Making Composite Time Trade-Off Sensitive for Worse-than-Dead Health States

The utilities elicited with the composite time trade-off (cTTO) method for health states worse-than-dead (WTD) often correlate poorly with other severity measures, indicating a poor sensitivity of cTTO. We aimed to explore modifications to cTTO to better understand this phenomenon and identify potential improvements.

A total of 480 respondents completed an online TTO interview, each valuing 12 EQ-5D-5L health states. The participants were randomized into four arms, A-D. Arm A followed the standard cTTO, serving as a reference. In arm B, we removed the sorting question comparing immediate death versus 10 years in a valued state. Arm C allowed for utility values $<-1$ by reducing the time in the valued state in the lead-time TTO (LT-TTO) part of cTTO. In arm D, we randomly selected the starting negative utility in LT-TTO. Utility value distributions, correlations between utilities and level sum score (LSS), and inconsistencies between Pareto-ordered states were analyzed.

Arm A replicated the lack of significant correlation between LSS and the negative utility observed in previous work. Of the experimental arms, only arm B exhibited a significant negative correlation. Compared with arm A, arm B produced a higher proportion of WTD states ( $46.5\%$ versus  $26.3\%$ ), less negative utility for WTD states on average ( $-0.571$ versus  $-0.752$ ), and a lower mean censored utility for 55555 ( $-0.486$ versus  $-0.406$ ).

The observed lack of correlation between LSS and utility for WTD states appears linked to the use of comparison with immediate death in the sorting question. LT-TTO is capable of eliciting utility values in a way that is sensitive to severity. Modifying the initial questions in cTTO to identify whether health states are BTD or WTD should be considered.