Purpose: To better understand and empirically test the multiplicative characteristics of utility scores. This objective is motivated by challenges in clinical decision analyses when average utility scores are not available for patient populations with multiple disease conditions.

Methods: Medical Expenditure Panel Survey (MEPS) has a nationally representative sample of the US civilian non-institutionalized population. Using the pooled 2001 and 2003 data, a sample of 40,846 individuals with their concurrent EQ-5D index scores reported were categorized into 238 clinical classification categories (CCCs by AHRQ). The average score for each disease condition was calculated first. The study focus was the difference between the estimated and the observed average EQ-5D scores in each comorbid pair, with the observed one presumed to be the true value. Due to the MEPS sampling property, comorbid pair categories with less than 100 individuals were excluded, which left us with 760 comorbid pairs in total. All statistics were achieved using the proper sampling weight and were age- and sex-adjusted to the national average.

Results: The scores estimated by multiplying the mean EQ-5D scores of the corresponding disease conditions had the largest absolute difference from the actual observed EQ-5D scores (0.105 on a scale of 1). The multiplicative estimator was statistically significantly worse than simply picking the smaller mean of the two paired conditions (absolute difference from the observed score = 0.028, P<0.0001), the larger mean of the two (absolute difference from the observed score = 0.062, P<0.0001), the average of the two means (absolute difference from the observed score = 0.042, P<0.0001), or the mean of the condition with smaller sample size of the pair, i.e., more specific sample (absolute difference from the observed score = 0.044, P<0.0001). Even when the mean scores of the paired conditions were calculated from samples excluding individuals with the other condition of the pair, the multiplication of the scores was still not a good estimate (average absolute difference of 0.094 from the observed score).

Conclusions: The multiplication is not a desirable estimate when the average utility score for patients with multiple disease conditions is not readily available. The recommendation is to use the smaller utility score of the comorbid conditions as a simple approach in clinical decision making and cost-effectiveness analysis.