Individual Utilities Are Inconsistent with Rationing Choices

Abstract

To test whether cost-effectiveness analysis and present methods of eliciting health condition "utilities" capture the public's values for health care rationing. Two surveys of economics students. The first survey measured their utilities for three states of health, using either analog scale, standard gamble, or time tradeoff. The second survey measured their preferences, in paired rationing choices of the health states from the first survey and also compared with treatment of acutely fatal appendicitis. The rationing choices each subject faced were individualized according to his or her utility responses, so that the subject should have been indifferent between the two conditions in each rationing choice. The analog-scale elicitation method produced significantly lower utilities than the time-tradeoff and standard-gamble methods for two of the three conditions (p < 0.001). Compared with the rationing choices, all three utility-elicitation methods placed less value on the importance of saving lives and treating more severely ill people compared with less severely ill ones (p < 0.0001). The subjects' rationing choices indicated that they placed values on treating severely ill people that were tenfold to one-hundred-thousand-fold greater than would have been predicted by their utility responses. However, the subjects' rationing choices showed internal inconsistency, as, for example, treatments that were indicated to be ten times more beneficial in one scenario were valued as one hundred times more beneficial in other scenarios. The subjects soundly rejected the rationing choices derived from their utility responses. This suggests that people's answers to utility elicitations cannot be easily translated into social policy. However, person-tradeoff elicitations, like those given in our rationing survey, cannot be substituted for established methods of utility elicitation until they can be performed in ways that yield acceptable internal consistency.