British Journal of Rheumatology 1996;35:574-578 ARE WE MAKING THE MOST OF THE STANFORD HEALTH ASSESSMENT QUESTIONNAIRE?

For many years, the Stanford Health Assessment Questionnaire (HAQ) has provided an effective measure of disability. Recently, some debate has emerged about whether or not the HAQ is an 'ordinal ' or 'interval ' scale. The opportunity to test its level of measurement arose when the scale was applied in a community survey which undertook a two-stage random sample using postal questionnaires to ascertain the health care needs of those with arthritis. The HAQ data are fitted to the Rasch model which tests for the presence of certain desirable characteristics of measurement, e.g. unidimensionality. The fit of the data to the model for those self-reporting rheumatoid arthritis (RA) was adequate. The transformed HAQ score, derived from the Rasch analysis, is compared with the ordinary HAQ (raw) score. This shows that, for those with RA, incremental units of the raw score at the margins of the scale reflect an increasing level of (dis)ability compared to similar units in the centre of the scale. Thus, the traditional HAQ score (range 0-3) is an ordinal score. The findings also indicate that scoring all 20 items may lead to greater sensitivity. Questions are also raised about the construct validity for those with other types of arthritis. For osteoarthrosis, the grip item does not appear to belong to the same underlying construct as the other items