This cluster randomized controlled trial was conducted in the Duanzhou district of Zhaoqing City, Guangdong Province, in southern China. This setting was chosen because Zhaoqing, one of southern China’s major cities, is readily accessible from Guangzhou, the provincial capital. The 2016 national census shows that Zhaoqing has a relatively stable population of 4,084,600, which is representative of the Chinese national urban population in terms of demographic and socioeconomic characteristics [20]. Smartphone use is very widespread in this area, as throughout the rest of China, where the number of mobile phone subscribers in December 2019 (1.6 billion) exceeded the total population of 1.43 billion [21]. China has mandated 9 years of compulsory education since 1986, and the enrollment rate for the 7th grade (12-14 years) is nearly 100% in Zhaoqing. Thus, the school-going cohort is representative of children in the general population.

The research protocol was approved by the Institutional Review Board of the Zhongshan Ophthalmic Center, Sun Yat-sen University, and the study was performed in accordance with the Declaration of Helsinki. Written informed consent was obtained from at least one parent or guardian of all participating children. The date of registration of the clinical trial was March 16, 2020.

The COVID-19 outbreak was recognized in China during the nationwide winter school vacation. On January 27, 2020, China’s Ministry of Education announced the postponement of the 2020 spring semester, and all schools were asked to recommend that students stay at home, avoiding group activities and large gatherings. As the outbreak continued, the Guangdong Education Department announced in February 2020 that secondary schools would begin formal online education beginning March 2, 2020. Teachers at the 14 public secondary schools in Duanzhou district (Figure S1 and Table S1, Multimedia Appendix 1) formulated an online curriculum for homeschooling, with standardized content and schedules for all students in the area (Table S2, Multimedia Appendix 1).

After excluding 2 secondary schools owing to an insufficient number of students per cluster, permission to conduct the study was requested from the remaining public secondary schools (n=12) in the district (Figure S1, Multimedia Appendix 1). Inclusion criteria were as follows: (1) grade 7 (12-13 years old) students in Duanzhou district and (2) under home confinement and enrolled in online learning courses during the COVID-19 outbreak. Exclusion criteria included the presence of disorders such as autism, pervasive developmental delay, and schizophrenia, which might interfere with participation in the intervention. The study followed the CONSORT (Consolidated Standards of Reporting Trials) guideline [22] (Multimedia Appendix 2). The reporting of the mobile-phone–based questionnaires followed the CHERRIES checklist (Checklist for Reporting Results of Internet E-Surveys) [23] (Multimedia Appendix 3).

Cluster randomization with schools as the cluster was adopted in order to maximize peer-to-peer support while avoiding contamination in the use of social media between study groups. A total of 12 schools with a block size of 4 were assigned (1:1) to receive either health education + download of the peer-to-peer live-streaming app (intervention group) or health education only (control). The randomization sequence was generated by an independent statistician using an online random number generator. In each school, 2 classes (on average, 40 students per class) were further randomly selected for participation.

Due to the nature of the intervention, students were not masked. However, masking of investigators was achieved through the exclusive use of electronic, self-administered questionnaires and masking the statistician to group allocation until completion of all analyses. In addition, participating students were informed that their responses to questionnaires would not be made available to their parents or teachers. Qualitative feedback on the app was not obtained from participants.

Recruitment and enrollment were assisted by trained teachers, who described the study in detail to participants and their parents over the phone. Participants and their parents were offered the opportunity to ask questions about participation. Before randomization, online seminars for parents were conducted through all 12 participating schools, during which the investigators answered questions and collected consent forms. Teachers were trained and were provided with a manual of operations, which included the study protocol and standard operating procedures for the study.

During the period of home confinement, the Chinese government had already issued a policy recommending a specific schedule for recess and physical activity breaks for students. In the control group, teachers delivered an online health information session covering the following topics and health advice to students: (1) an outline was provided on the recommended 20-20-20 rule during study and viewing of on-screen content; (2) during recess (15 mins for each recess; 4 times per day), participants in the control group received SMS text message prompts (≤50 characters) to participate in broadcast exercise programs at home, eye relaxation, or to stretch for 10 minutes. Students had access to at-home workout videos developed by exercise physiologists. Breaks were part of the online curriculum, and students were instructed by teachers (who were not aware of the study allocation) to rest and take exercise breaks according to government-issued recommendations.

