In partnership with

Hey Hospitalogists,

Today I’m sharing the story of why Privia's "boring" contract strategy is outlasting the VBC enablers who swung for the MA fences. I’m also recapping July's roundtable with Geisinger's Jonathan Slotkin on AI governance, trust, and why concealment (not the mistake itself) kills trust.

Before we get into it, a few things I’m looking forward to:

8/20 → Healthcare's Oppenheimer Moment: Thursday, I’m putting the one and only Eric Larsen in the hot seat to defend some of his biggest claims from his manifesto on AI in healthcare. RSVP to join the live event or get the recording.

9/4 → Community Roundtable: The September Roundtable with guests Mara Kilgore and Loren Anthes of Health Management Associates.

9/23 → Happy Hour in Vegas: At The Venetian, fresh off day two at the Vizient Connections Summit, presented by Hospitalogy and ShiftMed.

11/2-3 → Hospitalogy AI Retreat: A no-cost retreat for senior healthcare leaders. Join execs from Kaiser, Sutter, Baylor, Henry Ford, Mayo Clinic, UCHealth, Adventist, Ascension, etc.

Was this email forwarded to you?

Subscribe Here

Sponsored by Regard

I'm a big fan of stats that reframe an entire category, and here's one: physicians see roughly 3% of a patient's chart. The other 97% (old office notes, med lists, lab trends) is where the diagnosis that actually explains the admission tends to be hiding.

At Sentara Health, a hospitalist was treating a patient for uncontrolled heart failure and assumed medication nonadherence. Regard's AI flagged hyperthyroidism instead, buried in a primary care note, a diagnosis the hospitalist says he "wouldn't have even known to look for." That kind of catch, at scale, drove a 17% jump in CC/MCC capture and up to 4x ROI per user, with 74% voluntary hospitalist adoption during the pilot.

This case study is worth the read if you're thinking about where your CDI dollars are actually going.

Read the Case Study

BLAKE’S BREAKDOWN

Privia, Tony Gwynn, and the Death of the VBC Grand Slam

Most of healthcare wants to hit the grand slam. Full-risk Medicare Advantage. Vertically integrated payvider. Direct-to-employer disruption. Twelve hundred percent revenue growth pre-IPO. Big swing, bigger headline, bigger valuation.

Then 2024 happens. The MA bid cycle turns. CareMax and Cano Health implode in slow motion. Agilon's stock chart looks like a luge run. And suddenly the "value-based care is dead" takes start landing in my inbox at a clip of about two a week.

Jason Ross, Executive Vice President of Medical Groups at Privia, has heard this song before. He was doing what we now call VBC back in 2007 and 2008 across multiple markets. He's watched the cycle peak, valley, peak again, valley again. So when I asked him where Privia sits in this latest reset, he basically shrugged: we've been through this. The lust is gone for now. It'll come back. In the meantime, you build a book that doesn't depend on the cycle.

Which brings me to Tony Gwynn.

100 VBC Contracts, 100 FFS Contracts, Zero Apologies

Jason's favorite player growing up was Tony Gwynn. Hit for average. Career .338. Eight batting titles. Never hit more than 17 home runs in a season. Easy first-ballot Hall of Famer.

That’s the model. Jason didn't say this as a soundbite, but it's the cleanest way to summarize his pitch: Privia's model is structurally boring, and that's the whole strategy.

100+ VBC contracts. 100+ FFS contracts. Every Blues plan that matters. United. Cigna. Aetna. Humana. Across commercial, MA, traditional Medicare, employer, ACO, capitated, shared savings, upside-only, two-sided. Whatever the doctor needs to keep the lights on, Privia's at the table for it.

3,500 implemented providers when Jason joined. Now knocking on 6,000. Seven states to 15 for the PMG platform. Singles and doubles. A boring 100+/100+ contract portfolio while everyone else swung for the MA fences and ended up in the dugout. Aggressive prudence (his exact phrase) instead of aggressive growth.

The pure-play VBC enablers maxed out MA full-cap lives and bet on benders, stars, and RAF capture. When any one of those bets goes sideways, as it did in 2024 and 2025, there's nothing left to balance the book. Privia's diversification means a tough MA cycle hurts but doesn't kill. There's an FFS book underneath generating contribution margin while the MA side gets recalibrated. Not a sexy positioning slide. A survivable one. After the carnage of the last two years, survivable is starting to look really attractive.

