Communicating Uncertainty in Benefits and Harms: A Review of Patient Decision Support Interventions

Interventions designed to help people deliberate and participate in their healthcare choices frequently describe uncertainty in potential benefits and harms. This uncertainty can be generalized to aleatory, or first-order uncertainty, represented by risk estimates, and epistemic, or second-order uncertainty, represented by imprecision in the risk estimates.

The aim of this short communication was to review how patient decision support interventions (PDSIs) describe aleatory and epistemic uncertainty.

We reviewed PDSIs available online in five repositories and extracted all the uncertainty statements.

A framework was developed and each statement was classified by presentation of uncertainty (aleatory and epistemic).

Overall, we reviewed 460 PDSIs from eight main developers, which included 8956 uncertainty statements. When describing first-order, aleatory uncertainty, almost all PDSIs included at least one qualitative statement, such as 'treatment may cause side effects'. Forty-four percent of PDSIs included at least one natural frequency, such as '2 in 100 people have side effects'. Second-order, epistemic uncertainty was also most often communicated qualitatively; notably, nearly half of all PDSIs did not communicate epistemic uncertainty at all. Few PDSIs communicated epistemic uncertainty in quantitative terms conveying imprecision, e.g. risk ranges.

We found considerable heterogeneity in both the extent and manner in which aleatory and epistemic uncertainties are communicated in PDSIs. This variation is predominately explained by a lack of evidence and consensus in risk communication, particularly for epistemic uncertainty. This study highlights the need for more empirical research to understand not only the outcomes of communicating uncertainty in PDSIs but also the reasons for this variation in uncertainty communication.