Jack O'Meara

OpenEvidence went from a $1B to $3.5B valuation in 5 months. Last week, Doximity bought their competitor Pathways for $63M. OpenEvidence is positioning itself as the single source of truth for medical decisions, basically ChatGPT meets UpToDate. They've locked down partnerships with 13 major scientific journals (including JAMA) and claim 40% of US doctors already use their platform. Doximity—once "LinkedIn for doctors," now much more—just acquired Pathways for $63M ($26M cash, $37M equity). Pathways is essentially an OpenEvidence clone, and according to Kevin from Health Tech Nerds, they're generating maybe $1-2M ARR. That's a 30x revenue multiple. Doximity already has "Doximity GPT" for summarizing charts and reports. So Pathways fits their strategy. But what's amusing is that, as Brendan Keeler reported, OpenEvidence sued both Pathways AND Doximity for allegedly "stealing" their system prompt. So what happened is that Doximity realized they can't beat OpenEvidence, so they bought the company that also can't beat OpenEvidence (see visual), betting that their distribution moat can overcome OpenEvidence's first-mover advantage in medical decision support. Now we have two distribution powerhouses on a collision course. OpenEvidence owns 40% of physicians through product/marketing superiority. Doximity owns massive reach through platform dominance. Both want to capture the critical moment when doctors need answers. In Belgium we say "a cat in trouble makes weird jumps," but maybe this cat knows exactly where it's landing. PS: The above was co-written with Blake Madden from Hospitalogy, go subscribe as it’s my fav newsletter ❤️

193

64 Comments

🚨 Your “wellness app” might actually be a medical device 🚨 Take this example 👇 An app displays lab test results (such as vitamin levels or biomarkers) and then suggests supplements or lifestyle changes to restore those values to normal. The founders say: “We’re not treating or diagnosing, we’re just guiding. Sharing what other users have done, or what the latest science says.” But regulators in the EU and the US may not see it that way. 🇪🇺 EU (MDR) Under Rule 11 of MDR, software that monitors, prevents, or alleviates disease is considered a medical device. If your app interprets diagnostic data and recommends actions, it’s likely MedTech, possibly Class IIa or higher. 🇺🇸 US (FDA) FDA takes a risk-based approach. General wellness apps are fine if they stay generic: “eat healthy, exercise.” However, once recommendations are based on personal diagnostic results, the FDA may consider it a medical device. 👉 The lesson: you can’t “opt out” of regulation by calling your product lifestyle. If your app is designed to influence medical decisions, regulators will likely classify it as a medical device. ⚠️ Why this matters: Regulatory classification determines your evidence requirements, liability, and pathway to market. Misclassifying your product early risks years of rework and delays. Intended use defines reality, not what you call it. Luckily, it doesn't have to be that costly or troublesome to get through regulatory approval. You just have to know who to talk to... #MedTech #SaMD #DigitalHealth #MedicalDevices #HealthTech

