Aim
The primary aim of the present study was to investigate whether exposure to bullying at ages 15 or 18 was associated with the development of depressive symptoms at age 28.
The secondary aim was to investigate whether exposure to bullying at both ages 15 and 18 increased the risk of developing depressive symptoms at age 28.
Design and population
This is a longitudinal study using questionnaire data gathered as part of the ongoing West Jutland Cohort Study (VestLiv), which aims to investigate aspects of inequality in health and social differences in welfare from a lifelong perspective [23, 24]. Individuals born in 1989 who were living in the county of Ringkjoebing (West Denmark) in April 2004 (age 15) were invited to participate (N = 3681). Contact information for this complete regional cohort of young people was retrieved from the Central Office of Civil Registration and from public schools in the county of Ringkjoebing. Of the original source population, 3054 (83%) filled out the initial questionnaire at age 15 in 2004.
Information from two follow-up surveys were used at ages 18 and 28 with response rates of 65% (n = 2400) and 57% (n = 2102) respectively. Register information about the respondents was derived from national registers in Statistics Denmark using the personal identification number from the Central Office of Civil Registration (CPR number), which is given to every inhabitant in Denmark at birth (or upon entry for immigrants) [25]. To obtain information about parental educational level, gender and family type (split home), the respondents were also linked to their parents or guardians using the CPR number [25].
The study population consists of participants who provided information about depressive symptoms at age 28 and bullying at age 15 and/or age 18 (n = 1790). The response rates of the participants in years 2004, 2007 and 2017 are presented in Fig. 1.
Outcome
The outcome of the study was depressive symptoms, which was based on the participants' self-reporting at age 28. We included four items from The Center for Epidemiological Studies Depression Scale (CES-D). This scale is an abbreviated validated version of the original scale [26, 27] and is designed to measure the current level of depressive symptoms in a general population, with emphasis on the affective component “depressed state”. The CES-D scale was developed for use in epidemiological studies [27] and has been translated into several languages and validated for both young people and adults [26, 28]. Participants were asked the following 4 questions to assess their level of depressive symptoms: During the past week, how often have you had the following feelings?": (1) "I was happy", (2) "People were unfriendly", (3) "I felt sad", (4) "I could not get going" with the following response options: (1) "Not at all", (2) "A little", (3) "Some", (4) "A lot". The responses were subsequently awarded scores of 0–3 and generated into a sum-score ranging from 0–12, with high values corresponding to having more depressive symptoms. The scales were then dichotomized at 3 points and above into few depressive symptoms and more depressive symptoms, as suggested by Fendrich et al. [26], who found this cut-point relevant in relation to the prediction of major depressive disorders in the general population.
Exposures
Questionnaire information about bullying was obtained by the participants' self-reporting at age 15 and 18. At age 15, the participants were asked: "How much have you been bullied at school during the last six months?". At age 18, the participants were asked about bullying in a slightly different way: "How much have you been bullied in an unpleasant way at school during the last six months?". At both ages, the response options were: (1) "Never", (2) “Once or twice”, (3) “A few times”, (4) “Once a week”, (5) "Several times a week". These options were combined into the following three categories: "Not bullied" if they had answered (1), "Bullied" if they had answered (2) or (3) and "Often bullied" if they had answered (4) or (5), as suggested by Andersen et al. [29].
Potential confounders
Information about potential confounders was obtained from the participants' questionnaire responses at age 15 and 18 and from register information.
Information about the highest level of education in the household in 2003 was obtained from the educational registers [30] and divided into the following 4 categories: (1) ≤ 10 years (primary school), (2) 10–13 years (secondary school), (3) 13–15 years (short/middle tertiary education), and (4) > 15 years of school (long tertiary education). If the participants’ parents were divorced, information was taken from the household at which the participants had their postal address.
Information about close friends was measured as a question, at age 15, about whether the participants had a friend in whom they could confide (yes vs. no).
Information about family functioning was based on the participants' responses to questions regarding the general functioning of the family at age 15. The General Functioning Scale consists of 12 items that assess the overall health or pathology of the family and is one of seven scales from the McMaster Family Assessment Device (FAD) [31]. In this study, the variable was dichotomized at a cut-off of ≥ 2, corresponding to the 75%-percentile, with high scores indicating a problematic family function [29, 31].
Information about split home and gender was collected from registers [25]. The variable split home was dichotomized into whether the participant lived with one or both parents.
Statistical methods
The characteristics of the study population were presented by gender, and the distributions of the categorical variables were presented by number and proportion.
The distributions of being bullied at age 15 and 18 in relation to depressive symptoms at age 28 were presented by number (n) and proportion (%). Because only a limited number of participants reported that they were "often bullied" at age 18, the categories "Bullied" and "Often bullied" were collapsed into one category called "Bullied" in regard to all analyses using information about bullying at age 18.
The associations between being bullied at age 15 or 18 and the development of depressive symptoms at age 28 were analyzed using multiple logistic regression. Firstly, crude estimates between each exposure variable and outcome were performed. Secondly, estimates adjusted for parental educational level, close friends, family functioning, gender and split home were calculated. Thus, crude and adjusted odds ratios (ORs) were estimated with 95% confidence intervals (95% CI). As supplementary analyses, the associations between being bullied at age 15 or age 18 and depressive symptoms at age 28 was adjusted for depressive symptoms at age 15.
Finally, the association between being bullied both at age 15 and age 18 and developing depressive symptoms at age 28 was investigated by constructing the following three categories:
-
"Not bullied", if not bullied at age 15 or 18
-
"Bullied at one age point", if Bullied or Often bullied at either age 15 or 18,
-
"Bullied at two age points", if Bullied or Often bullied at both age 15 and 18
We furthermore carried out a sensitivity analysis using a multiple imputation chained model with 100 imputations. Seven chains, either logit or ologit dependent on the categorization of the variables, were constructed in order to impute missing data on the following variables: depressive symptoms at age 28, bullying at age 15 or 18, family functioning and close friends at age 15, split home and household educational level in year 2003. Information on gender was complete. Besides the above mentioned variables some additional were included: (1) depressive symptoms at age 15 and 18 were included when imputing depressive symptoms at age 28, (2) being bullied at age 28 was included when imputing bullying at age 15 and 18, respectively, The final estimates were found as the average of the m sets of estimates and the standard errors by applying a simple formula called Rubin’s rule (results not shown) [32]. All statistical analyses were carried out in STATA statistical package (V.15.0; State, College Station, TX).
Ethical considerations
This study is in accordance with the 1975 Declaration of Helsinki [33]. The study is approved by the Danish Data Protection Agency. According to Danish law at the time point of data collection, questionnaire and register-based studies did not need written informed consent nor approval by ethical or scientific committees [34].