Sample and procedure
The TRacking Adolescents’ Individual Lives Survey (TRAILS) is a prospective cohort study of Dutch early adolescents aged 10–12 years, who are followed biennially. The present study used data from the first (2001–2002) and second (2003–2004) assessment wave. The TRAILS target sample consisted of young adolescents from five municipalities in the North of the Netherlands, including both urban and rural areas. More details about the sample selection have been published elsewhere (de Winter et al.,
2005).
Of all subjects who were approached at wave 1 (
N = 3,145), 6.7% were excluded. The exclusion criteria were (1) adolescent incapable to participate because of mental retardation or a serious physical illness or handicap and (2) Dutch-speaking parent or parent surrogate not available, and not feasible to administer a part of the measurements in parent’s own language. Of the remaining 2,935 young adolescents, 24% did not want to cooperate, and 76.0% cooperated with the study at wave 1 (
N = 2,230, mean age 11.09 years,
SD .55, with 50.8% girls). Most frequent reasons for non-response were ‘not interested’ (33.8%), participation in other research or unfavorable experiences with research (15.4%), too much of a burden on the child (12.2%), lack of time (10.3%), concerns about privacy and confidentiality (8.0%), and the child’s refusal to participate because friend(s) did not participate (4.0%). In 34 cases (1.2%) we failed to contact anyone of the household (de Winter et al.,
2005). Responders and non-responders did not differ with respect to the proportion of single parent families, or the prevalence of teacher-rated problem behavior. Furthermore, no differences between responders and non-responders were found regarding associations between socio-demographic variables and mental health outcomes (de Winter et al.,
2005). To assess anxiety symptoms, the Revised Child Anxiety and Depression Scale (RCADS) (Chorpita & Daleiden,
2000) was used at wave 1, and also at wave 2. For 20 cases, RCADS data were not obtained at wave 1 because respondents were not present during the measurements that were conducted in the classrooms, and could not be reached afterwards. Hence, RCADS data of 2,210 pre-adolescents were available at wave 1.
At the second assessment wave, following similar procedures as at wave 1, RCADS information was obtained from 2,067 individuals. This was 95.5% of those for whom wave 1 RCADS information had been collected (51.4% girls). To examine possible selective attrition, a stepwise logistic regression analysis was performed with ‘wave 2 RCADS information available’ as a dependent variable, and wave 1 age, sex, and the wave 1 RCADS Total Anxiety score (that was constituted by summing scores on the five anxiety dimensions that were assessed with the RCADS in the present study, see below) as possible predictors. The RCADS Total Anxiety score and sex did not predict attrition. However, younger age predicted attrition significantly (odds ratio = .17, Wald = 93.1,
p < .001; Model chi-square = 109,551, df = 1,
p < .001). Cox and Snell
R-square of the regression model was .048, which indicated that the effect of age was small. Further, most importantly, the level of anxiety at the initial assessment did not influence cooperation at wave 2.
SAD
|
SoPh
|
Fears being alone at home | Worried when does poorly at things |
Scared to sleep alone | Worried when somebody angry |
Scared to sleep away from home | Worried will do badly at school |
Fears being away from parents | Worried about mistakes |
Worried in bed at night | Worried what others think |
Trouble going to school | Scared to take a test |
Afraid of being in crowded places | Worried might look foolish |
| Afraid to talk in front of class |
| Afraid to look foolish in front of people |
GAD
|
PD
|
Worried something awful will happen to family | Suddenly trouble breathing without reason |
Worried bad things will happen to self | When has a problem, feels shaky |
Worried something bad will happen to self | Suddenly trembling, shaking without reason |
Thinks about death | Suddenly dizzy, faint without reason |
Worried about things | When has a problem, stomach feels funny |
Worried about what will happen | When has a problem, heart beats really fast |
| Suddenly feeling scared without reason |
| Suddenly heart beats too fast without reason |
| Worried suddenly get scared without reason |
OCD
| |
Can’t get silly/bad thoughts out of head | |
Keeps checking | |
Has to think thoughts to stop bad events | |
Has to do things over and over again | |
Has to do things just right to stop bad events | |
Bothered by bad/silly thoughts or images | |
Measures
The Revised Child Anxiety and Depression Scale (
RCADS) (Chorpita & Daleiden,
2000) is a revision of the Spence Children’s Anxiety Scale (SCAS) (Spence,
1997). It is a self-report questionnaire with 47 items, that are scored on a 4-point scale (0 = never, 1 = sometimes, 2 = often, 3 = always). The questionnaire covers six scales, corresponding with DSM-IV dimensions of anxiety disorders and depressive disorder. The following five scales were used for the present study: separation anxiety disorder (SAD), generalized anxiety disorder (GAD), social phobia (SoPh), panic disorder (PD), and obsessive compulsive disorder (OCD) (see Table
1). The scale major depressive disorder (MDD) was not used. The internal consistencies of the scales that were used were (respectively at wave 1/wave 2) .66/.59 for SAD, .80/.72 for GAD, .78/.88 for SoPh, .75/.72 for PD, and .68/.66 for OCD. The factor structure—for all six scales together—that was originally based on data from 1,641 children and adolescents from a community sample from Hawaii (Chorpita & Daleiden,
2000), was confirmed by confirmatory factor analysis in the TRAILS sample at wave 1 (fit indices of NNFI = .96, RMSEA = .05, and SRMR = .05, indicating an adequate fit to the sample data) (Ferdinand et al.,
2006). The association of RCADS dimensions of anxiety with corresponding DSM-IV anxiety disorders was supported by previous research (Nauta et al.,
2004).
