Everyone wants to build agents for hospitals, but what do they actually do?

We walk through a few end-to-end examples in the latest post like someone with a smoking history that gets a CT for a lung cancer screening where a nodule gets found.

Think about what that means for ering, submitting to registries, etc. It's a lot of clinical admin work!

Company idea: World's Fair for healthcare

There used to be the World’s Fair which would allow people to see and experience a lot of advances in consumer technology (electronics, transportation, etc.). More recently there were the humanoid olympics in China for robots, a great eing what robots can do.

I think one of healthcare’s big issues is that people aren’t amazed and excited by healthcare technology. Have you watched a robotic surgeon do their work? Or seen what happens in the NICU to save a child? Or seen how the cutting edge physics of the MRI are used? They used to have premature babies in chicken incubators at the Coney Island fair to show how they could survive, which must have seemed amazing at the time.

There’s a big opportunity to create a new World’s Fair to show people all the advances in healthcare. Imagine seeing a person with a Neuralink implant live, or using one of these Wandercraft exoskeletons, or see a Trilobio automated lab in action, or see the internal medicine resident speedrun using dot phrases in Epic.

More than anything I think it would actually give people optimism and positivity of what’s possible in healthcare. And if you’re building robots in healthcare...that’s one of our hackathon’s request for startups.

you can sign up here to get it: outofpocket.health

we've also got some events we're announcing for healthcare people!

There's a race afoot to build AI agents for hospitals using different wedges

Tomorrow's post is going to walk through some of the tasks that an agent in a hospital is actually doing + the expansion into become more of an operating system within a hospital

When writing this it laces we are currently throwing highly trained bodies at an administrative problem. On top of that, it helps us rethink certain workflows if you imagine that every single specialist type is looking at the same picture, chart, etc.

In our free course tomorrow, we'll explain agents and see how they do with different back-office healthcare tasks like eligibility checks, prior auth, etc.

even if you can't make it, you'll get the slides and recording: outofpocket.health/courses/ai-age…

If you're running a healthcare company, where does it make sense to build agents?

Today it seems like people want to just throw the amorphous "agent" term at everything, but maybe not all of those workflows are good fits. Workflows that ARE good for agents:

1) You spend a ton There are pretty clear rules that you could write out
3) You can measure performance with a specific metric
4) Downside risk is low to test

writing out a series of other healthcare ideas someone should build leading up to our hackathon

first part is here, you can sign up to get part 2 coming out soon: outofpocket.health/p/healthcare-i…

Company idea - Instinct for healthcare workers

It’s clear to me that people are going to try to use personal AI tools like Instinct to do things at their job. Does your job require you to get data from different places and fill out a lot of forms? Or send a lot of follow up pdates to various google docs where you lowkey don’t know where that doc even lives?

This tool is EXTREMELY tempting to use for that because it’s excellent at that.

However this is also a security and privacy nightmare. These personal AI tools can basically mimic you as they go through your systems which is what makes them useful but also very scary. It even knows to switch to a YouTube short for like 10 seconds before switching back to the task at hand.

So any company that touches PHI, or sensitive data has to watch out because these tools will find it and accidentally expose it.

But again because it’s such a useful tool for employees you are 100% going to have people trying to use it for work at healthcare companies. So your options are:

1) Do your best to fight it and put in internal security that watches for actions that look bot-ish. However that takes internal resources and will always be an arms race against the personal AI tools that find workarounds.

2) Create compliant paths/tools that work like personal AI tools but have the appropriate security/privacy guardrails.

There’s probably opportunity in both of these buckets, but I’m more interested in the second. Something like an agent that gets its own login and virtual desktop on the org’s network instead of impersonating with cached passwords, the brain can be a cloud model under a BAA, memory is maintained locally, etc. Each employee’s agent gets different permissions/requirements of human-in-the-loop depending on their role.

There are some seedlings of companies attempting to do this. Contxt is one for regular enterprises, Heidi’s new agent is more healthcare specific. But there is a lot more to be done here and probably could be more specific to different roles in healthcare (e.g. Instinct for clinical research coordinators).

We've been doing our headhunting service for 2 months and just placed our first person!

We also have some ballers from the Out-Of-Pocket network that have joined our headhunting pool recently:

1. Staff engineer with 10+ years of experience building software and leading focused on hands-on healthcare data infrastructure, including member and claims data. Open to engineering, data or technical co-founder roles.

2. Growth marketer with 10+ years of experience building paid acquisition from scratch across healthcare and software. Looking for a Series B/C healthtech company selling to clinicians.

3. Strategy and operations leader with experience in value-based care, consulting and venture building. Open to chief of staff, strategy, partnerships, operations or business co-founder roles.

Can recycling waste be a new business model?

I think the concept of recycling waste in healthcare is interesting. An enormous amount of stuff that isn’t even used is discarded simply because there are no entities to check if it’s still usable. Currently redistribution is already markets via facebook/whatsapp groups with no oversight, or through non-profits that are relatively limited.

The dollar amount in drugs is non-trivial. Medicare alone is close to $1B in discarded drugs, and since 2023 manufacturers have to refund CMS if more than 10% of units are discarded.

I’m an investor in a company called Medicircle. It takes unused medications from patients that would normally be thrown out, certifies that they’re still good to be used, and then gives it to patients who need it for free. Plans reimburse them for less than they would normally pay for the drug. It’s a win all around! No one is a jerk, which is why they’re not called Medicirclejerk.

There are other areas you see recycling too, like surgery supplies. WestCMR, XS supply, RevMed, Hospital Overstock seem to do this by acquiring excess inventory from hospitals and then re-selling it. There’s also a program for pacemakers called My Heart Your Heart to recondition pacemakers from recently deceased patients.

I wonder if there are more places where this business model can exist. Durable medical equipment is an area that comes to mind. This happens with diabetes supplies - you’ll see these shady ass flyers posted around NY or weird Craigslist ads.

Could a company come in to provide more legit third-party re-certification for other forms of durable medical equipment like CPAP machines, power wheelchairs, unopened diabetes supplies, 60 walkers ordered to a patient in a South Florida fraud ring, etc.? Maybe slightly darker but what about implants?

I think re-certification marketplaces can be brought to more parts of healthcare - re-using materials is much more common in other countries. Maybe there’s an ideological conversation about whether patients should be allowed to try these recertified products at lower cost and absorb the risk themselves.

applications are due 11/19! outofpocket.health/build-at-deerf…

The gray area of medicine is getting pulled forward - things that are interesting in research but not yet implemented in care guidelines/covered by payers. A few reasons:

1) AI tools make it easier to sift through literature at scale, especially for patients. This is allowing al things that have promise.

2) Social media makes it easy for people to disseminate tricks and “hacks” that have worked for their health.

3) Cash pay and concierge healthcare has become more popular, with people willing to pay for interventions even if insurance doesn’t cover it. They don’t want to wait a decade until the guidelines are updated.

But the quality of papers is extremely variable and designed to be read by humans. Can we build better AI scientists to help assess papers, put research in context at point of use, or even rethink what papers are supposed to publish if you assume an agent is reading it?

If you have ideas around this, you should come to our upcoming hackathon with Deerfield Management.

We’re giving cash grants and prizes to teams with ideas on how to build AI scientists, and you’ll get mentorship/support to potentially turn this into a company.

like everyone else - we rented out a theater to watch You Can See Everything in NY last night with 80 other healthcare people

Did a little discussion afterwards about the movie, Theranos, an everything in between.

Actually had some polarizing view points around being a means, whether people had more sympathy for Elizabeth holmes or not, etc.

