Best of LinkedIn: Health Tech CW 08/ 09

00:00:00: This episode is provided by Thomas Alguyer and Frennis based on the most relevant LinkedIn posts on health tech in CW.

00:00:06: eight, nine.

00:00:07: Frennis equips product and strategy teams with market and competitive intelligence to navigate the health

00:00:12: tech landscape And honestly if you thought at the beginning of the year was loud With all the AI noise Yeah The signal we're getting from weeks eight and nine Is very different.

00:00:23: It's quieter But the stakes feel much higher.

00:00:26: They really do.

00:00:27: Welcome everyone to The Deep Dive.

00:00:29: You know, for a long time the headlines we were looking at where basically science fair projects it was all.

00:00:34: look at this shiny new bot or Look At This Generative Text

00:00:37: Exactly!

00:00:37: The whole wow factor

00:00:39: Right.

00:00:39: But Looking At The Sources This Week...the vibe has definitely shifted.

00:00:42: It feels less like A showroom and more Like A construction site..The party is over , the lights are on And now We Actually Have To Build The Infrastructure.

00:00:50: That's Perfect Analogy.

00:00:52: The honeymoon phase with concept of AI Is Officially Done.

00:00:56: The industry is pivoting hard to a theme of execution.

00:01:01: We aren't really asking, can we do this anymore?

00:01:03: We're asking how do we scale this without bankrupting the hospital or breaking the clinical

00:01:09: workflow?".

00:01:09: So the mission for today is to map that Execution Gap for you!

00:01:19: First, we're tackling organizational readiness.

00:01:21: Basically why your culture is likely the reason you pilot program failed not code.

00:01:26: Second were looking at next-gen hardware.

00:01:29: So we're talking about robotics and imaging, but specifically the economics behind them.

00:01:33: because it's not just

00:01:41: We

00:01:59: have to be honest here, when AI fails in a healthcare setting how often is it actually the algorithm's fault?

00:02:06: Rarely.

00:02:07: I mean its almost always human element or organizational structure.

00:02:11: Sigrid Berven-Roygen had brilliant breakdown on this and recent post.

00:02:16: She argues that organizations are rushing into installation.

00:02:19: They're literally just plugging box but they completely skipping readiness!

00:02:23: She breaks readiness down into three pillars right Psychological, structural and cultural.

00:02:27: She does in the order really matters here.

00:02:29: psychological readiness is first Are your clinicians terrified?

00:02:33: this tool will replace them or do they see it as an asset?

00:02:36: Because if they're scared They simply won't use it.

00:02:38: yeah though fine work around

00:02:39: exactly then.

00:02:41: Structural can you IT infrastructure actually handle the data load Or isn't going to crash the network?

00:02:46: And finally cultural Is It Safe To Fail?

00:02:50: If your hospital culture punishes every small error, no one is going to experiment with a new AI tool.

00:02:55: That cultural piece is huge!

00:02:57: It reminds me of that old saying you know... Culture eats algorithms for

00:03:01: breakfast.".

00:03:01: It absolutely does and Simon Philip Ross provided a really fascinating counter-example to this.

00:03:07: He's been observing the health tech landscape in the Middle East specifically Saudi Arabia and the UAE And he noted what we often call complexity in Western healthcare might actually just be bureaucracy.

00:03:19: And a lack of nerve, maybe?

00:03:20: Yeah exactly so.

00:03:21: he's saying they're moving faster not because the tech is better but Because The decision-making Is different.

00:03:27: He saw real willingness to decide.

00:03:29: in the Middle East.

00:03:30: he observed A shift toward tweeting Tech companies as partners rather than vendors Which is

00:03:35: a big shift from the West where we tend To write what four hundred page RFPs right.

00:03:39: We treat suppliers very transactionally But In the region he observed They co develop solutions.

00:03:44: That partnership mindset removes so much friction and it really accelerates execution.

00:04:04: How do you solve the people problem practically?

00:04:06: Well, Vladislav Puzhankov shared a strategy that I thought was genius in its simplicity.

00:04:10: The

00:04:11: shadowing strategy right.

00:04:12: yes his point is that You don't train people on the new machine after it arrives and it's plugged In.

