The use of confidence intervals for individual utilities: limits to formal decision analysis for treatment choice

This paper discusses the use of confidence intervals for utility measurements. Classic test theory is applied to estimate confidence intervals for utilities. The theory is enhanced to calculate confidence areas for combined utilities and confidence bands for the threshold line. As an example it is shown that, if confidence intervals are taken into account, the implied preferred treatment of T3-larynx carcinoma patients is uncertain for a wide range of utilities, considering the mediocre reliability of most methods of utility assessment. This implies that although utility measurement and formal decision analysis can be a useful way to look at the decision problem, ambiguity, which must be resolved by other means, will often remain.