Key points are not available for this paper at this time.
Introduction Healthcare professionals working in busy hospital environments are expected to make multiple back-to-back critical decisions related to patient assessment and treatment. Fatigue from a combination of complex decision-making over multiple patients can lead to less efficient care and an increased risk of error and harm. Artificial intelligence (AI) risk recommendation systems, hereafter referred to as AI risk recommenders, have the potential to reduce the impact of decision fatigue by prompting healthcare professionals with appropriate recommendations for patient care and management. A key barrier to the effective usage of such systems is the establishment of trust and subsequent acceptance among healthcare professionals. However, little is currently known about how trust and acceptance can be engendered. The aim of this review is to develop a theory explaining what influences healthcare professionals’ usage of AI risk recommenders and how trust and acceptance, facilitate their usage of such systems. Methods and analysis We will conduct a rapid realist review to develop a programme theory exploring how trust and acceptance of AI risk recommenders are established among healthcare professionals and how these mechanisms influence system usage. We will use the following databases—MEDLINE (Ovid), EMBASE (Elsevier), the Cumulative Index to Nursing and Allied Health Literature (CINAHL (EBSCOhost)), PubMed, The Cochrane Library, The Institute of Electrical and Electronics Engineers (IEEE) Xplore, The Association for Computing Machinery Digital Library, Scopus (Elsevier), Web of Science (Clarivate) and ProQuest Dissertation and Theses. The review will focus on identifying the resources and processes that stimulate trust and acceptance, leading to the actual use of the system in clinical practice. The review will be guided by the four steps of realist review described by Rycroft-Malone. Article searching and retrieval was conducted on 15 November 2025; full-text screening is ongoing and the review is expected to be completed by May 2026. Ethics and dissemination This study does not require formal ethics approval, as it does not involve primary research. Findings will be shared in peer-reviewed publications, conference presentations and engagement with relevant policy-makers involved in the development and integration of AI risk recommenders within hospital settings. Through these efforts, we aim to support the effective utilisation of such systems, leading to improved decision-making and patient care outcomes. PROSPERO registration number CRD420251155251
Glass et al. (Fri,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: