J.P. Morgan Healthcare Conference 2024: Key Takeaways – Day 3

Healthcare leaders are rapidly embracing artificial intelligence, signaling a potential turning point for an industry historically slow to adopt new technologies.

A Shift in Perspective

After years of cautious pilot programs, health systems are now deploying AI at an enterprise level, driven by financial pressures, regulatory changes, and a growing recognition that the status quo is unsustainable.

• The healthcare industry is moving beyond small-scale AI pilots to enterprise-wide deployments.
• Ambient listening technology is gaining traction, streamlining documentation and administrative tasks.
• Federated AI learning models are emerging as a solution to data privacy and security concerns.
• Neuroscience is poised to become a major growth area for health systems leveraging AI.
• Collaboration between humans and AI is seen as crucial for maximizing the benefits of the technology.

Initially skeptical, even I’ve become a believer, especially after attending the 44^th Annual J.P. Morgan Healthcare Conference this week. It reminded me of the old joke about a psychiatrist asked how long it takes to change a light bulb. His response? “Does the light bulb want to change?” The answer, increasingly evident at the conference, is a resounding “yes.”

Enterprise-Wide Pilots and a Willingness to Fail

Sarah Murray, Chief Health AI Officer at UCSF Health, articulated this shift during a panel discussion Wednesday with Nate Gross, M.D. (who leads OpenAI’s Health business) and others. “We’ve changed what a pilot looks like,” she explained. “We most recently deployed an inpatient stabilization pilot enterprise-wide. The idea was to turn it on across the entire enterprise and if it was causing harm we could turn it off. This is different from two years ago, when we would do a clinic-by-clinic pilot. Our queue has 90 tools in it right now, not including deploying [a secure, HIPAA compliant] ChatGPT Enterprise to everyone at UCSF.”

This willingness to scale up quickly represents a departure from the healthcare industry’s traditional “aviation disease”—numerous, long-running pilot programs that rarely achieve widespread implementation. UCSF’s approach fosters quicker implementation, real-time improvement, and a greater acceptance of “failing forward,” a Silicon Valley concept emphasizing rapid learning from both successes and setbacks.

Ambient Listening and Expanding AI Applications

A growing number of health systems are adopting ambient listening technology at scale, and its applications are expanding beyond basic transcription. Companies like AIDoc (clinical AI) and Abridge (ambient listening) now serve over two hundred hospitals and health systems. AI is also accelerating revenue cycle management, as discussed in a recent post.

The conversation also centered on the impact of AI on the workforce. While some roles, like billing and coding, may be threatened, the consensus is that the most successful outcomes will come from collaboration between AI and humans. Concerns that AI imaging interpretation would eliminate the need for radiologists were also dismissed, with speakers noting a growing demand for radiologists skilled in working with AI. As Murray succinctly put it, “AI will not replace you, but you could be replaced by someone who uses AI.”

Unlocking New Markets with AI

Dave Wessinger, CEO of PointClickCare—a software platform used by many skilled nursing facilities (SNFs) in the U.S.—described how AI is transforming their product. “We’ve done this [software work] for thirty years and added tremendous value. The value we will deliver in the next two years will be more than in the entire last thirty years.” Their AI tools automate the complex SNF admission screening process, improving occupancy rates, reducing litigation risk (by flagging potential issues like admitting sex offenders or habitual plaintiffs), and optimizing staffing and care processes. Wessinger believes AI could double or triple PointClickCare’s total addressable market (TAM).

Eron Kelly, CEO of ConcertAI, highlighted the importance of AI platforms that can access, manage, and transform both first-party and third-party data. ConcertAI, with over 2,000 clients, has become a leading agentic AI and real-world data company, particularly in oncology. Virta’s CEO also emphasized the value of large datasets, noting they possess the largest collection of human biomarkers for metabolic disease.

Gene Woods, CEO of Advocate Health, illustrated the sheer volume of data available, stating that Advocate has 90 terabytes of data—equivalent to three times all the books ever written. This data is becoming increasingly accessible as health systems move towards single, systemwide instances of EPIC.

The Power of Multimodal Data

The ability to integrate and analyze multimodal data—data from lab tests, imaging, genomics, electronic health records, remote monitoring, wearables, and even social media—is a key opportunity. This allows for a more comprehensive understanding of a patient’s health and the ability to detect subtle warning signs before a crisis occurs.

We’ve transitioned from a scarcity of data to an abundance, enabling a shift from low to high-resolution insights. This allows for a more detailed understanding of disease progression and the connection between diagnostic testing, treatment regimens, and previously unrelated factors like environmental exposures.

Addressing Data Challenges with Federated Learning

However, realizing the full potential of multimodal data requires addressing challenges related to privacy, security, and intellectual property. A promising solution is CAIA, the Cancer AI Alliance platform pioneered by the Fred Hutch Cancer Center, Dana-Farber Cancer Institute, Memorial Sloan Kettering Cancer Center, and the Johns Hopkins Sidney Kimmel Cancer Center (www.canceralliance.ai). CAIA uses a federated AI learning approach, allowing researchers to train AI models without sharing patient data.

Under this model, each participating institution trains a global model using its own data locally. Local parameters are then shared to create a new, improved global model, which is sent back for further training. This approach maintains data privacy and security while benefiting from the collective knowledge of multiple institutions.

The Future of AI in Neuroscience

Beyond oncology, neuroscience is emerging as a key area for AI innovation. Advocate Health is performing approximately 25,000 neurosurgeries annually and applying AI to imaging to reduce stroke-related harm. The Cleveland Clinic is building a 1-million-square-foot neurological institute, recognizing the potential of brain disease as a major driver of future healthcare growth.

The Importance of Conversation and Ambient Listening

Abridge CEO Dr. Shiv Rao emphasized that healthcare is fundamentally about conversations. Ambient listening technology captures these conversations, streamlining documentation and reducing administrative burdens. There was widespread enthusiasm for ambient listening at the conference, with companies like Abridge expanding its capabilities to include pre-charting, prior authorizations, and patient summaries. The AI models used in ambient listening are continuously improving as they learn from user edits.

Dr. Rao also noted a supply and demand mismatch in healthcare, with a growing patient population and a shortage of physicians. AI, particularly agentic and generative AI, could help bridge this gap by supporting existing physicians and increasing productivity.

Ultimately, as Dr. Rao observed, when anecdotes and data conflict, it’s often the anecdotes that hold the most weight—underscoring the importance of human judgment in interpreting AI-driven insights.

After all, a good story still resonates.