Do errors in the GHQ-12 response options matter?

Background: The twelve item General Health Questionnaire (GHQ-12) is a widely used measure of psychological wellbeing. Because there are seven different sets of response options across the twelve items, there is scope for transcription errors to occur when researchers assemble their study materials. The impact of such errors might be more important if they occur in the first set of response options than if they occur later in the questionnaire, once participants have become aware that options to the right of the GHQ-12 response sets always indicate worse wellbeing. Aims: To test the impact of introducing errors into the first and eighth set of response options for the GHQ-12 that render those response sets partially illogical. Methods: We used a double-blind randomised controlled trial, pre-registered with Open Science Framework (osf.io/syhwf). Participants were recruited by a market research company from their existing panel of respondents in Great Britain. Participants were randomly allocated to receive one of three versions of the GHQ-12: a correct version (n = 500), a version with a mistake in the first item (n = 502), or a mistake in the eighth item (n = 502). Mistakes replaced ‘better than usual’ (item one) or ‘more so than usual’ (item eight) with ‘not at all.’ Results: We found no differences between the versions in terms of number of participants with possible poor psychological wellbeing (χ2 = 0.32, df = 2, p = 0.85) or in mean GHQ-12 scores for the three groups (F(2, 1501) = 0.26, p = 0.77). Conclusions: Small deviations from the standard GHQ-12 wording do not have a substantive impact on results.