Psychometric Properties of EQ-5D-3L and EQ-5D-5L in Cognitively Impaired Patients Living with Dementia
Background: Assessing health-related quality of life in dementia poses challenges due to patients’ cognitive impairment. It is unknown if the newly introduced EQ-5D five-level version (EQ-5D-5L) is superior to the 3-level version (EQ-5D-3L) in this cognitively impaired population group. Objective: To assess the psychometric properties of the EQ-5D-5L in comparison to the EQ-5D-3L in patients living with dementia (PwD). Methods: The EQ-5D-3L and EQ-5D-5L were assessed via interviews with n = 78 PwD at baseline and three and six months after, resulting in 131 assessments. The EQ-5D-3L and EQ-5D-5L were evaluated in terms of acceptability, agreement, ceiling effects, redistribution properties and inconsistency, informativity as well as convergent and discriminative validity. Results: Mean index scores were higher for the EQ-5D-5L than the EQ-5D-3L (0.70 versus 0.64). Missing values occurred more frequently in the EQ-5D-5L than the EQ-5D-3L (8%versus 3%). Agreement between both measures was acceptable but poor in PwD with moderate to severe cognitive impairment. The index value’s relative ceiling effect decreased from EQ-5D-3L to EQ-5D-5L by 17%. Inconsistency was moderate to high (13%). Absolute and relative informativity increased in the EQ-5D-5L compared to the 3L. The EQ-5D-5L demonstrated a lower discriminative ability and convergent validity, especially in PwD with moderate to severe cognitive deficits. Conclusion: The EQ-5D-5L was inferior as a self-rating instrument due to low acceptability and discriminative ability and a high inconsistency, especially in moderate to severe dementia. The EQ-5D-3L had better psychometric properties and should preferably be used as a self-rating instrument in economic evaluations in dementia.