The necessity of applying Rasch measurement theory in questionnaire-based public health research

Abstract

Abstract Background Questionnaires are frequently used in public health research. In order to provide valid and reliable results to generate recommendations for practice and policies, scales with sound psychometric properties are required. Classical test theory such as factor analysis is most frequently used to assess the psychometric properties of scales. However, classical test theory may have limitations in confirming the validity of scales. Only Rasch measurement theory meet the requirements of fundamental measurement, such as additivity, invariance, sufficiency and specific objectivity. The objective is to exemplify how Rasch measurement theory can be used to evaluate the psychometric properties of a scale. Validation of the Hopkins Symptom Checklist-10 is used as an example. Methods This study is based on cross-sectional data from the Youth Data Survey. In total, 6777 adolescents responded to a web-based questionnaire. Data collection was carried out in lower and upper secondary schools in Norway during 2018. The data were analysed using the partial credit parameterization of the unidimensional Rasch model. Results Preliminary results indicated that the scale had acceptable reliability (person separation index: 0.82). However, one pair of items shows response dependence. The targeting could have been better (mean person location: -1.445). All items had ordered thresholds. Three items under-discriminated. Several items displayed differential item functioning with regard to gender and school level. Conclusions Applying Rasch measurement theory measurement problems that would go undetected using classical test theory approaches were observed. Scales used in public health research should be thoroughly validated applying Rasch measurement theory before the data are used to support claims about public health and used to provide recommendations for policy and practice. Key messages Public health practice and policy should be based on information from valid and reliable scales. Rasch measurement theory should be used to evaluate psychometric properties of scales used in public health research.