GPs' beliefs and decision making: is the message or the messenger more important?

Participants are presented with a series of standardized clinical vignettes describing ambiguous patient cases. The scenarios vary in contextual information and are used to study clinical decision-making under uncertainty. In some conditions, participants receive additional informational support alongside the clinical cases, designed to provide relevant clinical cues to assist decision-making. Participants are asked to evaluate each case by providing probabilistic judgments and indicating potential clinical actions, such as diagnostic testing or referral decisions. The order and type of information presented are systematically varied across participants as part of an experimental design.

2026-07-01

2026-12-31

Primary outcomes are clinicians’ diagnostic and management decisions in response to clinical vignettes. These include diagnostic probability estimates, perceived clinical severity ratings, and clinical management decisions (testing, referral, and disposition choices). Diagnostic uncertainty is derived from the distribution of probabilistic responses across alternative clinical options.

The primary outcomes are clinicians’ diagnostic and management decisions in response to clinical vignettes. These include (i) probabilistic clinical judgments regarding the most likely diagnosis, (ii) perceived clinical severity ratings, and (iii) behavioral clinical decisions, including ordering diagnostic tests, determining test urgency, and referral or disposition decisions (e.g., specialist referral, emergency referral, follow-up, or discharge). In addition, a measure of diagnostic uncertainty is derived from the distribution of probability allocations across alternative clinical actions.

Secondary outcomes include measures of clinical reasoning processes, decision-making patterns, and attitudes toward information sources and decision-support tools. These include probability allocations across alternative management options, self-reported frequency of encountering similar clinical cases, perceived usefulness and trust in decision-support information, and responses to questions regarding clinical practice patterns. Additional secondary outcomes include performance differences across experimental conditions and within-subject variation in responses to vignette-based scenarios.

This study uses a randomized controlled experimental design based on standardized clinical vignettes to examine clinical decision-making under conditions of uncertainty among general practitioners. Participants are recruited and complete an online survey consisting of multiple sections, including demographic and professional background questions, attitudinal measures, and clinical decision-making tasks based on hypothetical cases.

In the experimental component, participants are randomly assigned to different versions of clinical scenarios that vary in how clinical information is presented. Participants evaluate a series of vignette-based cases and provide clinical judgments and management decisions.

The order of scenarios and study conditions is randomized. The design includes both between-subject and within-subject components to compare decision patterns across different informational conditions and within individuals over multiple cases.

At the end of the survey, participants complete additional questions on clinical practice patterns, attitudes toward decision-making processes, and professional characteristics.

Not available

Randomization was implemented at the individual participant level. The software randomly assigned participants to one of the experimental conditions, randomly selected a balanced set of four clinical vignettes (according to AI support, clinical complexity, and diagnostic uncertainty), and randomized the order in which the selected vignettes were presented.

Individual participant. Each participant was randomly assigned to one of the three reporter-frame conditions. Within participants, the clinical vignettes, AI support condition, and vignette presentation order were randomized according to the pre-specified balancing constraints.

No

Individual participant. Randomization was not clustered.

500 general practitioner participants

200 participants – Female reporter (100 male patient, 100 female patient); 200 participants – Male reporter (100 male patient, 100 female patient); 100 participants – Neutral reporter (control group)

No formal ex-ante power calculation was conducted. The planned sample size of 500 general practitioners was determined based on feasibility and to ensure balanced allocation across the experimental conditions. Randomization is implemented at the individual level, with no clustering.