Características psicométricas de la Escala Center for Epidemiological Studies-depression (CES-D), versiones de 20 y 10 reactivos, en mujeres de una zona rural mexicana / Psychometric characteristics of the Center for Epidemiological Studies-depression Scale (CES-D), 20-and 10-item versions, in women from a Mexican rural area

ABSTRACT

Depressive symptoms constitute a common mental health problem, with a relevant social and personal impact. These symptoms are present not only among the urban population in more economically developed countries, but also in rural areas in poor and middle development countries. In order to obtain reliable information on the frequency of depressive symptoms, their risk factors or the impact of preventive and clinical measures, valid measurement instruments are needed. Radloff's Center for Epidemiological Studies - Depression scale (CES-D) was originally developed for the study of depressive symptoms in an open population. While the CES-D is not useful for the evaluation of depressive disorders according to psychiatric criteria, it can still yield useful information about the presence of depressed mood, feelings of guilt and worthlessness, feelings of helplessness and hopelessness, psychomotor retardation, and somatic complaints, which constitute dimensions of depression. The instrument has been shown to be valid in culturally diverse groups. It has also been shown to correlate with the clinical diagnosis of depression, with sensibility as high as 100%, while its specificity has been reported as 57-88%. Shorter versions of the CES-D have been developed. Their advantages include a more easy inclusion in ample questionnaires, and their being less tiresome for respondents. The main objective of this study was to evaluate the psychometric properties of CES-D, both the original, 20- item version, and the 10-item version by Andresen et al., in women living in the Mixteca, a poor rural area which includes part of the states of Guerrero, Oaxaca and Puebla, in southern Mexico. The instrument was applied as part of a comprehensive survey on health and migration in three rural municipalities in the Mixteca. The sampling design included cluster, proportional to size sampling of localities, and systematic selection of households. At each household, one woman of between 15 and 49 years of age responded a questionnaire which included the CES-D. A total of 468 women were included in the sample (median 35 years, interquartile range 28, 42). Of these, 89% were married or had a stable partner, 5% were single, 13% separated, and 12% widowed. The majority (65%) had only six years of schooling, while 1 6% had no formal education. The statistical analysis was conducted on the 343 questionnaires with complete answers to the CES-D (73% of the sample). The mean score in CES-D-20 was 11.3 (standard deviation 8.8). The mean score in CES-D-10 was 6.3 (standard deviation 5.0). According to the respective cut-off points, prevalence of depressive symptoms was 24.5% for the CES-D-20 and 22.3% for CES-D-1 0. A descriptive statistical analysis of the scores in each item and in the complete scales was conducted. In order to evaluate the internal consistency of CES-D, both 1 0- and 20- item versions, inter-item and item-total correlations were calculated. Cronbach's alpha coefficient was also obtained. Factor analysis was employed to determine if the actual aggregation of the items was coherent with the theoretical dimensions they were intended to measure. Another way to prove validity was through the analysis of the association between the score in the CES-D and the answers to questions about <>, an ethnical syndrome well recognized in the region and sharing characteristics with depression. Also, the association of scores with other variables known to be related to depressive symptoms, such as being chronically ill or the educational level, was investigated. In order to evaluate CES-D-10 capacity to identify depressive symptoms, taking the CES-D-20 as reference, Spearman's correlation coefficient between the scores in both scales was calculated. The kappa statistic was employed to evaluate the concordance between scales in the classification of individuals according to their respective cut-off points. For CES-D-20, Cronbach's alpha value was 0.84 and for CES-D-10 it was 0.74. For CES-D-20, four factors with eigen values over 1 were extracted, accounting for 50.6% of variance. The first one included items which, according to Radloff's original solution, are part of the dimensions depressed affect, retarded activity, and positive affect. The second one included items from the depressed affect, retarded activity, and interpersonal dimensions. The fourth factor included only two items, both from the positive affect dimension. A scree plot showed that a two factor solution could also be adequate. For CES-D-10, two factors were extracted, accounting for 46.5% of variance. As for convergent validity, women who reported having <> had a median CES-D-20 score of 13.5 (IQR 8.8, 22), while those who did not report the illness had a median score of 9 (IQR 4, 14). Women without a formal education had a median CES-D-20 score of 1 2 (IQR 8, 20), those who had completed elementary school had a median score of 10 (IQR 5, 15), and those with junior high or over had a median of 8 (IQR 3, 15). Those who reported having a chronic illness had a median score of 12 (IQR 8, 18), while those without a chronic illness had a median of 8 (IQR 4, 13). Similar results were observed for the CES-D-10. Spearman's correlation coefficient between CES-D-20 and CES-D-10 was 0.94 (p<.0001). Kappa value for concordance between both versions of the scale was of 0.80. In comparison to the longer version, CES-D-10 had a sensibility of 79.8% and a specificity of 97.3% for the detection of those over cut-off point. The results show that both scales had good reliability and validity in relation to measures of other variables related to depressive symptoms. The factorial grouping of the items was different from the original, as has been observed by other authors. A similar, unimodal distribution centered in 0 and with a positive skew was observed for the answers to all items, except for two items with a bimodal distribution. Those two items were also different to the rest in their presence (having the symptom at least on day during the past week) and persistence (having the symptom everyday during the past week). These differences suggest that items 4 and 8 of the CES-D could have validity problems in this population. In conclusion, both versions of the CES-D were found to have good psychometric properties in this sample, with the shorter one having the advantage of being easier to include in questionnaires for more comprehensive studies. However, further studies with the use of qualitative methods should clarify the true cross-cultural validity of the CES-D in rural areas in Mexico.

