DeLaval

Clinicians should be using Claude Cowork more than they are. I get a ton of physicians and health system leaders asking: how do I use AI when I can't integrate into my EHR? Fair question, consumer tools like Claude Cowork and other AI platforms aren't HIPAA-compliant (without enterprise license), and Doximity Ask and OpenEvidence have dominated clinician attention for HIPAA compliant work outside the EHR. However, there are SO many use cases that are applicable to clinicians. CME tracking. CV self-updates. Weekly literature digests. Conference and trip planning. Manuscript pre-review on a flight home. Triaging the half of your inbox that has nothing to do with patients. Drafting a talk before grand rounds. Pulling together a faculty dossier before promotion review. I think these use cases can compound across a career, and most clinicians are still doing them by hand (or paying someone to organize these things for them). I wrote a guide for the difference between dispatch, scheduled tasks, and live artifacts, the three modes that together make up my agentic-like workflow in Claude: — Dispatch: one-time sizeable tasks executed while you're away — Scheduled Tasks: the standing meeting that never cancels — Live Artifacts: pages that re-render from your connectors on open Full guide in comments with prompt guide! What other use cases are the doctors out there leveraging these tools for?

I had an incredible time on the Withings webinar, discussing what the CMS Access Model could be solving for and how hospitals and health systems should think about the model. With a little over a month to go until the program goes live, I think there is incredible potential in this program, as well as pitfalls to avoid. It has been encouraging for me to see the number of organizations deciding to work with us at MSK Access and RevelAi Health to offer this novel digital health benefit to their Medicare patients. We all need to have a more thoughtful customer-centric approach to care (yes, customer— to take the approach Baylor Scott White CEO Peter McCanna)— And the CMS Access model is a clear wedge to provide technology-enabled services that can fill this gap for health systems and ACOs. Looking forward to everyone else's thoughts on this program and seeing how it evolves over time.

On October 1, 2027, essentially every IPPS hospital in the United States that isn't already in TEAM will be mandatorily at 90-day episode risk for hip, knee, and ankle replacement. CJR-X is the most aggressive expansion of mandatory episode-based payment in the history of the Innovation Center. The original CJR model ran in 67 MSAs from 2016 to 2024 and generated $112.7M in net Medicare savings across 98,000 episodes at 323 hospitals while maintaining quality. CMS read that result and concluded, reasonably, that the intervention works. CJR-X takes it nationwide, bolts on outpatient LEJR and total ankle, keeps the 90-day window, adds a quality-first composite score gate, a 5% stop-loss for dual-eligible-heavy hospitals, 29 risk adjusters, and a waiver of the SNF 3-day stay rule. Projected 5-year savings: $725M. Read it alongside TEAM and the doubling down on accountable care seems more steadfast. TEAM holds 729 hospitals accountable for 30-day surgical episodes starting this year. CJR-X picks up almost everyone else for 90-day joint replacement episodes starting late 2027. Together they represent something CMS has never done before: near-universal mandatory episode risk for orthopedic surgery. This lands at a difficult moment. Primary TJA reimbursement is down 56% since 2000. The 2026 Physician Fee Schedule added another 2.5% efficiency cut on surgical procedures. Demand is projected to nearly triple by 2040. You cannot keep compressing procedural payment on the most cost-effective intervention in all of medicine and simultaneously expand episode accountability without building new infrastructure to capture value beyond the cut. Benjamin Schwartz, MD, MBA and I are going to dig in on the model Thursday: → What does CJR-X change for hospitals already thinking about TEAM — and what does it mean for the ones that assumed mandatory bundles wouldn't reach them? → Where does the 90-day window create real operational opportunity (post-acute navigation, PT partnerships, readmission prevention) versus where it creates brittle risk for hospitals without the infrastructure to manage it? → How do we think about gainsharing, physician alignment, and the role of ASCs and HOPDs as outpatient LEJR volume keeps migrating? Going live Thursday on the Techy Surgeon Substack with Benjamin Schwartz, MD, MBA to dive in on the model and discuss the shifting incentives of musculoskeletal care. If you're an orthopedic surgeon, hospital CFO, a service-line leader, or anyone trying to make sense of where MSK value-based care is actually heading — come join us. A reminder, public comment on CJR-X closes June 9. (Also if you haven't checked out the new Chat GPT Image module it's pretty incredible and made the anime portrait of us below).

