always interesting to read reactions healthcare tweets that exit the healthcare bubble I don't think this is unique to healthcare, but any system that currently relies on 1) payment based on gray area rules 2) the expectation that people will not do X thing because there's too much friction is going to blow up with AI tools and will need to be rethought from the ground up in healthcare the need is a completely different ruleset for payment + use of some form of ground truth dataset (e.g. scribe recordings) instead of proxies (e.g. CPT codes or what's in the EHR)

random thoughts on social norms and AI 1) as email gets flooded with AI slop, emails will need to "sell" themselves to agents instead of people. This will be both the quality of the email + context on your relationship with that person the agent has from other places 2) at what point would be feel comfortable with an agent in their group chat? I think there's a point at which having an agent both could be a lot of fun (e.g. prompt the group or play games within a group chat) but also be enormously helpful with logistics by getting more context on people. But I think we're pretty far from that culturally 3) One thing AI still lacks is feeling accountability. Whenever an agent asks me something, I don't really feel like I have to do it, vs. when a colleague does I'd feel bad if I didn't. But is there some point where that changes? Either feeling like the AI is more human or that the context from your interactions might pollute a given agent later and so you feel more obligation

Doulas, docs, and social media Let me start this by saying IMO doulas are hugely net positive, can be very helpful for people especially during the birth itself, and also improve outcomes especially for at-risk groups. But they also exist in a liminal space where they can be more...flexible in their opinions and have a pretty high variance even between doulas. We went to a doula-led birth class AND the hospital's birth class to prepare us for the birth. The hospital class was way more rigid, by the books, and started almost all answers with “it depends”. The doulas had much stronger opinions. One example was around “can I eat after my water breaks?” - the reason is because if you need to go into emergency surgery the food might cause you to aspirate and choke. The hospital was a strong “you should not, liquids only”, while the doula gave much more permission saying the labor was going to be extremely long and you should have energy if it's going to take 18-36 hours and eat well at the beginning. I'm not saying either is right, but it highlights the divide in medically correct vs. practically useful even if you're taking on additional risk yourself. Most of the opinions from the doula-led class were very useful and non controversial (positions to alleviate pain, things to pack for comfort during the hospital stay, etc.). But some of them seemed vaguely antagonistic towards hospitals. At one point they said the hospital makes most of their money from epidurals so they'll push you towards that, and my dumbass was about to “ackshuallyyyyy” but decided against it. Several other expecting parents in the class also came in with very tiktok-brained questions that came from a real place of curiosity. The doula also was able to correct/clear up some of the vague medical misinformation. This gray area and less formal space allowed people to ask questions about “things they heard” and feel like they might hear a more real answer. I'd bet they wouldn't ask their doctors those questions, so I think it's great the doula was able to course correct someone that could potentially have ended up down worse rabbitholes. For new parents who want answers and confidence, it actually feels more reassuring and engenders trust when someone has a strong stance. In some ways I wonder if more docs should have stronger stances on low-risk questions to make patients feel more confident. Though I'm sure another group of patients would also be annoyed at docs for doing that + docs still have to deal with liability if that low-risk question turns out to not actually be low-risk

It's crazy is how wide the variance of contracted prices can be between hospitals AND between payers. The Turquoise Health team graciously pulled the data so I could show you all just how wide that range is. Especially for C-Sections, where the swing can be $35- 40K between hospitals. Even WITHIN a hospital you could have the same C-Section costing twice as much depending on which insurance you have (e.g. NYU)! Obviously you have to layer outcomes on top of this. But damn...you could buy a Honda Civic and still save money if you went to Elmhurst (along with getting some of the best halal cart in the city). And that’s just for the facility component. Fwiw NYC is a bit of an outlier, most times it’ll look more like $10 - 18K ish in other cities

it's surprising how hard it is to find up to date studies on malpractice. A big part of this is data access. The National Practitioner Data Bank is public but only records paid claims and has no clinical detail which makes it pretty unhelpful. Below is a chart of malpractice claims for $1M+ and $2M+ per year Malpractice carriers seem to only give researchers access to their internal claims when the industry is going through some kind of crisis. The 2006 Studdert paper that everyone quotes had 5 malpractice carriers give rawdogged claims access to the researchers, which doesn't really seem to happen ever. Our current malpractice system creates an added issue that there really isn't a central repository to analyze claims, it's scattered across the carriers. Other countries collate it all into one place so you can actually use it for safety research + look deeper into doctors/institutions that might have more issues.

This is where agents could really change healthcare. So much of the market is fragmented that better matching is a much bigger unlock than it sounds. Agents can make those connections without everything needing to fit into a perfect taxonomy.

IMO AI enables the creation of new marketplace in healthcare in a few different ways 1) It makes inventory more discoverable - This can be knowing what's in stock at the pharmacy, what appointments are available, if a doctor still takes patients with X insurance, what kinds of patients are in the panel, etc. You can now verify this at scale with things like calls, structuring data in the EHR, etc. 2) Agents can match suppliers and buyers more easily - Agents can take requirements that buyers have match them to sellers even if it's a fuzzier match. Before you're need to create really specific tags to make that match happen, but now you can do it on pretty much anything. This becomes even more complex if the transaction requires 3+ entities. 3) You can bootstrap new networks quickly - If a company isn't in your network, you can automate a fax or call to transmit the same info. Using existing rails that anyone can participate in basically means anyone can be in your network. Examples of new marketplaces I'm personally invested in: Photon - better marketplace for providers, pharmacies, and patients by making pharmacy shoppable Turquoise Health - better marketplace between payers, providers, and employers by turning negotiated rate contracts into something more modifiable + able to be transacted on top of Double Blind Bio - better marketplace between sites and sponsors so they can find each other based on their needs I think there are many more to be built here

Malpractice reform needs to be part of any discussion on healthcare cost control. The data is all over the place, but it’s undeniable that doctors spend more on defensive medicine. I’m thankful to @nikillinit for letting me vent my ideas.

If you don’t find personal AI assistants to be useful the either 1) you are unimaginative 2) you have offloaded coordination to someone else who shoulders that burden on your behalf without you understanding that 3) you view doing administrative tasks as a core part of you identity

If you're a software engineer in healthcare, I bet you that we're going to be your favorite. Ship It is our new software engineering conference for healthcare. It's all workshops, you need to apply to get in so we only get serious people, and it's limited to 80 people. You can apply with a workshop idea you'd want to run! Some things I think would be good: 1) How are people budgeting and allocating for tokens 2) Wtf are harnesses and how do you use them well 3) How are you evaluating other AI tools as vendors vs. when you want to build internally But you can pitch whatever you want! Applications are due Wednesday, we'd love to see you there. We know how to throw a good conference that's actually educational, just ask anyone that's wearing a gender reveal colored bucket hat who's been to one

When AI starts doing clinical tasks, how should malpractice work? The subject of tomorrow's post - there are a few dimensions that AI is significantly changing how we think about what liability means IMO there's room for new malpractice carriers to be built that need different sets of skills (e.g. diligencing clinical agents) and provide different products (e.g. combo malpractice/professional liability/cybersecurity)