Anxiety and depressive disorders are among the most common psychiatric conditions experienced by youth (Lewinsohn et al.
1993). Current or short-term prevalence rates are approximately 2–4% and 5–6%, respectively, in recent reviews (Costello et al.
2004; Costello et al.
2003), while lifetime prevalence rates range from 6–15% and 15–20%, respectively, in epidemiological studies (Silverman and Ginsburg
1998). Given the significant functional impairment (Birmaher et al.
1996) and increased risk of continued psychopathology in adulthood (Pine et al.
1998) associated with anxiety and depression in youth, the accurate assessment of such difficulties is imperative in aiding both clinical and research efforts.
Traditionally, self-report measures have been the dominant method for assessing internalizing disorders in youth (March and Albano
1996; Southam-Gerow and Chorpita
2007) as they provide an efficient and cost-effective means of gathering information. Although several measures have been developed to assess anxiety and depression [e.g., the Revised Children’s Manifest Anxiety Scale (RCMAS) and the Children’s Depression Inventory (CDI), respectively], the Revised Child Anxiety and Depression Scale (RCADS; Chorpita et al.
2000) maps onto current
DSM nosology and indexes the main features of five prominent
DSM anxiety disorders [separation anxiety disorder (SAD), social phobia (SOC), generalized anxiety disorder (GAD), obsessive-compulsive disorder (OCD), panic disorders (PD)] as well as major depressive disorder (MDD). This is an advantage over measures developed prior to more contemporary diagnostic classification systems (e.g.,
DSM-IV; APA
2000), as for example, Chorpita et al. (
2005) found that the RCADS targets anxiety and depression more specifically than the RCMAS and CDI, respectively.
Despite youth self-report instruments demonstrating to be useful in measuring anxiety and depression, limitations of such child and adolescent self-reports have been noted (e.g., Kendall and Flannery-Schroeder
1998), speaking to the importance of incorporating parent reports in the assessment of youth internalizing problems (e.g., Jensen et al.
1999; Klein et al.
2005). Researchers have already begun to administer the RCADS to parents (coined the RCADS-P
1) to assist in the measurement of anxiety and depression among youth (e.g., Costa et al.
2009; Watts and Weems
2006; Weems and Costa
2005). Other similar parent-report measures of youth anxiety also exist, including the Multidimensional Anxiety Scale for Children parent version (MASC-P; Baldwin and Dadds
2007), and the Screen for Child Anxiety Related Emotional Disorders-Revised parent version (SCARED-R-P; Muris et al.
2004), both which target
DSM related anxiety problems and demonstrate the importance of gathering parent-reported data when assessing anxiety in youth.
Some advantages of the RCADS-P relative to the MASC-P and the SCARED-R-P, however, include the RCADS-P’s ability to concurrently assess both anxiety and depression—a useful feature given the high comorbidity between these disorders in youth (Brady and Kendall
1992). In contrast to the other measures discussed, the RCADS-P is also available for free,
2 thus supporting the feasibility of its use in a wider variety of settings. The RCADS-P also recently evidenced particularly strong psychometric properties in a clinical sample of youth (
N = 490) diagnosed with structured diagnostic interviews (Ebesutani et al.
2010b), including the ability for the RCADS-P anxiety subscales to discriminate between anxiety disorders—an advantage over the SCARED-R-P, which has so far failed to make such discriminations (Muris et al.
2004). Accurate classification percentages based on receiver operating characteristic analyses were also reported, ranging from 71.3% to 85.4%. Given these strengths, the RCADS-P (whether utilized alone or in combination with the RCADS) has the potential to be a comprehensive, efficient, and economical tool in the assessment of youth internalizing problems.
The present study thus sought to thoroughly examine the psychometric properties of the RCADS-P and all of its subscales in a more representative, community sample of school-based children and adolescents. An additional aim of the present study was to provide normative data, which have not yet been reported for the RCADS-P, to allow for the derivation of T-scores to increase the clinical utility and interpretability of the RCADS-P scale scores. Based on the results from the recent psychometric study on the RCADS-P based on a clinical sample (Ebesutani et al.
2010b), we hypothesized that similar results would be obtained in the present study based on a school-based sample of youth. Specifically, we hypothesized that (a) the RCADS-P 6-factor model would be supported via confirmatory factor analysis (CFA) and would evidence better model fit over a 5-factor model (treating GAD and MDD as a single “distress” factor; Lahey et al.