343

34 Comments

We are now starting Q4 2025 and quarter after quarter, category-shaping events show that HealthTech is maturing.... Q3 2025 was no exception. I wrote down a few thoughts. 1️⃣ 𝐑𝐚𝐝𝐢𝐨𝐥𝐨𝐠𝐲 𝐀𝐈: 𝐭𝐡𝐞 𝐛𝐥𝐮𝐞𝐩𝐫𝐢𝐧𝐭 𝐟𝐨𝐫 𝐜𝐥𝐢𝐧𝐢𝐜𝐚𝐥 𝐀𝐈 About 80% of FDA-approved algorithms are in radiology, making it the most advanced (and most widely adopted) clinical AI category so far (even as medical chatbots like OpenEvidence are catching up). Yet, there’s no clear winner. Will success go to: ◾Aggregator platforms like Deep, which bundle multiple AI tools? ◾Specialists developing full-spectrum algorithms, like Gleamer? ◾Combined approaches doing both aggregator and algorithms like AIDOC? ◾Next-gen foundation models in radiology, such as Raidium? ◾Or incumbents (RIS and PACS vendors) using targeted M&A to stay on top? Icometrix chose the safer path, selling to GE. 2️⃣ 𝐀𝐠𝐞𝐧𝐭𝐢𝐜 𝐀𝐈: 𝐧𝐞𝐰 𝐰𝐨𝐫𝐥𝐝 𝐯𝐬. 𝐨𝐥𝐝 𝐰𝐨𝐫𝐥𝐝 This is where the fiercest battle is happening. Software entered healthcare 20+ years ago, with Epic landing its first big deal in 2003 with Kaiser Permanente. Now AI is eating software... and incumbents are fighting. New AI-native operating systems are emerging, combining scribing, billing, clinical decision support, and workflow (like Commure in the U.S.). Incumbents like Epic and Doximity are trying to suppress challengers, while Cerner has launched a brand-new agentic AI EHR built from scratch. The battle is on. 3️⃣ 𝐁𝐢𝐠 𝐓𝐞𝐜𝐡 𝐡𝐚𝐬𝐧’𝐭 𝐥𝐞𝐟𝐭 𝐭𝐡𝐞 𝐫𝐚𝐜𝐞 Despite setbacks, Big Tech is still pushing hard. OpenAI: With OpenEvidence’s traction and U.S. healthcare making up ~20% of GDP, OpenAI sees a big opportunity. Microsoft: The most active so far, with cloud, data tools, Nuance/Dragon, and now bold clinical claims. In June 2025, Microsoft said its AI outperformed doctors on complex cases, scoring 85.5% vs. ~20% for a panel of 21 physicians. Apple: Trying again with its new hypertension feature. Tim Cook once said Apple’s biggest contribution to humanity would be in health..... we’re still waiting. 4️⃣ 𝐎𝐧 𝐭𝐡𝐞 𝐝𝐚𝐫𝐤𝐞𝐫 𝐬𝐢𝐝𝐞, 𝐦𝐚𝐧𝐲 𝐕𝐂𝐬 𝐚𝐫𝐞 𝐬𝐭𝐢𝐥𝐥 𝐰𝐨𝐧𝐝𝐞𝐫𝐢𝐧𝐠 𝐡𝐨𝐰 𝐭𝐡𝐞𝐲’𝐥𝐥 𝐫𝐞𝐭𝐮𝐫𝐧 𝐭𝐡𝐞𝐢𝐫 𝐟𝐮𝐧𝐝𝐬. For SaaS point solutions in #healthcare, life is tough. They’re caught in the middle: squeezed between incumbents with outdated tech that won’t integrate their tools, and a new wave of agentic AI platforms that bundle multiple SaaS functions into a single agent. Still, some #SaaS products bring valuable technology. Expect more M&A in the months ahead. ----- Follow HGM Advisory for daily reposts of the most relevant HealthTech news #ai

123

11 Comments

AI in healthcare: if you want adoption, start with the workflow that’s already breaking—not the model you’d love to train. Over the past year, the most credible signals we’ve seen at seed come from teams that: - pick a single, painful clinical task (not “reinvent radiology,” but e.g., “reduce time-to-lesion-tracking in MS follow-up”); - co-design with clinicians until the product saves minutes per exam every time; wire themselves into hospital data pipes early (governance + interoperability before the sales deck); - quantify value in operational terms (fewer repeats, faster discharge notes, shorter queues), then translate that into payor/provider ROI. One portfolio example: Raidium. Co-founded by Paul Herent, M.D., Msc and Pierre Manceron, the team is building an AI-native radiology viewer and focusing on concrete biomarkers that are slow and error-prone to read manually. What stood out wasn’t the model—it was the clinical scoping and the integration plan with reference centers. That’s what gets you from demo to deployment. For investors, this changes how we diligence: less model-benchmark theater, more data plumbing, consent pathways, and clinician time-saved. For founders: own one use case end-to-end, earn trust, then expand. Anne-Sophie Saint-Martin Pierre Frouin Constant Beroulle Henri Deshays Jean-David Zeitoun #AI #healthcare #breakthrough #data

30