Procedure
This self-report, cross-sectional study was carried out in 2014 in Singapore. The study design and questionnaires were adapted from Sourander and colleagues [33] as part of an international study investigating the impact of cyber environments on adolescents. In the current study, approval was obtained from the local research ethics committee. Permission was also obtained from the Ministry of Education prior to conducting the research.
A two-stage sampling strategy was employed. First, schools were randomly sampled from a list of local mainstream schools, which were stratified according to their types (i.e., degree of autonomy in school operations and amount of government funding) and geographical areas (i.e., north, south, east and west). Invitation letters were sent to the selected schools’ principals to explain the study rationale and design. Should any school declined to participate, the next school within the same strata was approached following the original stratification plan. A total of 28 schools (24 secondary schools, three junior colleges and one polytechnic) agreed to take part in the study. In the second stage, planned recruitment of approximately 120 students (four classes) from each participating school was conducted in order to achieve representative educational levels and academic streams (i.e., based on students’ performance in a national examination prior to secondary school entry) across the sample.
A passive consent procedure was used in this study and was approved by the ethics committee. All parents from the participating classes were informed about the study through written communication prepared by the investigaters and handed out by the schools at least two weeks prior to the scheduled questionnaire administration. Parents were requested to contact the school if they did not want their child to participate in the study. Assent was also obtained from the participants before they completed the online survey in their respective computer laboratories or classrooms. They were reminded about the anonymity and voluntary nature of the study participation. The overall participation rate was 85.8%. Data collection was conducted in school by study team members, which took approximately 30 to 45 minutes during school days.
Participants
A total of 3,329 students participated in the study. Ten participants were excluded from analyses as they provided invalid responses. The final sample consisted of 3,319 students with equal gender distribution (1665 males, 50.2%) and roughly proportionate educational levels. Their age ranged from 12 to 17 years (M = 14.42, SD = 1.48). Self-reported ethnic identification was as follows: 66.4% of the participants were Chinese, 22.1% were Malay, 6.0% were Indian and 5.5% endorsed Others (all other ethnic groups not listed). The distribution of ethnicity was proportionate to that of the Singapore population aged 10–19 years (34).
Data sources and collection tools
Cyberbullying
In order to improve the validity of responses, participants were provided with the definition of cyberbullying adapted from Hinduja and Patchin [35] who described cyberbullying as a phenomenon “when someone repeatedly makes fun of another person online or repeatedly picks on another person through email or text messages or when someone posts something online about another person that they don’t like”.
Participants were asked how often they have been cyberbullied and/or cyberbullied others in the past six months. These two items were rated on a 4-point Likert scale ranging from: 1 – ‘Never’, 2 – ‘Less than once a week’, 3 – ‘More than once a week’ and 4 – ‘Almost every day’. Similar to Sourander and colleagues [33], responses ‘2’ or more were referred to cyberbullying or being cybervictimised at least sometimes and hence were used for the subsequent categorisation. Based on their responses, participants were categorised into four groups: (1) Non-involved; (2) Cybervictim only; (3) Cyberbully only; and (4) Both cyberbully and cybervictim (cyberbully-victim).
Emotional and behavioral difficulties
The youth self-report version of the Strengths and Difficulties Questionnaire (SDQ) [36] was used to assess emotional and behavioral difficulties. It consists of 25 items divided into five subscales, which include emotional problems, peer problems, conduct problems, hyperactivity-inattention and prosocial behavioral subscales. Each scale comprises five items on a 3-point Likert scale ranging from: 0 – 'Not true', 1 – 'Somewhat true' and 2 – 'Certainly true'. To handle missing data, summation of each subscale was conducted if there are at least three valid items.
As participants in the current study were from the general population, broader internalising and externalising subscales were used in this study [37]. Internalising problem subscale was computed by summating emotional and peer problem subscales (Cronbach’s alpha = .70), while externalising problem subscale was a summation of conduct problem and hyperactivity/inattention subscales (Cronbach’s alpha = .65). Higher scores on these two subscales indicated more difficulties reported by the adolescents.
Self-harming and suicidal behavior
Three questions relating to self-harming and suicidal behaviours with yes/no responses were utilised in the study [33]. Participants responded whether they have ever hurt themselves deliberately (e.g., cutting), thought seriously about committing suicide and tried to commit suicide.
Help-seeking behaviour
Participants were asked if they have told someone about their bullying experience in order to receive help [33], and if so, who have they told: parent, siblings, friend, teacher, other adult at school (e.g., principal, counsellor) and mental health professional (e.g., psychiatrist, psychologist, social worker). They were allowed to indicate multiple individuals as their responses.
Data synthesis
Statistical Analysis
Analyses were performed using IBM SPSS Statistics, Version 20. Frequencies of cyberbullying status (i.e., non-involved, cybervictim, cyberbully and cyberbully-victim) were first explored to determine the prevalence rates of cyberbullying in Singapore. Pearson chi-square tests for independence and one-way ANOVA were conducted to investigate if there are any gender, ethnicity and age differences across cyberbullying groups. To further examine help-seeking behaviours, only cybervictimised participants (i.e., cybervictim and cyberbully-victim) were included. Patterns of help-seeking behaviours were explored with chi-square tests for independence to examine gender differences, if any.
Mean differences of the two SDQ subscales across cyberbullying groups were examined using one-way ANOVA tests with Bonferroni tests as post-hoc pairwise comparison. Furthermore, multinomial logistics regressions were conducted individually for self-harming and suicide-related behaviours to examine their associations among cyberbullying groups. Odds ratios (ORs) and 95% confidence intervals (CIs) were presented. Non-involved group was used as reference and all analyses were adjusted for gender, age and ethnicity. Given the large sample size and multiple comparisons, a more conservative p value (p < .001) was used in this study.