Statistical analyses
First, to obtain information regarding comorbidity between different types of anxiety problems in the study sample, correlations among wave 1 RCADS scale scores were computed for each sex. Then, Pearson correlations were computed between wave 1 and wave 2 RCADS scale scores, separately for each sex group. Correlations provide insight in the associations between measures. However, by just computing correlations, it can not be judged if continuity is homotypic or heterotypic. For instance, the magnitude of a correlation between wave 1 SAD and wave 2 SoPh scores depends on the correlation between wave 1 SAD and wave 1 SoPh scores. The higher correlations between wave 1 SAD and wave 1 SoPh are, the higher the correlation between wave 1 SAD and wave 2 SoPh will be. In other words, if assessment of continuity would solely be based on correlations, comorbidity at wave 1 would artificially inflate estimations of the extent of heterotypic continuity between wave 1 and wave 2.
To correct for the effects of wave 1 comorbidity rates, regression analyses were conducted. First, it was assessed which part of continuity in anxiety problems was typically homotypic. For this purpose, for scores on each of the five RCADS scales at wave 2, a set of regression analyses was conducted, with wave 2 RCADS SAD, GAD, SoPh, PD, and OCD scores as dependent variables. These analyses were conducted to investigate how much of the variance in a specific RCADS scale score at wave 2 was not accounted for by an overall elevation in different types of anxiety at wave 1, but instead, specifically by its own counterpart at wave 1. We will now describe the regression analyses that were conducted for wave 2 SAD. Those for GAD, SoPh, PD, and OCD were similar. In the first block of the analyses, wave 1 scores on GAD, SoPh, PD, and OCD were entered simultaneously as predictors. Then, in a second block, wave 1 scores on the SAD scales were added, to see how much of the variance in wave 2 scores was predicted specifically by wave 1 SAD scores, and not by scores on the other RCADS scales at wave 1. This variance reflects specific homotypic continuity. In the third block, sex was added. In the fourth block, an interaction between sex and SAD was added. If this interaction was significant, analyses were conducted for girls and boys separately. For each next block, the variance that was accounted for by the variable in this block was computed (R
2).
Second, it was assessed which part of continuity in anxiety problems was specifically heterotypic. For this purpose, for scores on each of the five RCADS scales at wave 2, a set of regression analyses was conducted, with wave 2 RCADS SAD, GAD, SoPh, PD, and OCD scores as dependent variables. These analyses were conducted to investigate how much of the variance in a specific RCADS scale score at wave 2 was not accounted for by its own counterpart at wave 1, but instead, by the other wave 1 anxiety scale scores. We will now describe the regression analyses that were conducted for wave 2 SAD. Those for GAD, SoPh, PD, and OCD were similar. In the first block of the analyses, wave 1 SAD scores were entered as predictor. Then, in the second block, scores on wave 1 GAD, SoPh, PD, and OCD scales were added, to see how much of the variance in wave 2 scores was specifically predicted by other RCADS scales at wave 1. This variance reflects specific heterotypic continuity. In the third block, sex was added.
To judge the magnitude of effects, Cohen’s rules for effects sizes can be used (Cohen,
1988). According to Cohen, R
2 between 1.0% and 5.9% is small, between 5.9% to 13.8% medium, and above 13.8% large.