基于项目反应理论的传染病健康素养问卷分析

Questionnaire on health literacy of infectious diseases based on item response theory

目的应用项目反应理论对《北京市居民传染病健康素养调查问卷》进行条目分析。方法数据来源于北京市海淀区居民传染病健康素养调查,调查问卷由传染病相关基本知识和理念、生活方式与行为、技能3个维度组成。采用三参数Logistic模型拟合样本,估计各条目的区分度参数、难度参数、猜测系数。结果调查问卷各维度均满足单维性假设,可以应用项目反应理论进行条目分析。3个维度中分别有70.0%、90.9%、0%的条目区分度参数处在(0.5,2)的合理取值范围内,剩余条目区分度参数偏大;所有条目的难度参数均在(-3,3)的合理取值范围内;3个维度中分别有70.0%、72.7%、100%的条目猜测系数处在(0,0.2)的合理取值范围内,剩余条目的猜测系数中最高者为0.352。结论调查问卷绝大多数条目区分度合理,少数条目区分度偏高,以技能维度最为严重;所有条目均难度合理,整体呈中等偏低水平;猜测行为仅存在于知识、行为维度的少数条目,对问卷整体答题情况的影响并不严重。

Objective To analyze the items of the questionnaire on health literacy regarding infectious diseases in residents in Beijing＇using item response theory（IRT）.Methods Data was collected from the survey on health literacy regarding infectious diseases of residents who lived in Haidian district,Beijing.The three-dimensional questionnaire identified conceptual knowledge and attitudes towards infectious diseases;lifestyle and health behavior;and health-related skills.Three-parameter logistic model was used to fit the sample and to calculate the discrimination,difficulty and guessing parameter of each item.Results Since each dimension of the questionnaire satisfied the unidimensional assumption,IRT could be applied to execute item analysis.Among the three dimensions,there were 70.0%,90.9% and 0% of the item discrimination parameters lying in the rational domain at（0.5,2）respectively,while rest of them had larger discrimination values.All the item difficulty parameters were within the reasonable range of（-3,3）.The reasonable value interval of guessing parameter was（0,0.2）.For different dimensions,there were 70.0%,72.7%and 100%item guessing parameter fit in this range.The maximum value of the guessing parameter was 0.352.Conclusion Most of the items in the questionnaire hadacceptable discriminations.Only few items had excessive discrimination levels and mostly existed in the health-related skills dimension.The difficulties of all the items were in a medium to slightly lower level,which was reasonable.Guessing behavior only happened to small number of items in the first two dimensions and had trivial effect on the overall survey.