AI Meets the Healing Arts

Lucy Family Institute for Data and Society

Available on-demand

Join Dr. Michelle Hermiston, an Iowa farm kid turned global oncology leader, as she bridges the gap between childhood leukemia survival rates in Vietnam and much higher survival rates in high-income countries. Witness how “train the trainer” models and AI revolutionize pediatric care through bidirectional learning between Western medicine and clinicians in settings where resources can be inconsistent. Explore a transformative vision of health equity that offers actionable hope for every child worldwide.

The Health AI Forum is sponsored by the Lucy Family Institute for Data & Society. Please visit the Health AI Forum website for more information.

The current landscape of global oncology presents a stark moral challenge: a child’s survival is often determined more by their zip code than by the limits of medical science. While survival rates for childhood leukemia have reached 90% in high-income countries, the rates in low- and middle-income countries (LMICs) fluctuate between 5% and 60%. Dr. Michelle Hermiston’s trajectory—from her roots in rural Iowa to her leadership at VinUniversity in Hanoi—provides a strategic blueprint for “leapfrogging” traditional developmental timelines to achieve global health equity.

The Vietnam Model: Bidirectional Innovation

Rather than a traditional, top-down colonial approach, the “Vietnam Model” centers on a “Train the Trainer” philosophy, partnering with all 14 pediatric oncology hospitals in the country to avoid creating inadvertent disparities. Crucially, this is a bidirectional partnership. Western medicine often “overtreats” patients, but by observing resource-varied settings where clinicians must modify protocols due to factors like antimicrobial resistance, US physicians are learning how to safely de-escalate therapies. This creates a global exchange where efficiency in Hanoi informs excellence in San Francisco.

AI as the Catalyst for Scale

In environments like Bach Mai Hospital, which manages a staggering 19,000 outpatient visits daily, AI is a strategic necessity. Dr. Hermiston identifies three primary “pain points” for technological intervention:

Diagnostics: Utilizing AI-assisted pathology and medical imaging to manage overwhelming volumes at public hospitals, providing a critical “first preview” of cases.
Operational Efficiency: At the National Children’s Hospital, staff often deal with 200 chemotherapy visits daily without knowing who will arrive. AI can forecast resource needs and track patient journeys to ensure life-saving drugs are ready.
Patient Navigation: In settings where the nursing ratio can be as lean as one nurse for 40 patients overnight, AI-driven avatars can provide essential symptom tracking and literacy support that staff simply cannot.

The Data Validity Warning

As a medical academic liaison, Dr. Hermiston warns of “Garbage In, Garbage Out.” AI tools lack genetic ancestry diversity, which carries lethal risks. For example, many Vietnamese children possess a specific genetic polymorphism; if given a standard Western dose of certain chemotherapies, they suffer profound neutropenia and die from toxicity. Furthermore, strategic implementation must navigate local regulations; new laws in China and Vietnam ban cloud-based data storage, necessitating a shift toward local, flexible tools like REDCap to maintain research momentum.

The Human Procedure

Ultimately, technology must serve the “Healing Arts.” Dr. Hermiston advocates for “talking as a procedure,” arguing that AI’s greatest value is “buying back time.” By automating administrative burdens, clinicians can return to the bedside to engage in the deeply human work of communication, empathy, and medicine.

These highlights provide a framework for integrating technology and ethics in 21st-century medicine:

Evaluate the “Leapfrog” Potential of Emerging Economies: We must move past the idea that LMICs must follow a slow, multi-decade developmental path. By implementing AI and modern training models today, these regions can bypass traditional infrastructure barriers.
Assess the Value of Bidirectional Data Flow: Global health is not a one-way street. We must actively study how resource-varied settings safely de-escalate toxic therapies, providing vital data that can reduce the burden of treatment for patients in high-income countries.
Critique the Limits of Western-Centric Data: We must address the “external validity” gap. AI tools trained without data on genetic polymorphisms—such as those causing fatal neutropenia in Southeast Asian populations—risk perpetuating dangerous clinical disparities.
Prioritize Culturally Competent Technological Design: For digital tools to be effective, they must be tailored to local norms. An example is Angelica Garcia Martinez’s work on AI chatbots for nutrition support, which respects the central role of food in Asian and Hispanic caregiving.
Synthesize Operational AI with Clinical Scarcity: The most immediate impact of AI is solving bottlenecks in settings with a 1:40 nurse-to-patient ratio. Operational AI for patient tracking and chemotherapy forecasting is a moral imperative to save lives in overwhelmed systems.

“Kids don’t choose their zip code to which they are born. And so then the question is, do we have an obligation to all children to improve therapy?” — Dr. Michelle Hermiston
“In the country of Vietnam… they had 183 inpatients that day. They only have 120 beds. Think about that—one child per bed is a luxury in many settings.” — Dr. Michelle Hermiston
“Half of what you learn in medical school will be shown to be dead wrong or out of date within five years of your graduation. The trouble is that nobody can tell you which half.” — Dr. Michelle Hermiston (Reflecting on the UCSF Inquiry Curriculum)
“Talking is the most important procedure of our profession. Our words can be medicine.” — Dr. Michelle Hermiston
“Partnering is key and there should never be anything about me without me. You have to have your partners at the table.” — Dr. Michelle Hermiston