Physicians’ prompting strategies for Generative Artificial Intelligence in clinical decision-making: a scoping review protocol

Clinical reasoning is one of the pillars of medical practice and, more recently, generative artificial intelligence (GenAI) has emerged as a tool to support clinical practice and decision-making. However, the incorporation of GenAI into clinical practice depends on multiple factors, including how physicians interact with these tools through user-generated prompts. This scoping review aims to map the available evidence on prompting strategies used by physicians when interacting with GenAI tools to support clinical decision-making across all levels of medical care. We will search on MEDLINE (via PubMed), Embase, Cochrane, Web of Science, LILACS (via BVS), Google Scholar and ProQuest. The main concept of this review is prompting strategies, defined defined as the ways in which participants structure prompts to interact with GenAI tools in the context of supporting clinical reasoning and clinical decision-making. We plan to describe the countries where the studies were conducted, GenAI tools used, and participants' professional or training level, including general practitioners, medical residents, and specialists. We will also report the prompting strategies used, prompt elements, previous user training, and the clinical questions addressed using GenAI, categorized into the following domains: basic science, risk factors or prognosis, diagnosis, treatment, communication, medical ethics, and others. The expected outcome of this review is a comprehensive mapping of how physicians interact with GenAI through prompting in clinical reasoning and decision-making, identifying the prompting strategies and prompt elements currently used, as well as gaps in the existing evidence. These findings may inform future research on physician–GenAI interaction and the development and evaluation of prompting approaches for clinical decision support.