Prevalence of sleep disturbances in Chinese adolescents: A systematic review and meta-analysis

Objectives To review cross-sectional studies on the prevalence of sleep disturbance in Chinese adolescents and use a meta-analysis to explore the factors that may explain the heterogeneity between estimates of the prevalence. Methods We followed the Preferred Reporting Items for Systematic Review and Meta-analysis and searched the PubMed, Embase, Cochrane Library, PsycINFO, Scopus, Web of Science, SinoMed, Chinese National Knowledge Infrastructure (CNKI), WanFang, and VIP Database for Chinese Technical Periodicals databases from their inception through June 30, 2020. Analysis of the abstract, full-text, and data were conducted independently with uniform standards. Sub-group analyses and meta-regression analyses were performed to explore the associations between prevalence and gender, sex ratio, mean age, area, studying stage, sample size, survey time, response rate, assessment tools, PSQI cut-off, and quality score of the study. Results A total of 63 studies (64 groups of outcomes) were included in our analysis, covering 430,422 adolescents across China, of which 104,802 adolescents had sleep disturbances. The overall pooled prevalence was 26% (95% CI: 24–27%). Adolescents in senior high school (28%, 95% CI: 24–31%, p<0.001) had a higher prevalence of sleep disturbances than those in junior high school (20%, 95% CI: 15–24%, p<0.001). Studies with effective sample size of more than 1,000 and less than 3,000 had the highest prevalence of 30% (95% CI: 24–35%, p<0.001). The prevalence of sleep disturbances was not affected by other factors. Conclusion This systematic review and meta-analysis revealed that sleep disturbances are common in Chinese adolescents, and effective psychological and behavior intervention may be needed to help adolescents solve their sleep problems.


Study selection
Studies were included in the analysis if they met the following criteria: (1) subjects aged from 13 to 19 years old in China; (2) cross-sectional epidemiology survey reported the prevalence of sleep disturbances with standardized assessment tools (Pittsburgh Sleep Quality Index/Athens Insomnia Scale/ICSD-2); (3) sample size �300; (4) full text written in Chinese or English. Studies were excluded for the following reasons: (1) only research articles were considered (conference abstracts, protocols, reviews, meta-analyses, citations, comments, and news release were excluded); (2) missing sampling method, sample size, or response rate; (3) based on experiments of clinical treatment or intervention; (4) focused on special population, such as patients with mental problems, students with post-traumatic stress disorder, and teenagers with medical conditions. Any disagreement in the screening procedure was discussed and resolved by a third reviewer.

Data extraction and study quality
Two reviewers independently screened the titles, abstracts, and full text. A duplicate extraction form was used to record information, which was tabulated using Microsoft Excel. The information included survey year, geographic location (area), effective sample size, response rate, sampling methods, male proportion, age, studying stage, assessment instruments, the cut-off for sleep disturbance, prevalence of sleep disturbance, the timeframe of sleep disturbance, and the quality score of study evaluation, which is provided in the Appendix in S1 File.
The quality of the included articles was assessed using the 11-item methodological checklist recommended by the Agency for Healthcare Research and Quality (AHRQ) [11] for cross-sectional studies (https://www.ncbi.nlm.nih.gov/books/NBK35156/). An item would be scored "1" only if it was answered as "Yes"; the item would be scored "0" when it was answered "No" or "Unclear". The quality of the study was assessed as (1) "low" when the score was 0-3; (2) "moderate" when the score was 4-7; and (3) "high" when the score was 8-11. Studies classified as "low" quality were excluded from further meta-analysis. Any discrepancy in assessment was also resolved by the third reviewer when necessary.

Data analysis
We used STATA version 14.0 (Stata Corporation, College Station, TX, USA) and the fixed/ random-effects models to calculate pooled estimates for sleep disturbance prevalence and corresponding 95% confidence intervals (CIs). Between-study heterogeneity was tested with I 2 statistic and Cochrane Q statistics, while random-effects models were used when I 2 >50% and p<0.10 indicated significant heterogeneity and I 2 >75% indicated high heterogeneity [12,13]. Subgroup analyses were performed to compare the differences in gender, area, assessment tools, effective sample size, and study stage with the chi-square test. Furthermore, the metaregression analysis, using the Knapp-Hartung modification method, was performed to explore the influence of moderators on heterogeneity. Publication bias was conducted using Begg and Egger tests and funnel plots [14,15]. Sensitivity analyses were conducted to test the consistency and quality of pooled results by removing each study individually. All tests were two-tailed with a statistically significant threshold of P <0.05.
from the American Academy of Sleep Medicine (2005). One study defined sleep disturbance using the Insomnia Self-Assessment Inventory (ISAI) from Social Development Trend Survey [19,20] in three scales: difficulty in initiating sleep (DIS), difficulty in maintaining sleep (DMS), and nonrestorative sleep (NRS). The participants from 11 survey studies were from junior high school, while 24 studies focused on students from senior high school. The other studies had students from both schools, see details in Table 1.

Methodological quality of studies
The quality of the included studies, as assessed by AHRQ, is shown in Table 2. The median study quality score was 5 (range 4-7). A total of 22 studies (34.92%) failed to indicate the exact time of the survey. Thirteen studies (20.63%) were scored 4, nine studies (14.29%) were scored 7, and the remaining were scored 5 and 6. None of the studies were scored less than 4 with low quality.

Prevalence of sleep disturbances
In total, the 63 eligible studies included 64 groups of outcomes (one study had two groups). For the meta-analysis, the prevalence of sleep disturbances in adolescence ranged from 8% in Gaomi to 54.7% in Taiwan, while the pooled prevalence was 26% (95% CI: 24-27%). Substantial heterogeneity existed between the studies, I 2 = 99.14%, chi-square = 1871.42, p <0.001, τ 2 = 0.02 (Fig 2). There were no significant changes in the results after removing each study sequentially.

