Online Program

How Do I Choose? Personalized Decision-Making in Multiple Attribute Settings (304231)

Keywords: precision medicine, multiple attribute decision making, personalized decision making, R Shiny

Decision-making is often 'personalized'. In healthcare this principle underpins precision medicine, with treatments tailored to each patient. Here, analyses typically assume all patients have the same objective, such as minimizing the risk of disease, and seek the decision rule that optimizes this single attribute across a population. In practice multiple attributes may matter, such as treatment efficacy, risk of side effects, or cost. The importance of each will vary by individual, which should be reflected in our analysis. This is described as Multiple Attribute Decision Making, and occurs in any context where preferences across multiple attributes may inform decisions. We explore this topic with a user-friendly, interactive R Shiny application. While multiple attributes may be processed through construction of utility functions, this may not be possible in practice. We show how lower-resolution information such as "I would pay double for a treatment three times as effective" or even simply "high efficacy is more important than low cost" may be used to inform decisions, and compare this to idealized settings when full utility functions are known.