A comparison of the EQ-5D-3L and EQ-5D-5L

Abstract

Introduction The EQ-5D-3L (3L) and EQ-5D-5L (5L) are both frequently used measures of health status. Previous studies have found the EQ-5D-5L to have superior measurement properties but no study has compared the two measures in a large general population survey using matched respondents. Methods Using data from the GP Patient Survey, coarsened exact matching was used to match individuals completing the 3L in 2011 with those completing the 5L in 2012. Measurement properties were assessed for a general population and multimorbid population (chronic conditions >= 2), with ceiling effects, informativity and distribution of response compared. Changes in the direction of response, as well as the impact on utility distributions, were quantified. Results Matching resulted in a cohort of 1,023,218 respondents (2011: 511,609; 2012: 511,609) for analysis. Ceiling effects for the 5L were lower than the 3L (43.8% vs. 54.4%). The 5L had improved informativity and broader spread of responses than the 3L (5L top 50 profiles: 77.4% vs. 3L: 98.8%). Overall, there was an upwards shift in utility values for the 5L versus the 3L as respondents using the 5L reported ill health more frequently but with less severity. Measurement improvements and effects on utility values were more pronounced for the multimorbid population. Conclusion The 5L had superior measurement properties than the 3L and should be preferred in general population surveys and for use in individuals with multimorbidity. At increasing levels of morbidity, the 5L is currently associated with higher utility values than the 3L.