Design
To investigate the research questions, we conducted a secondary data analysis using the Millennium Cohort Study (MCS), a nationally representative birth cohort of about 18,818 participants born in the years 2000-2002 across the nations in the UK [29]. Participants and their parents in the MCS provided informed consent before the data collection commenced and the current study received ethical approval for secondary data analysis. We extracted the data from the MCS when participants were at age 11, 14, and 17. The predictor was participants’ experience of age-11 victimisation, mediator was age-14 sleep quality and outcomes were age-17 self-harm and depressed mood (Sweep 7) (see measures for details).
Participants
In the MCS, 13,469 adolescents participated at age 11, with 11,872 at age 14, and 10,757 at age 17. Participants were included in this study if they have participated in these three time-points, and completed the measures on victimisation and sexual identity (Supplementary Figure 1). The final sample included 1922 SMA and 6900 non-SMA and their descriptive information was described in Table 1.
Measures
Sexual Identity
Participants were asked to select one of the following options to best describe how they currently think of themselves, “Completely heterosexual/straight; Mainly heterosexual/straight; Bisexual; Mainly gay or lesbian; Completely gay or lesbian; Other; Do not know; not applicable; Prefer not to say”. Similar to other cohort studies, e.g. [17, 30, 31], we coded those describing themselves as “Completely heterosexual/straight” as Non-SMA, while those selecting “Not applicable or Prefer not to say” were excluded from the analysis with the rest, who were not exclusively heterosexual as SMA.
Victimisation
At age 11, the adolescent participants reported the frequency of being hurt or picked on by other children from a scale of 1-6 where a smaller number representing more frequent experiences of victimisation. Participant’s parents were also asked if the participant has been picked on or bullied by other children and they responded “not true”, “somewhat true”, or “certainly true”. For both items, there was an option of “Don’t know/Don’t wish to answer/No answer” and those selecting these responses were not included for analyses in this study. Given our primary interests on SMA’s self-reported depressed mood and self-harm behaviours, we used participants’ self-reported victimisation as the predictor in the analyses and reported descriptive data on self-reported and parent-reported victimisation. (see Results). We also included the data of victimisation at age 14 for comparison analyses with non-SMA. At age 14, participants self-reported whether they had been exposed to a range of victimisation experiences (See Table 2). We coded victimisation at both ages as binary variables (Yes/No), where participants endorsing at least one victimisation experience were coded as having been victimised.
Self-harm
At age 17, the self-harm was measured based on the Edinburgh Study of Youth Transitions [32], where participants reported if they had self-harmed by different means (see Table 3 for details) in the past year. Participants endorsing any self-harm behaviour were coded as having performed self-harm. At age 14, participants were asked if they had hurt themselves on purpose in any way and they provided a response of yes or no.
Depressed mood
At age 17, depressed mood was measured by the following item in the Kessler 6 scale [33], “During the last 30 days, how often did you feel so depressed that nothing could cheer you up?” At age 14, depressed mood was measured by the short-version of the Mood and Feeling Questionnaire [34], where participants responded to 13 statements regarding their feeling on a likert scale (α = .94). A higher score indicated higher level of depressive symptoms in both time-points.
Sleep Quality
Sleep quality was measured by participants’ sleep onset latency (SOL), and nocturnal awakening at age 14, where they reported how long they usually took to fall asleep and how frequency they awakened during sleep at the last 4 weeks (See Table 1 for details). Participants were also asked about their typical bedtime and wake time during school days for
descriptive information regarding sleep patterns (Table 1).
Demographic variables
Participants’ age, sex, ethnicity and weekly family income were measured at age 14 (Table 1).
Data analysis plan
SPSS version 28 was used for all data analysis. Group differences were analysed using Chi-square or Mann-Whitney U test, given unequal variance. Correlational analyses were conducted to assess the cross-sectional relationship among victimisation, sleep quality, depressed mood and self-harm at age 14. Binary logistic and linear regression were used to analyse the temporal relationship between victimisation, poor sleep quality with self-harm and depressed mood. The variables were tested for multi-collinearity and their variance inflation factor were lower than 10, indicating no significant concern of multi-collinearity. The regression analyses were conducted by entering demographic variables and baseline (age-14) of the outcome measure at step 1, victimisation at step 2 and SOL and nocturnal awakening at step 3. In view of multiple comparison and risk of inflated type-1 error, all p-values were adjusted following the Benjamini-Hochberg procedure with a false discovery rate (FDR) of 5%. Statistical significance was determined by an adjusted p-value, pfdr <.05. For the mediation hypotheses, the PROCESS macro was used with a bootstrap re-sampling of 5000. Age-11 victimisation was the predictor, either age-14 SOL or nocturnal awakening as the mediator, and either age-17 self-harm or depressed mood as the outcome, with the corresponding measure at age 14 as the covariate. Significant mediation effect was inferred by a 95% confidence interval (95%CI) not containing 0.