The Workflow Doesn’t Stop at Submission

Short answer

For two years, skilled nursing was told its referral problem was a speed problem. That gap has largely closed.

AI built into the SNF EHR now helps teams assess referrals, surface documentation risk, identify missed billing opportunities, and — the part most coverage missed — flag authorization-readiness gaps and level-of-care mismatches before billing. A PDPM assessment assistant was announced as beta this summer.

That closes more of the referral-to-revenue chain than we credited it with in July.

But a readiness check is not a closed prior-authorization workflow. Once a payer issues a determination, a different operational loop begins: someone has to receive it, interpret it, respond to a coverage-end notice when appropriate, assemble evidence for an appeal, confirm submission, and learn from the result.

For many operators, that loop still spans multiple systems. And it looks different at every one of them — which is why a packaged product alone may not close it.

What shipped this year

In July we wrote that referral tools stopped at the accept/decline decision, and that authorization, assessment and appeal were "still someone's spreadsheet and someone's Friday afternoon."

The market moved faster than that map.

Row three is the one that changes the argument. Catching an authorization gap before billing isn't billing automation. It's the platform reaching upstream into one important source of post-acute denials: whether the documentation available to the plan supports the requested level of care.

And the case for staying native is a good one. An assistant built on the platform your teams already work in acts on notes, orders and assessments it already holds. Nothing bolted on from outside starts with that much context. Consolidating onto one vendor instead of managing five is not irrational.

For most of the workflow, that reasoning is simply right.

So where does the workflow fragment?

At the handoff from a provider-side readiness check to a payer-side decision.

The described capabilities are organized around the provider's EHR workflow — notes, orders, assessments, charge capture, pre-submission checks. Some include eligibility signals. What the public materials don't establish is that payer determinations, coverage-end notices and appeal outcomes are managed as one closed loop inside that same workflow.

Pre-submission isn't the only place a SNF loses money. Documentation gaps, missed charges and coding errors are real, and they're exactly what the new capabilities address. But payer determinations create a separate downstream exposure that pre-submission controls don't necessarily resolve.

The payer makes the coverage determination, using information that may be split across the provider record, payer criteria, authorization documentation and the plan's own review process. And it comes back on the payer's schedule — through portals, structured transactions, scanned documents, secure messages or a phone call, depending on the plan and the operator's connectivity.

The question isn't whether a vendor can reach that data. It's whether it's in the same loop as everything else.

What the announcements reviewed for this article do not describe:

• submitting a prior authorization through a payer's portal
• tracking pending determinations across plans in one place
• catching a mid-stay coverage termination in time to act
• assembling an appeal from the EHR, a therapy vendor's system and a scanned document
• filing that appeal and confirming receipt
• feeding the outcome back into the next referral from the same plan

For many operators, those steps still mean people moving information between systems under a deadline.

Why does that stretch cost so much?

Because in skilled nursing it isn't an edge case. It's the default path.

In KFF's 2026 analysis, 95% of Medicare Advantage enrollees are in plans requiring prior authorization for a SNF stay; 99% for at least some service. For many MA admissions, the payer conversation isn't a side process. It's part of the admission workflow.

Operators are already absorbing that cost. In an AHCA/NCAL survey of 363 nursing home providers, conducted May 2025: 66% saw MA denials or delays daily or weekly — 29% daily, 37% weekly. Sixty-seven percent had a plan pull coverage prematurely, against the advice of the resident's healthcare provider. And many reported limiting new admissions from certain payers in response.

That last one is worth sitting with. It turns reimbursement and administrative friction into a placement constraint — narrowing the set of facilities willing to accept a given plan.

One caveat on coverage terminations: these notices operate within Medicare Advantage utilization-management and continuity-of-care rules. The issue here is not that every notice is improper. It's that facilities need to receive, interpret and respond to it while there's still time to act.

What happens to denials that do get contested is a separate question, covered in The 82% That Never Reached Appeal.

What would close it?

Not necessarily a bigger appeals team. Staffing, managed services and process redesign can all be part of the answer — they just work better when the record moves with the decision.

For operators seeking a closed-loop post-decision workflow, a practical target architecture includes six capabilities:

• Receive and normalize determinations. Approvals, denials, partial approvals and coverage-end notices arrive in formats that are hard to compare. Turn them into structured events, not documents in an inbox.
• Treat a coverage-end notice as an operational event, not a later billing exception. While the resident is still in the building, there may be time to review, confirm the record, talk to the family and act.
• Assemble evidence from wherever it lives — therapy notes with a contracted vendor, assessments in the EHR, the determination in a scan. All of it, not the convenient subset.
• File, confirm receipt, track the deadline. An appeal drafted but not submitted is indistinguishable from one never written.
• Track outcomes by payer, denial reason and stage — subject to privacy, retention and contractual requirements. Not an aggregate appeal rate.
• Feed it back to intake. When the same plan sends a referral with a similar clinical pattern, the admissions team should be able to see comparable prior outcomes — not start from the packet alone.