The CMS ACCESS Model just announced its first cohort. 150+ organizations. The list is fascinating. This is the who's who of digital health meeting Medicare for the first time. Noom, Verily (Alphabet's life sciences arm), Whoop, Withings , Concert Health, Story Health Partners, Slingshot AI — alongside a few health systems, ACOs, and physician practices. CMS says the majority of accepted organizations have never previously served Medicare beneficiaries. That's probably the point. "Insurgents" innovating. This model is designed to pull technology-driven chronic care into the Medicare ecosystem at scale, not tweak what's already there. But there are some surprising absences too. Several major digital health players you'd expect to see aren't on the list (particularly in MSK as Benjamin Schwartz, MD, MBA pointed out). Whether that's by choice or by design is worth watching. RevelAi Health 's platform MSK ACCESS was accepted for the MSK track as was our partner Understood Care , and we'll be delivering musculoskeletal and behavioral health care to Medicare beneficiaries through our platform. We see this model as a patient engagement and chronic care management asset, not just for our direct work, but for all our ACO partners, orthopedic partners, hospitals and physical therapy practices looking to amplify the value based, consumer grade care their MSK patients need. The synergy with existing CMS models (TEAM, ACO REACH) is remarkable. A few reminders about ACCESS broadly: This is a 10-year model. Payments are contingent on measurable health outcomes, not volume. Rural beneficiaries get enhanced reimbursement. And CMS just extended the application deadline to May 15 because interest far exceeded expectations. Later applicants can still join the January 2027 cohort. Four clinical tracks are live: early cardiometabolic conditions (hypertension, obesity, prediabetes), established CKM conditions (diabetes, CKD, ASCVD), chronic MSK pain via non-pharmacologic approaches, and depression/anxiety through digital therapeutics. Want to see the applicants in detail? I built an interactive explorer mapping the entire ACCESS landscape — you'll find a link in the updated article on the AI-native approach to this model that I think is going to work best.

So relevant: "What if the future of human healthcare is already being shaped in veterinary medicine?" Ulrika Holmquist

I don't think there has been a more important moment for orthopedic surgeons to engage with value-based care. This last Friday (at 4PM right before the weekend...why do you do this to me CMS??) FY 2027 IPPS Proposed Rule was released. CMS detailed the Comprehensive Care for Joint Replacement Expanded Model CJR-X — a mandatory, nationwide bundled payment model for lower-extremity joint replacement (total hip replacement, total knee replacement, total ankle replacement), starting October 1, 2027. I spent the weekend poring over the 230 pages that pertain to it and its sibling, TEAM. A few things that jumped out: — Mandatory participation for EVERY eligible acute-care hospital in the country. No curated geographies, no voluntary opt-in. — 90-day post-discharge episode, and for the first time outpatient total joints are inside the bundle. — 100% regional benchmark from day one. No hospital-specific cushion. — A Composite Quality Score where 40% is patient experience and 7-day post-HOPD ED visits carry meaningful weight. — Three-layer financial arrangements that will require surgeons and hospitals to actually learn gainsharing. — Hip fractures that get a total hip replacement admitted through the ED trigger a 90-day episode. I believe this is a paradigm shift in how orthopedic care is delivered, and if this mirrors TEAM and CJR we can expand penalties of up to 20% on episode spend. The upside requires hospitals to deliver something that feels like a consumer-grade experience. Under this rule, patient experience comprises 40% of the composite quality score. For surgeons, this is the moment to understand gainsharing, collaborator structures, and what it means to co-manage an episode with the hospital you operate in. And for hospitals engage with us surgeons as drivers of experience, cost and outcomes. And invest in care navigation infrastructure today, not tomorrow. I think I went pretty deep on this one. Deep Substack article, a NotebookLM studio against the primary rule text, and the extracted 230 pages are all in the comments for anyone who wants to dig in. We'll have an academic primer on this model soon too. Make sure you check out Benjamin Schwartz, MD, MBA article as well, “Does CMS hate specialists?” — great perspective and insight!

Don’t be afraid to switch from ChatGPT to Claude. I know — you’ve spent months training GPT to understand how you think. Your prompts are dialed in. Your memory is loaded. Starting over feels like a loss. It’s not. And the switch is simpler than you’d expect. Here’s exactly how I did it: Step 1: Go to claude.ai/import . There’s a ready-made prompt sitting there. Copy it, paste it into ChatGPT, and let it run. GPT will generate a structured export of everything it knows about you. Copy that output and paste it into Claude’s import page. That alone gets you started. Step 2: Go deeper. In ChatGPT, go to Settings → Data Controls → Export Data. You’ll get a confirmation email, and a couple days later, a zip file with your full conversation history. Step 3: Create a new Project in Claude and drag that zip file straight in. Now Claude doesn’t just have a summary — it has your actual conversation history, your decision patterns, your communication style, the problems you keep coming back to. Step 4: Let Claude write its own project description and instructions. Give it the context and let it optimize how it understands you. I actually use Claude to refine the project setup itself. Your AI should compound over time. Don’t start from zero when you don’t have to. 🔗 Full video walkthrough coming on Techy Surgeon.