OBJECTIVES: To evaluate whether random utility maximization (RUM) or random regret minimization (RRM) is used when people make health preference decisions when faced with different survivals and risks.

METHODS: A 7-attribute discrete choice experiment (DEC) is designed using the five items of the EQ-5D-5L plus survival time (3 months, 1, 5, and 10 years), and out-of-pocket costs as percent of household income (5%, 10%, 20% and 50%). Bayesian D-efficiency is used to allow relatively large standard deviations for the priors. Ngene is used for the DCE design resulting two 2 blocks, each with 16 choice sets. Three alternatives are presented for each comparison. Preliminary data (N=200) were collected in April 2022 using the Amazon Mechanical Turk.

RESULTS: 109 respondents passed screening test and included in the analyses. Mean age is 36.5 years (SD ±10.8); 58% of them are male; 87% are white; 9% black; 64% have completed college; 64% are married and, over 90% of them have various health insurance coverages. All coefficients in both models (RUM and RRM) are significant (p<0.01). Health problems (p<0.01) and cost burdens (p<0.10) are negatively associated with loss of utility and regret reduction. On the other hand, increased survival time is significantly and positively associated with increased utility and decreased regret. The scale parameter µ from the µRRM was significantly larger than 1 (µ=4.99, se=0.01), suggesting that µRRM collapses into RUM in this pilot investigation.

CONCLUSIONS: Directions of all coefficients are consistent with prior expectations. The relatively high scale parameter µ suggested that, in this pilot study using the EQ-5D-5L, the traditional RUM and RRM yield the same behavior and health outcomes. Further investigation is warranted to validate the findings herein by assessing different scenarios with varying risk prospects.