OBJECTIVES: Measuring health state utility (HSU) for comorbid conditions has become increasingly valuable due to population aging and growing comorbidity prevalence. In the absence of values specific to comorbid conditions, researchers estimate joint HSUs based on single condition HSUs using a variety of methods. Performance of these estimators across specific pairs of comorbidities is not understood. In this study we compare the relative performance of four nonparametric estimators for joint HSUs across different types of pairs of conditions.

METHODS: Data came from participants in the Medical Expenditures Panel Survey with at most 2 conditions (N=127,544). Pairs of conditions represented by ≥ 20 subjects (Physical-Physical [PP]: 152; Physical-Mental [PM]: 69 pairs; Mental-Mental [MM]: 10 pairs) were selected. For each condition pair (A-B), we calculated mean age and gender adjusted HSUs for subjects with only A, only B, and both A and B. Joint HSUs were estimated using the additive, multiplicative, minimum and adjusted decrement estimator (ADE). Estimators’ performance was evaluated using weighted mean error, weighted root mean squared error and proportion of errors ≤ 0.02 (P0.02)

RESULTS: The multiplicative estimator was ranked first for all 3 performance measures for pairs where both conditions were mental, and ranked best in 2 measures when both were physical. Among physical-mental condition pairs the the multiplicative ranked best only for P0.02. Differences between estimators were generally small (within |0.01| or |10%|). The multiplicative estimator was ranked first in 6 of the 9 measures across the 3 types of comorbidity pairs.

CONCLUSIONS: For comorbidities of the same type (MM and PP pairs) the multiplicative estimator performed slightly better. Despite small differences the ADE estimator performed best among PM pairs. These results are in agreement with the recommendations on the preferred use of the multiplicative estimator.