Artificial Intelligence Shows Promise in Healthcare, but Faces Challenges in Adoption

Artificial intelligence (AI) has emerged as a potential game changer in the healthcare industry, with the potential to save the U.S. healthcare system up to $360 billion annually, according to the National Bureau of Economic Research. However, the adoption of AI in healthcare faces challenges due to the industry’s unique regulatory environment and resource constraints.

Jeffrey Sturman, senior vice president and chief digital information officer at Memorial Healthcare System, describes his approach to AI as “pretty bullish.” While AI is still in its early stages at Memorial, it has shown promise in various departments, particularly in clinical settings. The health system has successfully used AI to assist radiologists in evaluating scans, identifying abnormalities that may not be immediately clinically significant but could lead to future issues. This technology has already saved lives by enabling early intervention.

Dr. Thomas Maddox, a cardiologist at Washington University School of Medicine and vice president of digital products and innovation at BJC HealthCare, has also explored the use of generative AI in healthcare. A pilot program using ambient note-taking software allowed physicians to record conversations with patients, which were then summarized into clinical notes by AI. While some doctors embraced the technology, others found it less suitable for their needs, highlighting the importance of adapting both technology and workflow to maximize its value.

The use of AI in note-taking has also shown positive results in terms of patient and provider satisfaction. Software such as DAX and Nuance’s ambient note-taking co-pilot has lightened physicians’ workload and improved efficiency. Clinicians reported spending less time on computers during patient encounters and documentation, leading to a more personalized interaction with patients.

However, the effectiveness of large language models (LLMs), such as generative AI, in clinical use is still a subject of scrutiny. Independent reviews found that general-purpose LLMs often provided irrelevant or inaccurate answers due to their training on non-medical information. LLMs specifically designed for healthcare performed better but still faced limitations in answering questions related to novel treatments or underrepresented patient populations.

Despite these challenges, healthcare professionals and executives remain hopeful about the potential of generative AI. Sturman believes that AI’s time-saving capabilities can improve access to care and boost revenue. However, caution is necessary in selecting the right AI tools, as the market is flooded with various solutions claiming to be AI-driven.