Multiplicative rating scales do not enable measurement of vision-related quality of life

Purpose: Many questionnaires for the measurement of visual impairment exist. One, the Houston Vision Assessment Test (HVAT), takes a different approach: the patient is asked to rate overall impairment and the proportion attributed to vision, then through multiplication the visual and non-visual (physical) impairments are calculated. The purpose of this study was to determine whether the scores derived from this approach can be considered to be measures. Methods: The participants were 193 cataract patients awaiting surgery (mean age 74.1 +/- 9.8-years, 54 per cent female and 53.6 per cent were awaiting first eye surgery), who self-administered the HVAT, which consists of 10 questions, whereby impairment on each activity and the proportion attributable to vision is rated. Therefore, total, visual and physical impairments are calculated. For each question, multiplying the impairment (five response categories) by the proportion due to eyesight (five categories) gives 10 possible levels of visual impairment. Assessment of the multiplicative rating scales included frequency of category use and hierarchical ordering of response categories using category thresholds. Summary statistics of Rasch analysis were generated for the rating scale and overall questionnaire performance. Results: In the multiplicative scale, higher response categories were under-utilised and thresholds were disordered, indicating that the categories did not function as intended. Some of the dysfunction arose from disordered thresholds in the 'proportion due to eyesight scale', but repairing this gave little improvement to the multiplicative scale. The ill-defined nature of the disordered categories precluded further repair by combining categories. Measurement precision, as indicated by person separation reliability, was poor (0.70). Conclusion: Rasch analysis demonstrated that the categories of the multiplied rating scale of the HVAT were not ordered, as the user would expect; this precludes measurement. This provides evidence against the use of multiplicative rating scales in quality-of-life questionnaires. It would be better to use a single rating scale for each construct of interest.