Measurement properties of the Michigan hand outcomes questionnaire: Rasch analysis of responses from a traumatic hand injury population

PURPOSE: This study aimed to use Rasch analysis to test the content, scoring, and measurement properties of the Michigan Hand Outcomes Questionnaire (MHQ). METHODS: MHQ scores from 196 patients with hand and wrist conditions were collected in an outpatient hand rehabilitation facility. Rasch analysis was conducted to assess the fit statistics of MHQ to confirm the scaling structure of disability subscales, and to identify differential item functioning. RESULTS: The MHQ did not fit with the Rasch model (χ² = 2376.78, df = 74, p < 0.001), and most thresholds of item responses were disordered. The original scoring algorithm derived from 5-point Likert response options was adjusted to 3-point Likert (10 items) and 4-point Likert (11 items) based on the visual inspection of the thresholds map. Differential item functioning was present in the revised scale based on the age, sex, and dominant hand. Only 3 revised subscales of the MHQ including activities daily living (one hand), aesthetics, and satisfaction showed acceptable fit to the Rasch model. Unidimensionality was achieved in all revised subscales. CONCLUSIONS: The overall MHQ had a substantial misfit from the Rasch model. Despite efforts of item reduction and rescoring, we did not reach a satisfactory solution. This calls into question the validity of the statistical evaluations performed on this scale using the traditional scoring.Implications for rehabilitationThe MHQ was designed to measure different dimensions of pain and disability but demonstrates multiple measurement problems that undermine it use in present form.It is not appropriate to sum all 37 items of the MHQ into a single score.Three subscales of activities daily living (one hand), aesthetics, and satisfaction can provide unidimensional subscales scores with interval level scaling if scored with our proposed Rasch-based revised scoring.The 27-item version of the MHQ is shown to have strong psychometric properties for administration with patients with hand injuries; however, it requires further validation.