Making high‐quality measures available in diverse contexts—The psychometric properties of the Revised Child Anxiety and Depression Scale in a Norwegian sample

Abstract Objectives Recent initiatives have recommended the Revised Child Anxiety and Depression Scale (RCADS) for use in research and as patient‐reported outcome in health care globally. We aimed to investigate, for the first time, whether the psychometric properties of the anxiety and depression youth self‐report measures, RCADS‐47 and RCADS‐25, generalize to a Norwegian setting. Methods We examined gender and age differences in symptomatology among 592 children (mean age 10.7 years), and conducted a psychometric investigation of the internal reliability, structural validity, measurement invariance and convergent validity of the RCADS‐47 and RCADS‐25 youth versions. Results Girls reported higher levels of anxious and depressive symptoms than boys, but no age differences were observed. Reliability coefficients for the RCADS‐47 and RCADS‐25 scales indicated good internal consistency. Structural validity for RCADS‐47 and RCADS‐25 was supported by confirmatory factor analyses results. For both measures, strong gender‐based measurement invariance was present. Convergent validity of the RCADS‐47 and RCADS‐25 with other well‐established self‐report measures for anxiety (Multidimensional Anxiety Scale for Children) and depression (The Short Mood and Feelings Questionnaire) was supported. Conclusion The RCADS‐47 and RCADS‐25 youth versions are valid and reliable instruments for measuring symptoms of anxiety and depression in a Norwegian setting. The results add to the evidence supporting RCADS's cross‐cultural validity.


| INTRODUCTION
Anxiety and depression are common in children and adolescents (Polanczyk et al., 2015), frequently co-exist (Melton et al., 2016) and are associated with impairment in social, emotional, academic, and family functioning that can extend into adulthood if left untreated (Swan & Kendall, 2016). Early and targeted intervention is important to prevent further developments of problems (Giesen et al., 2007), but health systems have yet to find an adequate response to the burden of mental disorders. Despite substantial increases in the provision of treatment the last decades, the prevalence of mood disorders has not been reduced (Jorm et al., 2017), and the gap between the need for treatment and its provision is still large all over the world (World Health Organization, 2013).
There is a need for a more coordinated response to this problem from health, social and research sectors. One barrier is the large variation in measurement instruments used for mental health outcomes. Even when measuring the same constructs, results from different instruments are difficult to compare (Wahl et al., 2014). The competition between the many instruments cause data gaps, fragmentation, and inconsistencies, all of which hamper the potential to inform quality improvement efforts. An international standard set of health outcome measures may therefore be a necessary step to ensure that research leads to tangible improvements. Such standards have the potential to enhance and accelerate the understanding of mental health disorders and their treatments (Obbarius et al., 2017).
Recent initiatives therefore seek to attain global standardized mental health measurement (Krause et al., 2021). Some of the largest funders of mental health research worldwide, such as the National Institute of Mental Health (NIMH) and Wellcome Trust, endorse these initiatives, making funding obtainment for new research related to depression and anxiety in children and adolescents conditional on using two specific consensus-based measures (NIMH, 2020).
One of the two consensus-based measures recommended to use with children and adolescents worldwide is the Revised Child Anxiety and Depression Scale (RCADS; Chorpita et al., 2000). RCADS is available as a full 47-item version and as a 25-item short version. The full RCADS-47 provides an accurate indication of primary problems with its six subscales based on the Diagnostic and Statistical Manual of Mental Disorders, 4th edition (DSM-IV; American Psychiatric Association, 1994). The items can further be combined into aggregated anxiety, depression, or total internalizing scales. The RCADS 25-item short version (Ebesutani et al., 2012) includes 25 items from the full version, and can also be combined into anxiety, depression, and total internalizing scales. Although the full RCADS-47 version was initially chosen to be included in the set of consensus-based instruments, the shorter RCADS-25 version ended up as the final recommendation due to its brevity (Krause et al., 2021). In terms of feasibility, both versions of RCADS have several advantages over other measures.
They are freely available at no monetary cost, in contrast to many other instruments. The RCADS-25 is short. Although the RCADS-47 is longer, it is the briefest measure aligned with the DSM-IV (Southam-Gerow et al., 2008). Furthermore, both RCADS-47 and RCADS-25 assess anxious and depressive symptoms concurrently. This is useful given the high comorbidity of the two disorders and the increasing number of transdiagnostic interventions available (Ehrenreich & Chu, 2014;Hunsley & Mash, 2007). Previous studies in both community and clinical samples in several countries have reported good psychometric properties for RCADS-47 (e.g., Chorpita et al., 2005;Esbjørn et al., 2012;Mathyssek et al., 2013;Piqueras et al., 2017). Its discriminant validity is superior to some of the most common measures of anxiety and depression, such as the Revised Children's Manifest Anxiety Scale (RCMAS; Reynolds & Richmond, 1978) and Children's Depression Inventory (CDI; Kovacs, 1980) (see Chorpita et al., 2005). Although less widely validated than RCADS-47, good psychometric properties have thus far also been reported for RCADS-25 (Ebesutani et al., 2012).
To build confidence in the comparison of research findings, measures should be validated prior to usage in new cultural contexts.
It is especially important that translated versions that have undergone cross-cultural adaptation demonstrate their psychometric properties, as subtle changes could alter the instrument's psychometric properties (Chang et al., 1999). The  have demonstrated promising psychometric abilities in various countries in Europe, Africa, Asia, and the Americas (Piqueras et al., 2017). Western European data from the measures exist and a recent systematic literature review (Eidet et al., 2022) found that, in Scandinavia, there are existing Danish and Swedish peer-reviewed studies which have reported favorable psychometric properties of RCADS-47 (e.g. Esbjørn et al., 2012). To the best of our knowledge, however, no psychometric properties of the RCADS-25 have been reported in Scandinavia yet. And thus far, there is no documentation of the psychometric properties of the Norwegian RCADS full or short version. In the present study, we therefore aimed to investigate whether the psychometric properties of the RCADS-47 and RCADS-25 child self-report instruments could be generalized to a Norwegian setting. We examined the internal reliability and the structural validity of the RCADS-47 full version and the RCADS-25 short version.
Further, we explored measurement invariance across gender. Finally, the convergent validity of RCADS-47 and RCADS-25 with other wellestablished measures of anxiety and depression was investigated.