2008; Watson
2005), (b) the 6-factor structure of the RCADS-P would evidence factor invariance across boys and girls as well as across younger and older youth, (c) reliability of the RCADS-P subscales would be supported via adequate internal consistency and test-rest reliability estimates, and (c) the RCADS-P depression and anxiety subscales would evidence significant correspondence with criterion measures for MDD and corresponding anxiety problems.
Discussion
The RCADS-P demonstrated favorable psychometric properties in the present sample of school-based children and adolescents. Specifically, CFA results supported the six-factor RCADS-P model, which fit equally well across boys and girls as well as across younger and older youth. Consistent with findings from a recent RCADS-P psychometric study in a clinical sample (Ebesutani et al.
2010b), the present CFA results did not support combining the MDD and GAD scales into a single “distress” factor. These results, however, are not consistent with recent findings that support collapsing MDD and GAD into a single construct (e.g., Lahey et al.
2008; Watson
2005). Higa-McMillan et al. (
2008), for instance, found that GAD in children appears to have a stronger relationship to depression than to social phobia. More research is thus needed to determine whether MDD and GAD indeed constitute the same “distress” construct in youth, or whether, for example, these discrepant findings are due in part to differences between youth and parent reporting styles and/or due to differences in measurement strategies. The present findings, for instance, were based on
parent reports on the RCADS-P (a
self-report measure), while Higa-McMillan et al. (
2008) findings were based on
child reports on the ADIS-IV-C (a
clinician guided, semi-structured interview). Future studies controlling for these differences may help clarify this issue related to the GAD and MDD distinction.
Although the previous RCADS-P psychometric study (Ebesutani et al.
2010b) did not identify any significant problems with scale items, the present investigation evidenced a relatively low factor loading of one Panic item (“
When my child has a problem, he/she gets a funny feeling in his/her stomach”). Two other items on the SAD scale also evidence somewhat low item-total correlation coefficients [“
My child has trouble going to school in the mornings because of feeling nervous or afraid”; “
My child is afraid of being in crowded places (like shopping centers, the movies, buses, busy playgrounds)”]. Although inclusion of these items did not substantively degrade the psychometric performance of these scales in the present validity tests, future studies should give particular attention to these items and continue to examine how they may affect measurement precision.
Regarding the reliability of the RCADS-P scales over time, all scales evidenced satisfactory test-retest correlation coefficients. However, the Panic subscale was associated with the lowest test-retest reliability coefficients. This may be due to the episodic and transient nature of panic symptoms (e.g., rapid heart beating, shortness of breath), which may fluctuate and/or change over short periods of time. At the same time, the return rate of retest packets from parents was particularly low, which may have compromised the representativeness of the current sample. Consequently, while the RCADS-P scales appear to provide reliable estimates of MDD and the targeted anxiety disorders, additional research may focus on replicating these finding as well as addressing appropriate assessment intervals for panic symptoms.
The RCADS-P also demonstrated high convergent and divergent validity across all scales. The RCADS (child-report) was utilized as the convergent validity criterion when examining the RCADS-P subscales as both measures target the same
DSM-oriented MDD and anxiety problems. However, as child and parent reports are known to only moderately correlate with each other, inclusion of another parent-based measure that targets the same
DSM related problem areas would have allowed for a better convergent validity test. Although we did utilized the parent-report CBCL
DSM-oriented Affective Problems scale to evaluate the performance of the RCADS-P MDD subscale (given that this parent-reported scale was developed to target MDD and dysthymic disorder; Achenbach and Rescorla
2001), the CBCL’s
DSM-oriented Anxiety Problems subscale does not target anxiety disorders at the same level of specificity as the RCADS-P (i.e., the CBCL’s
DSM-oriented Anxiety Problems subscale was designed to target the cluster of GAD, SAD and specific phobia; Achenbach and Rescorla
2001). Future studies should thus consider examining the degree of convergence between the RCADS-P subscales and other parent-report measures that target comparable
DSM-oriented depression and anxiety subscales.
Although the results of the present study support the psychometric properties of the RCADS-P and the utility of this measure as a useful screen for identifying children and adolescents with depressive and anxiety problems in school settings, there were particular limitations as well as areas for future research and development worth noting. Although the present study was based on a large, ethnically diverse sample, including youths from under-researched populations (e.g., Pacific Islanders), this sample was based solely on a Hawaii youth population and did not include large numbers of specific minority populations that are more represented in several continental US regions (i.e., African American, Hispanic youth). This may pose a limitation to the generalizability of the present findings. Further, there was also a low response rate of parents consenting to participate in the current study, which also contributed to low return rates of our test-retest sample. Although low parent form return rates are typical for parent-based research in school settings (cf. Higa et al.
