Separate good and bad in Case 2 Best-Worst Scaling


Case 2 Best-Worst scaling, or BWS-2 for short, is one of the most popular methods for finding out what patients prefer. This is method is used to ask patients to rank different treatment attributes from best to worst. BWS-2 is a quite new method that preference researchers are still exploring. A recent PREFER study explores how the features of BWS-2 can lead to estimation problems when including both benefits and risks in scaling exercises. 

Vikas Soekhai
Vikas Soekhai

Although BWS-2 has the potential of being a valuable method for ‘eliciting’ patient preferences, PREFER researchers are clear that the experiments should only include positive or negative treatment attributes. Asking patients to rank treatment benefits against other benefits and risks or harms against other risks or harms. Otherwise, the authors of a recently published Journal of Choice Modelling paper state, there is a risk of dominance when mixing positive and negative attributes. According to them, the best way to use BWS-2 to find out what patients prefer is separating positive and negative attributes.

“Mixing positive and negative attributes leads to dominance, for example the positive attribute will also be selected as best. These situations will eventually lead to problems in the estimation phase, which makes the experiment not useful. This for example creates a problem when a BWS-2 experiment aims to find out if patients can consider living with a specific side effect to achieve a specific health benefit. When positives and negatives are mixed, strange results,” says Vikas Soekhai, research fellow at Erasmus University Rotterdam and one of the authors of this paper.

By Anna Holm

Soekhai V, Donkers B, Levitan B, de Bekker-Grob EW. Case 2 best-worst scaling: For good or for bad but not for both. Journal of Choice Modelling. Published online. 2021:100325. DOI: 10.1016/j.jocm.2021.100325

Explore PREFER publications

Last modified: 2021-11-10