Apply by 11/19, we’re giving $250K+ in non-dilutive cash prizes/grants.

outofpocket.health/build-at-deerf…

We did a hardware hackathon earlier this year and saw some very awesome ideas using robotics and wearables to help patients.

For example, here’s a very over-engineered robot that could see when a patient was getting out of bed, spot objects on the floor, and then shine a light think there’s more promise here, that’s why for our upcoming hackathon with Deerfield Management we’re giving cash grants and prizes to teams building with wearables, robotics, and prosthetics.

Think perioperative automation, assistive mobility, neuromodulation, patient-facing robot design, and simulation environments.

If you’re building something in this area that might improve the lives of patients, doctors, and front-line staff we’d love to help support you.

cool part is we basically force the companies to run workshops or show their demos! I negotiated a bill live with one of the voice agents, attendees build an agent from scratch in a workshop, and companies answered hard questions from the audience about their product

Thank you to our sponsors that made Knowledgefest possible - very relevant to people building in healthcare!

@cloudcruise_app is an AI platform that automates healthcare workflows inside payer portals and EHRs. It turns manual browser and desktop tasks like eligibility dical record downloads, and payer enrollment into repeatable, auditable outcomes.

Sakura Health is the infrastructure layer for governed AI agents in healthcare, giving teams the primitives to build, evaluate, and deploy agents that work inside EHRs, practice management systems, payer portals, and phone/fax channels. Every agent comes with policy guardrails, audit trails, and human-in-the-loop controls, and specialty practices, digital health companies, and RCM teams already run them across front office, back office, and revenue cycle workflows.

@candid_health automates revenue cycle management, ingesting data from any EMR to submit clean claims, reduce manual billing work, and collect more of the revenue providers have earned.

Leap is the healthcare company making specialty medicine more affordable and easier to access. We're building modern drug distribution infrastructure for specialty therapy. We bill each drug at its true acquisition cost with no markup, and we pair every patient with a dedicated Care Guide who coordinates treatment, insurance approvals, scheduling, and ongoing support.

@SierraPlatform is the leading partner for healthcare agents, already reaching over 50% of U.S. families. Our customers include leading health plans, providers, pharma and medical device companies, digital health companies, and other healthcare platforms. Our agents go beyond answering questions, helping patients understand their coverage and get their medications to proactively closing care gaps and accelerating revenue cycle management.

@GetHealthie is the world's leading API-first, ONC-Certified EHR for healthcare delivery outside of the hospital. We provide the powerful infrastructure every scaling organization needs: EHR, scheduling, patient engagement, billing, and more, all accessible via modern APIs and a white-labeled UI.

@Vapi_AI is the leading Voice AI platform that gives developers and enterprises control over building, deploying, orchestrating, and managing voice agents at scale. This makes AI voice agents easy to build, safe to operate, and flexible to evolve.

@myaccessone , a @phreesia company, is a patient financing platform that enables healthcare providers to offer patients low-interest monthly payment plans, with no credit check, no fees and no impact to their credit score. As a licensed lender, AccessOne accepts 100% of patients and integrates directly with leading EHRs. Providers receive cash acceleration from Day 1, reducing AR days to zero while driving 85% net yield on payment plans.

@champ_hq is an AI agents platform for healthcare operations teams. We build AI agents that automate your manual intake, credentialing, eligibility and claims follow up processes.

This is a crazy way to say you have Crohn’s disease

full post is here, go forth and build

outofpocket.health/p/healthcare-i…

First part in the series - healthcare companies I think someone should build

1. DME recycling marketplace: A third-party marketplace that re-certifies used durable medical equipment (CPAPs, power wheelchairs, unopened diabetes supplies) so it can be resold legit instead of for healthcare: A new World's Fair where people can actually see and touch cutting-edge healthcare tech, like a live Neuralink demo or a Wandercraft exoskeleton. The point is to make people amazed by healthcare again, the way the old fairs did for consumer tech.