2006), we compared the demographic data of the youths and families of our sample to the most recent demographic data provided by the U.S. Census Bureau for the Honolulu County - the county of the schools surveyed in the present study—in order to assess how well our sample is representative of students and families of the general Honolulu County population. Based on the most recent and available U.S. Census Bureau data (U.S. Census Bureau
2010), the median household income for the Honolulu County in 2008 was $70,010, and the three most represented ethnicities in 2009 in this county were Asian American (43.9%), White (26.6%) and Multiethnic (16.2%). The median income from the Honolulu County ($70,010) fell in the median income range of our sample ($60,000–$89,000); further, Asian American, White and Multiethnic were also the three most represented ethnicities in our sample. It is notable, however, that there was a smaller percentage of White youths in our sample (4.4%) compared to the percentage of “White persons” reported living in the Honolulu County in 2009 (26.6%). Although our sample nonetheless appears somewhat representative of our targeted population, additional research appears needed with larger and more inclusive samples of youths and families from an increased variety of regions, ethnicities and backgrounds to better understand the generalizability of the present findings. Notably, results of the present study did not identify differences in RCADS-P depression scores between boys and girls, despite girls typically evidencing more depressive symptoms than boys (e.g., Glambos et al.
2004). The degree to which this finding and others inconsistent with previous research is due to characteristics specific to this sample deserves future attention. Relatedly, it is also possible that parent reports of their children’s internalizing experiences are limited in certain ways (given that parents do not have direct insight into their children’s internal states), and this should also be considered when further examining the RCADS-P. Exemplifying this point, Weems et al. (
2005) recently found that child report—but not parent report—on the RCADS was significantly related to fear induced physiological response. The degree to which parent reports on the RCADS-P are (or are not) related to such internal states of children warrants future attention.
Interestingly, parent-child agreement conducted on Boys-only and Girls-only subsamples revealed that agreement fell in the expected (moderate) range for the majority of analyses. However, a closer investigation of the data provides additional insight into the complex nature of parent-child (dis)agreement research with regard to age, gender, and type of symptoms. For example, consistent with previous findings on parent-child agreement specific to anxiety subtypes, such as on the SCARED anxiety questionnaire (Birmaher et al.
1997) and on the RCADS-P in a clinical sample (Ebesutani et al.
2010b), parent-child agreement was greatest for SAD. On the other hand, parent-child agreement was smallest for GAD (based on the total sample), demonstrating that the nature of parent-child agreement differs by anxiety type.
With regard to youth sex, parent-child agreement based on the Boys-only subsample fell in the very low range (near zero correlation) for the GAD, panic and social anxiety subscales. De Los Reyes and Kazdin (
2005) recently highlighted the complexity of discrepancies between child and parent reports. For example, these authors suggest that informant characteristics, such as youth sex, may affect parent-child agreement, but that the evidence regarding this relationship has been mixed. Some studies found youth sex to be significantly related to parent-child agreement (Grills and Ollendick
2003; Verhulst and van der Ende
1992), whereas other studies have not supported this finding (Choudhury et al.
2003; Christensen et al.
1992). The present results add to this debate by supporting the notion that parent-child agreement may vary somewhat as a function of youth sex (i.e., greater cross-informant agreement among girls). Future research should incorporate these reporter qualities when further evaluating the RCADS-P. Specifically, a related implication worthy of further exploration is that the integration of child and parent reports on the RCADS may thus not be able to rely on a simple additive approach.
Despite the noted limitations and areas for future research, the present study broadened the psychometric support for the RCADS-P to a wider population of school-based youth, and also provided normative data to allow for identifying youths who are clinically-elevated in the targeted areas of depression and anxiety problems. The current findings also provided insight into a variety of theoretical implications regarding the assessment of psychopathology (i.e., models of psychopathology, sampling characteristics, reported agreement). Continued research will benefit from examinations of both the RCADS and RCADS-P in a greater diversity of contexts (e.g., other communities, other languages), investigation of possible adaptations for purposes of briefer assessment models (e.g., Chorpita et al.
2010), and evaluation of the suitability of the measurement structure in light of the ongoing evolution of the clinical and psychiatric nosology (e.g.,
DSM-V).