«Fonseca-Pedrero, Eduardo; Sierra-Baigrie, Susana; Paino, Mercedes; Lemos-Giráldez, Serafín; Muñiz, José Factorial structure and measurement ...»
International Journal of Clinical and Health
Asociación Española de Psicología
Fonseca-Pedrero, Eduardo; Sierra-Baigrie, Susana; Paino, Mercedes; Lemos-Giráldez, Serafín;
Factorial structure and measurement invariance of the Bulimic Investigatory Test, Edinburgh across
gender and age
International Journal of Clinical and Health Psychology, vol. 11, núm. 1, 2011, pp. 109-123
Asociación Española de Psicología Conductual Granada, España Available in: http://www.redalyc.org/articulo.oa?id=33715423007 How to cite Complete issue Scientific Information System More information about this article Network of Scientific Journals from Latin America, the Caribbean, Spain and Portugal Journal's homepage in redalyc.org Non-profit academic project, developed under the open access initiative © International Journal of Clinical and Health Psychology ISSN 1697-2600 2011, Vol. 11, Nº 1, pp. 109-123 Factorial structure and measurement invariance of the Bulimic Investigatory Test, Edinburgh across gender and age1 Eduardo Fonseca-Pedrero2, Susana Sierra-Baigrie, Mercedes Paino, Serafín Lemos-Giráldez, and José Muñiz (Universidad de Oviedo and Centro de Investigación Biomédica en Red de Salud Mental, CIBERSAM, Spain) ABSTRACT. The main purpose of this instrumental study was to examine the factorial structure and measurement invariance of the Bulimic Investigatory Test, Edinburgh (BITE) across gender and age in a community sample of nonclinical adolescents.
The sample consisted of 1,794 adolescents (50.2 % males), with a mean age of 14.7 years (SD = 1.72). The results showed that the BITE is a measurement instrument which presents adequate psychometric properties. The level of internal consistency for the Symptom subscale was.95 whereas for the Severity subscale it was.70. The analysis of the dimensionality of the BITE using exploratory factor analysis revealed a one-factor solution. In addition, the confirmatory factor analysis showed adequate fit indices for the one-factor solution. This dimensional structure of the BITE proved to be invariant across gender and age. Statistically significant differences were found as a function of age and gender when the groups were compared in the latent means.
Future studies should incorporate the new advances in psychological and educational assessment pertaining to Computerized Adaptive Testing as well as examine the measurement invariance of the tests which assess bulimic symptomatology across cultures.
KEYWORDS. BITE. Psychometric properties. Measurement invariance. Bulimic symptoms. Instrumental study.
This research was funded by the Spanish Ministry of Science and Innovation (BES-2006-12797, SEJ 2005-08924, PSI 2008-03934-PSIC, PSI 2008-03934) and by the Centro de Investigación Biomédica en Red de Salud Mental (CIBERSAM) (Instituto Carlos III).
Correspondence: Facultad de Psicología. Plaza Feijoo, s/n. 33003 Oviedo (Spain). E-mail:
RESUMEN. El objetivo de este estudio instrumental fue examinar la estructura factorial y la invarianza de medición a través del sexo y la edad del Bulimic Investigatory Test, Edinburgh (BITE) en una muestra comunitaria de adolescentes no clínicos. La muestra la formaron 1.794 adolescentes (50,2% varones), con una edad medida de 14,7 años (DT = 1,72). Los resultados mostraron que el BITE es un instrumento de medida que presenta adecuadas propiedades psicométricas. El nivel de consistencia interna para la subescala Síntomas fue 0,95, mientras que para la subescala de Gravedad fue de 0,70.
El estudio de la dimensionalidad del BITE mediante análisis factorial exploratorio mostró una solución esencialmente unidimensional. Los índices de bondad de ajuste para el modelo unidimensional sometido a prueba en el análisis factorial confirmatorio fueron adecuados. Más aún, esta estructura dimensional del BITE se mostró invariante en función del género y la edad. Se encontraron diferencias estadísticamente significativas en función del género y la edad cuando se compararon las medias latentes. Futuros estudios deberían incorporar los avances relacionados con la evaluación psicológica y educativa como la construcción de tests adaptativos computerizados, así como examinar la invarianza de medición de la sintomatología bulímica a través de las culturas.
PALABRAS CLAVE. BITE. Propiedades psicométricas. Invarianza de medición.
Sintomatología bulímica. Estudio instrumental.
