Sample size calculation
To calculate the sample size, we used the following formula [18]:
$$N=\frac{{Z}^{2}\times \left(1-P\right)P}{{\delta }^{2}}\times deff$$
Where N = sample size; Z = Z statistic for confidence level; P = expected prevalence rate; \(\delta\)= allowable error. We used the design effect deff to calculate the sample size due to multistage sampling methods in the present study.
In this study, the sample size was calculated to meet the following conditions: (1) the preliminary estimate of ADSs prevalence was 23.6% in the pre-survey study; (2) the allowable error was taken to ensure accuracy by using 3%; (3) for a 95% confidence level, which was conventional, the Z-value was 1.96; (4) the investigators present their results with 95% confidence intervals (CI) and deff = 2. In this study, the sample size was calculated as 1,539. Additionally, the calculation of sample size needed to consider students who were lost in the follow-up survey, student rejection rate, sampling error, and stratification factors. The final sample size was satisfied in our study as 2,160.
Study methods
This study was a school-based, two-center cross-sectional study, in which six schools were recruited in Hangzhou City. We used a random number table, and a stratified cluster sampling method to select participants. In total, a sample of 360 students aged 13–18 years at each of the enrolled schools were randomly selected from Jiande County and Fuyang District of Hangzhou City [i.e., two School Health Surveillance System (SHSS) centers]. To be included, participants had to be mainstream adolescents (i.e., had no intellectual disability), aged 13–18 years, and could converse in Chinese. Participants were excluded if they had a history of psychosis or neurocognitive deficits or received a secondary mental health service. Figure 1 provided further details of this study.
In the present study, ADSs were estimated in a two-stage appraisal procedure. In the first stage, ADSs were screened by the Adolescent Dissociative Experience Scale (A-DES) [19]. In the second stage, subjects whose responses to the A-DES suggested they might have ADSs were further evaluated by three psychiatrists to obtain a final diagnosis. The investigators included uniformly trained psychiatrists, medical students, and school health care personnel. About 60 trained investigators performed 2 questionnaires to participants that were estimated as having ADSs. The investigators administered face-to-face evaluations and investigations in each school. In addition, 3 trained investigators were present in each school in order to guarantee the quality of our research process.
Measures and procedures
ADSs measurement
We used a two-stage identification procedure to appraise ADSs. In the first stage, ADSs were screened for using the A-DES, which contained 30 items [19]. The A-DES is a broadly applied self-reporting dissociative symptoms scale discussed by Likert as 11-point counts ranging from "never" to "always, where each symptom is scored 0–10 [19]. A dissociation score was calculated from this scale by dividing by 30 from the summing item scores of A-DES, with higher scores indicating greater severity [12,19].
Participants whose responses to the A-DES suggested that they might have ADSs (i.e. a medium or high dissociation scores ≥ 3), which were further estimated by three psychiatrists (such as three chief physicians) to obtain a final diagnosis as follows. First, they carefully reviewed the diagnostic criteria of ADD and related studies on DSs due to ensuring the diagnostic consistency and accuracy of the three psychiatrists. Next, the psychiatrists used the Diagnostic and Statistical Manual of Mental Disorders-5 and the Chinese classification of mental disorders to eliminate all patients with DSs and other mental or personality disorders. Then, to ensure the appraisal consistency of the different psychiatrists, they utilized a mutual evaluation form; thus, ADSs were only confirmed when DSs were consistently diagnosed by all three psychiatrists. In addition, ADSs were considered if a subject had (1) the exclusion of organic diseases, or other mental or personality disorders; (2) no history of neurological/psychiatric illness; and (3) no intelligence deficit (i.e. an intelligence quotient > 70).
The reliability of the questionnaire was tested in the pre-survey study. The responses were analyzed using the Cronbach’s alpha. The result revealed an internal consistency was 0.97 (Cronbach alpha).