00:04:18: you send your team to a site That's already using it effectively.

00:04:23: You let them shadow the workflow not just push the buttons.

00:04:26: that distinction Is crucial.

00:04:28: you're building the mental model of how the day flows before you introduce the physical tool And Puzonkov added a kicker that I think is even more important.

00:04:36: Bring a proctor back three months post-go live

00:04:39: because That's when the bad habits set in.

00:04:41: training isn't a one time event It's a curve, Three months in.

00:04:45: your team has likely found workarounds that compromise The data or the efficiency.

00:04:50: you need someone to come Back In and tighten the screws on the process?

00:04:55: was saying about the skills gap.

00:04:58: We tend to think, oh we need to modernize let's just hire ten data

00:05:01: scientists.".

00:05:02: But

00:05:02: that creates a silo of smart people who don't understand clinical care at all.

00:05:07: She argues for digital literacy For clinicians and operational teams.

00:05:12: If your chief medical officer can't assess risk of an AI tool They cannot govern it.

00:05:17: You cant outsource understanding tools you use

00:05:20: And Tessa Vanden Rook reinforced this too.

00:05:23: She said, digital growth is a strategy problem not a marketing one.

00:05:27: you can't just market your way into digital maturity.

00:05:29: You have to structurally change how your leadership thinks.

00:05:32: so the takeaway for that first cluster Is pretty clear fix the mindset train?

00:05:36: The workflow and then buy the

00:05:38: robot which brings us perfectly to theme To hardware robotics and multiplication effect.

00:05:43: This is where the physics meets the economics

00:05:45: and we have some incredible examples from last two weeks to dive in.

00:05:49: We

00:05:49: have to start with the quail egg.

00:05:51: Dr Martha Buchenfeld shared a video of a surgical robot from Beijing.

00:05:55: Now picture this, it's peeling a raw quail eggs.

00:05:59: It removes the hard shell, but leaves incredibly thin membrane underneath completely intact.

00:06:05: That's absurd precision!

00:06:07: And here is the kicker... it costs two hundred thousand dollars.

00:06:10: It rivals The Da Vinci System which cost around TWO MILLION DOLLARS.

00:06:14: so we're talking ten percent of that for the same level of precision.

00:06:19: My immediate thought was game over.

00:06:21: everything gets cheaper access explodes.

00:06:23: But thats the trap right?

00:06:24: Buckenfeld references the futurist Ravinha Suthassan to make a critical point.

00:06:28: A cheaper robot in a broken system changes nothing.

00:06:32: This introduces the concept of the multiplication effect.

00:06:34: Walk

00:06:34: us through that because it's really important for you guys listening to Grasp.

00:06:38: Think about operating room, if you swap human surgeon but you keep the same administrative bloat, the same slow turnover time between surgeries and The same lack of post-op monitoring.

00:06:50: You don't get ten X productivity Just got a slightly lower bill for the hardware.

00:06:54: to get the multiplication effect You have to redesign the work itself.

00:06:58: So the robot does the cutting?

00:07:00: But this system around it has to change too

00:07:02: exactly.

00:07:03: the robot cuts the human judges the AI handles the documentation and the vitals.

00:07:09: The innovation isn't the machine.

00:07:11: It's the ecosystem around it that allows the machine to run continuously.

00:07:16: That reminds me of FEGUO's analysis of United Imaging, they seem to be terrifying Western competitors right now because they understand this ecosystem play deeply.

00:07:25: They didn't build a scanner and then try to bolt some AI on top for later?

00:07:28: They built the AI into the physics of hardware from day one.

00:07:32: And look at results!

00:07:33: They drop PT scan times from fifteen minutes down to one minute.

00:07:36: That's an order of magnitude change.

00:07:38: if you can scan in one minute, You can clear a waiting room and hour.

00:07:41: And their moat isn't even the speed.

00:07:43: it is data partnership.

00:07:45: They have relationship with Zangshan Hospital where they aren't just vendor but act as co-developers.

00:07:50: They feed real world clinical data directly back into R&D.

00:07:55: It s feedback loop the transactional vendors just can't match.

00:07:58: We're seeing these technical leaps everywhere now, validating this integrated approach.

