A six-location Tennessee orthopedic practice put AI on surgical coding and denials instead of hiring more staff
November 4, 2026 · 5 items
A six-location Tennessee orthopedic practice put AI on surgical coding and denials instead of hiring more staff
Healthcare IT News · November 3, 2026Regenerative medicine
Bone and Joint Institute of Tennessee — six locations outside Nashville, roughly 350 employees spanning orthopedic physician care, rehab and DME, with surgeons in both hospital and ASC settings — deployed vendor Ember AI against two targets: surgical coding and denial resolution.
CEO John Reynolds framed the problem as an arms race, not a staffing gap: 'payers' increasing use of AI and automated claim review was creating an imbalance… Claims could be reviewed and denied at scale, but responding still required our team to investigate individual accounts, review documentation, interpret denial reasons, and prepare corrections or appeals.'
The stated goal was accuracy 'comparable to an expert surgical coder at a substantially lower cost per case' — and they ran an actual comparative evaluation, benchmarking Ember against expert surgical coding on both accuracy and cost per coded case, with the director of operations leading implementation and the revenue cycle manager involved. The headline and dek report an 80% cut in surgical coding costs with expert-comparable accuracy.
It runs alongside their athenahealth AthenaOne EHR, with weekly AthenaOne coordination — explicitly positioned as not replacing experienced employees.
Why it matters for an independent practice: This is the clients' specialty and the clients' size — an independent multi-site ortho group, not a health system. The replicable part isn't the vendor, it's the method: before you buy, run your own cases through both the AI and your expert coder and compare accuracy AND cost per case. That benchmark is what makes the ROI defensible to a partner group.
Athenahealth now lets patients pipe their own chart into Claude — and your patients will arrive at visits holding its answers
Healthcare IT News · November 3, 2026Practice operations
Announced Monday: the athenaPatient MCP server connects a patient's record to Anthropic's Claude over the open Model Context Protocol. Patients sign in with athenahealth credentials, choose what to share, and can revoke access at any time; Claude Pro and Max users in the U.S. can add the connector.
What Claude can reference once connected is the whole chart: lab results, medications, health conditions, immunizations, visit history, diagnostic reports, allergies, procedures and care plans.
Anthropic's head of healthcare Syed Mohiuddin positions it around the visit itself — 'Patients often have questions about their health records between visits, from what a lab result means to what to ask at their next appointment.' Athenahealth's Paul Brient calls it 'an important milestone in responsibly integrating AI into the patient experience.'
Healthcare IT News flags the counterweight directly: these tools 'hold the potential to deliver inaccurate medical information and lack clinical liability.' Michael Abrams of Numerof & Associates: 'Responsible boards and executives must recognize the risks associated with these new tools and take steps to protect the institution and staff.' Context: Elation Health integrated Claude into its EHR Clinical Insights platform earlier this year, and OpenAI launched ChatGPT for Healthcare at AdventHealth, Boston Children's, Cedars-Sinai, HCA, Memorial Sloan Kettering and Stanford Medicine.
Why it matters for an independent practice: If a practice is on athenahealth, this is live for patients now and requires no action from the practice — which means the first encounter with it will be in an exam room, with a patient quoting an LLM reading of their own labs. Decide this month how staff respond to 'Claude said my result means…', and consider getting ahead of it with physician-authored explainers on the common labs your panel asks about. Patient-held AI is a trust surface, and whoever answers the question first owns the authority.
Anthropic is building an open-source way to prove which model actually produced a given AI output
TechCrunch · November 3, 2026Buildable AI
Per a blog post Anthropic published Tuesday, the system is called FlashProve and is described as open source — built to deliver what researchers call 'provable inference.'
The stated function, in TechCrunch's summary: let anyone check that an AI response really came from the model it claims to be from.
Reported by Russell Brandom, Nov 3, 2026 at 2:41 PM PST. Note the limits of what's confirmed here — this is reported as a system Anthropic 'is working on,' not a shipped product with availability or pricing.
Why it matters for an independent practice: Attribution is the unsolved problem in every AI-assisted content workflow a practice runs. If model identity becomes cryptographically checkable, the near-term use is vendor accountability — a scribe or marketing vendor saying 'we use the frontier model' becomes a verifiable claim rather than a sales line. Worth tracking as infrastructure for the thing Atiba already insists on: content whose provenance a physician can stand behind.
ONC's HTI-5 final rule landed smaller than proposed — and it punted the AI information-blocking rules entirely
Healthcare IT News · October 30, 2026Practice operations
HTI-5 — formally 'Health Data, Technology, and Interoperability: ONC Deregulatory Actions to Unleash Prosperity' — was published in the Federal Register on Oct. 28. It removes six annual reporting measures for EHR developers.
It also preserves legacy data exchange standards for public health and safety net providers, pushing their transition to FHIR out to 2030.
The draft earlier this year proposed removing 34 and revising seven of the program's 60 criteria; ONC left many unchanged and — the line that matters here — 'notably delayed rulemaking on artificial intelligence-related information blocking policies.'
On this HIMSSCast episode, HIMSS government relations' Alana Lerer and Jonathan French attribute the slowdown to industry feedback and political sensitivities around the midterm elections, with information blocking and HTI-6 among the talking points.
Why it matters for an independent practice: The practical read for an independent practice: the federal rules governing how AI-driven decision support interacts with information blocking are now on a later clock than expected, so vendor contracts — not regulation — are what governs your exposure in the interim. When a scribe or coding vendor tells you a feature is 'compliant,' ask which rule, because the AI-specific ones haven't been written yet.
UCHealth's 2018 sepsis-alert AI is why it can now build its own virtual nursing tool, its informatics director says
HIMSS TV · November 3, 2026Practice operations
Short interview clip with Brittany Cyriacks, UCHealth's chief informatics program director, posted Nov 3, 2026 at 9:26 AM.
The framing in the publisher's description: because the health system started using AI in 2018 to help identify sepsis and patient deterioration, it is now in a position to develop its own AI tool for virtual nursing.
Disclosure on thinness: this is the channel's own description, not a transcript — no runtime, no outcome numbers and no detail on the virtual nursing tool are stated. Treat it as what the video covers, not what it concludes.
Why it matters for an independent practice: The transferable idea is compounding, not sepsis: the capability to build your own AI comes from years of operating someone else's. For a practice, that argues for starting with one boring, measurable deployment — coding, intake, recall — and keeping the institutional muscle, rather than waiting for a perfect flagship tool.