Social adjustment status measurement
We used the Chinese Adaptation Scale for Adolescents (CASA) to assess the social adjustment status of adolescents [16,20]. The CASA comprises three factors: emotional adaptation (6 items), social adaptation (6 items), and study and life adaptation (5 items) [20]. Previous studies from our group showed that the CASA has high reliability and validity, with a Cronbach coefficient of 0.88 [16]. The internal consistency was 0.92 in the present sample.
School environment status measurement
Teacher-student relationship
We used the Student-Teacher Relationship Scale (STRS) to evaluate the relationship between teachers and students [21]. The STRS is a 28-item self-report instrument, which includes four domains: intimacy, conflict, support, and satisfaction [21]. Our previous study suggested that the Chinese version of the STRS (STRS-CV) has good internal consistency and acceptable test-retest reliability (with a Cronbach coefficient of 0.87) [16]. The internal consistency was 0.83 in the present sample.
Peer relationship
We used the Chinese version of the Peer Relationship Inventory (CPRI) to estimate adolescents’ peer relationships [22]. The CPRI is a 20-item self-report instrument, which measures three analytically derived dimensions of peer relations: social maturity, aggression, and independence [22]. In our previous studies, the scale has high reliability and validity [16]. The peer relationship was analyzed using a five-point Likert count for a total score of 20–100 points. The internal consistency was 0.90 in the present sample.
Family environment status measurement
We used the Chinese Version of Family Environment Scale (FES-CV) to assess relationships, personal growth, and system maintenance in one’s family environment [23]. The FES-CV is a 90-item self-report instrument [23]. Prior studies indicated that ten subscales of the FES-CV showed moderate to excellent internal consistency (ranging from 0.63–0.75) and acceptable test-retest reliability at 0.55–0.92 [16,23]. The internal consistency of this scale was 0.85 in the present sample.
Mental health status measurement
We used the Chinese Version of Symptom Check List-90 (SCL-90-CV) to estimate the mental health status of adolescents [24]. The SCL-90-CV is a 90-item self-report instrument, which includes ten factors (i.e., objective-comprehensive, somatic complications, interpersonal sensitivity, depression, anxiety, hostility, phobic anxiety, paranoid ideation, psychoticism, and diet/sleeping) that respond to an individual’s specific aspect of the symptomatic situation [24,25]. Based on the scores of each factor, it is possible to initially determine which factors represent problematic psychological symptoms [24]. Related researches demonstrated that the SCL-90-CV had good internal consistency and acceptable reliability [24,25]. In the present sample, the internal consistency was 0.99.
Statistical analysis
All data were double entered and verified by using EpiData 3.1 software. We analyzed the data using IBM SPSS Statistics 26 software (SPSS, Inc. Chicago, IL, United States). In addition, AMOS 24.0 (Amos Development Co. Greene, MD, United States) was used for performing structural equation model (SEM) and mediation analyses. Multiple imputation was used to handle missing data.
We used the Kolmogorov-Smirnov test to assess the normality of quantitative variables. We calculated the frequencies and percentages for categorical variables. Continuous variables was presented as the mean ± standard deviation.
Univariate analyses were used to separately analyze all the variables that were potentially associated with ADSs, which they were included in multiple regression models at a significance level of 0.05. The one-way analysis of variance test and independent samples t-test were used for normal distributed data. Pearson’s chi-squared test was used for categorical variables.
We used multiple linear regression models to evaluate whether the variables, including demographics, environmental factors, and individual psychological factors, were associated with ADSs. A stepwise procedure was used to select further the variables related to ADSs at a significance level of P > 0.05 for removal and a significance level of P ≤ 0.05 for reentry. We performed the hypothesis testing to indicate statistical significance by using a two-sided test with an alpha value of 0.05.
AMOS 24.0 was selected to construct the SEM. Beta values were used to report direct, indirect, and total effects, and P-value<0.05 was considered statistically significant. Model fit indices, including Chi-square (CMIN), Cardinality of freedom ratio (CMIN/DF), Root Mean Square Error of Approximation (RMSEA), Goodness of Fit Index (GFI), and Comparative Fit Index (CFI), were used to determine the best SEM.