![](/sites/default/files/2024-06/loss-manisha-1083832877.jpg)
Monthly AI Seminar Synopsis: Getting Real with Gen-AI at Johns Hopkins Medicine
Meeting Summary for HBHI Workgroup on AI and Healthcare
Quick Recap
The Hopkins Business of Health Initiative Workgroup on AI and Healthcare met to discuss GenAI implementations in clinical practice. Dr. Manisha J. Loss, Associate Professor of Dermatology and Associate Chief Medical Information Officer at Johns Hopkins Medicine, discussed the implementation of generative AI and other technologies in the context of patient care at Johns Hopkins Medicine. She highlighted their potential to increase efficiency, reduce physician burnout, and improve the patient experience. Dr. Loss emphasized the importance of responsible deployment, addressing issues such as user adoption, quality, patient safety, and regulatory compliance. She also highlighted the growing demand for AI tools within the institution and suggested the formation of subcommittees to oversee their implementation.
Summary:
HBHI Workgroup AI Healthcare Seminar Series
This seminar, led by Dr. Tinglong Dai, Bernard T. Ferrari Professor of Business, and Dr. Risa Wolf, Associate Professor of Pediatric Endocrinology, marked the season finale of Spring 2024 HBHI-AI Seminar Series. Dr. Manisha Loss introduced the session's topic on implementing generative AI in patient clinical journeys, focusing on problem definition before solutions. She outlined Johns Hopkins Medicine's approach to responsibly and efficiently integrating AI tools, exploring the unique opportunities and challenges in enhancing patient care through AI.
Changing Healthcare Communication and Patient Portal Usage
Dr. Loss discussed the evolving dynamics of healthcare communication and care delivery, especially post-COVID-19. She noted a significant increase in patient portal usage for minor concerns, leading to a 60% rise in patient messages at Johns Hopkins Medicine. Certain specialties, like primary care and dermatology, reported even higher usage rates, exceeding 95%.
Addressing Patient Care Delivery Concerns
In response to the surge in patient messages requesting new types of care, Johns Hopkins Medicine allowed clinicians to bill for medical advice requests starting July 2023. Despite receiving over 1 million message threads from July 2022 to May 2023, only 0.8% were converted into billable services. Dr. Loss highlighted the need to balance patient care with workforce demands and evolving care delivery models.
Exploring Generative AI Tools for Efficiency
Dr. Loss presented three key generative AI tools aimed at improving operational efficiency: Co-Pilot, which assists in responding to messages; Ambient Listening, which enhances face-to-face encounters by reducing the time physicians spend on documentation; and Message Categorization, which helps prioritize and manage the high volume of non-clinical messages, thereby improving workflow efficiency.
AI and Machine Learning for Clinical Messaging Management
Dr. Loss elaborated on the use of AI and machine learning to categorize and manage clinical messages, collaborating with external partners to develop large language models. Initial testing in the primary care department showed promise, with ongoing evaluations in surgical departments. The 'Co-Pilot' project, which drafts responses to patient messages, demonstrated potential in reducing burnout but not significantly improving response times or efficiency.
Implementing AI Medical Tools at Johns Hopkins Medicine
Dr. Loss discussed the implementation of AI scribes and auto-draft responses, highlighting their potential to decrease documentation time and improve efficiency. She addressed concerns about costs and emphasized the need for a strong return on investment. The search for a long-term vendor is ongoing, with the goal of finding the most suitable partner from the market.
New Tools in Healthcare and Patient Safety
Dr. Loss emphasized the importance of understanding the impact of new tools on user adoption, quality, and patient safety. She introduced the concept of deskilling and its implications for medical education, highlighting the need for patient consent and ethical considerations in AI deployment.
Rapid Deployment of AI Tools at Johns Hopkins Medicine
Dr. Loss discussed the rapid deployment of new AI tools by a team of health clinical operations and executive leaders, stressing the importance of responsible deployment, considering automation bias, equity, data privacy, and regulatory compliance. A task force, led by Dr. Peter Green, developed a responsible deployment framework to ensure continuous updates and accountability.
AI Tools in Health System Management
Dr. Loss addressed the increasing requests for AI tools, categorizing them into clinical delivery, imaging, and operational tools. In 2023, there were 255 requests for new software, excluding AI-enabled devices. She proposed that AI tools should be overseen by sub-councils that bridge the software intake process and the data model.
Evaluating New Tools and Implementation
Dr. Loss stressed the importance of thoroughly evaluating new tools, focusing on their intended purpose, implementation, and potential risks.
Discussion Led by Professor Mario Macis
Professor Mario Macis, Professor of Economics at the Johns Hopkins Carey Business School and Core Faculty at HBHI, provided an insightful discussion on potential evaluation studies, highlighting the importance of conducting a wide range of studies to assess the impact of AI tools. First, he emphasized the need for Clinical Outcome Studies to evaluate the direct impact of AI tools on patient health outcomes. Second, he discussed the importance of Operational Efficiency Studies to analyze improvements in clinical workflows and overall efficiency. Third, he mentioned the need for Technology Adoption Studies to understand the patterns and factors influencing AI tool adoption among clinicians. Additionally, he underscored the significance of Patient Satisfaction and Experience Studies to measure patient perceptions and experiences with AI-enhanced healthcare services. Fourth, he highlighted Economic Impact Studies to assess the return on investment, including indirect benefits like reduced clinician burnout and improved job satisfaction. Fifth, he noted the value of Comparative Studies to compare AI implementation strategies and outcomes with peer institutions. Finally, he stressed the importance of Ethical and Privacy Studies to ensure the ethical deployment of AI tools while maintaining data privacy and security.
Q&A Session
During the Q&A session, Dr. Loss addressed several important questions from the audience. First, Dr. Kevin Frick asked about the potential for AI-generated responses to become repetitive and less empathetic over time. Dr. Loss explained the importance of tailoring AI prompts to match individual clinician styles to maintain the humanistic aspect of communication. Next, Dr. Risa Wolf inquired about the ethical considerations and integration of AI systems with existing EMR platforms. In response, Dr. Loss emphasized the need for fully integrated systems to ensure data privacy and reduce the risk of errors. Finally, members of the audience raised concerns about the impact of AI on medical education and the potential for deskilling. Dr. Loss highlighted ongoing studies to assess the effects of AI tools on training and competency development among medical residents.
The session concluded with a reminder about the next seminar series starting in September 2024, with announcements and updates to be sent over the summer.
Thanks again to all of you! We wish you a productive, enjoyable, and AI-augmented summer! As always, if you know of anyone who should be part of our workgroup, please let Liana know at [email protected].
Cheers,
Risa Wolf, MD and Tinglong Dai, PhD
Co-Chairs, HBHI Workgroup on AI and Healthcare