I pushed Jason on how Privia decides where on the risk spectrum to land each market and each physician group. The answer: actuarial rigor first, contract walk-away discipline second. They underwrite every population, model out the guardrails physicians would need to win, and if the math doesn't work, they pass, even on attractive-sounding MA full-risk plays.

That discipline is the whole ballgame. A lot of enablers signed pure-play full-risk deals where the docs had, in Jason's words, "absolutely zero chance of success." Aggressive growth plus weak underwriting is how you get the 2024 enabler death pile. Privia mostly avoided it through some luck, but mostly discipline.

The philosophical anchor Jason kept returning to is skin in the game, for everyone in the risk loop: payor, provider, enabler, even the patient. If anyone's fully insulated, you get moral hazard, the wrong incentives flow through the system, and the deal eventually breaks. It's why Privia walked from deals other enablers grabbed, and why Privia is still here writing earnings releases instead of restructuring debt.

The health system physician subsidy thesis

Jason is one of the rare operators who has run a managed care function inside an integrated health system and now sits across the table from those same systems on partnership conversations. So I wanted him to lean into the health system subsidy thesis.

I've written for years that the per-employed-physician loss numbers MGMA and Kaufman Hall publish (the $300K+ annual loss per physician at most systems) are economically indefensible at scale. The whole edifice depends on downstream specialty and surgical revenue covering up the loss. But the specialty revenue is also under pressure, the site-of-service shift is real, payors are pushing back on facility fees, and at some point the math breaks.

Jason essentially confirmed the thesis, and then introduced a wrinkle I appreciated. He said he has guessed wrong every single time on when health systems would finally cry uncle on the subsidy model. So whatever you think the timing is, push it further out.

His read is that there are two camps. There are systems that have privately told Privia they cannot keep doing this, the model is broken, they need an alternative. And there are systems that still have the capital and the market position to keep acquiring practices and absorbing the per-physician losses because their downstream revenue still supports it. The systems with the capital aren't the ones rethinking the model. The systems rethinking the model often can't muster the courage to walk back 20-plus years of acquisition strategy and admit they were wrong.

I pushed back here because I keep hearing the "doctors are lining up to be employed" framing from health system folks at conferences, which directly contradicts the "the pendulum is swinging back to independence" framing from enablers like Privia. Jason's answer: both can be true. There's a bell curve of physician preferences. Some want W-2 stability and will keep walking into employment. Some want PE platform optionality. And there's a meaningful and growing portion of the curve that wants to stay independent and just needs a partner to make it economically viable.

Forget scale; density is the whole game

A word that kept coming up in our conversation was density. Most enablers and health systems sell scale. Total physicians. Total markets. Total lives. The leaderboard everyone tracks is national. Jason's point is that none of that matters operationally unless you have provider density inside specific geographies. Enough PCPs across family medicine, internal medicine, peds, and OB-GYN, plus the downstream medical and surgical specialty coverage, all operating under the same ethos in the same metro area.

When you have that density, you can quote-unquote manage the care pattern of a population. You see your attribution across 3, 4, or 5 payor panels in the same market. You outperform on quality and total cost not because you're a magic ACO operator, but because the patient's downstream care is mostly happening inside the same group of physicians who share an information backbone and a value system. The conflict of interest that haunts hospital-employed primary care — where the system's economic incentive runs against the payor contract's incentive — largely disappears when the PCP isn't a cost center for somebody's facility-fee book.

This matters because the next leg of VBC isn't going to be won by the enabler with the most attributed lives nationally. It's going to be won by the enablers and groups that can credibly say "in this metro, we are the primary care infrastructure for X% of the commercially insured population." That's a density (not scale) argument. And it's why Privia's market-by-market growth (7 states to 15 for the PMG platform, plus 9 more through the Evolent Care Partners deal) matters more than the topline provider count.

Sponsored by Navvis

I've been tracking CJR-X since the model dropped this month, and one thing's clear: the health systems set up to succeed in CJR-X aren’t starting from zero. They're building on proven capabilities:

• Episode management

• Physician alignment

• Post-acute optimization

• Care coordination

• Performance analytics

Navvis has spent more than a decade helping health systems develop and operate exactly these capabilities across bundled payment programs. Curious where your organization stands? Take 2 minutes and use this quick form to request your CJR-X Readiness Assessment.

