COMPARISON BETWEEN THE EQ-5D AND THE SEVEN DERIVED HEALTH UTILITIES IN VISUAL IMPAIRMENT IN A U.S. NATIONAL REPRESENTATIVE SAMPLE

Purpose: 1) To assess the associations of EQ-5D index score, EQ-VAS, and seven estimated utilities with self-reported visual impairment (VI) in a U.S. nationally representative sample. 2) To compare the differences between derived utilities and EQ-5D/EQ-VAS across severity of VI. 

Method: Data were extracted from the 2000 and 2002 Medical Expenditure Panel Survey. Seven estimated utilities were derived from the SF-12v1®, including HUI3/VAS item models (IM) and categorical models (CM) from the Sengupta-Nichol, Brazier SF-6D, Lundberg VAS, and Sullivan EQ-5D algorithms. An analysis of covariance was used to determine differences in mean utility scores across severity of VI (none, mild, moderate, and severe/blind). Covariate adjusted effect sizes (ES) were calculated to contrast the magnitude of effect on utilities between each severity level of  VI and no VI, as well as between estimated utilities and EQ-5D.

Result: A total of 34,313 individuals completed both SF-12 and EQ-5D. Mean age was 44.9 years, 54% were female, and 6% had VI. Individuals with VI had significantly lower covariate adjusted mean utility scores (0.04 to 0.10 points difference) or showed small (ES=0.28 for VAS-IM in mild VI) to moderate effect (ES=0.56 for EQ-5D in severe/blind VI) in utility scores when compared to no VI (all p<0.0001). Utility scores were not statistically different among different levels of visual impairment. When comparing estimated utilities with EQ-5D across subgroups of VI, the ESs decreased with increasing severity of VI; Sullivan EQ-5D (ES=0.02 for none to 0.13 for severe/blind VI) showed no effect.  Both VAS-IM and VAS-CM showed moderate effect for any level of VI and large effect for no VI (ES=1.2); other estimated utilities showed small effect for presence of VI, except for Lundberg VAS which showed no effect in moderate or severe/blind VI. 

Conclusion: EQ-5D, EQ-VAS and estimated utilities in self-reported VI individuals displayed small but clinical meaningful differences from those with no VI. The discriminative abilities were similar between EQ-5D/EQ-VAS and estimated utilities. However, they were not sensitive to severity level of visual impairment.