| Participants and procedure
The current study was a part of a larger trial of an indicated transdiagnostic preventive intervention, the ECHO study (clinicaltrials.gov NCT04263558; see Neumer et al., 2021). Children and parents at participating schools received oral and written information about the study, inviting children who experienced symptoms of anxiety or depression to participate in the study. Children with valid parental consent completed electronic surveys at school. The surveys were mostly forced choice to minimize the amount of missing data. Altogether 43 public schools in urban and rural areas, from 24 municipalities and boroughs across three regions in Norway (southeast, central, and north), took part in the present study. The participating children (n = 592) had a mean age of 10.7 years and attended fourth through sixth grade.
To investigate psychometric properties, the RCADS youth selfreport was included as a pre-intervention measure for a subsample of the participants in the larger trial (all participants from two out of five data collection waves; fall and spring term 2020/2021) alongside the anxiety measure Multidimensional Anxiety Scale for Children (MASC; March et al., 1997) youth report and The Short Mood and Feelings Questionnaire (SMFQ; Angold et al., 1995) youth report. The current article thus presents data from this subsample.

| The Revised Child Anxiety and Depression Scale (RCADS-47 youth version)
The full RCADS-47 (Chorpita et al., 2000) youth version is a 47-item self-report anxiety and depression rating scale for children and youth aged 8-18 years. Though a parallel caregiver version exists, the present study examined the youth self-report only. Respondents rate how often each item applies to them on a 4-point Likert scale. The RCADS-47 is based on the Spence Children's Anxiety Scale (SCAS; Spence, 1997), and includes six subscales which corresponds with the DSM-IV classifications for anxiety and depressive disorders: separation anxiety disorder, generalized anxiety disorder, panic disorder, social phobia, obsessive-compulsive disorder (OCD), and major depressive disorder. Scores can also be aggregated into Total Anxiety (sum of the five anxiety subscales) and Total Internalizing (sum of all six subscales). Although obsessive-compulsive disorder has been recategorized and excluded from the anxiety disorders category in the DSM-V (APA, 2013), the manual still recognizes the high comorbidity between OCD and anxiety disorders. As both often share treatment approaches (such as cognitive behavioral therapy with exposure), it is useful to continue to assess OCD in line with anxiety disorders (Ebesutani et al., 2017). Chorpita et al. (2005) found good internal consistency for the six RCADS subscales, with α ranging from 0.78 (SAD) to 0.88 (GAD). Furthermore, they found good convergent and discriminant validity, and RCADS demonstrated greater correspondence to specific diagnostic syndromes than traditional measures of anxiety and depression. The RCADS is freely available from the developers in 19 languages, including Norwegian (www.childfirst.ucla. edu/resources).
The abbreviated RCADS-25 youth self-report is a 25-item scale developed by shortening the five separate anxiety scales of the RCADS-47, keeping three items from each original anxiety subscale, resulting in a broad 15-item anxiety scale (Ebesutani et al., 2012). The

