#91 – Eric Topol, M.D.: Can AI empower physicians and revolutionize patient care?
Dr. Eric Topol, founder and director of the Scripps Research Translational Institute, discusses AI's impact on medicine, emphasizing its potential to restore the patient-doctor relationship by offloading data tasks. He also shares his experience with the Vioxx controversy and highlights the need for physicians to organize for systemic change.
Deep Dive Analysis
15 Topic Outline
Early Career and Interest in Cardiology
Evolution of Mobile Health Monitoring
US Healthcare System's Resistance to Change
Mission and Transformation at Cleveland Clinic
The Vioxx Controversy and Personal Impact
Founding Scripps Research Translational Institute
Artificial Intelligence and Deep Learning in Medicine
Deep Learning for Image Recognition and ICU Monitoring
Continuous Glucose Monitoring and Personalized Nutrition
The Role of the Gut Microbiome in Health
Blind Spots in Medical Diagnosis and Treatment
Human vs. Machine: Restoring the Doctor-Patient Relationship
Physician Burnout and the Need for Activism
The Dream Experiment: Planetary Digital Health Infrastructure
The Value and Limitations of Genomic Data
3 Key Concepts
Deep Learning
A specific subtype of Artificial Intelligence (AI) that uses neural networks to process millions or billions of data points, such as images, speech, or text. It distinguishes features through layers of artificial neurons, often achieving superior accuracy in tasks like image recognition compared to human experts.
COX-2 Inhibitors
A class of non-steroidal anti-inflammatory drugs (NSAIDs), such as Vioxx and Celebrex, designed to selectively block the cyclooxygenase-2 enzyme. The promise was to relieve pain and inflammation more potently than traditional NSAIDs while sparing the stomach from adverse effects, though their selectivity was not as high as advertised.
Digital Twins in Medicine
A concept where a comprehensive digital representation of an individual is created, incorporating all their health data (genomic, phenotypic, metabolic, microbiome, etc.). This 'twin' would be continuously updated, allowing for personalized prevention and treatment strategies by drawing insights from billions of similar 'twins' globally.
8 Questions Answered
The major limitation is developing an 'ice pick view' of the patient, where specialists focus narrowly on their area (e.g., arteries for interventional cardiologists, arrhythmias for electrophysiologists) rather than the patient's overall cardiovascular health.
Medicine exhibits a general resistance to change, often likened to a 'sclerotic or ossified nature' of the medical community. Incentives are often tied to improved reimbursement rather than genuine advancements, leading to an unwillingness to move forward even with proven technologies.
Vioxx, a COX-2 inhibitor, was found to significantly increase the risk of heart attacks and strokes. Merck, the manufacturer, was accused of concealing data and denying the risks, leading to its abrupt withdrawal from the market and a subsequent period of personal and professional difficulty for those who exposed the issues.
Studies show that doctors are wrong in 40% of cases when asked about the cause of death, with autopsy revealing a different reason than what was determined pre-mortem.
The book's unique premise is that by leveraging machines more for data analysis and diagnosis, doctors can be freed up to spend more time with patients, thereby restoring the essential human doctor-patient relationship and fostering 'deep empathy'.
Currently, the Apple Watch's deep learning algorithm is primarily cleared for detecting atrial fibrillation. While it can pick up heart rate irregularities, it cannot yet diagnose other ventricular or atrial rhythms, though future algorithms are expected to expand its diagnostic capabilities.
The DNA of the gut microbiome is generally very stable from day to day, unless there is a significant change in diet (e.g., fiber content) or the use of antibiotics.
While the cost of a full genome sequence is now below $1000, its value is currently limited due to an inadequate number of sequenced individuals with accurate phenotyping. However, studies like Geisinger Health System's have found important pathogenic findings in about 5% of individuals.
11 Actionable Insights
1. Utilize Continuous Glucose Monitor
Wear a Continuous Glucose Monitor (CGM) for at least 90 days, even if not diabetic, to gain personalized insights into how your body responds to food, exercise, sleep, and stress. This continuous data helps you understand your unique physiological responses and learn about your body’s fuel partitioning, which is a constant learning process.
2. Prioritize Human Connection in Medicine
Seek out and support medical professionals who prioritize deep, empathetic conversations and truly listen to your life history and non-verbal cues. This restores the essential human bond in medicine, which machines cannot replicate, and is crucial for understanding a person’s real, deep issues.
3. Use Portable 6-Lead ECG
Consider acquiring a portable six-lead electrocardiogram (ECG) device, which is credit card-sized and connects to your phone, to monitor your heart rhythm and conduction system at home. This provides high-quality data, can be used for self-stress tests, and offers remarkably insightful information beyond a simple pulse check.
4. Question Primary Prevention Statins
Engage in an informed discussion with your doctor about the actual benefits of statins for primary prevention, recognizing that current clinical trial data suggests a small number of people (e.g., 12 out of 1000 over five years) derive benefit. Understand the limitations of long-term data and the heterogeneity of patient response when considering this widely prescribed medication.
5. Be Wary of Executive Physicals
Avoid unnecessary ’executive physicals’ or excessive medical screenings, as these often lead to ‘overcooked’ individuals and incidental illnesses without clear benefit. Such extensive testing in affluent populations can result in bad outcomes and unnecessary procedures.
6. Understand Gut Microbiome Utility
Approach gut microbiome testing with a clear understanding of its current utility: it’s primarily a readout of your stable bacterial DNA, useful for predicting individual food responses, but not yet a malleable state for direct manipulation outside of specific conditions like C. difficile colitis. Be cautious of companies that only report bacterial density or make unsubstantiated claims about probiotics.
7. Use Apple Watch for AFib (Cautiously)
If at risk for atrial fibrillation, consider using the Apple Watch’s ECG feature, which can detect irregular heart rhythms based on your resting heart rate. However, be aware that for individuals not at risk, it can lead to false positives, anxiety, and unnecessary medical workups.
8. Engage in Medical Activism
Physicians and patients should organize and become activists to advocate for a healthcare system that prioritizes the patient-doctor relationship and addresses issues like physician burnout. This collective ‘gravitas’ can challenge administrative pressures and restore humanity to medicine.
9. Embrace Challenging Personal Projects
Continuously seek out and engage in ‘big projects’ or challenges that are a ‘reach’ to keep yourself motivated and growing. This personal strategy helps maintain engagement and fosters innovation.
10. Empower Self-Diagnosis (Common Conditions)
Explore tools and knowledge to enable ‘doctorless diagnoses’ for common, benign conditions like ear infections, urinary tract infections, or skin rashes. This shift towards patient autonomy can decompress routine medical visits and change the landscape of specialty care.
11. Inquire About AI Colonoscopy
If undergoing a colonoscopy, inquire about the availability of deep learning machine vision, which has been shown to pick up polyps (e.g., 20% more) that are routinely missed by human operators. This technology can improve the thoroughness and accuracy of polyp detection.
5 Key Quotes
We could be facing a public health disaster.
Eric Topol
This is my fucking lane.
Judy Mellinick (quoted by Eric Topol)
The biggest blind spot is how poor we are in diagnosis, no less in treatment.
Eric Topol
The point being is that machines, that's what we don't want machines. They don't have context. They're not going to be able to truly understand all the non-verbal communication and the real issues of a person that are deep. And that's where the humanity, we need to bring it back.
Eric Topol
The EMR was created as a billing solution, not a clinical solution.
Peter Attia (attributing to Eric Topol)