3. Instinct for healthcare workers: A personal AI agent built for healthcare employees, with its own login on the org's network, a BAA-covered cloud model, and role-based permissions instead of impersonating you with cached passwords. People are going to use these tools at work no matter what, so someone should build the compliant version.

3 companies collectively worth $6 billion...excited to be doing a happy hour at HLTH with @doximity and @WHOOP

If you're going to be there - come join us Tuesday evening, it's gonna be fun

you can sign up to get it here: outofpocket.health

last time I wrote a post like this there ended up being 3 multi-billion dollar companies that were exactly the ideas in it

for the next few weeks, I'm going to be writing out healthcare company ideas I think should exist

A few I'll write out and send in tomorrow's newsletter

- Instinct for healthcare employees
- The World's Fair for healthcare
- Refurbished DME marketplace

You can sign up to get why I don't start these companies myself...it's because I lack CONVICTION and Out-Of-Pocket is a real business I swear

Knowledgefest number 5 is in the books - honestly who would have thought a conference for healthcare builders would be so fun

we brought 120 people who applied to be there, on a weekend, to chat about how to build healthcare companies. We chatted about everything from how to ilot was financial successful to "how stop doing heroics at your company" to live building AI agents for eligibility checks

A huge thank you to everyone that came - this was only possible because good people come together to share real talk about the ops of healthcare

Grindr CEO had some interesting thoughts on how the dating app and health clinic could be integrated

(both of these would be only if users consented)

1) With a new PrEP pill that's fast acting (within an hour), they can remind people before meeting up with someone to take it
adding a verified HIV status or "taking PrEP" status on profiles (with user consent). Today it's mostly self-attestation.

there's some interesting (sad) data around this

Among nursing home residents who initially paid out of pocket, 16.4% spent down into Medicaid during their stay, increasing to 61.8% among those who remained in a nursing home at least four year.

source, Trilliant Health: trillianthealth.com/hubfs/Trillian…

i guess I should use this opportunity to shill something? I write a newsletter that's pretty fun about trends in healthcare and stuff relevant to people building in the space (how AI fax works, dating apps moving into healthcare, etc.)

It's pretty good, people seem to like it outofpocket.health

You’ll probably improve your longevity more if you just get married than all this other shit

@stableshaman @_TomHoward He does talk about this

More hackathon details here, applications due 11/19: outofpocket.health/build-at-deerf…

I interviewed George Arison about why Grindr would pay $250M to buy a telemedicine company. It was a really interesting conversation, some highlights:

1) The company believe healthcare could be at least half as big as their core dating business if not more

2) There's a lot of under the hood between their partner pharmacies + clinics

3) Some interesting ideas around how the dating app and healthcare pieces could be tied together - e.g. with fast-acting PrEP in trials, you could notify someone before they meetup to remind them to take it

Surprising to me, according to the Hint Health report 60% of DPC subscriptions are now paid for by employers - though I assume there's some selection effects here on who uses Hint

the new Trilliant Health report has a lot of interesting charts/graphs - one I thought was particularly interesting was this chart showing ~27% of CMS quality measures are actually tied to a patient outcome

I think there's an interesting debate around value-based care. Should if they did all the right processes but the outcome was still negative? Maybe things happened outside of their control, in which case we should still give them payments for doing the things we consider the right process.

On the flipside, it's easy to overrate and have process be more important than the actual outcomes. Worse, you might reduce the options providers have to EXPERIMENT with process to achieve an outcome even if there might be other way to get to the same goal

Weird side effect of online shitposting is sometimes important people end up reaching out to you. Anyway tomorrow's newsletter is an interview with Grindr CEO George Arison on what the company is planning to do in healthcare

It was a really interesting conversation, we're going t works and some of the benefits of the dating app <> telemedicine integration

You can sign up to get it here: outofpocket.health

Who knew making hyper specific healthcare memes would lead to public company CEOs reaching out

Broker commissions for the indemnity plans are really good

Brokers get commissions when selling different insurance products. Health insurance tends to have much lower % commissions per member but a much higher total dollar amount.

