JAAPA Podcast: "AI in Cardiovascular Practice"

Date: July 2, 2025
Podcast: Journal of the American Academy of Physician Assistants
Guests: Marci Farquhar Snow, Amy Simone, Sheil Singh
Guest Co-host/Data Scientist: Ben

Overview

This episode explores how artificial intelligence (AI) is transforming cardiovascular medicine, both in clinical practice and research. Hosted from the 2025 AAPA conference in Denver, the panel features the authors of the JAAPA article "Artificial Intelligence and Cardiovascular Practice" and guest data scientist Ben. The discussion defines key AI concepts, reviews real-world use cases in cardiology, highlights the opportunities and limits of current technology, and dives into ethical, educational, and practical implications for advanced practice providers (APPs).

Defining AI & Key Concepts (00:06–04:22)

Summary:
Ben offers foundational explanations of AI, machine learning, linear regression, and predictive analytics.

Key Points:

• Machine Learning (ML): Computers learn patterns from data to make predictions/decisions without explicit programming.
  • Example: “A machine learning model trained on thousands of electronic health records might be able to flag a case where somebody may be at high risk for a heart attack.”
    — Ben (01:40)
• Linear Regression: A statistical tool to understand relationships between variables.
  • Example: “Converting Celsius to Fahrenheit is a good example—a linear equation applied to an input.”
    — Ben (02:19)
• Predictive Analytics: Uses current/historical data to predict future outcomes.
  • “Any level of insight or information we can get to act sooner—that’s the real value.”
    — Ben (03:44)

The Authors’ Backgrounds & Motivation (04:22–08:34)

Summary:
The authors share their varied experience in cardiovascular medicine and their common goal to clarify AI’s role for advanced providers.

Highlights:

• Marci: Explains AI's varied use in diagnostics, decision making, and efficiency.
  • “By writing this article, we wanted to show that as AI evolves, advanced practice providers need to understand general AI terms as they're being used in our practice.”
    — Marci (05:29)
• Amy: Discusses disruption as positive; self-education was key to participating in AI adoption.
  • “AI has become a daily disruption... and I'm by no means inferring disruption is a negative thing. Right. It's fuel.”
    — Amy (06:40)
• Sheil: AI is new but exciting; responsible research is critical.
  • “I was really excited... to learn a little bit more about how I can use AI in my daily life, but also use it responsibly.”
    — Sheil (08:05)

Case Study: Real-World Impact (10:10–13:10)

Case Overview:
A 78-year-old Black woman in a rural area with aortic stenosis benefited from AI-driven diagnostic tools, predictive analytics, tailored education, and decision support, enabling timely, effective care and better outcomes.

Key Insights:

• Access to Care:
  • “AI helped to mitigate disparities and improve her access to care... AI applications can certainly increase reach and ensure patients who need treatment are sought out with proper representation.”
    — Amy (10:49)
• Predictive Analytics & Personalization:
  • AI enhanced operational efficiency, care timing, and produced patient education at appropriate literacy levels to improve shared decision-making.
• Quality and Outcomes:
  • “Treatment plans were changed... based on predictive modeling... Kind of astounding if you think about the impact downstream.”
    — Amy (12:34)

Clinical Perspectives: Machine Learning in Cardiology (13:10–16:45)

Summary:
The group discusses how ML tools currently assist clinicians.

Machine Learning Tools:

• Risk Scoring:
  • TIMI, HEART, and Edax scores help triage chest pain, facilitate shared decision-making, and streamline care.
  • “Machine learning is a powerful tool that can analyze automatically the data that the electronic medical record provides, and that leads to efficiency.”
    — Marci (14:37)
• Deep Learning:
  • Integrates multi-modal data (e.g., EKG, echocardiograms) for advanced analysis.

Clinical Utility:
Supports novice providers, reduces burnout, and increases efficiency.

AI Subfields Explained: Deep Learning, NLP, Generative AI (17:18–20:59)

Summary:
Ben clarifies subcategories of ML and their healthcare applications.

• Deep Learning: Inspired by neural networks, captures complex patterns (e.g., ChatGPT).
• Natural Language Processing (NLP):
  • “A spam filter on your email... That is a form of natural language processing.”
    — Ben (18:47)
• Generative AI:
  • “Generative AI... is a form of AI that can create new content like text or images or audio... based upon previous information.”
    — Ben (19:32)
• Key Point:
  • “AI is not new, but what's really changed here is... the amount of computing power and data.”
    — Ben (20:06)

Specific AI Applications in Cardiology (21:09–29:22)

Machine Learning:

• Used in existing cardiac risk scores, hospital readmission screening, and large-scale registry data.
• Global Health Example:
  • AI algorithms increase EKG interpretation access in rural Kenya for LV dysfunction detection.
    — Marci (21:56)

Deep Learning:

• Consumer wearables (Apple Watch) detect arrhythmias, but data management/interpretation remains a challenge.

Natural Language Processing (Amy, 23:28):

• Reads physicians’ dictated echo reports, improves clinical documentation, suggests billing adjustments, and creates personalized patient education materials.

Generative AI (Amy, 25:06):

• Enables risk prediction for surgical complications, creates individualized 3D cardiac models for surgical planning, enhances patient-provider communication.

Consistency (Marci, 27:15):

• “With this machine learning, it is usually a more consistent manner... that can leverage the benefits versus the potential negative of AI not interpreting appropriately.”