Bulimia nervosa (BN), one of the two major eating disorders included in the DSMIV-TR, is characterized by the presence of binge eating and maladaptive compensatory methods along with an excessive influence of body shape and weight on self-evaluation (American Psychiatric Association, 2000). Individuals with BN suffer from significant distress, role impairment and medical and psychological consequences which severely affect their quality of life. Although epidemiological data place lifetime prevalence of BN in 1-3% (American Psychiatric Association, 2000; Hudson, Hiripi, Pope, and Kessler, 2007; Preti et al., 2009), many individuals with BN do not seek help, therefore, many cases may go undetected (Hoek, van Hoeken, and Katzman, 2003). Moreover, research shows that the percentage of individuals with subclinical forms of the disorder, that is, individuals suffering from bulimic symptoms associated with great distress but that may not meet the diagnostic criteria for BN is high (Chamay-Weber, Narring, and Michaud, 2005). In addition, bulimic symptomatology is not restricted to individuals with eating disorders but rather studies have shown that bulimic symptoms are also present in adults and nonclinical adolescents (de Souza Ferreira and Valeria da Veiga, 2008; Poyastro Pinheiro, Bulik, Sullivan, and Machado, 2008; Rodríguez et al., 2001; Sierra-Baigrie, Lemos-Giráldez, and Fonseca-Pedrero, 2009). Therefore, as BN represents an important public-health concern, the understanding of the nature of this phenomenon, as well as finding ways of early detection of at-risk individuals, are vital for the prevention of the disorder.
The field of psychological assessment of eating disorders has advanced considerably in the last few years (Peterson and Mitchell, 2005). In this regard, in the scientific literature we can find numerous self-reports developed for the assessment of behaviors, symptomatology and beliefs related to BN. Specifically for the assessment of bulimic symptomatology we find the Bulimia Test (BULIT) (Smith and Thelen, 1984), and the Int J Clin Health Psychol, Vol. 11. Nº 1 FONSECA-PEDRERO et al. Bulimic Investigatory Test, Edinburgh Bulimic Investigatory Test, Edinburgh (BITE) (Henderson and Freeman, 1987), or the Eating Disorders Examination-Questionnaire (EDE-Q) (Fairburn and Beglin, 1994) Since Henderson and Freeman first developed the BITE, it has been widely used in different studies and cultures (Kiziltan, Karabudak, Unver, Sezgin, and Unal, 2006; le Grange, Louw, Russell, Nel, and Silkstone, 2006; Miotto, De Coppi, Frezza, and Preti, 2003; Rivas, Bersabé, and Jiménez, 2004; Sierra-Baigrie et al., 2009), and its psychometric properties have also been widely investigated (Orlandi, Mannucci, and Cuzzolaro, 2005;
Ricca et al., 2000; Ricciardelli, Williams, and Kiernan, 1999; Rivas, et al., 2004; RuedaJaimes, Camacho, and Rangel-Martínez-Villalba, 2008; Waller, 1992). In a pioneer study conducted by Henderson and Freeman (1987), where the construction and validation of the BITE was first presented, its psychometric properties were examined in different groups of patients with eating disorders and controls. When the reliability of the subscales of the BITE was analyzed using Cronbach’s alpha coefficient, they found for all samples a value of.96 for the Symptoms subscale and.62 for the Severity subscale.
Moreover, the test-retest reliability levels ranged from.86 to.68. Ricciardelly et al.
(1999), using a sample of 427 Australian girls and 350 boys (age range 12-17 years), analyzed the internal structure of the Symptom subscale of the BITE. A general factor of bulimic symptomatology was found in the female sample which explained 20.85% of the variance; however, in the male sample two factors were found, namely: Emotional and Rigid/Disruptive Eating Style and Food Preoccupation and Binging, which respectively explained 13.8% and 8.4% of the total variance. On their part, Orlandi et al. (2005), in a more recent study using a general population sample (N = 995), another sample of 388 eating-disordered females and a third sample of 710 patients with obesity (575 female), found Cronbach’s alpha coefficients which ranged from.82 to.93. When the internal structure of the BITE was analyzed, they found a general factor which explained 26-38% of the total variance depending on the group used in the analysis.
