coefficient = 0.120, SD = 0.037, P < 0.05). This conclusion is consistent with the extant literature, especially in the East Asian countries (32, 33). Of interest, these countries were defined characteristically by a high prevalence of online learning; accordingly, online learning might bring an inevitable disadvantage and impair the eyesight of college students.

As shown in the middle two columns of Table 2 (Model 2), a negative correlation between PA and SB could be found. Holding all else equal, an increase of 10% in SB was associated with a 4.9% decrease in PA—statistically significant (Coefficient = −0.491, SD = 0.207, P < 0.05)—while the correlation between ST and PA was statistically insignificant (Coefficient = −0.116, SD = 0.208, P > 0.05). These conclusions are rarely found in previous literature because the previous studies rarely took ST and SB into the relationship between ST and myopia.

Model 3 reveals the correlation between ST, SB, PA, and myopia, simultaneously. First, the correlation between PA and myopia was statistically significant (Coefficient = 1.05, SD = 0.45, P < 0.05). Holding all else equal, college students with high PA had a 5.1% lower relative risk of developing myopia compared with those with low PA. Similar conclusions had been found in Denmark, Australia, and other countries (34, 35). Second, the correlation between SB and myopia was statistically significant and positive (Coefficient = 0.573, SD = 0.182, p < 0.05). When all else is constant, an increase of 10% in SB was clearly related to a 5.32% increase in myopia. The conclusion corroborated the findings in the Pearson correlation test. Third, the results indicated no significant association between myopia and ST (Coefficient = 0.172, SD = 0.183, p > 0.05). We reached the opposite conclusion compared with a representative meta-analysis which indicated that ST significantly predicted an increased risk of myopia (28). A possible explanation was that the confounding influence of ST and PA was not considered in this meta-analysis, we isolated the net influence of ST by controlling the SB and PA and eventually drew a more accurate result.

We set up an SEM to reveal the potential association between ST and myopia. According to Figure 3, the serial multiple mediator model showed acceptable fit statistics: X² = 135.289, p < 0.05, CFI = 0.944 > 0.90; TLI = 0.971 > 0.90, SRMR = 0.059 < 0.08.

As shown in Table 3, ST was positively associated with SB, statistically significant (Coefficient = 0.120, 95% CI = 0.036; 0.204, p < 0.05); thus, hypothesis 3 was verified. While there was no statistically significant relationship between ST and myopia (Coefficient = 0.148, 95% CI = 0.135; 0.432, p > 0.05), ST and PA (Coefficient = −0.091, 95% CI = −0.387; 0.204, p > 0.05), simultaneously. Meanwhile, a 10% increase in SB and PA was significantly associated with a 4.41% increase (Coefficient = 0.441, 95% CI = 0.158; 0.725, p < 0.05) and 5.49% decrease (Coefficient = −0.549, 95% CI = −0.829; −0.268, p < 0.05) in myopia, respectively; thus, hypothesis 1 and hypothesis 2 were simultaneously verified. It is worth noting that there was a significant negative relationship between SB and PA (Coefficient = −0.520, 95% CI = −0.855; −0.186, p < 0.05); so, hypothesis 5 was verified. Additionally, there is no significant difference in myopia between men and women (Coefficient = 0.056, 95% CI = −0.234;0.346, p > 0.05). Similarly, there were no significant differences in myopia among college students between different grades (Coefficient = 0.153, 95% CI = −0.077; 0.383, p > 0.05), different residence (Coefficient = 0.142, 95% CI = −0.149; 0.432, p >0.05), and different self-perceived family wealth (Coefficient = −0.015, 95% CI = −0.287; 0.258, p > 0.05). However, the better the subjective health status, the lower the probability of getting myopia (Coefficient = −0.229, 95% CI = −0.457; −0.002, p < 0.05).

The last five rows of Table 2 showed the indirect effects including specific and total indirect effects (the combination of specific indirect effects), and total effects (the combination of indirect and total direct effects) in addition to the direct effects. The total effects of the model were 0.286 and statistically insignificant (Coefficient = 0.286, 95% CI = −0.065; 0.636, p > 0.05). This is acceptable in our model because there were two opposite-direction mediating paths in the relation between ST and myopia, simultaneously. Additionally, 51.75% of the effects originated through the direct effects between ST and myopia, statistically insignificant (Coefficient = 0.148, 95% CI = −0.135; 0.432), which indicated that there was no direct correlation between ST and myopia. However, the total indirect effects were statistically significant (Coefficient = 0.137, 0.042; 0.317, p < 0.05). On the one hand, specific indirect effects through SB accounted for 38.6% of the indirect effect (Coefficient = 0.053, 95% CI = 0.003;0.103, p < 0.05). On the other hand, the serial multiple mediator consisting of SB and PA, showed significantly positive indirect effects (Coefficient = 0.034, 95% CI = 0.004–0.065, p < 0.05). Eventually, hypothesis 6 was verified.

This model supported that serial mediation, including SB and PA sequentially, existed in predicting the relationship between ST and myopia. Since the coefficients indicated opposing directions, our results revealed a suppressive and promoted the role of SB and PA, respectively. More specifically, an increase of 10% in ST was associated with a 2.86% increase in myopia, though fully mediated by SB and PA. Copious studies supported that there was no significant relationship between ST and myopia, because myopia prevalence increased before the massive use of digital devices in some countries and territories (36). However, we found that a serial multiple mediator consisting of SB and PA, completely mediated the relationship between ST and myopia. In other words, the rise in ST increases SB and reduces PA, and accordingly indirectly increases the risk of myopia.