Supplemental insurance has higher commissions 15-25% premiums vs. 5% in medical).

They’re sometimes “heaped” commissions, which stacks more commissions upfront like 50%+ of collected premiums in the first year, and then much lower for every subsequent year.

You see this in Medicare Advantage too, where brokers will often tell clients to get a hospital indemnity plan in addition. Many of the broker sites talk about how much more they can make in commissions offering these plans (double your earnings, 55% year 1 commission).

reference-based pricing was all the rage for a while - but my understanding is that you'd run into a lot of issues like this

the employer gets caught in the middle of a negotiation between the hospital, the vendor, and the reference based plan

Which basically results in a ch of these high cost interactions

Curious if I'm way off base with my understanding of it, or if there are ways it's done that work well

I’m really excited to announce something months in the making…announcing a 5 week hackathon with $250K+ in total grants and prizes!!!

Build @ Deerfield (powered by Out-Of-Pocket) is for early stage teams that want a chance at non-dilutive capital, support in building their o present what they’ve built.

You can see more details below, but in general Deerfield Management and Out-Of-Pocket are joining forces to do something big for the New York healthcare community.

We’re selecting 12-15 teams who want to build for 5 weeks. You’ll get some cash upfront to start, mentors you can access during the build-time, and then a pitch showcase and reception with the broader New York healthcare community on January 20-21, 2027 at Cure in NYC.

Judges at the showcase will pick a few winning teams to receive additional grants and prizes to continue their work. And we’ll have a really good time 🙂

when looking into indemnity insurance, one thing that surprised me is how active UnitedHealth is in it via 80+ subsidiaries

they seem to have nearly 50% of the hospital indemnity market and it's a $1B+ business line with great loss ratios

Coming to SF this week for our Knowledgefest conference!!! Excited to bring 110 of the best healthcare builders in one room

And excited to announce the final two sponsors that ops people should really get to know, @cloudcruise_app and Leap

See everyone this week for Knowledgefest!!!

pretty interesting - in one of the new condo buildings in Miami they're including access to a membership primary care clinic (Baker Health)

It does feel like many of the amenities in these buildings don't actually get used - this feels like something much more attractive

If that sounds like you, sign up for the newsletter (plus you'll get some very fun healthcare memes/trend writeups): outofpocket.health

for the last couple of months we've been working behind the scenes on something awesome. It's for teams that really want to take the the jump and start something (or just did)

It'll be announced in the newsletter this week, but I'm really excited for it. And it may or may not ng 👀

It's your sign to NOT be the below meme

there's a common belief in healthcare among providers that most people would rather trust a doctor than AI - but I think we're not far from a tipping point here

Was interesting to see this Oliver Wyman survey of consumers and see how close we are to people trusting AI more than f that the number of times where the AI and clinician disagree is increasing, with people MORE often doing something other than what the clinician directly said

This is heading for a collision course at some point

guess what? We are renting out a movie theater in NY for the new Elizabeth Holmes + Nathan Felder doc

It'll be fun! 10/22 in NY

We're prioritizing friends of Out-of-Pocket, not saying we're gonna look up if you're on the newsletter and how much you read it but...we could

got in next tweet or tell your NY homie to come

one thing that surprised me was how good the loss ratios are for hospital indemnity/critical illness plans

I was reading this NAIC report where all the plans are filed. The loss ratios for these plans are WAY lower than basically every other kind of supplemental benefit (let ratios are basically the amount paid out in claims for every $1 they get in premiums. Most health plans are taking L+ratio for 85-90% of the premiums they collect.

full essay here, it's a very interesting niche
outofpocket.health/p/fixed-indemn…

new essay - the weird world of indemnity plans

I went down the rabbithole of these indemnity plans, which essentially give you direct cash pay outs if you get certain kinds of diseases