| Demographics variables
Parents provided demographic information (age, gender, grade) upon completing parental consent forms.
-3 of 13 differences in symptoms of anxiety or depression. Pearson correlations were used to investigate the association between age and symptoms.
Reliability, for both the full sample and by gender, was assessed with the internal consistency coefficient McDonald's omega (ω; McDonald, 1999) and corresponding BCa bootstrapped 95% confidence interval (CI) (10,000 replications). Omega can be interpreted the same way as Cronbach's Alpha, but it does not depend on items being tau-equivalent and provides a more accurate approximation of a scale's internal structure than Cronbach's Alpha (Dunn et al., 2014;Revelle & Zinbarg, 2009).
CFAs were carried out to investigate construct validity. As the required sample size depends on values in the covariance matrix unknown before the study, there is no consensus on estimation methods for CFA sample-sizes (Fayers & Machin, 2007). With the large RCADS-47 model, it is reasonable to assume that a few hundred participants would be required and that 589 participants were sufficient to ensure power and low bias in parameters interest (see Myers et al., 2011;Wolf et al., 2013). Given the ordinal metric nature of the items, the robust weighted least square (WLSMV) estimator, using a diagonal weight matrix, and theta parameterization was chosen for all analyses. Latent variables were scaled by fixing the first loading of each factor to 1. There was no missing data. Error variances were kept uncorrelated in all analyses. Factor loadings above 0.5 were considered acceptable (Hulland, 1999). For RCADS-47, the CFA was based on the theorized 6-factor structure equivalent to the RCADS subscales (Chorpita et al., 2000). For RCADS-25, the CFA was based on an a priori 2-factor structure of the two subscales (anxiety and depression). To evaluate the models' goodness-of-fit, we considered the CFI, TLI, and the RMSEA with its corresponding 90% CI. Higher CFI and TLI values indicate a better fit, and we considered CFI and TLI values above 0.90, and 0.95 as indicative of acceptable and good fit, respectively (Hu & Bentler, 1999). With RMSEA, lower values indicate a better fit, and RMSEA below 0.08 and 0.06 were considered as indicative of acceptable and good model fit, respectively. Limitations of the χ 2 -test includes test sensitivity that increases with sample size. This problem is compounded when accompanied by many degrees of freedom, which means that model size also affects the χ 2 -test (Moshagen, 2012). Due to these issues, we largely relied on alternative fit indices.
Measurement invariance tests across gender were performed following a 4-step procedure (Bowen & Masa, 2015). First, CFAs for each gender were run separately using the same approaches as in the CFAs previously described. Then we ran multigroup models with both genders to investigate configural, metric (weak) and scalar (strong) invariance. In the configural model, factor loadings and thresholds across groups were free. In the metric model, factor loadings and the first threshold were held equal across groups. In the scalar model, factor loadings and thresholds were held equal across groups. Like the χ 2 -tests, χ 2 -difference testing is sensitive to sample size and model size. Alternative fit indices are much less sensitive to sample size and more sensitive to lack of invariance than χ 2 -based tests of measurement invariance (Meade et al., 2008). In terms of the model comparison, we therefore reviewed CFI change (ΔCFI) and RMSEA change (ΔRMSEA). The ΔCFI and ΔRMSEA cut-offs used were 0.002 and 0.007, respectively, as suggested by Meade et al. (2008).
Pearson correlations (r) were used to evaluate convergent validity of RCADS anxiety and depression scales with the MASC and the SMFQ. In addition, partial correlation analyses were employed to Within each instrument, children had completed all the items and there was no missing data.

| Gender differences in symptoms
Mean scores for all measures are shown in Table 1. Girls reported higher levels of both anxious and depressive symptoms. Bootstrapped independent samples t-tests showed that the gender differences in means were significant for all subscales, although the MASC subscale Harm Avoidance was approaching a non-significant level. The effect sizes were small to moderate (Cohen, 1988

| Age differences in symptoms
Children's age did not appear related to symptoms of anxiety and depression. Pearson correlations revealed no significant associations between age and the total internalizing scores for RCADS-47

| RCADS-47 reliability
The internal consistency of the RCADS-47 subscales (Table 2) was good, with omega ranging from 0.81 (Obsessive-Compulsive) to 0.92 (Panic Disorder and Social Phobia). Omega was also estimated for each gender separately. Good internal consistency was indicated for both genders on all subscales. There were some gender discrepancies in the omega point estimates, particularly for the subscales Separation Anxiety and Obsessive-Compulsive disorder, but there was considerable overlap between the confidence intervals for boys and girls also for these two subscales.

