Background: Healthcare systems in the United States are navigating a complex landscape of financial strain, high clinician burnout, and workforce instability. The emergence of artificial intelligence (AI) tools in the clinical space offers the potential to improve clinical care and productivity but faces challenges like cost, clinician acceptance, and bias. Use of AI tools in the clinical arena is not yet well understood or described. We aimed to understand the current state of AI use in hospital medicine, explore challenges and barriers to AI use, and to explore future use cases of AI in the future.
Methods: Using an exploratory, embedded mixed methods approach that combined a brief survey with semi-structured virtual interviews we conducted 7 virtual concurrent focus groups on May 10th, 2024. Participants were part of the Hospital Medicine Reengineering (HOMERuN) Research Network, a national organization of hospitalists and healthcare team members dedicated to enhancing patient care in hospital settings. Data were collected through video-recorded Zoom sessions and survey responses, followed by a rapid qualitative analysis using templated summaries and matrix analysis. Reflective practices and participant validation ensured trustworthiness and minimized researcher bias.
Results: Forty participants from 28 different institutions participated in the focus groups and 32 participants participated in the surveys (80% response rate). Survey participants included physicians (81%), an advanced practice provider (3%), leaders (9%) and an administrative professional (3%). Fifty-six percent of participants reported utilization of AI tools. Thirty-four percent reported using AI tools in their personal life, while 25% reported using AI tools for clinical applications and 25% for educational applications. The main barriers to use were around concerns regarding data quality and systemic bias in the data (59%), privacy, data stewardship and consent (50%), lack of education regarding AI use (44%), difficulty integrating AI into current workflows (44%) and liability and regulatory issues (44%). Four main themes were identified including: (1) wide variation exists in the use of AI tools; (2) numerous barriers hinder adoptions ; (3) tension exists between clinician goals, such as spending more time with patients, and corporate goals, such as profit-driven motives, when using AI tools; and (4) there are many hopes and dreams for AI to enhance care and work processes.
Conclusions: AI technologies show promise in enhancing clinical workflows, provider efficiency, and patient outcomes. However, barriers such as lack of standardized guidelines, trust issues, and the financial costs of implementation pose challenges to AI adoption among hospitalists. While participants are optimistic and excited about AI’s future in healthcare, they worried about organizational motives. Addressing these barriers and aligning clinician and corporate goals will be critical to unlocking AI’s full potential in healthcare.