Using the EuroQol 5-D in the Catalan General Population: Feasibility and Construct Validity

Spanish and Catalan versions of the EuroQol 5-D (EQ-5D) were included in the Catalan Health Interview Survey (CHIS) and administered to a randomly selected cross-section of 12,245 individuals from the Catalan general population. This paper analyses the feasibility, convergent validity and construct validity of three parts of the EQ-5D (the descriptive system, the visual analogue scale (VAS) and the Spanish tariff) using the results obtained in the CHIS. The feasibility was assessed by the number of missing responses. The convergent validity was based on the correlations between the EQ-5D scores and the scores on the General Health Questionnaire (GHQ) and on an index of self-perceived overall health. The construct validity was assessed by analysing the degree to which lower scores on the EQ-5D correlated positively with increasing age, being female, being in a lower social class or having a lower level of education and with increasing levels of disability, co-morbidity, restricted activity, mental health problems and poor self-perceived health. A low number of missing responses on the descriptive system and the VAS (1.5%) indicated a high level of acceptance. A marked ceiling effect was found, with 67% of the sample reporting no problem in any EQ dimension. The convergent validity with the GHQ was generally low, though moderate on the mood dimension. Self-perceived overall health correlated moderately to strongly with the mean VAS and tariff values. The positive correlations between lower scores on all three elements of the EQ-5D and increasing age, increasing levels of disability, co-morbidity, restricted activity, mental health problems and poor self-perceived health provide some evidence of the instrument's construct validity, as does the fact that women reported more problems than men. Multivariate analyses using the VAS and tariff values as dependent variables and all of the sociodemographic and health variables as independent variables reached R2 values of 0.45 and 0.81, respectively. The Spanish and Catalan versions of the EQ-5D have proved to be feasible and valid for use in health interview surveys.