Request Your Assessment

ROUNDTABLE RECAP

AI Transformation Expert Session with Jonathan Slotkin of Geisinger

July’s roundtable featured Dr. Jonathan Slotkin, Geisinger neurosurgeon, AI/innovation leader, and Hospitalogy AI Retreat (in November!) keynote, speaking about the surprising throughline between his advocacy work on autonomous vehicle safety and how health systems should actually be governing and deploying AI. The conversation ranged from Capitol Hill to Geisinger's internal AI roadmap, with sharp pushback from community members on governance paralysis, vendor hype, and where the real deployment bottlenecks live.

Key Takeaways

• Autonomous vehicle safety has become a public health story and clinicians have standing to own it. Slotkin's analysis of Waymo's public safety dataset went viral, led to a New York Times essay, and culminated in an open letter signed by 400+ clinicians calling for smarter AV safety reporting. His argument: road deaths kill roughly 40,000 Americans a year (the equivalent of a fully loaded plane crashing daily), and clinicians bring public-health expertise to that problem that traffic researchers simply don't have.

• Healthcare already knows something the AV industry is still learning: trust is infrastructure, not marketing. Slotkin's framework (measure "felt safety" separately from actual safety, insist on independent verification, and engineer systems for the worst day) comes directly from healthcare's patient experience movement. The gap between objectively safe and subjectively trustworthy is real, and paving over it with better PR doesn't work in either industry.

• Concealment kills trust. The Cruise AV division wasn't shut down because its vehicle dragged a pedestrian; it was shut down because Cruise didn't disclose the dragging in its federal incident report. Slotkin drew a direct line to healthcare's own reckoning with proactive, candid error disclosure (University of Michigan's model was cited as the gold standard). Don’t hide the miss. Radical transparency is what protects an organization's downstream trust.

• AI is a method for executing strategy and Geisinger's governance is built to enforce that discipline. Their intake funnel is intentionally wide (a "Catalyst" group vets ideas early), but a senior leadership group still has to sign off on directional bets, and execution routes through innovation, IT/informatics, or partnerships depending on the ask. The organizing question at every stage: is this solving a problem we actually have, or are we just excited about the tech?

• Geisinger's 18-24 month AI roadmap centers on 5 horizontal capabilities that all want to integrate with each other: LLM-in-the-EHR for chart summarization, next-best-action engines (individual and population-level), knowledge management/CDS, Epic Cosmos for cross-organizational insight, and ambient scribing. The real work isn't picking one. It's stitching all 5 into something usable rather than 5 more disconnected point solutions.

• The health system should own the central "brain," with vendors plugging in as specialized modules, not the other way around. A community member’s framing: every AI vendor wants to expand into being all 4 of the other categories, but health systems are best positioned to understand their own workflows and should build the integration layer themselves, sourcing specialty capabilities rather than resources they don't need to build in-house.

• Foundational HIPAA and data governance questions are largely solved, and using them as a reason to move slowly is now more excuse than obstacle. Multiple participants pushed back on governance-as-blocker, particularly for non-patient-facing use cases (admin, back-office workflows) where the risk of "moving fast and breaking things" is low and the cost of continued AI-strategy paralysis is high.

• Deployment architecture is a significant bottleneck right now. One guest flagged the tension between safely test-driving new AI tools in a sandboxed environment and building the sustainable, enterprise-wide infrastructure needed so those experiments don't calcify into 50 disconnected silos. Another guest added that most orgs still lack real change management or a rigorous way to measure ROI, pointing to MD Anderson's Watson failure as the cautionary tale of buying the demo instead of the outcome.

• Cost and build-vs-buy remain real constraints despite blistering AI adoption rates. One guest noted healthcare is adopting AI 2.2x faster than any other US industry, but many systems still can't justify $300K for an unproven point solution, raising the practical question of whether to lean on EHR-native tooling (Epic, Cerner) versus an early-stage startup with no track record.

• The likely labor outcome is creative destruction, not wholesale replacement, but the burden won't land evenly. Slotkin's framing, borrowed from this year's Nobel Prize in Economics: innovation typically creates more new roles than it eliminates, and "medicine is an accountability relationship, not just a task" — meaning clinicians aren't going anywhere. But job losses that do occur will hit the most vulnerable, lowest-paid workers first, making deliberate workforce planning essential rather than optional.

Catch the replay on YouTube. And keep the conversation going by joining us for the Hospitalogy AI Retreat in November, including a keynote from Jon Slotkin.

Thanks for the read! Let me know what you thought by replying back to this email.

— Blake