Resumen:

El Eating Disorder Inventory (EDI) es un instrumento psicométrico creado por Garner y cols. en 1983 con el fin de evaluar las características psicológicas en los trastornos de la conducta alimentaria. Su propósito es conseguir una evaluación integral de estos padecimientos, que no se limite solamente a las manifestaciones conductuales relacionadas con la alimentación o el deseo de adelgazar. A lo largo de los años, el EDI ha sido uno de los instrumentos más utilizados en la exploración de actitudes y conductas alimentarias de riesgo tanto en población clínica como en pobla ción abierta. Sin embargo, el uso de escalas psicométricas en poblaciones diferentes de aquella para la cual fueron originalmente diseñadas es una práctica que requiere un cuidadoso trabajo de traducción y adaptación. Los reactivos deben expresarse con frases culturalmente significativas, y es posible que algunos constructos no lograr expresar el significado original al cambiar el contexto de su aplicación.

A la fecha existe el informe de los puntos de corte, y el análisis de la sensibilidad y especificidad de los mismos para cada una de las subescalas del instrumento en pacientes diagnosticadas con

TCA; sin embargo, no se ha obtenido la validez ni la confiabilidad del mismo.

El objetivo del presente trabajo fue obtener la confiabilidad y validez de constructo y concurrente del Eating Disorder Inventory en una muestra de pacientes mexicanas con trastorno de la con ducta alimentaria. El inventario junto con una batería de pruebas psicológicas que incluyó entre otras la Symptom Check List (SCL90) y el Inventario de Autoestima de Coopersmith, se aplicó a una muestra de 523 mujeres con edad promedio de 19.9 años y diagnóstico de trastorno de la conducta alimentaria según los criterios del DSM-IV, que se hallaban en tratamiento en la Clínica de Trastornos de la Conducta Alimentaria del Instituto Nacional de Psiquiatría Ramón de la Fuente en el período 1997-2002. A partir del análisis de consistencia interna del inventario se eliminaron 12 reactivos, los restantes se incluyeron en el análisis factorial. Una vez eliminados los reactivos se obtuvo un valor de confiabilidad alfa de Cronbach de .93. El resto de los reactivos se utilizó para llevar a cabo un análisis factorial de componentes principales con rotación Varimax, el cual arrojó una estructura de 11 factores de los cuales sólo seis cumplieron con los requisitos necesarios para formar un factor. El resultado fue la eliminación de 12 reactivos. Al proceder a un segundo análisis factorial se obtuvo una estructura de seis factores que explicó 56% de la varianza total y se obtuvieron valores del alfa de Cronbach de .76 a .86 para cada uno. La versión final del inventario constó de 36 reactivos y obtuvo una consistencia interna de .92. Se obtuvieron correlaciones positivas y significativas entre las subescalas del EDI con las subescalas del SCL90, mientras que éstas fueron negativas con el Inventario de autoestima de Coopersmith. El EDI es adecuado para evaluar actitudes y conductas relacionadas con el comer, la imagen corporal y las características psicológicas de pacientes con trastorno de la conducta alimentaria en México. Al emplear la versión original se debe actuar con cautela en virtud de que se encontraron reactivos con correlaciones bajas con la escala completa, otros que se incluyeron en dos factores y los que no se incluyeron en ninguno. Es interesante observar que la escala de Perfeccionismo desaparece en su totalidad, hallazgo que confirma lo encontrado por otros autores en México y en el extranje ro. Asimismo, se observa que el mayor porcentaje de la varianza total en el análisis factorial se explica con la subescala Bulimia. Esto sugiere que el EDI es un instrumento que distingue adecuadamente las conductas de tipo bulímico, lo que permite evitar el sesgo a favor de la anorexia nerviosa restrictiva que presentan otros instrumentos, como el Eating Attitudes Test, de acuerdo con sus autores.

Palabras clave: Eating Disorder Inventory; México; reproducibilidad de resultados; trastornos de la conducta alimentaria

Abstract:

The Eating Disorder Inventory (EDI) is a psychometric instrument developed by Garner et al. for the evaluation of psychological traits in patients with eating disorders. The questionnaire offers an integral evaluative approach that includes other psychological characteristics besides those of fear of fatness. It consists of 64 items in 8 subscales: 1) Drive for Thinness, 2) Bulimia, 3) Body Dissatisfaction, 4) Ineffectiveness, 5) Perfectionism, 6) Interpersonal Distrust, 7) Interoceptive Awareness, 8) Maturity Fears. The EDI is an instrument widely used for the exploration of eating disorders (ED). Nevertheless, the use of psychometric scales in a population other than the one in which they were originally developed, requires careful adaptation. Items should be phrased in a culturally significant way, and even after that, constructs still could have different meanings in different contexts. The factorial grouping of the EDI in open populations has been reported as different from the results in clinical samples. Also, Lee and cols. reported a low correlation of the EDI subscales that measure weight preoccupation and the General Health Questionnaire - 12, concluding that fear of fatness is not related with emotional distress in their sample. In Mexico City, Alvarez and Franco conducted a validation study, finding good reliability and discriminative power, and a factorial grouping close to the original. On the other hand, in a sample of teenage girls from a semi-rural area, the factorial grouping was very different. Although we already have data about the EDI's specificity and sensitivity in Mexican ED patients, there are no studies of the validity and reliability of the test in this population. Therefore, our main objective in this work was to validate the EDI in a sample of Mexican ED patients. Also, the score information provided could be used for comparison purposes with other clinical samples.

