Direct Mapping Between Oswestry Disability Index Scores and EQ-5D-3L Utilities: The Need for Utility Scores to Perform Cost-Utility Analysis.

Abstract

Health-related quality of life is most often measured by using disease-specific questionnaires rather than using preference-based generic questionnaires, which are necessary for cost-utility analysis. The availability of algorithms to map the score of a specific questionnaire to the utilities of a generic one would solve this problem. Our objective was to develop an algorithm to directly map Oswestry Disability Index (ODI) scores into EuroQoL five-dimension-three level (EQ-5D-3L) utilities in Spanish patients with low back pain. The sample of this retrospective cross-sectional study comprised 592 pairs of questionnaires from 296 patients. The three functional forms evaluated were ordinary least squares, Tobit, and adjusted limited dependent variable mixture models. The explored specifications consisted of all possible combinations of either total ODI score or individual ODI item score and the covariates of age, gender, and pain intensity, including squared and cubic terms for quantitative covariates and interaction terms with gender. A five-fold cross-validation technique repeated 10 times was used. The best model was selected according to the best goodness of fit (mainly higher adjusted R2) and accuracy (mainly lower root mean squared error). Total ODI scores and EQ-5D-3L utilities showed an excellent correlation (Spearman rank coefficient = -0.910). The final model selected used ordinary least squares as a functional form and was specified by total ODI scores and their squared and cubic terms as covariates. It accurately (root mean squared error = 0.1125) predicted EQ-5D-3L utilities and explained 90.56% of variance. EQ-5D-3L utilities can be accurately predicted from ODI total scores in patients with low back pain.