«The factorial structure of the 41-item version of the Screen for Child Anxiety Related Emotional Disorders (SCARED) in a Spanish population of 8 to ...»
© International Journal of Clinical and Health Aberration Scale and Magical Ideation Scale
FONSECA-PEDRERO et al. Perceptual Psychology ISSN 1697-2600
2009, Vol. 9, Nº 2, pp. 313-327
The factorial structure of the 41-item version of
the Screen for Child Anxiety Related Emotional
Disorders (SCARED) in a Spanish population of 8
to 12 years-old1
Andreu Vigil-Colet2 (Universidad Rovira i Virgili, Spain),
Josepa Canals (Universidad Rovira i Virgili, Spain), Sandra Cosí (Universidad Rovira i Virgili, Spain), Urbano Lorenzo-Seva (Universidad Rovira i Virgili, Spain), Pere Joan Ferrando (Universidad Rovira i Virgili, Spain), Carmen Hernández-Martínez (Universidad Rovira i Virgili, Spain), Claustra Jané (Universidad Autónoma de Barcelona, Spain), Ferran Viñas (Universidad de Girona, Spain), and Edelmira Doménech (Universidad Autónoma de Barcelona, Spain) ABSTRACT. On this instrumental study we intend to analyse the factorial structure of the Screen for Child Anxiety Related Emotional Disorders (SCARED) in a Spanish sample using exploratory and confirmatory factorial analysis. As a second objective we intend to develop a short form of it for rapid screening and, finally, to analyze the reliabilities of both questionnaires. The SCARED was administered to a community sample of 1,508 children aged between 8 and 12 years. The sample was randomly split using half for the exploratory analysis and the other half for the confirmatory study.
Furthermore a reduced version of the SCARED was developed using the SchmidLeiman procedure. Exploratory Factor Analysis yielded a four factor structure comprised of Somatic/panic, Generalized anxiety, Separation anxiety and Social phobia factors.
This research was partially supported by a grant from the Catalan Ministry of Universities, Research and the Information Society (2005SGR00017) and by two grants of the “Fondo de Investigaciones Sanitarias» (PI07/0839 and PI04/0978), Spanish Ministry of Health and Consumption.
Correspondence: Research Center for Behavior Assessment. Dept. Psicología. Universidad Rovira i Virgili. Ctra. de Valls, s/n. 43007 Tarragona (España). E-mail: firstname.lastname@example.org Int J Clin Health Psychol, Vol. 9. Nº 2 314 VIGIL-COLET et al. Screen for Child Anxiety Related Emotional Disorders This structure was confirmed using Confirmatory Factor Analysis. The four factors, the full scale and the short scale showed good reliabilities. The results obtained seem to indicate that the Spanish version of the SCARED has good internal consistency, and along with other recent results, has a structure of four related factors that replicates the dimensions proposed for anxiety disorders by the DSM-IV-TR.
KEYWORDS. Anxiety assessment. Anxiety disorders. Anxiety screening. Instrumental study.
RESUMEN. El presente estudio instrumental pretende analizar la estructura factorial del Screen for Child Anxiety Related Emotional Disorders (SCARED) en una muestra española utilizando tanto análisis factorial exploratorio como confirmatorio. Como objetivo secundario se pretende desarrollar una version reducida utilizable como instrumento de cribaje y, finalmente, analizar las propiedades psicométricas de ambos cuestionarios. El SCARED fue administrado a una muestra comunitaria de 1.508 niños de entre 8 y 12 años. Dicha muestra fue subdividida de forma aleatoria utilizando la primera mitad para el análisis exploratorio y la segunda para el confirmatorio. Además se desarrolló una versión reducida utilizando el procedimiento de Schmid-Leiman. El análisis factorial exploratorio reveló una estructura de 4 factores: Somático/pánico, Ansiedad generalizada, Ansiedad de separación y Fobia social. Esta estructura fue confirmada mediante al análisis factorial confirmatorio. Los cuatro factores, la escala completa y la escala reducida mostraron fiabilidades satisfactorias. Los resultados obtenidos parecen indicar que la version española del SCARED, al igual que algunos estudios recientes, presenta una estructura de cuatro factores relacionados que replican las dimensiones propuestas para los transtornos de ansiedad del DSM-IV-TR.
