Choices in health are everywhere around us, each with their own characteristics: from treatment choices in the doctor’s office to choices on a policy level about vaccination strategies. The research of PhD candidate Vikas Rogier Soekhai (Erasmus School of Health Policy & Management) shows how choice modelling can help understand choices in health. He addresses several challenges of choice modelling in health and opportunities to overcome these challenges.
Choices about the Dutch vaccination strategy for COVID as well as the use of contact tracing applications late 2020 and begin 2021 led to a fierce debate in Dutch society. This might have come from a societal feeling that citizens were not as much involved in the decision-making process as they would have liked. By using choice experiments, decision-makers could have gotten more insights into societal preferences and uptake predictions when making choices for vaccination or the use of contact tracing applications.
A variety of methods
There is a variety of methods available to provide insights into preferences in health, which was shown in a systematic literature study in which 32 methods were identified. In his PhD dissertation Vikas Rogier Soekhai specifically focuses on two methods that gained a lot of attention in health economic literature as another literature study showed: discrete choice experiments (DCE) and case 2 best-worst scaling (BWS-2). These are two different survey-based preference elicitation methods in which individuals are asked to select their preferred (or least preferred) option from a set of hypothetical options. Soekhai: “By asking individuals to answer several of these questions, we can learn how they make choices by analyzing their preferences: how do they trade-off between vaccination effectiveness and side effects for example?”.
Challenges and opportunities
Although DCE and BWS-2 have been quite popular in health, there are several challenges regarding design and analysis. In his dissertation Soekhai shows, by using simulation studies, that when predicting with DCEs the type of prediction analysis should correspond with the econometric model used for estimation. He also conducted another simulation study and empirical study that showed that mixing positive (e.g., benefit) and negative (e.g. harm) characteristics in BWS-2 leads to statistical problems and therefore non-useful results for decision-making. Soekhai: “However, this problem can be solved by using one specific framing in the choice experiment: either all positive or all negative”.
Impact for decision-making
Outcomes from an empirical study among patients showed both similarities as well as differences in preferences between DCE and BWS-2. This indicates that both methods differ from each other and the choice for one of both depends on the decision-making context. Soekhai: “Both methods have the potential to provide useful evidence for decision-makers, but they can also complicate the decision-making process as an additional source of information. As a health economist it is however very useful to see what determines health choices and what challenges we need to overcome to provide useful information for decision-making.”