Comparing SF-6D and EQ-5D utilities across groups differing in health status

Objectives  To compare the EQ-5D and SF-6D within socio-demographic and clinical groups in a representative sample (n = 1,005) of the Greek general population and to examine mean utility differences across groups differing in health in this population and in a highly morbid disease sample (diabetes, n = 215). Methods  Association and level of agreement between instruments were estimated with Pearson’s r and the intraclass correlation coefficient (ICC), respectively. Paired-samples t-test was used to identify significant score differences, which were regarded as minimally important differences (MID) when they exceeded 0.03. The EQ-VAS was used to classify individuals into health status groups, covering the range from very poor to very good health, and the same classification was used in the diabetes sample. Results  EQ-5D and SF-6D were in agreement and strongly correlated over the entire sample (ICC = 0.536, P < 0.001 and r = 0.662, P < 0.001), but correlation varied according to socio-demographic factors and clinical conditions. In healthier responders, EQ-5D scores were significantly higher than SF-6D scores (P < 0.001) and differences constituted MIDs. Contrarily, in individuals with clinical conditions, SF-6D scores were predominantly higher than EQ-5D. The pattern of results was replicated in the disease sample as well. Conclusions  The hypotheses that EQ-5D generates higher scores in healthier populations and the SF-6D in less healthier groups were confirmed. Based on the evidence provided here, EQ-5D and SF-6D measuring discrepancies generate utility differences across VAS-based health groups, which warrant further within-sample investigation.