Development of United States population-based preference weights for the EQ-5D health states

Abstract

The EQ-5D is a brief, multi-attribute, preference-based health status measure. This dissertation describes the development of a statistical model for estimating U.S. population-based preference weights for the EQ-5D health states. A multistage probability sample was selected from the adult U.S. population. Using the time trade-off (TTO) method, each respondent valued 13 of the 243 health states described by the EQ-5D. The valuations were linearly transformed to lie on the interval [-1, 1]. Numerous model specifications were investigated, and a modified split-sample approach was used to evaluate the predictive accuracy of the models. All statistical analyses took into account the clustering and disproportionate selection probabilities inherent in our sampling design. The best model proved to be one based on a conceptual notion of the effect of movements away from perfect health. This model, which we have named D1, included ordinal terms to capture the effect of departures from perfect health as well as interaction effects due to increasing health problems. Relative to other models tested, a random effects specification of the D1 model provided a good fit for the observed TTO data. This model yielded an overall R² of 0.38, a mean absolute error of 0.02, and a correlation between mean observed and predicted valuations of 0.99. We also examined differences in health state valuations among the three major racial/ethnic groups in the U.S., i.e., Hispanics, non-Hispanic blacks, and others. In general, non-Hispanic blacks valued health states more highly than Hispanics or non-Hispanic non-blacks. Non-Hispanic blacks appeared to perceive extreme health problems to be associated with less disutility than did members of the other racial/ethnic groups. Differences in valuations did not appear to be related to differences between groups in education, income, or self-reported chronic conditions. The D1 model predicts the values for observed health states with a high degree of accuracy. This model's predictions provide a set of EQ-5D preference weights specifically developed for use in the U.S. population. Within the U.S. population, there exist differences among the major racial/ethnic groups in the perceived desirability of the EQ-5D health states. These differences cannot be readily explained by socioeconomic disparities.