Responsiveness of the anxiety/depression dimension of the 3- and 5-level versions of the EQ-5D in assessing mental health

Katelynn Crick, Fatima Al Sayah, Arto Ohinmaa, Jeffrey A. Johnson

Anxiety and depression disorders are associated with significantly lower health-related quality of life (HRQL). The EQ-5D is a commonly used generic measure of HRQL; it captures mental health through a single domain-the anxiety/depression dimension. Evidence on the responsiveness of this measure in assessing changes in mental health changes is limited. To examine the performance of the anxiety/depression dimension (A/D) of the 3- and 5-level (3L and 5L) versions of the EQ-5D in assessing changes in mental health. Data from two patient populations were used: 495 adults post-discharge from general internal medicine ward (EQ-5D-3L), and 225 type 2 diabetes patients who screened positive for depressive symptoms (EQ-5D-5L). Anchor-based approach along with effect sizes (ES) and ROC analysis was used. Anchors included patient health questionnaire 9-items "PHQ9" and generalized anxiety disorder 2-item questionnaire "GAD2" for EQ-5D-3L, and PHQ9 and SF-12 mental composite summary scores (MCS) for EQ-5D-5L. A/D change was quantified as the difference between follow-up and baseline levels. The A/D dimension of the EQ-5D-3L showed limited responsiveness to changes in depressive symptoms measured by PHQ9 and for anxiety symptoms measured by GAD2, whereby in those who improved or deteriorated in either symptom, more than half of the patients did not have an A/D change. In the ROC analysis, the A/D dimension of the EQ-5D-3L showed weak performance with C-indices ranging from 0.58 to 0.63 and probability of detection of depressive or anxiety symptoms ranging between 20 and 40%, which are all well below acceptable ranges. Similar results were observed for the A/D dimension of the EQ-5D-5L; although the performance was slightly better, it was still below acceptable range. In patients who improved or deteriorated based on the PHQ9 or MCS, around a third had no changes on the A/D dimension. The performance of the A/D dimension of the EQ-5D-5L was also very limited with C-indices ranging between 0.67 and 0.76, and probability of detection between 50 and 67%, slightly better than that of the 3L, yet unsatisfactory. Although the A/D of both EQ-5D-3L and 5L was limited in capturing changes in mental health in these populations, the 5L was slightly more responsive than the 3L. While the performance was better for depressive than anxiety symptoms, it varied by the direction of change. Further research using other measures of mental health in other populations is warranted.