This is a market that's adjacent to healthcare, but has a lot of interesting aspects to it

- mics (they have shockingly good loss ratios and interesting broker commission structures)

- Who buys and sells them (United seems to have like 80+ subsidiaries that sell some version of these products)

- some open questions I have that I couldn't understand when doing research (why is pregnancy an option? why do they all use parents with kids as their stock photos?)

full essay in the comments

After the posts in my pregnancy series, I got some interesting emails...about indemnity insurance

People were telling me that they were getting $3-5K payouts for it despite paying pretty little in premiums

Which then sent me down the rabbithole of what this insurance product post is about the weird world of fixed indemnity insurance (hospital, critical illness, etc)

software budgets in healthcare are tiny, gotta go for services

@shailinthomas i think that's reasonable and it seems like in other areas we're seeing new liability products come up

wrote a longer thing about how malpractice and AI works together

outofpocket.health/p/how-ai-is-re…

Excited to be moderating a panel about wearables at HLTH this year

I think wearables have a very bright future in healthcare. We have tools that can help us interpret the data coming out of them, more consumers are wearing them, and we now have more regulatory paths for imentation.

Three requests for you all:

1) What are questions you'd want to know the answer to?

2) Let me know if you're going to be at HLTH, Out-Of-Pocket will likely be hosting something

3) Send any recs for docs that do neck shortening procedures

@justindross new york is def somewhat of an outlier - even in the full post i pulled the data within new york and outside and it's closer to 10-18K in other cities

outofpocket.health/p/pregnancy-co…

crazy to read how much practices and hospitals are paying for e-fax - below is from a solo practice

I've been told from some outpatient clinic chains they're paying $30K/mo and apparently some hospitals are paying 7 figures a year!

So how much does it cost to have a baby in the US? The Europeans are really gonna shit themselves at this one.

(caveats - based in new york, good commercial coverage, etc.)

We were billed ~$150K, our insurance paid ~$79K, and our out-of-pocket costs were ~$4300. And this was l birth cost wise - NICU babies can end up as multi-hundreds of thousands of dollars to occasionally $1M+!

Why are we sharing all of this? IMO healthcare wins because patients are afraid to share their experiences with each other. We gained a lot by having friends who would share their stories with us. This is our family's small part of trying to break that information asymmetry.

I didn’t realize Grindr was making a big push into healthcare - they just acquired a telemedicine company for $250M!
That's one of the biggest telemedicine acquisitions and I haven't seen anyone really talking much about it?

Although these estimates seem really high? Unless cts the combined U.S. telehealth and pharmacy business will generate more than $400 in monthly revenue per active patient, or more than $4,800 annually for a patient who remains in care for a full year."

If you know who these people are you should ask your doctor if a statin is right for you

"I want interoperability"

Anatomy of a referral

So what happens during a referral. You go to the doctor, doctor says “yeah I’m good actually”, and refers you to a specialist. Now what?

A few things have to happen.

1) The doctor picks a specialist for you to see. The doc might know which doctor they sed on who was best at remembering extra long mnemonics in med school. Or they have a medical assistant or referral coordinator that uses different back-office interfaces to find someone that’s in-network.

2) They send your details, why they’re referring, what they’ve done/measured, etc. A majority of the time this is done via fax since it’s the one ubiquitous standard across all doctors’ offices. I have written about this with tears in my eyes in the past. Faxing is typically done through e-fax - which uses the telephone rails to send the info but lives in a digital inbox.

3) The fax is sent to the doctor’s office, where it sits in the queue along with all the other faxes that they get. Requests for medical records, lab results that come in, love letters from the war, etc. Someone goes through the faxes and prioritizes which ones are important.

4) The receiving practice now needs to check all the insurance details. Is the patient’s insurance valid? Are they in-network? Do they need a prior auth to have this visit/procedure? This needs to be sorted before the patient is accepted.