Research Applications (Sheil, 28:14):

• Automates literature/data review for eligibility screening, increasing speed and accuracy, but requires precise prompting and cautious code-writing.

Rule-Based Systems in Cardiac Care (30:47–36:21)

Summary:
Rule-based “if-then” systems form the foundation of many cardiology decision tools.

Examples:

• Risk calculators for statin initiation, AFib stroke prevention, medical/device/surgical management decisions.
• Mobile telemonitoring systems generate alerts based on patient inputs (e.g., weight, HR).

Limits:

• “...they don't learn. So in a complex patient, you still need that provider... to consider before you make that decision making.”
  — Marci (32:14)
• Clinical judgment remains critical; algorithms are a guide, not a replacement for hands-on care.

Robotics, AI Integration & Medical Education (37:01–42:20)

Summary:
AI and robotics are increasingly visible in procedural and educational settings.

Clinical Use:

• Robotic Surgery: Da Vinci system for cardiac, urologic, gynecologic, and general surgery.
• Imaging: AI improves diagnostic interpretations—detects cancer, fractures, lung disease with high accuracy; flags incidental findings.
• Wearables: Implantable sensors for real-time patient monitoring (e.g., predicting fluid overload in heart failure).

Education:

• AI can assess learning styles and progress, deliver tailored education, track performance, and may soon play a role in individualized medical education.

Challenges & Limitations (Ethical, Legal, Practical) (42:55–63:33)

Job Security & Provider Role:

• “It's already working. It can enhance our jobs, it could promote efficiency, decrease workload... But you still need a real person to discern or interpret did the AI work appropriately?” — Marci (43:20)
• “Nothing can ever replace the patient in front of you... the clinician, the humanistic element.” — Amy (35:30)

Limits/Challenges:

• Learning Curves: Integrating new technology is always an adjustment, as seen with EHRs.
• Bias & Data Quality: Algorithms may amplify biases from flawed or limited datasets.
  • “Is this going to create biases or are they going to class you in a different category?” — Marci (47:46)
• Privacy & Security: Risks of adversarial attacks and privacy breaches go beyond de-identification.
  • “You are trying to protect your participants from adversarial attacks that might trick models into giving up information that they were never meant to share.” — Sheil (53:54)

Cost/Ethical Considerations:

• Initial investments are high, benefits might favor larger organizations first.
• Use of AI for predictive risk (e.g., sports physicals) may cause discrimination (scholarship eligibility, insurance, employment).
• No overarching federal guidelines/regulation (yet).

Data Sharing & Research:

• “Sharing healthcare data at scale means we have to be incredibly careful. Data is a fuel for AI...”
  — Sheil (56:04)
• Large national/regional cardiovascular registries can power more robust analytics but raise questions about consent and privacy.
• “Garbage in is garbage out.” — Marci (58:17)

Patient Empowerment

• “AI empowered patients”: Anecdotes of families and lay people using generative AI tools to address dietary needs or rare disease research; patients can now access and act on more information than ever.
  — Ben (64:20)

Notable Quotes & Highlights

• “AI has become a daily disruption... and I'm by no means inferring disruption is a negative thing. Right. It's fuel.” — Amy (06:40)
• “Machine learning is a powerful tool that can analyze automatically the data that the electronic medical record provides, and that leads to efficiency.” — Marci (14:37)
• “These are not science fiction. These are capabilities within our reach today.” — Amy (12:58)
• “Nothing can ever replace the patient in front of you…that humanistic element to help really guide patients along the right care pathway.” — Amy (35:30)
• “Data is the fuel for AI. The better the data... the more sophisticated our models can be. But when it comes to data sharing, we're talking about protecting against things like jailbreaking.” — Sheil (56:04)

Timestamps for Key Segments

• 00:06: Opening; AI ubiquity and definitions (Ben)
• 04:22: Authors’ backgrounds & article purpose
• 10:10: Case Study—AI in real cardiovascular care
• 13:10: Clinical AI/ML tools (risk scoring, deep learning)
• 17:18: AI subfields: deep learning, NLP, generative AI (Ben)
• 21:09: Cardiology use cases—ML, deep learning, wearables
• 23:28: NLP and generative AI in cardiac imaging, documentation (Amy)
• 28:14: Generative AI in research (Sheil)
• 30:47: Rule-based systems: telemonitoring, risk calculators
• 37:01: Robotics and AI: procedures, imaging, monitoring, education
• 42:55: AI’s limits, ethical, cost, and workforce implications
• 56:04: Data sharing, quality, privacy (Sheil/Marci)
• 64:01: Ben on patient empowerment and clinician advice

Tone

The discussion was collegial, candid, and enthusiastic, driven by optimism for AI’s promise but grounded by caution and realism about its limitations, ethics, and dependence on quality human oversight.

Summary Takeaways

• AI is already foundational in cardiovascular practice, from diagnostics to decision support to research.
• Machine learning, NLP, and generative AI are all distinct but increasingly interconnected and impactful.
• Providers must understand AI’s basic concepts to leverage its strengths and guard against pitfalls (bias, privacy, reliance).
• AI and rule-based systems augment—not replace—clinician expertise and the crucial human connection in medicine.
• Ethical, legal, and implementation challenges remain and will require regulatory attention, especially as AI tools’ power and data requirements grow.
• Patient and provider empowerment, efficiency, and improved access are among the greatest payoffs, provided healthcare professionals remain vigilant about data quality, privacy, and the irreplaceable role of human judgment.