OpenEvidence in the EHR: Vertical Research Tools Win by Embedding Into the Institution's System of Record — and Now the Hosts Are Embedding Too
Revised 2026-09-27: adds the September 22–24 wave — the quiet $250M raise at a $15B valuation, the Memorial Sloan Kettering Epic + OncoKB partnership, the Anthropic 100-country access deal, the Darwin model family, and the pivot into oncology drug development — plus the host-layer countermove: OpenAI's read-only Epic integration for ChatGPT Health (announced Sept 1, 2026). Prior cycle's core evidence (the UTMB embed, >50% clinician adoption, SSO) stands.
The vertical keeps compounding
The distribution endgame first demonstrated at UTMB (Sept 22: OpenEvidence embedded in the EHR with SSO) is now a multi-front strategy:
- Deeper Epic integration + proprietary data: the new Memorial Sloan Kettering deal works both ways — "MSK is plugging OE into its Epic workflows, and OE is integrating MSK's OncoKB precision oncology database into its engine to let physicians access the treasure trove of data on genetic variants of different cancers and their treatments." (Digital Health Wire) That is the vertical moat in one sentence: institution-embedded workflow plus exclusive data integration.
- A quiet mega-round: "OE quietly raised $250M at a $15B valuation last week" — after $250M at $12B in January and a reported weighing of $200M at $20B in July. "Not necessarily a down round, but the silence was deafening."
- Distribution via foundation-model partner: a Reuters-reported partnership "with Anthropic to give physicians in 100 lower-income countries open access to OE's clinical decision support."
- Model-layer credibility: a new family of four AI models featuring "Darwin, the first AI in history to score a perfect 100% on the MedQA benchmark."
- A pivot beyond research tools entirely: Nadler told Forbes OE will develop its own oncology therapies — "the first drug entering clinical trials before year-end" — going after "the stuff that they are not doing, and maybe can't do" for lack of OE's "economies of scale in finding these patients." A research-tool company is becoming a pharma company on the strength of its evidence network.
Competition is compounding from both incumbents (Wolters Kluwer, Elsevier) and new heavyweights (Abridge with Lilly and NVIDIA ties; Doximity's "blowout quarter" on directly competing products) — and one rumor on Doximity's call of a "major system" banning a competitor product over shadow-AI enforcement shows institutional distribution can be revoked as well as won.
The host-layer countermove: OpenAI wires Epic into ChatGPT
On September 1, 2026, OpenAI integrated ChatGPT Health with Epic — "which holds data for over 325 million patients — to let clinicians import patient data and use AI to ask questions," with read-only access ("AI doesn't write anything back"), and "in some deployments, clinicians can directly access ChatGPT for getting pre-visit review and building clinical timelines without leaving a patient chart." It also shipped a Healthcare Public Data plug-in (ClinicalTrials.gov, CMS Coverage, RxNorm, DailyMed, PubMed) and opened BAA-covered use of ChatGPT Work, Codex, apps, and connectors for compliant workflows. (TechCrunch)
OpenAI's safety evidence is self-reported — "over 4,300 responses from physicians across 27 clinical use cases... 99.1% of responses were safe" — against pending lawsuits alleging harmful medical advice, and with the company maintaining AI "is not suitable for diagnosis or treatment."
What this means for the market map
The embed-into-the-system-of-record play is now contested from both directions: the vertical deepens it (workflow + exclusive data + downstream monetization), while the horizontal host replicates it at consumer scale (read-only chart import + public-data plug-ins + 300M weekly health queries). For vertical research tools the defensible layer is the exclusive-data integration and the write-back/clinical-grade workflow — for orchestrators like Hey, Lefty, the lesson is that institutional embedding, not destination traffic, is where vertical distribution settles (Destination vs. Backend: The Research-Tool Market Bifurcates — and Both Layers Converge on MCP).