PALABRAS CLAVE. Evaluación de la ansiedad. Trastornos de ansiedad. Cribaje de la ansiedad. Estudio instrumental.
Several epidemiological studies indicate that anxiety disorders are one of the most prevalent categories of psychopathology among children and adolescents (American Academy of Child and Adolescent Psychiatry, 1997; Anderson, Williams, Mc Gee, and Silva, 1987; Ollendick, King, and Muris, 2002; Spence, 1998). Available data from the community suggest that 2.8% to 27% of children and adolescent might be affected by some form of broadly conceptualized anxiety disorder (Costello, Egger, and Angold, 2005; Costello, Mustillo, Erlanki, Keeler, and Angold, 2003; Krain et al., 2007) and half of these estimates put the figure above 10% (Costello and Angold, 1995). Separation anxiety disorder (SAD), generalized disorder (GAD), and specific phobias seemed to be the most common anxiety disorders in childhood (Bernstein, Borchardt, and Perwien, 1996).
Anxiety disorders, as with other internalizing disorders, are often underdiagnosed, because anxious children do not cause problems. However, there are reasons why it is important to identify childhood anxiety disorders: their relation to psychosocial difficulties, Int J Clin Health Psychol, Vol. 9. Nº 2 VIGIL-COLET et al. Screen for Child Anxiety Related Emotional Disorders (Brent et al., 1998; Clarke, Hops, Lewinsohn, and Andrews, 1992; Messer and Beidel, 1994), the risk of developing other comorbid psychiatric disorders (Curry and Murphy, 1995; Thapar and McGuffin, 1997) and their continuity into adulthood (Keller et al., 1992; Pfeffer, Lipkins, Plutchik, and Mizruchi, 1988). Taking into account the high prevalence and the negative consequences of childhood anxiety disorders, Birmaher et al. (1997) pointed out that a reliable and valid self-reported instrument needed to be developed to screen anxiety disorders symptoms and to provide diagnostic information about the different types of anxiety disorders. For this reason they developed the Screen for Child Anxiety Related Emotional Disorders (SCARED).
The SCARED is a 38 item self-report questionnaire. According to its authors, it measures five child and adolescent symptom dimensions, specifically: SAD, GAD, panic disorder (PD), social phobia symptoms (SP) and school anxiety. Due to the difficulties in discriminating between the social phobia factor and other anxiety disorders, Birmaher et al. (1999) re-examined the questionnaire adding three new items for SP, thus developing the final 41-item version of the SCARED (Birmaher et al., 1999). Several studies have shown that SCARED is reliable, most of them obtained reliabilities in the range α =.70 –.85 depending upon the scale, and showed good convergent validity when it was related to other anxiety scales such as the State–Trait Anxiety Inventory for Children or the Revised Children’s Manifest Anxiety Scale (Birmaher et al., 1999; Boyd, Ginsburg, Lambert, Cooley, and Campbell, 2003; Hale, Raaijmakers, Muris, and Meeus, 2005; Muris, Merckelbach, Gadet, and Meesters, 2000; Muris, Merckelbach, Ollendick, King, and Bogie, 2002; Wren et al., 2007; Wren, Bridge, and Birmaher, 2004). Thus, a sufficient number of studies have given support to the reliability and validity of the SCARED.
Referring to the factorial structure of SCARED, the seminal papers of Birmaher et al. (1997, 1999) reported a five factor structure, but other studies, such as Boyd et al.
(2003) or Wren et al. (2007), found in the first case a four factor structure, the school phobia factor being integrated into the generalized anxiety factor, and a three factor structure in the second case. In Birmaher et al.’s original paper in 1997, four of the anxiety scales represented anxiety disorders corresponding to DSM categories. The fifth, the school phobia scale, according to Birmaher et al. (1997), could be best considered as a separate anxiety category.
Analyzing the revised 66 item version of SCARED (SCARED-R) developed by Muris, Merckelbach, Schmidt, and Mayer (1999), they reported a unidimensional solution, only finding other subscales when the factorial analysis was performed on parts of the SCARED-R. The same authors analyzed the initial 38 items version of SCARED finding also a strong first factor, although a confirmatory factorial analysis (CFA) reported a good fit for a five factors solution. Nevertheless, other CFA analyses performed on the SCARED showed a bad fit with the five factors solution and a good fit for both one and five factors solutions (Boyd et al., 2003; Hale et al., 2005).