Subgroup analyses
The estimates of sleep disturbance prevalence from the subgroup analysis, determined for gender, area, sample size, studying stage, survey time, instruments, PSQI cut-off, and quality score of study, are shown in Table 3. Compared to other sample sizes, effective samples of more than 1,000 and less than 3,000, indicated the highest pooled prevalence of 30% (95% CI: 24-35%, p<0.001). The prevalence of sleep disturbance in senior high school students (28%, 95% CI: 24-31%, p<0.001) was higher than junior high school students (20%, 95% CI: 15-24%, p<0.001). The subgroup of the sample size and studying stage indicated statistically significant heterogeneity with Q11.99 (p = 0.01) and Q7.77 (p = 0.01). There were no statistically significant differences in heterogeneity between the other groups.

Meta-regression analyses
The meta-regression was performed using the Knapp-Hartung modification method. Further details are presented in Table 4. The sample size between 1,000 and 3,000 (τ 2 = 0.006994, p = 0.004<0.05) and studying stage (τ 2 = 0.005651, p = 0.009<0.05) were statistically significant, which accounted for 9.78% and 20.83% of between-study heterogeneity, respectively. There were no statistically significant contributions to heterogeneity in mean age, sample size, survey time, the quality score of study (all p-values >0.05). No model was created that included all factors to explain the heterogeneity of this meta-analysis.

Publication bias
Publication bias was found using visual inspection of the funnel plot and Egger's test (t = 2.04, p = 0.046), as shown in
Age is commonly associated with the degree of disturbance in Chinese adolescents [35,38,81]; however, our findings show that the prevalence increases with the studying stage, rather � Article quality was assessed as follows: low quality = 0-3; moderate quality = 4-7; high quality = 8-11.
1. Define the source of information (survey, record review). 2. List inclusion and exclusion criteria for exposed and unexposed subjects (cases and controls) or refer to previous publications.
3. Indicate time period used for identifying patients.
4. Indicate whether or not subjects were consecutive if not population-based.
5. Indicate if evaluators of subjective components of study were masked to other aspects of the status of the participants.
6. Describe any assessments undertaken for quality assurance purposes (e.g., test/retest of primary outcome measurements).
7. Explain any patient exclusions from analysis. 8. Describe how confounding was assessed and/or controlled.
9. If applicable, explain how missing data were handled in the analysis. 10. Summarize patient response rates and completeness of data collection. 11. Clarify what follow-up, if any, was expected and the percentage of patients for which incomplete data or follow-up was obtained.

PLOS ONE
Prevalence of sleep disturbances in Chinese adolescents than the mean age. Senior high school students had a higher pooled prevalence of sleep disturbance compared to junior high school students (28%, 95% CI: 24-31% vs. 20%, 95% CI: 15-24%) in China, which may be attributed to academic pressure, insufficient sport duration, and increased anxiety/depression levels [33,34,56]. The difference in the studying stage accounted for 20.83% of the variation for between-study heterogeneity (τ 2 = 0.005651, p = 0.009). Gender did not affect the pooled prevalence in our results, which is inconsistent with other studies [31,36,52,55,66]. There were also no differences in the sex ratio, geographical location, survey time, or quality evaluation of the studies. In total, 84.13% of studies used PSQI and 79.19% used PSQI, with a cut-off �8, and there were no significant differences between PSQI and other instruments and between PSQI cut-off �8 and others through the sub-group analyses. This was likely due to representation issues. Statistically significant differences were also found in the sample size (<500 (26%, 95% CI [24-29%]) vs. 500-100 (21%, 95% CI [18-24%]) vs. 1,000-3,000 (30%, 95% CI [24-35%]) vs. >3000 (27%, 95% CI [24-29%]). Furthermore, the sample size of 1,000-3,000 contributed to 9.78% (τ 2 = 0.006994, p = 0.004 � ) of the heterogeneity. All of the included studies with "moderate" quality pooled 104,802 adolescents with sleep disturbances, covering 14 administrative provinces and 28 cities across China, and showed a good representation of Chinese adolescents. Approximately 79.37% of the included studies response rates were more than 90%. In this meta-analysis, we used sub-group analyses and meta-regression analyses to explore the sources of heterogeneity, but we could not find an overall explanation of the heterogeneity sources. Thus, several limitations should be addressed. First, some of the studies did not report the exact survey time, which is important for students. Sleep quality is different between school days and non-school days, and between exam periods and non-exam periods [82]. Second, most studies failed to report issues of study quality control that could affect the reliability and validity of the study results. Third, some detailed characteristics of the subjects were not presented. The meta-analysis was based on published data, which could lead to difficulties in further sub-group analyses and meta-regression analyses. In addition, we believe that sub-group analyses of different types of sleep disturbances in Chinese adolescents by severity degree should be taken into consideration in future research, which will be useful to get to the point., Fourth, we could not eliminate the high level of heterogeneity in the prevalence estimate due to many different epidemiological survey studies.
In conclusion, the pooled prevalence of sleep disturbances in Chinese adolescents was 26% (95% CI, 24-27%), higher than the pooled prevalence in the general population (15%, 95% CI: 12.1-18.5%) in China. The sub-group analyses and meta-regression analyses indicated that higher prevalence was associated with studying stage and effective sample size, especially for adolescents in senior high school (28%, 95% CI: 24-31%) and sample sizes between 1,000 and 3,000 (30%, 95% CI: 24-35%). Further research is needed to better understand the sleep disturbances of adolescents to develop effective psychological and behavioral interventions.