Students in the intervention group received the identical health information session, online curriculum, workout videos, and breaks as described above. Additionally, at the beginning of the study, students in the intervention group were asked to log on and download a peer-to-peer live-streaming app (the Recess and Exercise Advocacy Program [REAP]). REAP is a live-streaming platform that allows users to capture short videos and photographs with their smartphones related to their physical exercise or eye relaxation activities (eg, looking outdoors through the window). The app has been optimized for both iOS (Apple) and Android operating systems. During each recess from homeschooling (15 mins for each recess; 4 times per day), participants in the intervention were prompted by SMS text messaging (≤50 characters) to log in to the REAP app to participate in live streaming or posting their workouts and outdoor videos and photographs (Supplementary Method S1, Multimedia Appendix 1).

The outcomes were measured by a self-assessed survey through mobile-phone–based questionnaires after signing the informed consent form. The questionnaires were surveyed at the beginning of the study and at the 2-week follow-up. At the start of the questionnaire, participants were given the following information: the purpose of the study, the expected length of time of the mobile phone survey, the names of the investigators, and the way and duration the data were stored. To ensure completeness checks, the participants were asked to complete the mandatory items before the questionnaire was submitted. The study investigators ensured that the confidentiality of the participants’ data was preserved. Individual participant data will not be disclosed to outside personnel and will not appear in any publications. Data assistants entered the questionnaires into a database and pseudonymized the data before it passed to the study statistician.

The primary study outcome was a change in self-reported anxiety score between baseline and the 2-week follow-up, as measured using the 45-item Chinese version of the Spence Children’s Anxiety Scale (SCAS). In addition to the total SCAS score, this measure assesses 6 domains of anxiety: panic/agoraphobia, separation anxiety, social phobia, fears of physical injury, obsessive compulsive problems, and generalized anxiety. Details relating to the SCAS can be found in Multimedia Appendix 1 [24]. This instrument has been well validated in Chinese children [25]. A higher total score (range 0-114) summing the point values of all items indicates greater anxiety.

Secondary outcomes included changes from baseline to the 2-week follow-up in the following items: children’s self-reported eye strain, sleep quality, and time spent on different near-work activities. Eye strain was measured with the self-reported Computer Vision Syndrome Questionnaire (CVS‐Q), which evaluates the frequency (never, occasionally, or often/always) and intensity (moderate or intense) of 16 symptoms. Details relating to the CVS-Q scale are available in Multimedia Appendix 1 [26]. The 4-item Patient-Reported Outcomes Measurement Information System (PROMIS) pediatric sleep disturbance questionnaire was used to assess sleep disturbance (Multimedia Appendix 1) [27]. Self-reported information regarding the average daily time spent on near-work activities was collected, including each of the following: reading; writing; and use of computers/tablets, smartphones, television, and video games (Multimedia Appendix 1).

In addition, parentally reported anxiety scores on behalf of children were obtained using the SCAS for Parent (SCAS-P) questionnaire (Multimedia Appendix 1). This variable was not prespecified but was used to assess the reliability of self-reported scores from children.

The sample size was calculated based on a cluster randomized design, assuming that children’s self-reported raw anxiety score in the intervention group would be reduced by 2.5 points, with a standard deviation of 5 points, and the anxiety score of the control group would not change. The average cluster size was about 80, and an intracluster correlation coefficient of 0.02 was estimated based on the Refractive Error Studies in Children [28]. A sample size of 8 schools (4 in each group) at a two-sided significance level of .05 would give a power of 90%. Assuming a participation rate of 90% and attrition of 20%, a total of 12 schools was required. The sample size was calculated using PASS 16.0 (NCSS, LLC).

The distribution of baseline characteristics was presented as the mean (SD) for continuous variables and frequency (percentage) for categorical variables. The unadjusted mean differences between study groups in 2-week change and the 95% CI for primary and secondary outcomes were calculated using linear regression. The adjusted intervention effect on the primary outcome and 95% CI were estimated using linear regression models, adjusting for baseline measures. The study group and all variables with P<.20 in univariable regression analyses were included in the multivariable regression analysis. Histograms and quantile-quantile plots were used to verify the normality assumptions of the t test and linear regression models. Square-root transformation was applied to all outcomes due to lack of normality. To satisfy intention-to-treat criteria, we conducted missing data imputation. Multiple imputation was used to create 20 copies of the data, and final results were obtained by averaging these data sets using the rules developed by Rubin [29]. A two-sided P value less than .05 was considered statistically significant. All analyses were performed using Stata 15.0 (Stata Corp). The study protocol was registered at ClinicalTrials.gov (NCT04309097) prior to enrollment of the first participant and is available for online access.

