Comparing the psychometric properties of the EQ-5D-5L between mental and somatic chronic patients populations

Abstract

The validity and reliability of the EQ-5D-5L in comparison with the standard 3L has been tested through the analysis of psychometric properties making use of different samples of patients. However, it is likely that the condition of the illness may affect the power of the 5L version with respect to the 3L one. Here we report on parallel testing of EQ-5D-5L and 3L administered to a sample of chronic patients of both somatic and mental illness. The aim of this study is to check some psychometric properties in both subsamples. Methods: We check for the usual psychometric properties: feasibility, (in)consistency, ordinality (and transitivity), informativity, face validity and convergent validity. Also, we perform new analysis for checking transitivity and the Cronbach-? for convergent validity. Finally, we proposed a complementary way for looking at the property of informativity through three different indexes (effective, absolute and overall) based on the statistical discriminatory power. Data: We have a total of 1002 questionnaires finally collected. 444 (46.25%) chronic mental patients, 516 (53.75%) have somatic chronic illnesses; 42 observations of unknown origin of the illness have been dropped to perform this analysis. Results: The mean value reported in the VAS for the full sample is 60.93. Somatic patients report a mean of 64.42 points in this scale and mental patients report 56.83 points in the VAS. Analyzing the distribution of the responses to problems on each dimension we found, for all cases, a highly skewed distribution. Moreover, the distribution of responses changes significantly between subsamples, as expected. In all dimensions, it seems that somatic patients take more advantage of the extra levels introduced by the EQ-5D-5L. This group reduces to a greater extent the missing response rate, commit less (and of lower importance) inconsistencies, get a higher correlation of the 5L scale and the VAS within the 3L levels, complements better the dimensions to get an overall score (measured through the Cronbach’s alpha), reduces in a more significant way the “no problem” response and the Informativity gain is also superior (for both the Shannon Evenness Index and our Absolute Index). This higher performance of the EQ-5D-5L on somatic patients is endorsed by a higher preference of somatic patients toward the 5L version of the questionnaire than to the 3L one. Conclusion: Results show the suitability of the 5L version in both subsamples, but it is much more effective for somatic patients. These subsamples’ differences may be of concern when aggregating and comparing different data.