Evidence-informed language: interpretation and impact on intentions to treat - results of an online survey of medical students and specialists in German-speaking countries.

objectivesCurrently, there is no generally accepted consensus on how to translate strength of evidence into language. Against this background, we here investigated how widely used verbal descriptors of evidence grades and clinical practice recommendations, respectively, are understood, interpreted and transferred into intentions to treat. We explored differences between medical specialists and undergraduate medical students.

designCross-sectional, anonymous online survey. Assessment was based on publicly available proposals for wording to characterise strength of evidence from randomised versus non-randomised trials and of clinical practice recommendations, respectively.

settingThe online survey was conducted between September 2021 and March 2022 and promoted by several professional organisations in German-speaking European countries (Germany, Austria and Switzerland).

participantsMedical students, trainees and medical specialist (open to all medical specialties). OUTCOME: The survey was composed of two sections: (1) Aim of the first survey section was to assess if the linguistic differentiation between results from randomised versus non-randomised studies is correctly understood as put forward by the proponents. To this end, participants were asked to grade the relative weight of the expression for the results of a randomised trial versus two proposals for how to express the results of non-randomised studies. (2) Next, strong positive, weak positive, weak negative and strong negative clinical recommendations were presented, and participants were asked to answer in a forced-choice format if they would treat all/no patients, selected patients or only consider treatment. Additionally, the number of eligible patients who would be treated was assessed.

resultsN=1081 physicians and N=539 medical students completed the survey. (1) Less than half of the participants (48.5%) interpreted use of 'associated with' as linguistic differentiation between results from randomised versus non-randomised trials in a sense as put forward by the proponents. However, use of the subjunctive mood ('could') resulted in 87.3% correct differentiations. (2) Even with only four types of clinical practice recommendations (positive/negative and strong/weak), interpretation and translation into intention to treat, respectively, showed a heterogeneous picture: while the presentation of a strong clinical recommendation led to largely congruent responses, the interpretation of weak recommendations showed a high variability, with no clear response pattern for intentions to treat. Responses from physicians and medical students were largely comparable.

conclusionThis study demonstrates limitations in the currently used linguistic expressions of strength of evidence and clinical practice recommendations and supports the need to prospectively test effects of language on intentions to treat prior to implementation of a certain wording. Study results cast doubt on that linguistic means alone will lead to optimally targeted intentions to treat.