Assessing the Readiness of Health Care Organizations for Safe AI Integration: Perspectives From Quality and Safety...

Artificial intelligence (AI) technologies hold great promise for improving patient outcomes, reducing clinician workload, and enhancing patient engagement. However, improper design, implementation, and monitoring can introduce significant safety risks. Health care quality and safety leaders play a critical role in mitigating these risks. As AI adoption accelerates, understanding how these leaders perceive their institutions' progress in assessing and managing AI safety is critical for identifying gaps, addressing potential risks, and guiding safer clinical integration.

Semi-structured interviews were conducted with 22 quality and safety leaders from 19 US health care organizations between March and April 2024. Participants included leaders from both single hospitals and multi-hospital systems, with an average of 16 years of experience. None had formal AI training, but some reported practical exposure. Interviews focused on participants' knowledge of AI, organizational structures for AI governance, and barriers to safe AI implementation. Thematic analysis was used to identify common themes and knowledge gaps.

Most organizations (78.9%) reported using steering committees for AI oversight, with some combining this with IT, research, or innovation teams. Barriers to AI implementation included interoperability challenges (78.9%), lack of AI expertise (68.4%), and difficulty evaluating AI effectiveness (52.6%). Participants highlighted the need for stronger governance and evidence-based tools but noted variability in their organizations' preparedness to adopt AI.

Health care organizations lack standardized approaches to AI safety and often rely on fragmented governance structures. Leaders emphasized the need for enhanced expertise, solutions to barriers that affect implementation, and alignment of AI tools with organizational priorities.

Strengthening organizational knowledge, governance, and solution generation to barriers of implementation is essential to safely integrate AI into clinical care. Addressing these gaps will support patient safety and optimize AI's potential benefits.