Background
Anxiety disorders frequently occur in children, with lifetime prevalence of 15 to 20 % [
1], and although depression is less prevalent in this age group (0.4 to 2.5 %), its prevalence rises quickly during adolescence [
2]. Anxiety and depression are closely related. Children with an anxiety disorder are at increased risk for developing a depressive disorder [
1,
3], while childhood-onset depressive disorder increases the risk for anxiety [
3]. In addition to this, anxiety disorders in adolescence are associated with substance abuse/dependence and academic underachievement [
1]. Moreover, childhood and adolescent depression are associated with poor outcomes in later life, including suicidal behavior, substance abuse, increased risk for other psychiatric disorders (e.g., bipolar disorder and personality disorders), and psychosocial, academic and work-related problems [
3].
Despite the fact that anxiety and depression significantly hamper children’s current and future well-being, only a small percentage of children with these problems receive mental health care [
4‐
6]. Children are dependent on parents or teachers if professional help is needed [
6,
7]. Unfortunately, identifying children in need of treatment for anxiety or depression is difficult for others without information from the children themselves. Other informants may notice behavior that suggests emotional problems, such as crying or sad posture, but may be not well-informed on what children actually think or feel. This most likely explains the low correlations between children’s self-reported and teacher- or parent-reported internalizing problems found in an extensive meta-analysis [
8]. Comparable results were reported in a Dutch general population sample of 10–11 year-old children [
9]. When parent- and teacher-report of emotional problem behavior was examined in detail, it was found that parents and teachers did not recognize several of the symptoms reported by children. In practice, mental health professionals underline the usefulness of children’s report [
10]. An instrument that can identify children in need of help, based on children’s own reporting, that can be used next to proxy-informants is therefore of utmost importance.
An instrument that may be useful for obtaining reliable information about anxiety and depression in youths is the Revised Child Anxiety and Depression Scale (RCADS). The RCADS is a self-report questionnaire with scales corresponding to the
Diagnostic and Statistical Manual of mental disorders (DSM-IV, [
11]) diagnostic criteria for anxiety and depressive disorders [
12]. It has been shown to be a reliable and valid instrument in general population- and school-based samples in Australia, the Netherlands, Denmark and the United States [
12‐
17], and in clinical- and school-based samples in Hawaii, USA [
12,
14,
18]. The RCADS measures anxiety and depression symptoms separately, and is, in addition, the only self-report questionnaire for youth that measures symptoms of five different types of anxiety disorders. This is important because of the close relationship between these disorders. In contrast to other instruments, such as the Revised Children’s Manifest Anxiety Scale 2 [
19] or the Children’s Depression Inventory 2 [
20], the RCADS is freely available in various languages [
21]. Due to budget constraints in both health care and research, free-of-charge questionnaires could enhance their use and therefore increase the possibility of identifying children with problems, as well as promoting research in the field of childhood anxiety and depression.
In this study we examined the factor structure, internal consistency, short-term stability, construct validity of the RCADS in school-aged children. Our study was conducted in a multi-ethnic urban sample of Dutch children, whereas samples in previous studies mainly consisted of Caucasian children from the more rural parts of the Netherlands [
16,
22].
Construct validity, which refers to the extent to which the RCADS correlates with other similar constructs, was assessed in five ways [
23]. Our first hypothesis was that the RCADS anxiety and depression scales would correlate positively, since anxiety and depression are closely related. A correlation around
r = 0.7 was expected (e.g., [
24]). The second hypothesis was that there would be moderate agreement between children scoring in the 90
th percentile of the RCADS scales and those scoring in the 90
th percentile of teacher-reported anxiety and depression. Children scoring in the 90
th percentile represent the most anxious or depressed children in the sample. Although research has shown that teachers may not identify every child with elevated anxiety and depression symptoms [
8,
9], we expected that they are able to recognize the most anxious or depressed children. Mesman and Koot [
9] found correlations of 0.30 for anxiety and depression between teacher and child report in a general population sample. As we only compared children with scores above the 90
th percentiles, we expected stronger agreement between child and teacher report. As a third aspect of construct validity, we investigated gender differences in RCADS scores. Female gender constitutes a risk factor for anxiety and depression. For depression however, gender differences start to occur in adolescence [
1,
2]. Therefore, we expected that the girls in our sample would have significantly higher levels of anxiety symptoms, but because of the age of our sample (pre-adolescence), we expected smaller gender differences in depression than in anxiety. The fourth hypothesis was that children willing to participate in a prevention program addressing anxiety and depression would have significantly higher RCADS scores than children not willing to participate. Fifth, and finally, we examined sensitivity to change. An instrument for assessing emotional problems is regarded more useful if it not only indicates individual or group differences in symptom level or severity, but also is sensitive to changes in symptom levels (e.g. due to targeted interventions). Sensitivity to change has however not been examined in several other RCADS related papers. Mathyssek et al. [
25] established in a general population adolescent sample longitudinal measurement invariance, which implies that changes in anxiety scores over time most likely reflect true changes. However, this study did not investigate the major depressive disorder (MDD) scale. Therefore, in this study we established sensitivity to change of the instrument using data from children participating in an indicated preventive intervention program, including the MDD scale.
