Quality-adjusted life years (QALYs) are commonly used to evaluate the impact of new medical technologies on patient mortality and morbidity. However, use of QALYs

Many authors have discussed the fact that standard assumptions result in utility functions that are unrealistically linear and separable, overlooking the diminishing marginal utility a fundamental property of utility functions in economics.

A paper by Johnson et al. (2024) conduct a discrete-choice experiment (DCE) in order to quantify patient preferences over treatments for major depressive disorder (MDD) in order to test for diminishing marginal utility and path dependence. In the study, they offered respondents different symptoms sequences of severe, moderate and mild depression and estimated a willingness to pay for each. A deductive latent-class mixed-logit analysis was used to evaluate survey data from 751 respondents with MDD.

Using this approach, they found that:

Consistent with theoretical expectations, we found evidence of diminishing marginal utility,
holding sequence timing constant, as well as evidence of path dependence, holding symptom improvement constant. Our experiment isolated the effects of path dependence from the influence of conventional time discounting by focusing on short 6-week treatment durations

The authors argue that the short time frame they present (6 weeks) means that discounting isn’t an issue. If individuals did not fully appreciate the timing was weeks–instead just considered generic time periods–then discounting could have played a role here. However, assuming that respondents did properly understand the nature of the question, discount rates would have a very small impact over 6 weeks and the findings can be attributable to non-linear utilities. This may be particularly relevant for treatments with fast-acting mechanisms of action. You can read the full paper here.