Patterns of responses in time trade-off studies: a latent class analysis of health states worse than dead.

AIM: The study was to characterize participant responses in a time trade-off (TTO) task involving states worse than dead (WTD) to determine patterns of congruence and incongruence.

METHODS

The online TTO task involved 4 hypothetical health states describing a rare paediatric condition. Participants completed the task from a parental perspective with a 10-year time horizon. Mean health utilities were derived for the health states (HS). Congruence patterns were defined a priori as "perfect", i.e. completely logical sequence of health state utilities, "incongruent", reversal of the most and least severe HS utilities, and "mixed" congruence. A multinomial regression and latent class analysis (LCA) were applied to elucidate congruence patterns.

RESULTS

A total of 322 participants completed the TTO (49% females; average age 43 years). Perfect congruence response patterns were observed in only 11.5% (37) participants; 99 participants (30.8%) had incongruent response patterns and 186 (57.8%) a mixture. Two (most severe) HS were rated WTD by the "mixed", but only one HS was rated WTD by the "perfect" congruence group, who rated the other 3 HS better than dead. The multinomial regression identified age and gender as potential factors influencing congruence patterns; the most impactful factor was the number of stated rated WTD. The LCA was not able to identify a single class of perfectly congruent responders.

LIMITATIONS: This online TTO did not allow further exploration of individual choices through post-task participant interviews.

CONCLUSIONS

Facets of TTO tasks such as states WTD may introduce an additional cognitive burden on participants and impact on congruence. However, incongruence was observed even within ostensibly perfectly congruent responses; conversely incongruent response patterns showed some internal coherence. Removing the latter in favour of the former may therefore also introduce bias. A balance may need to be struck between sacrificing perfect congruence and the inclusion of some incongruence to achieve greater representativeness of health state utilities.