Health Data Ethics

Part two of my breakdown of the White House National AI Policy Framework — what it says about workforce, what it leaves out, and what it would take to become law. The workforce section has good instincts but no mandates, no funding, no timelines. In healthcare, this can create a patient safety problem, if our health systems don't fill this gap thoughtfully. The legislative road is crowded and uncertain, but even if codified into law, this federal posture is a light touch with the rest landing on health systems.

This week's episode of the Health Data Ethics Show: The White House just released a four-page legislative framework asking Congress to pass national AI policy this year. Not a law. A wish list, but one that tells us a lot about where federal AI policy is heading. In this episode I break down what it means for healthcare governance: the preemption debate, FDA as the designated health AI gatekeeper, and the notable absence of HIPAA from the entire document. The governance responsibility has always sat with health systems. This framework confirms the federal government intends to keep it that way.

In this week's episode of the Health Data Ethics Show, I dig into a new Nature Medicine paper that stress-tested ChatGPT Health across 960 clinical vignettes. ChatGPT Health performs remarkably well in the middle of the acuity spectrum, correctly triaging semi-urgent and urgent cases at rates that rival clinical judgment. At the extremes, though, it struggles — undertriaging more than half of true emergencies and triggering crisis resources for suicidal ideation more reliably when patients had no plan than when they did. I walk through the methodology, the results, what they reveal about the limits of benchmarks like HealthBench, and what I think health systems and patients should take from it.

Last week, Anthropic published new research on the labor market impacts of AI. In this week's episode, I break down what the paper actually says, why the gap between theoretical and observed AI exposure matters, and which workers and which sectors are most impacted. Those workers, and the people reading these articles, are also your patients. They're walking into your hospital carrying headlines, anxiety, and ChatGPT conversations from the parking garage — and most governance frameworks have no place to put that.

In this episode, I invite special spousal guest Brad Owens to help explore a landmark federal court case involving Workday’s AI-powered hiring tools and potential implications for AI governance in employment and healthcare. Join us as we dissect the legal, ethical, and technological facets of AI decision-making and bias mitigation. EEOC Four-Fifths Rule - https://www.eeoc.gov/laws/guidance/questions-and-answers-clarify-and-provide-common-interpretation-uniform-guidelines NYC Bias Auditing Law - https://www.nyc.gov/site/dca/about/automated-employment-decision-tools.page

In this week's Health Data Ethics podcast, I dissected a 2023 JAMA study on using a voice-based AI tool to coach patients through insulin titration in type 2 diabetes. Big thanks up front to Aida McCracken who listened to me work through this one! The results were encouraging—patients hit optimal dosing in 15 days vs. 56+ for standard care. But governance questions kept nagging me. → Self-titration works for insulin because it's clinically established. Try the same with other drugs and you're in much riskier territory. → Study was 8 weeks. Diabetes is lifelong. What happens when engagement drops or the smart speaker isn't supported anymore? I'm all for empowering patients to follow validated protocols. We need scalable chronic disease solutions. But let's innovate with eyes open about equity, sustainability, and regulatory frameworks.

This week, I recorded a new episode of Health Data Ethics about FDA and EMA’s joint principles on AI in drug development. I used ChatGPT to help generate a draft. I loved it, until I started fact checking. It cited a section of a review paper that didn’t exist. Referenced a tool that never appeared in the text. When I called it on its hallucinations, it gave me fake quotes with fake page numbers. I was relying on this to build the backbone of an episode about AI transparency and instead, I scrapped the whole thing and started over from scratch. I texted my husband mid-edit: “If I'd recorded this I'd have seriously undercut my credibility with anyone who wanted to check.” He sent back: "The difference between enterprise and demo AI in two texts." In this episode, I talk about that failure—mine, and the model’s—and what it tells us about algorithmic bias and creating a culture of transparency.

Just had a brilliant conversation with Sahar Hashmi MD-PhD—one of my favorite people to talk AI-in-healthcare with. Dr. Hashmi made a powerful case for customized AI workshops tailored to clinical teams. We also unpacked: - What healthcare can learn from other industries - Why assessing AI readiness is foundational - The value of an AI Center of Excellence as a coordination engine - A Flexner moment for AI in healthcare If you're building or scaling AI in a provider org and haven’t asked “who owns readiness?”, this conversation is for you.

I recently spent time unpacking the FDA’s clinical decision support software guidance and what it really means for healthcare organizations deploying AI. At the center of the guidance is a simple question: when does software cross the line into being regulated as a medical device? The FDA lays out specific criteria that hinge on how recommendations are generated, how they are presented, and whether a human can independently review and understand the basis for those recommendations. If clinicians cannot reasonably evaluate or challenge an AI’s output, organizations may find themselves in regulated territory whether they intended to be there or not. Understanding where human judgment sits in the loop is essential for compliance, trust, and responsible scaling of AI. If you are deploying or governing clinical AI, this is guidance worth revisiting with both legal and clinical stakeholders at the table.

In this week’s Health Data Ethics episode, I break down four proposed bills in Ohio that aim to regulate artificial intelligence. These aren’t laws (yet). But they are early indicators of where state-level AI governance is headed, and the healthcare sector is definitely in scope. We also zoom out to look at where these proposals fit within wider state and federal activity on AI.

