Seemingly minor changes to a questionnaire can make a big difference to mean scores: a cautionary tale