Estimating Seven Health-Related Utilities in Chronic Disease Conditions using a U.S. National Representative Sample
Presenter: Michael Nichol, USC
Objectives: 1) To assess the associations between seven estimated utilities and the most prevalent chronic diseases in a U.S. national representative sample. 2) To compare sensitivity of seven estimated utility methods on their discriminative ability to detect differences between individuals with and without disease. Methods: The 2005 Household Component Full-Year Files in Medical Expenditure Panel Survey (MEPS) was used. Seven estimated utilities were derived from the SF-12v2™, 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 estimated utilities among individuals with and without disease conditions (depression, anxiety, back disorder, diabetes, joint disorder, arthropathy, hypertension, and other conditions). Covariate adjusted effect sizes (ES) were calculated for estimated utilities between individuals with and without disease. Covariates included age, gender, education, employment status, marital status, and social economic status using family income as percentage of poverty line as the proxy.
Results: We analyzed a total of 19,475 individuals who completed SF-12v2™ survey. Mean age was 45.5 years (range 18 to 85), 45% were male. Mean estimated utilities ranged from 0.71 (VAS-IM and VAS-CM) to 0.88 (EQ-5D). Individuals with disease condition had significantly lower covariate adjusted mean estimated utility scores than those who did not have condition (all p<0.0001). Individuals with hypertension showed no effect (ES=0.18 for SF-6D) to small effect (ES=0.45 for VAS-IM) in all estimated utilities as compared to those who had no disease condition. When compared the ES within each estimated utility across different diseases, VAS-CM (ES=0.45 for arthropathy/joint disorder to ES=0.82 for diabetes), VAS-IM (ES=0.55 for arthropathy/joint disorder to ES=0.91 for diabetes), SF-6D (ES=0.57 for arthropathy to ES=1.14 for depression) showed small to large effects across different disease conditions compared to those who did not have condition; Lundberg VAS (ES=0.5 for diabetes to ES=1.0 for depression), EQ-5D (ES=0.57 for arthropathy to ES=1.14 for Depression), HUI3-CM (ES=0.6 for arthropathy to ES=1.2 for depression), and HUI3-IM (ES=0.7 for arthropathy, diabetes, and joint disorder to ES=1.2 for depression) showed median to large effects across different disease conditions compared to no condition. When compared the ES within each disease condition among different estimated utilities, HUI3-IM had largest ES for four conditions (ranged from ES=0.45 for hypertension to ES=1.20 for depression), and EQ-5D had largest ES for two conditions (ES=0.71 for Joint disorder and ES=0.86 for other conditions). SF-6D had smallest ES for two conditions (ES=0.45 for Diabetes, and ES=0.18 for hypertension).
Conclusions: Estimated utilities were significantly associated with disease conditions and can discriminate between individuals with and without condition. The discriminative abilities varied from different estimated utility algorithms across different diseases. However, the algorithm that derived HUI3 from the SF12 items model displayed the most discriminative ability. These findings may help researchers in the selection of derived utilities in cost-effectiveness studies for different disease conditions when direct preference based health utility is not available.
Authors: Joanne Wu, Ning Yan Gu, Michael Nichol
Session: Chronic Disease
Time: Tue 10 a.m.-11 a.m.