In this week's podcast episode, I had the chance to chat with Nicholas Woo of AlignAI about AI governance by design at multiple stages of AI development and implementation. We also talked about how easily great technological solutions can get stalled when they don’t fit into existing workflows or when there’s no clear operational champion. If you're thinking about AI governance, this episode is for you

In this week’s episode of the Health Data Ethics Podcast, I dive into a recent JAMA Network Open study that looked at ambient AI scribes in primary care settings — what worked, what didn’t, and where we still need better evidence. I also tackle what makes a successful scribe deployment, and end by musing about what ROI really means. Is it always dollars? Or is reducing burnout enough?

We’re officially in a new era: the Joint Commission has released guidance on the responsible use of AI in healthcare, developed alongside the Coalition for Health AI (CHAI). The guidance lays out a clear framework for responsible healthcare AI that includes: 🏛️ Strong governance structures 🔐 Patient privacy & data security safeguards 📊 Ongoing quality monitoring 🧑‍🏫 Workforce education & training ⚠️ Risk and bias assessments 📝 Anonymous safety event reporting In the episode, I share some strategies orgs can use to operationalize these principles. If you’re in healthcare IT, compliance, clinical leadership, or just AI-curious, I highly recommend giving this guidance a read. And if you’d rather listen your way through it with me, the link to the full episode will be in the comments!

This week on the Health Data Ethics Podcast, I sat down with Doug Meil to talk about his new book, "The Rise and Fall of Explorys and IBM Watson Health." What started as a discussion about a decade-long journey in health tech turned into a reflection on ambition, acquisition, and the hard-earned lessons of analytics. Doug talked about missed opportunities, broke down the reasons why they were missed, and gave great advice for what future innovators can take away from Explorys's story. If you’re working in healthcare data at all, it’s essential listening. One key theme that stuck with me: Iterate relentlessly, but don’t ignore the fundamentals—especially data access. Also loved his advice for the next generation of health tech leaders: Build for impact, not just headlines. Check out our conversation if you’re thinking about where health AI is headed, or just want a better understanding of how big bets like IBM Watson Health can shape the industry.

Just had a fantastic conversation with Kylen Bailey, Executive Director of Growth at Clarium, on the latest episode of the Health Data Ethics Podcast. We dug into a topic that doesn’t get nearly enough attention in the AI-in-healthcare hype cycle: the supply chain. More specifically, how AI can help us connect the dots across healthcare systems in a way that’s actually usable and secure. What really stood out in our conversation: The importance of co-developing AI tools with health systems, not just for them Why data integration isn’t just a tech challenge, it’s a collaboration challenge The ongoing need to treat data safety and security like mission-critical infrastructure And how AI can deliver real operational value if we build it with the right people in the room Kylen brought such clarity to the conversation—especially on the role of trust, transparency, and partnerships in getting these tools from pilot to production.

In this week's episode of the Health Data Ethics Podcast, I dig into something that doesn’t get enough airtime in healthcare AI conversations: what we can learn from the projects that don’t work. A recent MIT study suggests that up to 95% of AI pilots don’t deliver a measurable return on investment. That statistic may feel discouraging, but it’s also a great jump off point for investigation. What's working, and what's not, and why? In this episode, I talk through two recent experiments with generative and agentic AI and try to pull out the lessons: Where agentic AI tends to break down in real-world workflows How governance and oversight need to evolve when AI systems are making decisions And why it may be more useful to think of AI as a new class of labor rather than a new type of software Healthcare has always been a hard place for innovation, not because we don’t want change, but because the consequences of failure are high, and the systems are complex. That makes it even more important to treat each failure, false start, or stalled pilot not as a sunk cost, but as feedback.

This week on the Health Data Ethics Podcast, I unpack the recent announcement from CMS about the Health Technology Ecosystem and what it could mean for patient access to data. I dive into: Why CMS is using a pledge instead of regulation The current state of HIEs and where TEFCA fits in The real risk: that this pledge ends up helping flagship systems while leaving rural hospitals and safety-net clinics behind One zillion thanks to Kara Justi for sharing her wealth of knowledge about HIEs and TEFCA as I prepped for this podcast!

On this week’s Health Data Ethics Podcast, I break down the AI Action Plan released in July 2025—a big move in how the U.S. is approaching artificial intelligence policy. I walk through the plan’s three central pillars and what they could mean for the future of AI in healthcare and beyond. We talk about: 🏛️ Executive orders that now mandate ideological neutrality in federal AI systems 🧱 The massive infrastructure investments needed to support trustworthy AI 🩺 The potential impact on healthcare One theme I keep coming back to: transparency and accountability. If you're working on AI implementation, governance, or policy in a health system, this episode is for you.

In this week’s episode of the Health Data Ethics Podcast, I dove into one of my favorite recent papers: "Your Brain on ChatGPT: Accumulation of Cognitive Debt When Using an AI Assistant for Essay Writing." I loved reading about how the researchers explored the mental trade-offs when using ChatGPT for writing tasks. In the episode, I unpack: The study’s experimental design and takeaways What this could mean for AI in healthcare How Dune quotes are the key to my heart

I sat down with John Valutkevich from Drummond for the latest Health Data Ethics Podcast to talk through their new report on AI adoption in healthcare. We dug into the gap between curiosity and actual deployment. Everyone wants AI, but fewer are ready for the governance, risk management, and infrastructure it really takes to do it well. John shared insights on what’s working, where trust and explainability still fall short, and what healthcare orgs can actually do right now to move from buzz to meaningful adoption. If you're thinking about how to integrate AI into clinical settings, without skipping the hard stuff, this episode’s for you.