Abstract: Objective To reevaluate the Preschool Children Eating Behavior Questionnaire by using multidimensional item response theory(MIRT),and to explore the feasibility and effectiveness of MIRT in the development of medical scale,in order to make Preschool Children Eating Behavior Questionnaire more completed and concise,thereby providing basic tools for the assessment of children's eating behavior and related research in China. Methods On the basis of extensive review of relevant literature,data were from the results of the Preschool Children Eating Behavior Questionnaire from children aged 3 to 6 in Xi'an city in June 2017. The MIRT were used to evaluate and select the items,such as degree of differentiation,difficulty and guessing coefficient,etc. Reliability and validity of the questionnaire were evaluated by result of Cronbach's α coefficient,split-half reliability coefficient,retest reliability coefficient and factor analysis. And the results were compared with those of the original scale. Results Three items were deleted according to the differentiation degree and difficulty under MIRT,and the dimension of one item was adjusted. The Cronbach’s α coefficient of revised questionnaire was 0.94,the Guttman split-half reliability was 0.81,and the test-retest reliability was 0.73. Exploratory factor analysis showed that the cumulative variance contribution rate was 59.04%,and confirmatory factor analysis indicated that the values of NFI,NNFI,CFI,RMSR were 0.86,0.92,0.93,and 0.04,respectively. There were significant differences on the scores of satiety responsiveness,food responsiveness,food fussiness,and eating habit dimensions among different weight groups. Finally,the reliability and validity of the revised scale were better than the original one. Conclusions The item screening method based on MIRT could improve the reliability and validity of the scale,making the scale more concise. The newly revised Preschool Children's Eating Behavior Questionnaire is more appropriate for evaluation and study of children's eating behavior problems.

Key words: preschool children, eating behavior, questionnaire, multidimensional item response theory

摘要： 目的 采用多维项目反应理论(MIRT)理论对“学龄前儿童饮食行为量表”进行进一步的条目评估,探索MIRT理论在医学量表编制中的可行性和有效性,编制出更加完善精简的“学龄前儿童饮食行为量表”,为我国儿童饮食行为评价和相关研究提供基本工具。方法 广泛查阅国内外相关文献,2017年6月以西安市城、郊3～6岁儿童“学龄前儿童饮食行为量表”测试结果为基础,采用MIRT下的条目区分度、难度和猜测系数对各条目进行评价和筛选,采用Cronbach’s α系数、分半信度系数、重测信度系数以及因子分析等对修订后的量表的信度、效度、反应度进行再检验,并与原量表进行对比。结果 依据MIRT下的条目区分度、难度和维度载荷共删除3个条目,并调整了1个条目的维度归属。修订的量表同质信度Cronbach’s α系数为0.94,Guttman分半信度系数为0.81,重测信度系数为0.73,探索性因子分析显示累积方差贡献率为59.04%,验证性因子分析显示范拟合指数(NFI)=0.86,非范拟合指数(NNFI)=0.92,比较拟合指数(CFI)＝0.93,均方根残差(RMSR)=0.04。不同体重分集在过饱响应、食物响应、挑食、不良进食习惯维度得分上差异均有统计学意义(P<0.05)。修订后量表的信度、效度和区分度各类指标中大部分都优于原量表。结论 MIRT理论下的条目筛选方法可进一步调高量表的信度和效度,使量表更加精简,新修订的学龄前儿童饮食行为量表更适合于儿童饮食行为问题的评估和研究。

关键词: 学龄前儿童, 饮食行为, 量表, 多维项目反应理论