A new model for category-scaling data with an application to the development of health-status measures

Category rating is used to assess patient, family-member, provider, and societal preferences for health outcomes. Often, it is of interest to compare ratings obtained from different groups. Standard methods for making comparisons, such as regressions, correlations, and multiple t-tests, do not account for the dependency among ratings. The authors propose a new model for category ratings that does consider their relative dependency. This model provides a profile of ratings for a single group and facilitates comparisons across groups. It was applied to category ratings for four levels of Bending and Lifting function as defined by the recently developed Functional Capacity Index (FCI). Differences in ratings were observed across groups with different personal experiences of functional limitations and across groups with different degrees of clinical knowledge. These differences were not observed when standard methods were used. Thus, ignoring the relative nature of category ratings can lead to different conclusions about group preferences for health outcomes. When the ratings are being used to scale a health-status measure, this discrepancy has implications for the application of the measure in resource allocation.