In a study by Rivas et al. (2004), the BITE was translated and adapted into Spanish using a sample of 1,122 nonclinical adolescents, with an age range of 12-19 years. They found an internal consistency (Cronbach’s alpha) for the Symptom subscale of.82 and.63 for the Severity subscale. The analysis of the dimensional structure of the Symptom subscale revealed the presence of a general factor which explained 19.89% of the total variance. Finally, Rueda-Jaimes et al. (2008) further examined the psychometric properties of the BITE in a sample of 261 Colombian university students. In this study They found an internal consistency (Cronbach’s alpha) for the Symptom subscale of.86 and.52 for the Severity subscale. The analysis of its internal structure indicated the presence of two factors which explained 29.80% of the total variance. The first factor was composed of aspects related to diet and the second factor was composed of aspects pertaining to loss of control.
The BITE is also a useful measurement instrument for examining the prevalence rates of bulimic symptomatology, and in particular binge eating, in nonclinical and clinical populations (Rodríguez-Cano, Beato-Fernández, and Belmonte-Llario, 2005).
Furthermore, it allows us to deepen our comprehension of the phenomenology of these symptoms across gender, age, or culture of origin. In this sense, when the observed scores on the Symptom and Severity subscales are compared as a function of gender, Int J Clin Health Psychol, Vol. 11. Nº 1 112 FONSECA-PEDRERO et al. Bulimic Investigatory Test, Edinburgh we find that females score higher than males in bulimic symptoms measured with this instrument (le Grange, Telch, and Tibbs, 1998; Miotto, De Coppi, Frezza, Petretto et al., 2003; Preti et al., 2007; Ricciardelli et al., 1999). Very few studies have used the BITE to analyze the role that age plays in the expression of bulimic symptomatology, being age a less frequently studied variable than gender. In general terms, the studies which have employed the BITE in different samples, have not found statistically significant differences in the BITE scores by age of the participants analyzed (Preti et al., 2007;
Ricca et al., 2000). However, when mean scores are used to compare groups (e.g., male/ female; youths/ adults) it is important that the scores have the same meaning in each group; that is, the assessment is invariant across groups. In a classic study, Horn and McArdle (1992) defined measurement invariance as: «whether or not, under different conditions of observing and studying phenomena, measurement operations yield measure of the same attribute» (p. 117). When comparisons between groups are made, it is typically assumed that both the measurement instrument and the psychological construct underlying said instrument behave in the same manner and have the same significance across the groups being compared. If measurement invariance does not hold, the validity of the inferences and interpretations extracted from the data may be erroneous (Byrne, 2008). Therefore, it is crucial to examine measurement invariance of the assessment tool so that findings based on comparisons of the groups can be valid.
Within this field of research, the main purpose of this instrumental study (Carretero-Dios and Pérez, 2007; Montero and León, 2007) was to examine the dimensional structure and measurement invariance of the BITE across gender and age in a community sample of nonclinical adolescents. The present investigation is relevant for various reasons as it allows us to: a) understand and examine the dimensional structure (number and content) of bulimic symptomatology at an age of special risk for the development of eating disorders; b) test whether bulimic symptomatology holds invariant across gender and age without the confounding effects of medication and stigmatization frequently associated to clinical samples; c) determine the psychometric properties of a measurement instrument for screening purposes and for its use in detection and early intervention programs for participants at risk; and d) provide epidemiological data to better understand the nature and phenomenological expression of this psychological phenomenon in community adolescent samples.
Method Participants One thousand seven hundred and ninety-four adolescents enrolled in ten different Secondary Education Centers in the Principality of Asturias, a northern region in Spain, participated in the study. The schools were selected to ensure the heterogeneity of the sample with participants recruited from both urban and rural areas as well as from different socioeconomic statuses. The age of the participants ranged from 12 to 19 (M = 14.67; SD = 1.72), 50.2% males (n = 900) and 49.8% females (n = 894). The age distribution of the sample was the following: 12 years (n = 185), 13 years (n = 347), 14 years (n = 362), 15 years (n = 285), 16 years (n = 315), 17 years (n = 212), 18 years (n Int J Clin Health Psychol, Vol. 11. Nº 1 FONSECA-PEDRERO et al. Bulimic Investigatory Test, Edinburgh = 74) and 19 years (n = 14). With the aim of conducting the pertinent statistical analyses, a cross-validation study was performed where the total sample was then randomly split into two subsamples of 897 participants. The first sample consisted of 457 boys and 440 girls with a mean age of 14.66 years (SD = 1.72) and the second consisted of 443 boys and 454 girls with a mean age of 14.68 years (SD = 1.71). Neither age (t = -.193; p.05) nor sex rates (χ2 =.43; p.05) differed across subsamples.