A non-probabilistic sample was obtained of all subsequent patients attending the ED Clinic at the National Institute of Psychiatry Ramón de la Fuente (INPRF) in the period 1997 2002 (n=523). Patients were diagnosed according to DSM-IV criteria in a clinical interview. They also completed other questionnaires, such as the Symptom Check List (SCL90) and the Coopersmith's Self-esteem Inventory. According to diagnosis, the sample was composed of compulsive/purging type anorexia nervosa, 5.7%; restrictive anorexia nervosa, 8%; purging type bulimia nervosa 45.1% and eating disorders not otherwise specified 41.3%. Mean age was 19.9 years (s.d.=3.9), within a rank of 13 to 39 years. Mean age at the beginning of ED was 16 years (s.d.=3.1). Mean educational level was 12 years (s.d.=3), i.e. high-school level. The sample included single women 93.9%, married 4.8% and divorcees, 1.4%. Mean Body Mass Index was 21 (s.d. = 5.5). Participants completed the EDI, SCL90, and Coopersmith's self-esteem inventory during their first visit to the Clinic. They were assured of the voluntary nature and confidentiality of their participation. Completing the tests took them about 60 minutes. An internal reliability analysis was conducted, followed by a factorial analysis of main components with Varimax rotation. Pearson correlations were made to assess the concurrent validity of EDI and other instruments. Analysis of variance was employed to compare between diagnostic groups. Data were captured and analysed in the SPSS software, versión 10.0.

The first step of the analysis was the item-total correlation, considering as valid correlations equal or over 0.28. This step eliminated 12 items that were not included in further analyses. Cronbach's alpha was 0.93. Most of the items in the Perfectionism subscale disappeared in this step. Second step was factorial analysis. We found 6 factors with a minimum of 3 items included with factorial charges equal or over .40. Then a second analysis was conducted with only the 40 items that had been grouped in the 6 factors. Factor 1 included items from Bulimia and Interoceptive Awareness; factor 2, from Drive for Thinness and Body Dissatisfaction; factor 3, from Interoceptive Awareness; factor 4, from Ineffectiveness; factor 5, from Maturity Fears; and factor 6, from Body Dissatisfaction. The resulting factorial structure explained 56% of total variance. Cronbach's alpha of the final version was 0.92. Correlation analysis showed a positive and significant correlation of EDI with SCL-90, and a negative and significant correlation of EDI with Coopersmith's self-esteem inventory. Comparisons between diagnostic groups showed that bulimia nervosa patients had the highest scores in the EDI. Patients with restrictive AN had the lowest scores in all sub-scales except for Maturity Fears. Bulimia nervosa and compulsive/purging type AN patients were different from restrictive AN and EDNOS patients in the total score of Interoceptive Awareness and Ineffectiveness subscales. Bulimia nervosa was different from the other groups in Bulimia and Drive for Thinness subscales.

The results show that, in this sample, many of the EDI items have a poor correlation with the scale, and factorial grouping is different from the original. However, once non-correlated items are eliminated, a version of the EDI remains that is valid and reliable. Items from the Perfectionism subscale were eliminated because of low correlation with the rest of the EDI. This supports the findings in Bulgaria, rural Mexico, and Mexico City. Maturity Fears, which also had dubious results in other studies, grouped correctly in this sample, although it did not distinguish among diagnostic groups. Analysis of variance showed that subscales were able to differentiate the Bulimia nervosa patients. Also, most of the variance explained corresponded to the Bulimia subscale, suggesting that EDI can detect bulimic attitudes, and so is a useful complement to instruments that are more capable of detecting anorexia nervosa, such as the Eating Attitudes Test. However, this could also be an effect of the sample's composition, with more than half of it being bulimic patients. Another important segment of variance was explained by Interoceptive Awareness, Ineffectiveness and Maturity Fears subscales, psychological traits that are not necessarily related to ED. In this sample, psychological subscales correlated with eating and weight attitude subscales, showed that Mexican patients do present ED according to the way they are conceptualized in the DSM-IV. Our results show that the EDI is adequate for the evaluation of psychological traits of ED patients in Mexico. Perfectionism and Interpersonal Distrust subscales are an exception, that requires further investigation.

Key words: Eating disorders; Eating Disorder Inventory; Mexico; reproducibility of results