Premium
Communicating Uncertainties About the Effects of Medical Interventions Using Different Display Formats
Author(s) -
McDowell Michelle,
Kause Astrid
Publication year - 2021
Publication title -
risk analysis
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.972
H-Index - 130
eISSN - 1539-6924
pISSN - 0272-4332
DOI - 10.1111/risa.13739
Subject(s) - transparency (behavior) , psychology , recall , credibility , point estimation , perception , computer science , scientific evidence , cognition , icon , bar chart , applied psychology , social psychology , cognitive psychology , statistics , mathematics , neuroscience , political science , law , programming language , computer security
Abstract Communicating uncertainties in scientific evidence is important to accurately reflect scientific knowledge , increase public understanding of uncertainty, and to signal transparency and honesty in reporting. While techniques have been developed to facilitate the communication of uncertainty, many have not been empirically tested, compared for communicating different types of uncertainty, or their effects on different cognitive, trust, and behavioral outcomes have not been evaluated. The present study examined how a point estimate, imprecise estimate, conflicting estimates, or a statement about the lack of evidence about treatment effects, influenced participant's responses to communications about medical evidence. For each type of uncertainty, we adapted three display formats to communicate the information: tables, bar graphs, and icon arrays. We compared participant's best estimates of treatment effects, as well as effects on recall, subjective evaluations (understandability and usefuleness), certainty perceptions, perceptions of trustworthiness of the information, and behavioral intentions. We did not find any detrimental effects from communicating imprecision or conflicting estimates relative to a point estimate across any outcome. Furthermore, there were more favorable responses to communicating imprecision or conflicting estimates relative to lack of evidence, where participants estimated the treatment would improve outcomes by 30–50% relative to a placebo. There were no differences across display formats, suggesting that, if well‐designed, it may not matter which format is used. Future research on specific display formats or uncertainty types and with larger sample sizes would be needed to detect small effects. Implications for the communication of uncertainty are discussed.