00:08:03: Dr.

00:08:03: Filippo Catamartiri was discussing photon counting CT.

00:08:07: for those of you who aren't radiologists why does that specific tech

00:08:11: matter?

00:08:12: Think a standard CT scan like black and white photo.

00:08:15: it measures intensity.

00:08:17: Photon Counting Measures The Individual Particles Of X-ray Light.

00:08:20: It Gives You Pixels As Small as .

00:08:21: Point

00:08:23: one, one millimeters.

00:08:24: That's microscopic it

00:08:26: is.

00:08:26: It allows you to see things like bone marrow edema or differentiate crystals and joints?

00:08:30: It's not just a sharper image It's a new type of diagnostic data entirely

00:08:35: And speed as the currency here too.

00:08:37: Laura Giordiello mentioned cardiac MRI scans becoming three times faster with dual AI Reconstruction.

00:08:43: Jeremy Whitters and Faster for our Philips are talking about AI standardizing The data quantification.

00:08:48: that Standardization is key because it stops the garbage in garbage out problem.

00:08:53: If the AI ensures that data is cleaned first time, you don't have to call patient back for a repeat scan.

00:09:00: That efficiency is business model!

00:09:02: It's not about can we see the heart, it is.

00:09:04: Can We See In Ten Seconds?

00:09:06: So We Can Treat Twenty Patients Today Instead of Ten?

00:09:08: Speaking

00:09:09: Of Volume The US Market Is Really Heating Up!

00:09:11: We Saw Updates From Jeff Martha and Kamal Dave Regarding Hugo Robotic System Getting Clearance And Starting Surgeries At Cleveland Clinic And Atrium Health

00:09:21: And Fabian Vellings & Yvonne Middle State Highlighted The Medtronic Stealth XIC System For Spine Surgery.

00:09:27: That'S Interesting Because It Combines Planning Navigation And Robotics Into One Platform.

00:09:32: Again, it's about the ecosystem not just the drill.

00:09:35: Before we move off hardware I have to mention one thing that made me laugh but is actually kind of genius.

00:09:39: Steven G shared a post about smart underwear.

00:09:42: i saw That they called The Fitbit for farts.

00:09:44: They Actually Do.

00:09:45: It Tracks Gut Hydrogen To Monitor Digestive Health!

00:09:48: It sounds funny and no people will chuckle.

00:09:50: But think About Data Implications.

00:09:53: We Are Moving Toward A Human Atlas Of Continuous Data.

00:09:57: We Aren't Just Taking Snapshots At The Doctor's Office Once A Year Anymore.

00:10:01: We are monitoring everything from heart rate to flatulence, two hundred four seven.

00:10:06: And that flood of continuous data leads us directly to our third and most futuristic theme.

00:10:11: If we have robots peeling eggs an underwear tracking gas where does all this state ago?

00:10:17: This brings us to agentic AI and the potential death of the EHR.

00:10:22: This is probably the biggest structural shift on the horizon for health tech.

00:10:25: Nir Berensweig framed it perfectly in his post, we are moving from assistive to autonomous.

00:10:31: Let's define that clearly.

00:10:32: what does the difference between an AI assistant and a AI agent?

00:10:35: Assistive AI as co-pilot sits next.

00:10:37: you says hey doc look at this x-ray.

00:10:39: I think i see a nodule.

00:10:40: You still have do work.

00:10:42: Agentec AI walk away workflow.

00:10:44: The agent notices the issue, schedules follow-up updates chart orders labs and maybe even drafts insurance claim.

00:10:51: It acts on your behalf to achieve a goal.

00:10:53: But you can't run that kind of software in a hospital system built in two thousand five?

00:10:56: No definitely not.

00:10:58: And thats why Mario Amaro is arguing that electronic health record as EHR is dying its too clunky it's siloed.

00:11:06: he envisions more like a web three wallet where The patient or the provider holds the keys to data and these AI agents interact with it directly.

00:11:15: Kumur

00:11:15: Makalit took this idea even further, he's talking about sovereign agents... Right!

00:11:19: Instead of just having more dashboards collecting data which Makala points out creates some noise and burnout we need context brokers.

00:11:26: an agent that works for you synthesizes your life.

