Using Rasch-models to compare the 30-, 20-, and 12-items version of the general health questionnaire taking four recoding schemes into account

Summary Background The present study compares the 30-, 20-, and 12-items versions of the General Health Questionnaire (GHQ) in the original coding and four different recoding schemes (Bimodal, Chronic, Modified Likert and a newly proposed Modified Chronic) with respect to their psychometric qualities. Methods The dichotomized versions (i.e. Bimodal, Chronic and Modified Chronic) were evaluated with the Rasch-Model and the polytomous original version and the Modified Likert version were evaluated with the Partial Credit Model. Results In general, the versions under consideration showed agreement with the model assumption. However, the recoded versions exhibited some deficits with respect to the Outfit index. Conclusions Because of the item deficits and for theoretical reasons we argue in favor of using the any of the three length versions with the original four-categorical coding scheme. Nevertheless, any of the versions appears apt for clinical use from a psychometric perspective.