Requests for anonymized individual participant data and study documents will be considered on a case-by-case basis and on scientific merit by the principal investigators and approved by the relevant institutional review boards.

Our findings from a geographically representative region in urban China showed that a digital behavior change intervention effectively reduced anxiety and eye strain during children’s homeschooling, without increasing overall screen time. This digital tool might be useful during the COVID-19 pandemic, which has led to widespread school closures, with over a billion children around the world receiving online homeschooling, potentially increasing risk of anxiety and eye strain [10,30,31].

A recent review suggested that children’s mental and visual health may suffer during home confinement and online schooling [32]. These effects may result from reduced face-to-face social interaction, persistent and intense screen exposure, prolonged time spent on near-work activities, and the disruption of normal life rhythms, all of which may contribute to increased anxiety, eye strain, and sleep disturbance [33]. Children with preexisting mental health conditions such as depression may be especially susceptible to such problems [7].

The app only permitted users to post content related to exercise and activities promoting eye relaxation. This focus on activities highly relevant to both parents and children during the COVID-19 lockdown may explain the high study participation rate of over 95% and the large counts of children who shared photos or video content on a daily basis (Figure 2). We did not record any adverse events related to our intervention, nor did we observe any significant increase in screen time in the intervention group (Table 1). This latter finding is significant in view of concerns that children’s use of smartphones and social media may increase the risk of attention-deficit disorder, sleep disturbance, obesity, and other conditions [18,34,35]. Further studies are needed to investigate the long-term impacts of our digital intervention on children’s social media behavior.

Strengths of the current study include the fact that participants were recruited from a population-representative sample of schools in a typical Chinese urban region, allowing our findings to be generalized to a wider range of Chinese children affected by the COVID-19 lockdown. Second, because of social media’s reliance on peer support, our protocol followed a cluster randomized design, which minimized the risk of cross-contamination in social media use between study groups. Thirdly, children in both study groups received the same online courses, ensuring intergroup homogeneity except with respect to the intervention. Finally, the results of primary, secondary, and sensitivity analyses were in full agreement; for example, there was concordance between self- and parentally reported anxiety scores (Supplementary Table S7, Multimedia Appendix 1), indicating the robustness and reliability of our findings.

Study limitations must also be acknowledged. First, masking of participants was not feasible due to the nature of the intervention, so the possibility of placebo effects cannot be excluded. However, good agreement between parental and self-report of anxiety makes this somewhat less likely. Second, data for the main outcome were obtained by questionnaires and were thus based on an evaluation of symptoms, rather than a clinical diagnosis of anxiety disorder. Regardless, the reproducibility of this questionnaire has been well validated [21], and, as mentioned above, parental and self-reported scores were in accord (Supplementary Table S7, Multimedia Appendix 1). Third, the study design allowed for only 2 weeks of follow-up. Previous studies have shown that interventions of this length can be effective [36-38], and such brief periods of engagement may be particularly applicable to unpredictable, short-term school closures [39], which may become more common in the future with additional waves of the COVID-19 pandemic. For example, a brief (2-3 week) lay counsellor–delivered, problem-solving intervention was found to be effective for adolescents with diverse mental health problems [38]. Nevertheless, it is certain that a longer follow-up will provide additional information on the robustness and stability of this intervention effect. Fourth, only Chinese children were included in the study, and the effectiveness of live-streaming interventions to reduce children’s anxiety and eye strain must be validated in other settings. Fifth, due to the nature of our behavior intervention, we could not exclude the possibility that the changes in children’s anxiety and eye strain may have been caused by other factors (eg, the increased interactions between children and parents), rather than a change in the behavior itself. In addition, the questionnaires were not fully validated for online use, which is the only means that can be used during COVID-19–related home learning. Finally, the urban sample used in the study may not be representative of China, not only because of the lack of smartphones in less developed parts of China but also due to the higher probability of separation between children and their parents due to labor migration. Our results therefore need to be interpreted with caution. 

In summary, this study demonstrated that a novel digital behavior change intervention can significantly reduce anxiety and eye strain in homeschooled children. Further work is needed to determine whether these results can be extended to other aspects of pediatric and adult health, and to different geographic and cultural settings.