Cultural adaptation and predictive modelling of the eating disorder examination questionnaire (EDE-Q) using hybrid statistical approaches among Bengalee female adolescents

The abnormal eating and related lifestyle attitudes are termed as eating disorders (ED), which often arise among adolescent females and become the third most common chronic illness. The Eating Disorder Examination Questionnaire (EDE-Q) is a widely used screening tool for ED, but has limited cultural validation, which could be a linguistic barrier across diverse populations. Appropriate translation and cultural adaptation are crucial for accurately estimating the prevalence of ED. Therefore, the study aimed to evaluate the dimension structure and appropriateness of the EDE-Q items in the Bengali version, assess its validity and reliability, and identify the important items contributing to ED risk. A cross-sectional school-based study was conducted among 472 Bengalee adolescent female students aged 13–19 years from both rural and urban areas of North 24 Parganas district of West Bengal. EDE-Q was translated into Bengali language following standard forward translation procedure and administered to participants. Incomplete responses were excluded from the analysis. ED risk was defined using the global EDE-Q score based on established cut-off criteria. Principal component analysis (PCA) was performed to examine dimensionality and ensure cultural adaptation of the items. The reliability and validity of the questionnaire were assessed to ensure its consistency and accuracy. Artificial Neural Network (ANN) modeling was used to evaluate the relative contribution of each dimension and item in predicting ED risk and model performance was assessed using metrics such as sensitivity and specificity. PCA produces three dimensions (subscales) named as ‘Cognitive-preoccupation and self-evaluation’, ‘Restrained-eating attitudes’, ‘Weight and shape concerns’, comprising all twenty-two items. The Bengali version has a high internal consistency score of 0.941, along with 0.868, 0.898, and 0.871, respectively, for the three dimensions. The items also have an acceptable to good convergent validity. ANN reveals that the ‘Restrained-eating attitudes’ dimension contributed most significantly to the prediction of ED risk with acceptable sensitivity and specificity. The Bengali EDE-Q demonstrates a three-dimension structure encompassing all items, supporting their relevance in predicting ED risk among Bengalee adolescent females. This hybrid analytical approach highlights the novel contribution of the study in integrating statistical validation with predictive modelling, which could improve the practical utility of the tool. It also provides a foundation for future translation into other Indian languages.