The empirical evidence on its internal structure is one of the key aspects in test selection in psychological research (Carretero-Dios and Pérez, 2007). Given the diversity of factorial structures reported for the SCARED, we believe that some methodological issues must be studied if the lack of consensus on the factorial structure is to be Int J Clin Health Psychol, Vol. 9. Nº 2 316 VIGIL-COLET et al. Screen for Child Anxiety Related Emotional Disorders explained and the internal structure of SCARED determined. To begin with, most of the structures have used principal components analysis as the extraction method including both the seminal papers by Birmaher et al. (1997, 1999) and more recent studies such as those by Linyan, Kai, Fang, Yi, and Xueping (2008) and Muris et al. (2006). This method is questionable when, as in this case, a latent variable is intended to explain the relationships between the items. Furthermore, most of the exploratory factor analysis (EFA) studies have applied an orthogonal rotation which is only appropriate when the different factors are supposed to be independent. It seems difficult to suppose that different factors related to anxiety disorders are totally independent. There are three more facts that question this independence. One refers to the fact that those CFAs which reported a good fit with the five factor structure tested models with correlated factors. The second refers to the high internal consistencies reported for the full scale of the SCARED (around α =.90), which are difficult to reach if the items implied in the scale belong to five independent subscales. Thirdly, many studies analyzing the factorial structure of the SCARED have found that the first factor to be extracted explains much more variance that the other factors, which may indicate the presence of an overall factor that explains the relationships between the subscales. Moreover, most of the EFA have been done using product moment correlations when, usually, a 3-point response format, such as the SCARED format, implies that the distribution of the items is usually non-normal, with asymmetric distribution and/or with an excess of kurtosis. In these cases polychoric correlation instead of Pearson correlation is advised (Muthén and Kaplan, 1992).
From the point of view of the sample size, some studies analysed moderate or large samples (around 500 individuals) whereas others analyzed small samples (around 200 individuals) and so do not provide a stable solution for a questionnaire of 41 items.
Furthermore, when CFA has been performed, the sample sizes meant they could not be split in order to test the structure found by EFA in a different sample. Thus, the structure found in CFA may be contaminated by the characteristics of the sample used in the EFA. Finally, as Ferrando and Lorenzo-Seva (2000) pointed out, when personality or psychopathology questionnaires obtained by EFA are tested using CFA, the model proposed is usually rejected, although a series of different exploratory studies have previously replicated the same factorial structure. This case is especially usual in multidimensional questionnaires with a moderate or high number of items in each subscale. In these cases they propose that unrestricted models are more appropriate for testing the model fit in CFA. Moreover, children are less accurate than adults in assessing their own behaviours so less restrictive models are more suitable than restrictive ones.
The purpose of the present reseach, which may be considered an instrumental study according to Montero and León’s (2007) research classification, is to analyze the factorial structure of the SCARED in a sample sufficiently large so as to allow it to be split into two subsamples, one for the EFA and the other for the CFA. In the first subsample an EFA analysis using polychoric correlations and oblique rotation procedures will be performed. Results obtained will be confirmed with a restricted CFA in the second subsample. On the other hand, we expect that the different subscales of the SCARED Int J Clin Health Psychol, Vol. 9. Nº 2 VIGIL-COLET et al. Screen for Child Anxiety Related Emotional Disorders are not independent, therefore we expect there is a second order factor that explains the interrelationships between them. To test this we will introduce a second order factor in the CFA to evaluate this possibility. If we find this structure of correlated factors we think that it may be useful for developing a short form of the SCARED choosing the items that are more closely related to the overall factor. We think that such a short form may be useful as a fast screening test for anxiety disorders. Finally, previous research has shown that anxiety is related to sex and age. In this respect, females have a higher risk of anxiety disorder symptoms while age is related to a decrease in separation anxiety and an increase in generalized anxiety in early childhood (Essau, Muris, and Ederer, 2002; Hale et al., 2005; Wren et al., 2007). Taking into account these results, we will test whether these age and sex differences are also found in the Spanish version of SCARED.