00:11:28: context only shares what is necessary in a clinical system privacy first Agent driven model.

00:11:34: This sounds very sci-fi.

00:11:35: is anyone actually building the infrastructure for this agentic future today?

00:11:39: They are.

00:11:40: Christian Schmidt and Jesper Kongsdahl-Surenzen were discussing the Oracle Life Sciences AI data platform.

00:11:46: This is real stuff happening now!

00:11:48: They have a hundred twenty nine million longitudinal records embedded in this platform.

00:11:52: We're

00:11:52: a hundred and twenty nine millions?

00:11:54: Yes,

00:11:55: And they are using agentic reasoning to do things like generate synthetic control arms for clinical trials.

00:12:00: Okay... Synthetic Control Arms That sounds complicated.

00:12:04: Let's unpack that.

00:12:05: For you guys listening

00:12:06: It's actually a brilliant concept for saving money.

00:12:09: Usually In a drug trial, you need a group of people taking the new drug and a group for people.

00:12:13: Taking up placebo sugar pill finding recruiting and managing those placebo patients takes months in cost millions

00:12:21: And it delays the drug getting to market

00:12:23: exactly.

00:12:24: Oracle is using those one hundred twenty nine million records To simulate a control group based on real-world historical data.

00:12:31: They essentially say we know what happens to a fifty year old male with diabetes who doesn't take this drug because We have data on ten thousand.

00:12:40: So you don't need to recruit real people to take a sugar toe?

00:12:42: Precisely.

00:12:43: You simulate them, this saves massive amounts of capital and time but you can only do that if you have a unified, massive data platform!

00:12:50: Can't do it with fragmented records in file cabinets.

00:12:53: That efficiency changes the business model too.

00:12:55: If we could simulate trials or automate workflows how would pay for all these?

00:13:00: Syrendra Agarwal discussed SAP brim and shift paper use.

00:13:04: This is an economic flip.

00:13:05: Hospitals might stop buying scanners entirely.

00:13:08: Instead of spending two million dollars on a scanner, they pay per scan.

00:13:12: Or better yet — per successful diagnosis— it aligns the cost with outcome.

00:13:17: Jan Beger shared a simulation study on coronary artery disease that perfectly illustrates why this outcome-based thinking is better.

00:13:25: That was a fascinating study!

00:13:26: They used an AI decision support system to analyze patients.

00:13:30: The AI shifted seventy-two percent of treatment decisions and interestingly It often recommended more invasive surgery like a bypass over simpler options.

00:13:39: We isn't bypass surgery more expensive, I thought AI was supposed to save money?

00:13:43: It is more expensive up front but the AI predicted that by doing the surgery now you prevent repeat heart attacks and re-hospitalizations for next five years.

00:13:51: so in long run it saved many lives.

00:13:54: That's a big take away right there.

00:13:56: we are moving from cheapest intervention now To best intervention for lifetime value

00:14:01: Exactly.

00:14:01: But humans are bad at this long term math.

00:14:05: AI is great at predicting the bill in five years.

00:14:08: That's the value shift

00:14:09: Incredible.

00:14:10: So let's wrap this up, we are seeing hardware getting faster and cheaper like the one minute PET scans but only if you have the workflow to support it.

00:14:19: We're seeing software becoming agentic and potentially replacing old record systems And organizations that are struggling update their culture To match speed of tech.

00:14:29: That cultural lag is the biggest risk.

00:14:31: We've got twenty first century tools operating in twentieth-century organizational structures.

00:14:36: And as we leave you today, I want to leave with a final thought from the sources.

00:14:40: We've talked about efficiency and robots but Krista Calpe raised a concern regarding mental health AI that really stuck with me.

00:14:47: This was openAI data she referenced?

00:14:49: Yes

00:14:50: OpenAI's data shows over one million users are talking chat Jpt who potentially considering suicide or showing signs of psychosis.

00:14:57: That is staggering number

00:14:58: It is.

00:14:59: So while we celebrate robots peeling eggs and synthetic control arms, We have to ask are we ready for the ethical responsibility of AI becoming The first line of defense from mental health crises?

00:15:11: but a

00:15:14: necessary one to keep in mind as we build these systems.