Bayesian versus diagnostic information in physician-patient communication: Effects of direction of statistical information and presentation of visualization

Abstract

Background Communicating well with patients is a competence central to everyday clinical practice, and communicating statistical information, especially in Bayesian reasoning tasks, can be challenging. In Bayesian reasoning tasks, information can be communicated in two different ways (which we call directions of information): The direction of Bayesian information (e.g., proportion of people tested positive among those with the disease) and the direction of diagnostic information (e.g., the proportion of people having the disease among those tested positive). The purpose of this study was to analyze the impact of both the direction of the information presented and whether a visualization (frequency net) is presented with it on patient’s ability to quantify a positive predictive value. Material and methods 109 participants completed four different medical cases (2⨯2⨯4 design) that were presented in a video; a physician communicated frequencies using different directions of information (Bayesian information vs. diagnostic information). In half of the cases for each direction, participants were given a frequency net. After watching the video, participants stated a positive predictive value. Accuracy and speed of response were analyzed. Results Communicating with Bayesian information led to participant performance of only 10% (without frequency net) and 37% (with frequency net) accuracy. The tasks communicated with diagnostic information but without a frequency net were correctly solved by 72% of participants, but accuracy rate decreased to 61% when participants were given a frequency net. Participants with correct responses in the Bayesian information version without visualization took longest to complete the tasks (median of 106 seconds; median of 13.5, 14.0, and 14.5 seconds in other versions). Discussion Communicating with diagnostic information rather than Bayesian information helps patients to understand specific information better and more quickly. Patients’ understanding of the relevance of test results is strongly dependent on the way the information is presented.

How general is the natural frequency effect? The case of joint probabilities

“…However, highest performance levels can be observed, if information is presented in natural frequencies and participants also work with natural frequencies instead of translating them “back” into probabilities ( Weber et al, 2018 ; Feufel et al, 2023 ). It also has an impact on the performance, whether the given information and the question are “aligned”, which means that the presented and requested information should be attached to the same subset ( Tubau et al, 2019 ; Tubau, 2022 ; Brose et al, 2023 ). Furthermore, the performance also improves if the task format is formulated “explicitly” (the intersecting set is explicitly named, i.e., “How many of the positive tested women are ill and test positive?”) instead of “implicitly” (i.e., “How many of the positive tested women are ill?”; Böcherer-Linder et al, 2018 ).…”

“…The communication quality in Bayesian situations is a further aspect worth to consider. Since Bayesian situations often occur in medical contexts in which a physician is supposed to advise patients, the way of (verbally) communicating the meaning of a positive test result is very important ( Gigerenzer et al, 1998 ; Brose et al, 2023 ). Unfortunately, counselors are not always communicating the results in a correct and comprehensible way ( Gigerenzer et al, 1998 ; Ellis and Brase, 2015 ; Prinz et al, 2015 ) and medical students cannot even identify a high-quality communication with the correct value when it is presented as one out of several short video clips ( Böcherer-Linder et al, 2022 ).…”

Understanding and Implementing Diagnostic Stewardship: A Guide for Resident Physicians in the Era of Antimicrobial Resistance

“…Aimed at optimizing comprehension of diagnostic stewardship principles, educational efforts at the pre-analytical stage should involve a thorough understanding of the concept of pre-test probability, which is the likelihood of a patient having a disease before a diagnostic test result is known [ 29 ]. Although healthcare providers are more statistically literate than the general population, they still struggle to understand, incorporate, and convey Bayesian reasoning [ 30 ]. Given its integral part in understanding pre-test probability and predictive values [ 31 ], as well as its overall critical role in medicine, there is a compelling need for effective educational interventions to foster comprehension of Bayesian logic among medical professionals.…”