Further, we investigated age differences in RCADS scores. As anxiety and depression increase in adolescence, we expected more symptoms of anxiety and depression in older children. These hypotheses are tested in the largest Dutch sample to date.
Discussion
In the present study, the structure, reliability and validity of the RCADS were investigated in a large urban, multi-ethnic sample of Dutch school-aged children.
First, we investigated whether the original factor structure of the RCADS (Chorpita et al., 2000) could be replicated in our sample. The results were not univocal. In general, factor loadings were good and comparable to previous research [
17]. However, whereas one fit index (i.e. the RMSEA) indicated a close fit, the TLI was slightly below the cut-off value of a good fit. The present study is not the first that found lower values for one or more of these fit indices [
13,
16,
17]. The close relation between the RCADS subscales may make it difficult to establish a clear factor structure. Previous research has found high comorbidity rates between different types of anxiety (e.g., [
34]). Further, Ferdinand et al. [
22] found no distinction between GAD, SP, SAD and PD in a general population sample of pre-adolescent children. For depression, which has a low prevalence in the age category of our sample, the MDD scale may reflect anxiety symptoms rather than symptoms of depression. The strong correlation between the anxiety and MDD scale in the present study as well as in previous (e.g., [
35]) research also points in this direction.
The internal consistency was good for all RCADS scales. Cronbach’s alphas were comparable with other samples [
12,
13,
17], and were comparable between ethnic groups, indicating reliability in these groups. The ICCs indicated good stability over three months and were fairly comparable to the ICCs in the study of Muris et al. [
16] over a four-week period.
The RCADS scores of children who were willing to participate in a prevention program for anxiety and depression were significantly higher than the scores of children who did not want to participate in such a program. This may indicate that the RCADS is capable of identifying children who feel the need to participate in such a program.
In contrast to our expectations, we found low agreement between child- and teacher-reported 90
th percentile scores. Apparently, even children with very high self-reported anxiety or depression scores are not easily identified by their teacher. As other measures in the present study provide confirmation for the validity of the RCADS (CFA, reliability, remaining validity measures including sensitivity to change), these results seem to indicate relative insensitivity of teacher reports rather than low validity of the RCADS child reports (cf. [
9]). An important message from our findings as well as from previous studies is that, when screening for childhood anxiety and depression in the school context, child reports are essential to include next to reports from teachers [
10].
Gender differences between RCADS subscales were as expected. Girls reported more anxiety and depression symptoms than boys. Again, these findings are indicative of the validity of the RCADS.
Sensitivity to change analyses showed that the RCADS detects change in anxiety and depression symptoms in children who participated in a preventive intervention up to 12 months after participation. The biggest changes were detected between the pre-intervention measurement and the last one 1.3 years later. Changes between consecutive measurements became smaller over time. These results are in line with the expected course of symptom decrease after participation in an intervention. Therefore, the RCADS is suitable for screening purposes as well as for evaluating change.
In general, scores of anxiety and depression were lower with increasing age. Three studies report lower RCADS scores with increasing age in childhood as well [
16,
17,
36]. The latter two studies – which included a broader age range – reported a decrease of symptoms until middle or late adolescence, after which mean scores increased. The age category of our sample – childhood/early adolescence – was probably too young to detect a decrease in symptoms.
Most studies on the RCADS reported means and standard deviations. However, in our study the RCADS scores were not normally distributed but skewed to the right, with most children scoring low, as was to be expected in a general population sample. Our results show that, by using means, levels of anxiety and depression in children are being overestimated. Therefore, medians are the preferred descriptives. To enhance comparison between studies, we also reported the mean values.
Strengths and limitations
The present study is the first to investigate the structure, reliability, stability and validity of the RCADS in a large and ethnically diverse sample of children from the general Dutch population. Although internal consistency of the RCADS subscales was comparable between the various ethnic groups, the applicability of the RCADS in different ethnic groups should be studied in more detail. For instance, multi-group confirmatory factor analyses can be used to investigate whether RCADS items are answered comparably by various ethnic groups (measurement invariance; [
17]). However, the extensive report of measurement invariance is beyond the scope of the present study.
One of our methods to investigate the validity of the RCADS was to compare teacher and child reports of the presence of symptoms of childhood anxiety and depression. Ideally, the validity of the RCADS should also have been studied by comparing the RCADS scores with another child report instrument. However, because of time constraints at schools, we could not administer more questionnaires to the children.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
MK was responsible for the data collection and analysis and generated the first draft of this manuscript. MC, MZ, and HK were the principal investigators of this study and supervised the data analysis and interpretation. All authors contributed to the writing and revision of the manuscript. All authors read and approved the final manuscript.