PUB - Publikationen an der Universität Bielefeld

**Background**
There is evidence to suggest that the proportion of missing values is slightly higher in the older population resulting in lower completion rates of the EQ-5D. However, existing studies rarely provide a within-sample comparison of feasibility properties across age groups to quantify this difference. Hence, this study examines feasibility properties of the EQ-5D-3L and 5L in the general population and explores the impact of age on the completion of EQ-5D instruments. **Methods**
We pool five waves from the English GP Patient Survey, where respondents self-report their health in either EQ-5D-3L or 5L. Descriptive analysis was undertaken to analyse the distribution and proportion of missing values and completion rates stratified by age and EQ-5D version; logistic regression models were specified to quantify the impact of age, gender and potential long-term conditions on the completion of each of the EQ-5D instruments. **Results**
The total sample comprises ~ 4.36 million observations, of which 2.88 million respondents report their health in 5L and 1.47 million in 3L, respectively. Respondents over 64 years have slightly more missing values in each dimension than younger respondents. The highest share was observed for the oldest age group in the dimension anxiety/depression (3L 9.1% vs. 5L 7.6%), but was otherwise below 5%. Consequently, completion rates (observed and predicted) decreased with older age and at a higher rate after the age of 64; this was more pronounced for the 3L. **Conclusion**
Evidence from our study suggests that both the EQ-5D-3L and 5L have good feasibility properties. In comparison to younger populations there appears to be a higher proportion of respondents with incomplete responses beyond the age of 64 years. Overall, the 5L version compares more favourably in terms of missing values, completion rates as well as with regard to the expected probability of an incomplete descriptive system.