RESUMEN

Los síntomas depresivos son un problema de salud mental frecuente e importante en cuanto a sus consecuencias personales y sociales, que afecta no solamente a la población urbana de los países más desarrollados, sino también a los habitantes de zonas rurales en los países pobres. Para obtener información confiable acerca de la frecuencia de síntomas depresivos, así como de sus factores de riesgo o el éxito de las medidas preventivas y de atención, es necesario contar con instrumentos de medición confiables y válidos. El instrumento Center for Epidemiological Studies - Depression (CES- D), de Radloff, fue desarrollado originalmente para el estudio de síntomas depresivos en población abierta. Si bien no es útil para evaluar la presencia de trastornos depresivos del estado de ánimo tal como son definidos en la nosología psiquiátrica, este instrumento permite estudiar la de un rango de manifestaciones basadas en dimensiones de la depresión consideradas en la bibliografía clínica. El objetivo principal de este estudio fue evaluar las propiedades psicométricas del CESD en mujeres de una zona rural de alta marginación, tanto en su versión original como en la versión de 1 0 reactivos de Andresen et al. La información para este estudio se recabó en la zona mixteca, en tres municipios rurales en los cuales se llevó a cabo un muestreo por conglomerados de localidades y sistemático de hogares. Se aplicó el CES-D a 468 mujeres de entre 1 5 y 49 años en los hogares seleccionados. Se hizo un análisis descriptivo de los resultados de puntuación en cada reactivo, así como de los de la escala completa. Para evaluar la consistencia interna del CES-D en sus versiones de 10 y 20 preguntas, se calcularon las correlaciones entre reactivos, y de cada reactivo con la puntuación en la escala completa, así como el coeficiente de alfa de Cronbach. Se llevó a cabo un análisis factorial con el fin de determinar si la agrupación de los reactivos correspondía a sus dimensiones teóricas. Otra medida de la validez de constructo consistió en analizar la relación entre la puntuación en el CES-D y los resultados en preguntas acerca de los <>, un padecimiento reconocido en la tradición étnica de la región. Se observó también la asociación con la presencia de enfermedades crónicas y con el nivel educativo, dos variables que han mostrado estar asociadas a los síntomas depresivos. Para evaluar la capacidad del CES-D-10 de medir los síntomas depresivos, en comparación con la versión de 20 reactivos, se calculó el coeficiente de correlación de Spearman entre las puntuaciones en ambas escalas. Se calculó también el estadístico kappa para evaluar la concordancia entre las versiones larga y corta en la clasificación de individuos por encima del punto de corte. El valor de alfa de Cronbach del CES-D-20 fue de 0.84 y el del CES-D-1 0 de 0.74. Para el CES-D-20, se extrajeron cuatro factores con valores propios mayores a 1, que explicaron en conjunto 50.6% de la varianza. El gráfico de sedimentación mostró que una solución en dos factores también hubiera sido adecuada. Para el CES-D-1 0, se extrajeron dos factores que explicaron en conjunto 46.5% de la varianza. La correlación de Spearman entre el CES-D-20 y el CES-D-10 fue de 0.94 (p<.0001). El valor del estadístico kappa para la concordancia entre ambas formas de la escala fue de 0.80. Se observó un comportamiento similar en la distribución de las respuestas a todos los reactivos, a excepción de dos, los cuales difirieron también en su presencia y persistencia en comparación con el resto, por lo que se sugiere que estos dos reactivos podrían estar presentando problemas de validez.