#5 - Doctor's Notes: When AI Writes Your Medical History

Code & Cure

What if an AI could write your medical chart—and what happens when it gets it wrong? Doctors have long lamented the paperwork that comes with every patient encounter. “Charting was the bane of my existence,” admits Dr. Laura Hagopian, an emergency physician who’s spent countless hours piecing together fragmented notes and outdated records. Could artificial intelligence finally lift this administrative weight?

Recent advances in large language models promise to generate discharge summaries as accurately as seasoned clinicians, potentially returning precious time to the bedside. By training on thousands of patient encounters and lab reports, these systems can stitch together coherent narratives of care—micro-diagnoses, treatment plans, and follow-up recommendations—at a speed no human chart-writer can match.

Yet with speed comes risk. When an AI hallucination slips into a diagnosis and becomes enshrined in a patient’s record, who is accountable? Dr. Hagopian highlights the stark difference between human and machine error: “I feel very different about a human making a mistake compared to an AI making a mistake.” As trust in automated documentation grows, so too do questions about responsibility, oversight, and patient safety.

In this episode, AI researcher Vasanth Sarathy and Dr. Hagopian peel back the layers of these complex issues. They explore the nuts and bolts of AI summarization algorithms, discuss promising clinical trials, and weigh the ethical dilemmas of delegating clinical judgment to code. How do we ensure that efficiency doesn’t override accuracy when every data point can mean life or death?

Whether you’re a clinician craving relief from chart fatigue, an AI developer pushing the boundaries of what’s possible, or a patient curious about who’s really recording your health story, this conversation offers a vital look at the future of medical documentation. Join us as we navigate the promise—and the pitfalls—of letting machines tell our most critical health narratives.

References:

Physician- and Large Language Model–Generated Hospital Discharge Summaries
Christopher Y. K. Williams, et al.
JAMA, Internal Medicine, 2025

Credits:

Theme music: Nowhere Land, Kevin MacLeod (incompetech.com)
Licensed under Creative Commons: By Attribution 4.0
https://creativecommons.org/licenses/by/4.0/

Розділи

1. AI in Medical Documentation Intro (00:00:00)

2. The Challenge of Clinical Charting (00:01:22)

3. Types of Medical Summaries (00:03:07)

4. Who Uses Medical Summaries? (00:05:43)

5. LLMs vs Human Summarization (00:09:10)

6. Accountability and Trust Issues (00:16:25)

7. Making LLMs Better at Summarization (00:21:30)

8. Risks of Overreliance on AI (00:28:27)