The Psychometric Properties of the EuroQol 5D Five Level in Survivors of Critical Illness

Objectives: The EuroQol 5D five level (EQ-5D-5L) instrument is a standardized measure of health-related quality of life and is routinely used in survivors of critical illness. However, information on its psychometric properties and minimal clinically important difference (MCID) in this patient group is lacking.

Design: Secondary analysis of data from the previously published PREDICT (a registry in critically ill patients to determine predictors of disability-free survival) study, a prospective, multicenter cohort study.

Setting: Six ICUs in the state of Victoria, Australia.

Patients: Four hundred fifty adult patients admitted to the ICU and ventilated for over 24 hours.

Interventions: None.

Measurements and main results: The EQ-5D-5L was administered by telephone at 6 months following ICU admission. Internal consistency (inter-item correlations, Cronbach's α, and split-half reliability coefficients), construct validity (against age, body mass index, and other outcome measures), responsiveness (observing change over time and effect sizes), percentage of participants presenting no change, and MCID (triangulation of distribution-based and anchor-based estimates) were evaluated. The EQ-5D-5L showed high internal consistency, Cronbach α coefficients of 0.82 (between dimensions) and 0.79 (between the EuroQol-Visual Analogue Scale [EQ-VAS] and utility score), and average split-half coefficients of 0.79 each (between dimensions and between EQ-VAS and utility score). Construct validity was confirmed with a strong correlation between the EQ-5D-5L and the World Health Organization Disability Assessment Schedule 2.0 (EQ-VAS: r = 0.72; p < 0.001 and utility score: r = 0.81; p < 0.001). Effect sizes for change over time for EQ-VAS and utility score were low. The final MCID estimates were 10 (EQ-VAS) and 0.11 (utility score).

Conclusions: The EQ-5D-5L, using the Australian value set, demonstrated evidence of good internal consistency and validity, but poor responsiveness in a critically ill population.