| RCADS-47 measurement invariance across gender
Models fitted for boys and girls separately showed good fit with the data (Table 4). No modifications were made. Model fit for the configural model with both groups was also good, indicating configural T A B L E 1 RCADS-47, RCADS-25 MASC and SMFQ means and SD for the full sample, and by gender, along with ttest significance value and standardized effect size (Hedges g) supported. This finding implies that the differences in the proportion of responses in each category was caused by factor mean differences.

Measure/Subscale
RCADS-47 scores can therefore be interpreted the same way across gender.

| RCADS-47 convergent validity
The Pearson correlation between RCADS-47 total anxiety and MASC total was 0.81 (p < 0.001), and the partial correlation controlled for gender was 0.80 (p < 0.001). The convergent validity, then, was in the "excellent" category (≥0.75) according to the EFPA guidelines on congruent validity coefficients. The correlation between RCADS-47 major depression and the SMFQ total scores was 0.75 (p < 0.001), and the partial correlation was also 0.75 (p < 0.001). There was thus evidence of excellent convergent validity also for the RCADS-47 depression scale. As expected, there were also associations between anxiety symptoms and depressive symptoms. The Pearson correlation between RCADS-47 total anxiety and SMFQ was 0.71 (p < 0.001), and the partial correlation controlled for gender was 0.70 (p < 0.001). The correlation between RCADS-47 depression and MASC was 0.68 (p < 0.001), and the partial correlation controlled for gender was 0.67 (p < 0.001).

| RCADS-25 measurement invariance across gender
At the first step in the RCADS-25 gender invariance testing, separate models for boys and girls demonstrated acceptable fit with the data (    We observed gender differences in symptoms of anxiety and depression, with girls reporting higher levels of anxious and depressive symptoms than boys. The effects sizes for gender differences were larger for the anxiety scales than depression. There is abundant evidence that gender differences, with girls having increased risk for anxiety and depression, increase during adolescence (e.g., Larsson et al., 2016;Leikanger et al., 2012), although the evidence for gender differences in internalizing disorders among prepubescent children is inconclusive in the literature (see Costello et al., 2005;Thapar et al., 2012;Zahn-Waxler et al., 2008). We did not observe significant age differences in symptoms, but this may be due to the limited (and prepubescent) age range in the current study.
The means observed in the present study were higher than the means for population-based Icelandic (Olason et al., 2004) and American (Chorpita et al., 2000) samples among children the same age. This was expected and has a natural explanation in the recruitment strategy for the main study of an indicated intervention where the aim was to recruit children with more anxious or depressive symptoms than their peers. As such, the participating children's symptomatology should not be seen as representative of the general population.
Strengths of the study included negligible dropout, high response rate, a decent sample size in which both genders were represented, and sufficient variation in scores. An important study limitation to consider is that the sample is recruited as part of a larger study of an indicated preventive intervention for children with symptoms of anxiety or depression. While the response rate for the children with parental consent was high (99.8%), the participants represented 14% of the total children attending the participating schools. Children who signed up for the intervention will, on average, most likely have experienced more symptoms of anxiety and depression than the general population (e.g., Kösters et al., 2015). It is also possible that the parents who consented to their child's participation differed from the population on demographic or clinical characteristics. Another important limitation to consider was the narrow age-range of the sample (9-12 years), which meant that we could not explore the psychometric properties for the entire age-range for which the In conclusion, findings from the present study indicated that RCADS-47 and RCADS-25 are valid and reliable instruments for measuring symptoms of anxiety and depression in Norwegian settings. In combination with other cross-cultural research on these instruments, the results contribute to the evidence of a robustness of the construct validity and a stable performance of the RCADS-47 and RCADS-25 across cultures and settings (Piqueras et al., 2017;Young et al., 2020). investigating psychometric properties across a wider age-span in a population-based sample would also be useful.

ACKNOWLEDGMENTS
We thank all participating children and their parents for contributing to the study. We also thank the larger part of the ECHO project group, health services and collaborating schools, with special thanks to our research coordinators and our schools' project liaisons for aiding in the data collection. We are also grateful to Tore Wentzel-Larsen for critical manuscript revision. Finally, we thank the Kavli Foundation for funding the study.