Reimagining healthcare delivery and public health with AI
Dr. Umair Shah, former Washington State Secretary of Health, and Dr. Gianrico Farrugia, CEO of Mayo Clinic, discuss the practical applications and strategic imperatives of AI in public health and clinical care. They explore how AI can shift public health from a reactive to a proactive model, the necessity for a new data-centric architecture in healthcare systems, and the transformative potential of generative AI beyond mere optimization of current workflows.
The Untapped Potential of AI in Public Health
Dr. Umair Shah, former Washington State Secretary of Health, illustrates the frequent disconnect between public health and healthcare delivery using a football analogy. Healthcare delivery is the “quarterback” (like Tom Brady), who is highly visible and well-understood. Public health, conversely, is the “offensive line”—the invisible workforce doing the essential blocking and tackling to ensure the quarterback’s success. A critical difference is that in society, unlike in football, the offensive line (public health) is defunded after a success, leading to a resurgence of the very problems it solved.
Dr. Shah argues that public health practitioners must be “at the table” as AI technologies are developed. Otherwise, solutions will be designed for them rather than with them, failing to meet the unique needs of the field.
From Reactive to Proactive: AI Use Cases
Dr. Shah emphasizes that AI is not a future technology but a present-day necessity for public health. He points out that a recent survey by the National Association of County and City Health Officials (NACCHO) found that two-thirds of local health agencies report their AI capacity is low or less than ideal. He outlines several key areas where AI can have an immediate impact:
- Data Analysis: An epidemiologist sifting through mounds of data could have AI handle 80% of the “grunt work,” freeing them to focus on pattern recognition and strategic oversight. AI can help clean and organize the vast amounts of incomplete and inconsistent data that plagued public health during the COVID-19 pandemic, where agencies often relied on fax machines and dealt with messy, non-standardized patient information.
- Strategic Communication: Communications teams can use AI to rapidly synthesize best practices and generate effective messaging during public health crises, such as a measles outbreak.
- Predictive Modeling: During the pandemic, leaders had to make critical policy decisions (e.g., curfews, bans on large gatherings) with incomplete data. Dr. Shah notes, “what AI has the potential to do is to help complete the incomplete data.” It can synthesize global information and model the potential impacts of different policy choices, enabling more informed, proactive decision-making.
- Addressing Misinformation and Inequities: AI is a double-edged sword. It can amplify existing biases and accelerate the spread of misinformation. However, it can also be a powerful tool to close trust gaps and combat these issues, underscoring the need for a responsible AI framework.
Dr. Shah concludes that AI enables a critical shift from reactive to proactive systems. He states, “AI doesn’t just save time; it saves lives.”
Re-Architecting the Health System with AI
Dr. Gianrico Farrugia, President and CEO of Mayo Clinic, describes how his institution began its deep investment in AI around 2018, long before the generative AI boom. The driving motivation was a recognition that the fundamental architecture of the healthcare system was flawed and incapable of delivering on future technological promise. The goal was not to pursue low-hanging fruit like administrative efficiency but to tackle the hardest problem: fundamentally improving clinical outcomes.
The Mayo Clinic Platform (MCP)
The genesis of the Mayo Clinic Platform was a strategic plan focused on three pillars:
- Cure: Shifting the mindset from managing chronic disease to actively seeking cures.
- Connect: Connecting people with data to generate new knowledge.
- Transform: Using Mayo’s resources to transform healthcare for everyone.
This led to the creation of a new data-centric architecture. A key strategic decision was to decentralize AI. Instead of a central team, Mayo empowered individual clinical departments to hire their own AI scientists and integrate AI as they would any new surgical or radiological technique. This created a diffused base of expertise and a culture of ownership, making the organization remarkably prepared for the arrival of generative AI.
When generative AI came, for us, it’s like, I wouldn’t say we told you so, but it’s like, ah, there you go. Here’s another tool. This is what we’ve been talking about. Now we can do it even better… It truly never was disruptive. It truly immediately became enabling.
From Predictive to Generative: Real-World Applications
Dr. Farrugia details a wide range of AI algorithms already integrated into practice, moving beyond predictive models to embrace generative AI.
- Early Proof of Concept: An early success was an AI algorithm that could diagnose left ventricular dysfunction from an electrocardiogram (ECG) with high accuracy, making nurses with the algorithm more effective than physicians without it. This provided a powerful internal and external proof of concept.
- In-Hospital Monitoring (CEDAR): In patient rooms, a system of cameras and sensors feeds data to algorithms that predict patient decompensation. A dedicated team of intensivists monitors these alerts, leading to documented decreases in hospital length of stay, ICU transfers, and mortality.
- Digital Pathology at Scale: Mayo has digitized 21 million pathology slides. In partnership with other companies, they have developed an LLM for pathology (
Atlas) and are now deploying AI-enabled digitized frozen sections. This allows for near-instant, detailed analysis of tissue samples while a patient is still in surgery, providing information far beyond what is currently possible. - Genomic Transformers: The clinic is developing an “exomic transformer” to predict patient response to therapies, such as for the 40% of rheumatoid arthritis patients who don’t respond to first-line treatment.
- Education and Research: AI is now a core part of the curriculum for all medical students, with master’s and PhD programs available. The ultimate goal is to coalesce all of Mayo’s vast research data—from registries to multi-omics data—into a unified “atlas” to spur research at an unprecedented scale.
Dr. Farrugia’s vision is not to simply use AI to optimize a broken system. He advocates for using AI to build a fundamentally new and better one, stating, “Healthcare is not perfect by any means, therefore let’s run and embrace AI.”