Participants and procedures
The present study is part of a controlled trial in which the effectiveness of an indicated school-based prevention program for childhood anxiety and depression is being evaluated [26].
All 265 primary schools in the Amsterdam area, the Netherlands, were asked to participate in the trial, 45 (17 %) of which expressed willingness to participate. The main reasons for declining participation were time constraints, other priorities or participation in other studies. Children and parents in participating schools received an information letter with a passive informed consent form. If parents or children did not wish to participate, they were free to withdraw. Children and teachers completed questionnaires (see Measures) in the classroom during school time. Researchers or research assistants explained how to complete the questionnaires and were available for additional clarification during completion of the questionnaire. Three months later, the RCADS was completed again by a subsample of children (children in control schools, see [26]) to assess its short-term stability. In addition, RCADS data were obtained from the intervention sample at baseline (T1), immediately after the intervention 10 weeks later (T2), and at 6- (T3) and 12-month (T4) follow-up. Children with elevated RCADS scores at T1 in intervention schools were invited to participate in an intervention targeting anxiety and depression. In control schools, parents of children with elevated anxiety and depression symptoms were notified after the trial (see [26]). The Medical Ethics Committee of the VU University Medical Center in Amsterdam, the Netherlands, approved the study protocol.
In total, 3890 children from grades 4, 5 and 6 of 45 primary schools were invited to participate in the trial. Parents of 3775 children consented to participate. No questionnaires were available for 139 children because they had left school, were ill during data collection or due to unknown reasons. The remaining sample of 3636 children (93 %) consisted of 1733 boys and 1898 girls (5 unknown), aged 8–13 years (M = 10.6, SD = 0.9) (21 unknown). Age was divided into four categories: 8/9 (10 %, n = 360), 10 (35 %, n = 1267), 11 (40 %, n = 1447) and 12/13 (15 %, n = 541). Ethnicity was based on the mother’s country of birth, or, if the mother was born in the Netherlands, the father’s country of birth [27]. Children were from diverse ethnic backgrounds: Dutch (40 %, n = 1438), Turkish (8 %, n = 302), Moroccan (15 %, n = 550), Surinamese and Antillean (11 %, n = 395), other Western (10 %, n = 354), other non-Western (13 %, n = 472), or unknown (3 %, n = 125). The Western group consisted of children from Europe (excluding Turkey), North America, Oceania, Japan and Indonesia (including former Dutch East Indies). The non-Western group consisted of children from Africa, Latin America and Asia (without Japan and Indonesia) [27]. These percentages are comparable to the distribution of ethnic groups among Amsterdam primary school children (37 %, 8 %, 18 %, 11 %, 11 %, 15 %, and 2 % in the 2011/2012 school year) [28]. The short-term stability subsample consisted of 1019 children (54 % girls), and the subsample in which self-report as well as teacher questionnaires were available consisted of 841 children (55 % girls).
Measures
Children’s self-reported symptoms of anxiety and depression were assessed by the Revised Child Anxiety and Depression Scale (RCADS) [12]. The RCADS assesses symptoms of anxiety and depression through child self-report. It consists of 47 items, corresponding to childhood and adolescent anxiety and depressive disorders as defined by the DSM-IV. Items are summed into six subscale scores, i.e. generalized anxiety disorder (GAD), social phobia (SP), separation anxiety disorder (SAD), panic disorder (PD), obsessive compulsive disorder (OCD), and major depressive disorder (MDD). A total anxiety score consisting of all GAD, SP, SAD, PD, and OCD symptom scores can also be computed. Examples of items are: “I worry when I think I have done poorly at something”, “I have trouble sleeping” and “I feel scared if I have to sleep on my own”. Children rate how often each item applies to them on a 4-point Likert scale, ranging from 0 (never) to 3 (always).
Socio-demographic information was assessed by means of a questionnaire. Children were asked to fill out their own date and country of birth, and the country of birth of their parents. Children were also asked to indicate whether they wanted to participate in a prevention program addressing anxiety and depression.
Teacher reports of anxiety and depression symptoms were assessed by means of the Problem Behavior at School Interview (PBSI) [29]. The PBSI is a 42-item instrument that measures internalizing and externalizing problems in children as perceived by teachers. In the present study only the 12-item internalizing scale, consisting of a 5-item anxiety scale and a 7-item depression scale, was presented as a questionnaire, as has been done in previous research [30]. Teachers rated children on a 5-point Likert scale, ranging from 0 (never) to 4 (often). Examples of items are: “This child is nervous or tense”, “This child has a lack of energy” and “This child is unhappy or depressed”. In previous research, Cronbach’s alphas ranged from 0.79 to 0.81 for the anxiety scale and from 0.78 to 0.83 for the depression scale [31]. In the present study, Cronbach’s alphas were 0.90 for the overall PBSI internalizing scale, 0.83 for the anxiety scale, and 0.86 for the depression scale.
Analyses
When one item was missing on a particular RCADS subscale, the item median of the total sample was imputed. Medians were chosen, as data were skewed to the right. Per RCADS subscale, approximately 100 of the 3636 children had one missing item.
To test whether the six subscales as defined by Chorpita et al. [11] could be replicated in our sample, a confirmatory factor analysis was performed using Amos version 19 (James L. Arbuckle, 2010). A maximum likelihood estimation procedure was used. Model fit was evaluated by means of two indices. The Root Mean Square Error of Approximation (RMSEA) represents the average difference between correlations observed among variables and those expected on the basis of model assumptions. It also takes into account model parsimony. A value < 0.05 is considered to indicate a close fit and a value ≤ 0.08 indicates an acceptable fit (see [32]). The Tucker-Lewis coefficient (TLI; Tucker & Lewis, 1973) indicates the overall fit of the proposed model relative to a null model, adjusting for model complexity. A value > 0.90 indicates an acceptable fit and a value > 0.95 indicates a close fit of the model [33].
All other analyses were carried out with SPSS Statistics version 19 (IBM SPSS Statistics, 2010). Cronbach’s alphas were used to assess internal consistency of the RCADS. Alphas were computed for the total group as well as for ethnic subgroups. Short-term stability was assessed by calculating intraclass correlation coefficients (ICCs) in a subsample of 1019 children who did not receive the intervention (children in control schools, see [26]). To test the sensitivity to change of the RCADS, change in scores between baseline and three follow-ups were tested in 339 children who participated in an intervention study of an indicated prevention program targeting anxiety and depression (see [26]), by means of paired sample t-tests, followed by calculation of Cohen’s d.
Since all RCADS scales were skewed to the right, gender differences are reported in medians and interquartile ranges. However, to enable comparisons with previous studies, we also report means and standard deviations. Correlations were calculated as Spearman’s rho. Agreement between 90th percentiles was calculated as kappa. Differences between scores for help-seeking, wanting to participate, gender and age were analyzed with linear regression analyses with the anxiety or depression scores as dependent variable. Linear regression analyses were chosen since they show the exact differences in scale scores between groups. The data allowed us to conduct these analyses, as residual plots revealed no large deviation from normally distributed residuals.