SENSITIVITY OF THE EQ-5D FOR MEASURING MENTAL HEALTH IN AFRICAN-AMERICAN TEENAGERS

Purpose: A number of studies have shown the EQ-5D to be insensitive to mental health (MH) domains. It is unclear whether this insensitivity jeopardizes the derivation of utilities from non-preference based measures. In this study, we test for the sensitivity of the EQ-5D for measuring MH in African-American teenagers.

Method: The SF-36, EQ-5D, and Center for Epidemiologic Studies Depression Scale (CES-D) instruments were completed by 472 teenagers (age: mean=17.5, range=16.2-19.9) engaged in the Rural African American Families Health Program, an intervention focused primarily on substance abuse prevention. Five mapping studies were identified that provide algorithms for determining a utility value from SF-36 responses through the EQ-5D. Utilities based on EQ-5D responses were established using the US valuation. For the CES-D, high depressive symptoms were indicated by a score of 16 or higher. Comparisons were made between MH domains using the Spearman rho and within survey responses between genders using the Wilcoxon rank-sum test.

Result: The correlation between the MH domains of the SF-36 and EQ-5D was weak (rho=0.281). This is due in part to the 25% of respondents that reported high depressive symptoms on the CES-D and low SF-36 MH scores yet had no self-reported depression or anxiety on the EQ-5D. Women with high depressive symptoms on the CES-D were significantly more likely to report elevated levels of anxiety or depression (EQ-5D = 1 or 2) than their male counterparts (p=0.038). When these data were used to derive utilities, the mean absolute error (MAE) from mapping the SF-36 onto the EQ-5D ranged from 0.113 to 0.378. For the model with the lowest MAE, 15% of the participants had errors greater than 0.2, with individual errors as high as 0.66.

Conclusion: Our results validate that the correlation between the MH component of the EQ-5D and the SF-36 is weak. This may be due in part to the insensitivity of the EQ-5D to identify depressive symptoms, which may cause the derivation of utilities from mapping to be problematic in situations where MH is an important outcome. Our results indicate that African-American males are less likely than their female counterparts to indicate depression on the EQ-5D when high depressive symptoms were reported on the CES-D, and therefore deriving utilities using a mapping algorithm may be even more compromised in this population.