Item Response Theory-based validation of a short form of the Disordered Eating Attitude Scale (DEAS-s) to a Brazilian sample

This study aimed to validate a short version of the Disordered Eating Attitude Scale (DEAS-s). To this end, 2,902 adult individuals answered the original DEAS and informed age, weight, and height. Data were analyzed using the full-information factor analysis and Item Response Theory (IRT) analysis. Exclusion criteria retained items with adequate values of commonality and factor loadings. Estimation of IRT parameters, the Item Characteristic Curve (ICC), and test information guided the selection of the best quality items. The final model adjustment was evaluated using Root Mean Square Error of Approximation (RMSEA), Standardized Root Mean Square Residual (SRMSR), Comparative Fit Index (CFI), and Tucker-Lewis Index (TLI). The eating attitudes on each level of DEAS-s were described. The analyses were performed on R software and Microsoft Excel version 2013. As results, six items were excluded because of the low communalities and factor loadings, and one more was excluded because of an overlapping on the ICC. The remaining 17 items explained 0.53 of the total variance and had an adequate goodness-of-fit (RMSEA = 0.05; SRMSR = 0.05; CFI = 0.98; TLI = 0.98). The information test is more accurate between the scores 0 and +3. Scores higher or equal to 1.5 identified individuals with disordered eating attitudes. Women, older individuals, and those with a higher body mass index presented more disordered eating; thus, the one-dimensional and short version of DEAS showed a suitable adjustment and may contribute to properly evaluate disordered eating in diverse populations.

本研究旨在对饮食失调态度量表简版（Disordered Eating Attitude Scale, DEAS-s）进行效度验证。为此，共招募2902名成年受试者填写原版饮食失调态度量表，并报告自身年龄、体重与身高信息。本研究采用全信息因子分析（full-information factor analysis）与项目反应理论（Item Response Theory, IRT）对数据进行分析，以条目共同性与因子载荷达标作为剔除标准，结合项目反应理论参数估计、项目特征曲线（Item Characteristic Curve, ICC）与测验信息函数完成高质量条目筛选。最终模型的拟合效果通过近似误差均方根（Root Mean Square Error of Approximation, RMSEA）、标准化残差均方根（Standardized Root Mean Square Residual, SRMSR）、比较拟合指数（Comparative Fit Index, CFI）与塔克-刘易斯指数（Tucker-Lewis Index, TLI）进行评估，并对DEAS-s各得分水平对应的饮食态度特征进行描述。所有分析均通过R软件与Microsoft Excel 2013完成。结果显示，因共同性与因子载荷偏低，共有6个条目被剔除；另有1个条目因项目特征曲线重叠被排除。最终保留的17个条目可解释总方差的53%，且模型拟合效果良好（RMSEA=0.05；SRMSR=0.05；CFI=0.98；TLI=0.98）。该简版量表的测验信息函数在得分0至3区间内准确度最高，得分大于等于1.5即可筛查出存在饮食失调态度的个体。分析显示，女性、年长群体以及体重指数较高的个体表现出更显著的饮食失调态度。综上，该单维简版DEAS-s具备良好的模型拟合性能，可用于不同人群的饮食失调态度评估。