Study Design

This survey study explores acute and emergency medicine physicians’ views, opinions, and attitudes towards artificial intelligence (AI) and its implications for clinical practice. To this end, we developed a survey to assess three domains: (1) attitudes towards AI in EM, (2) trust in AI, and (3) attribution of responsibility in AI-assisted patient care. Ethical approval was obtained from the Institutional Review Boards (IRB) from Ankara Yıldırım Beyazıt University, Ankara, Turkey (No: 2022-1120) and the University of Groningen, Groningen, The Netherlands (PSY-2223-S-0042).

Survey Development

Survey development occurred in two phases. In December 2022, two focus groups were conducted with researchers interested in AI integration into acute and emergency medicine. Focus groups are qualitative methods designed to elicit attitudes, perceptions, and ideas in a structured setting (Stewart & Shamdasani, 1992; Denscombe, 2007). Each session lasted approximately 3 hours and was moderated by two authors (GS and SV). Participants were selected through purposive sampling and included acute and emergency physicians (specialists, residents, academics) and researchers in EM-related fields. The first focus group involved four acute and emergency physicians from Turkey, who discussed perceptions and anticipated impacts of AI on ED operations. The second group included six researchers from the Netherlands, representing fields such as emergency medicine, engineering, and data science, who discussed the intersection of AI with research and clinical practices.

Transcripts of the sessions were analyzed thematically following Clarke and Braun’s (2006) approach. We then transformed these themes into potential scale items, creating an initial pool of 74 statements. In doing so, we followed the following criteria (Brinkman, 2009; DeVellis & Thorpe, 2021; Hinkin, 1998; Hinkin et al., 1997; Johanson & Brooks, 2010): clarity and comprehensibility of the items, avoidance of double-barreled or double-negative items, avoidance of jargon, avoidance of leading items, avoidance of the acquiescence and social desirability biases, and avoidance of value-laden items.

In the second phase, the item pool was evaluated by an expert panel consisting of 10 social psychologists and 10 acute and emergency physicians via an online survey. These experts rated items on clarity and relevance using a three-point Likert scale. Based on their feedback, the research team refined the survey through minor edits (e.g., restructuring, addition of prompts). The moderator (GS) or co-moderator (SV) made minor changes such as formatting, restructuring of questions, and addition of question prompts. The resultant scales are as follows:

• Attitudes Towards Artificial Intelligence in Emergency Medicine Scale: A 20-item balanced scale assessing positive and negative attitudes toward AI in acute and emergency medicine; responses are on a 5-point Likert scale.
• Trust in Artificial Intelligence Scale: 13-item scale measuring cognitive and emotional trust towards AI in acute and emergency medicine; responses on a 5-point Likert scale.
• AI Responsibility Attribution Scale: Six-item scale evaluating responsibility attribution in AI-assisted care towards various stakeholders (patient, physician, hospital administration, regulatory authorities, technology developers, Ministry of Health); responses on a continuous scale from 0 to 100.

Survey Translation

The final survey, originally developed in English, was translated into Turkish, German, French, Spanish, Italian, Dutch and Chinese using DeepL. To ensure accuracy, each translation was verified by a native speaker who compared the translated version with the original English survey. Additionally, a back-translation procedure was performed, in which the translated surveys were independently translated back into English to identify and resolve any inconsistencies or deviations from the original meaning.

Survey distribution

To gather a diverse and comprehensive range of responses globally, we formed a team of healthcare professionals with strong networks in acute and emergency medicine. This team helped identify national coordinators through social media and email outreach. We also created a dedicated website (www.arise-study.com) for national coordinators to register and access detailed instructions on how to share the survey in their countries. They were encouraged to distribute the survey mainly through professional societies and direct mail to reach physicians involved in acute and emergency care. To boost engagement and improve response rates, we actively tracked the number of completed surveys per country and regularly updated this information on the study website, giving national coordinators a clear and motivating performance benchmark.