5) If they are accepted, data from the faxed referrals now needs to be input into the electronic health record. As you can imagine, it is typically done manually. Painfully, clicking through menus as the carpal tunnel slowly builds.

6)Now the practice has to actually get the patient to come. Practices will reach out to patients, call them, beep them, if they want to reach them. If they get a hold of them, they schedule the patient and start the process of getting all their records, bringing them into the office, etc.

7) The receiving practice is supposed to also send back the details of what happened to the original practice that referred them as well. Which as you can guess, also happens through fax.

I'm sure I missed things, but this process is really important because it both dictates a ton of how healthcare spend happens AND it's a frequent area that patient's fall through the cracks

@neejkapoor 3 magicians! And it’s honestly not that hard to host stuff like that

trying to tab out of the compliance training

I learn something every time @healthapiguy gives a talk - at healthcare 101 yesterday there were a few things that really made me think

1) Can AI doctors access data through the health information exchanges? what counts as an AI doctor?

2) Healthcare actually has it better than es because we have regulations that force data out. For example CCD-A exports, or EHRs allowing robotic process automation if an API doesn't exist.

In other industries those rules may not exist which allow systems of record to box out third-party applications

3) The sheer number of places data is sent between stakeholders is just enormous, I mean look at this chart of just a small subset of networks that already exist. There's so much work to do

full post is here - they also have some interesting side quest products they've launched that digital health companies might particularly find interesting

outofpocket.health/p/referrals-st…

Referrals still run on fax and that's bad

Today's post is a deep dive into how the process of referrals work. The core process still runs on fax, which runs into all sorts of issues in terms of triaging patients, not closing the loop with referring providers, and patients cks.

The reason referrals typically run on fax is that all practices can participate in the fax network(as Brendan Keeler has talked about many times). But the problem is that you basically need to hire people to manage the loads of unstructured data that comes out it.

Typically practices have hired people just to manage these fax queues for referrals., but today AI is a perfect tool to bring clinical expertise + data structuring to this problem.

Medsender has built out a bunch of AI + fax workflows (an unholy union in healthcare but here we are) including managing referrals for practices. We walk through their product and the pain points of referrals today.

Anyone who's had a referral as a patient can probably attest to how clunky the process is. Hopefully we can improve it.

i think this would fix me

Tomorrow's post is a deep dive into referrals - how they work, why do they use fax, and some interesting new approaches to solving this problem.

You can sign up to get it here: outofpocket.health

full study from above here: link.springer.com/article/10.100…

Data about how referrals are made still feels like the most underutilized dataset in healthcare.

Being able to understand the referral patterns from a certain doctor or to a certain doctor/hospital can help inform a lot about how the healthcare system gets used, understand why e, and potentially intervene if the referral is a wrong fit.

But referral data is largely dark matter that exists in a liminal space between the two providers. It’s inferred from other datasets like claims or EHR notes, but it misses a lot of the nuance that’s actually sent in the referral transmission. But if this data can be captured and structured easily now, I think it can be a really valuable dataset.

Here is an interesting study. Within a health system, when primary care makes a referral to a specialist only 35% had a documented completed appointment. But looking at the methods, every out-of-network referral automatically shows as "no status" (~10% of referrals) because Duke's Epic can't see what happens at a non-Duke specialist. Another massive chunk (~20%) of tickets had no date because schedulers booked appointments without linking them back to the original order.

Just goes to show how hard it is to do analyses with our current available data, especially across providers.

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 r 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 N 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

this was obviously what was going to happen

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 + 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.

Healthcare needs more engineers

You'll have a real impact and discover some hilarious quirks to working in the space

@cremieuxrecueil I had some people send in their experiences with MAID and thoughts - was interesting to hear others people perspective even if I disagreed with it

outofpocket.health/p/a-few-more-t…

Come through, 10/27 - 10/29: outofpocket.health/courses/ai-age…