None of that is exotic. What makes it hard is that it crosses systems whose ownership tends to fragment — differently at every facility. The exact integration pattern depends on the EHR, payer connectivity, document formats, delegated entities, therapy vendors, and whatever revenue-cycle and utilization-management tools are already in place.

Why this is a build, not a subscription

Buy the native capabilities for what they cover. Where a facility's process matches the model those products were built on, buying is correct and building is waste. That's most operators' right answer.

The post-decision loop is different — and not because of anyone's roadmap. It resists packaging. A single template rarely fits every operator's payer mix, contract structure, therapy arrangement and documentation history without meaningful configuration.

So the question stops being which product covers this. It becomes what do we build, and around whose systems.

That's the work, and it's worth naming who it's for. Not every operator. The one who already went through the marketplace, priced the monthly subscriptions, and found nothing on the shelf matched the building they actually run — because therapy documentation sits on a separate platform, or two facilities came in through an acquisition and never converged, or the plans they deal with most still send determinations as scans.

For that operator, interoperability isn't a feature. It's the deliverable: determinations received and normalized, coverage-end events caught in time, appeal evidence pulled from wherever it lives, outcomes tracked by payer and returned to the people making admission decisions.

Specified around their systems. Built for their configuration. Owned by them, not licensed back monthly.

And because it's specified rather than subscribed to, it changes when the operator needs it to — not when a roadmap says so. That's what matters in year three.

One honest qualifier: some of those six steps may already exist in a given environment through modules or connectivity a public product page doesn't show. Finding out what's actually connected comes first. That's a discovery conversation, not a proposal.

The short version

EHR-native AI in skilled nursing got better this year. It can surface authorization-readiness gaps before a claim goes out.

A readiness check is not a closed payer-decision workflow. Receiving the decision, acting on a coverage-end notice, assembling evidence, tracking the outcome, learning from it — those remain architecture questions.

They need an owner.

When a payer determination comes back on one of your residents, who sees it first — and how long until it reaches the person who can act on it?

Itirra builds custom integration for healthcare organizations whose workflow doesn't fit a packaged product — the interoperability layer between clinical systems, billing systems and payer-facing processes. Specified around your systems, built for your configuration, owned by you rather than licensed back monthly.

Let's talk about your project.

Product scope described here reflects public announcements reviewed in August 2026.

FAQ

Does EHR-native AI replace a denial-management workflow?

Not on the evidence of what's been announced. The described capabilities cover provider documentation, referral assessment, billing integrity and pre-submission authorization readiness. The announcements reviewed for this article don't describe a closed loop for payer determinations, coverage-end notices, appeals and payer-level outcome tracking. In many organizations those workflows still span payer portals, EHR documentation, therapy systems, scanned records and manual follow-up.

Should a SNF buy an AI suite or build its own integration?

For most of the workflow, buy. Where referral, documentation and billing processes match the model a packaged product was built on, buying is faster and cheaper, and building would be waste. Custom integration becomes worth evaluating at the handoffs a package doesn't cover — and specifically when a configuration differs from the standard one: separate therapy platforms, two documentation standards after an acquisition, determinations arriving as scans. It isn't automatically right either; an existing module, a payer-connectivity vendor or a process redesign may fit better. Start with a discovery pass on what's already connected.

How many Medicare Advantage plans require prior authorization for a SNF stay?

In KFF's 2026 analysis, 95% of MA enrollees are in plans requiring prior authorization for SNF stays and 99% for at least some service — versus 97% for acute inpatient, 94% for Part B drugs, 90% for home health. Two caveats matter: KFF's estimates exclude employer group health plans and special needs plans, and they measure the share of enrollees in plans with the requirement, not the share of SNF admissions for which a request is ultimately submitted.

What does a mid-stay coverage termination cost a SNF?

It creates financial exposure for the remaining stay plus an immediate operational burden: clinical review, documentation, family communication and, where appropriate, an appeal or alternative discharge plan. The financial effect depends on the payer contract, the coverage decision, the appeal outcome and available discharge options. In the AHCA/NCAL survey, 67% of providers reported coverage pulled prematurely against the advice of the resident's healthcare provider. These notices operate within Medicare Advantage utilization-management and continuity-of-care rules — the issue isn't that every notice is improper, it's receiving, interpreting and responding to it in time to act.

Sources

• Medicare Advantage in 2026: Premiums, Out-of-Pocket Limits, Supplemental Benefits, and Prior Authorization — KFF. These estimates exclude employer group health plans and special needs plans, and measure the share of enrollees in plans with a prior-authorization requirement — not the share of SNF admissions for which a request is ultimately submitted.
• Provider Insights: Medicare Advantage — AHCA/NCAL (n=363, conducted May 2025, published August 2025)
• Medicare Advantage post-acute nursing home denials — Healthcare Dive
• MAOs Overturned Nearly All Appealed Prior Authorization Denials for SNF Admission — HHS-OIG
• Contract Year 2024 Policy and Technical Changes to the Medicare Advantage Program (CMS-4201-F) — Federal Register
• Public product materials from a leading SNF EHR vendor: a January 2026 referral-AI release, a June 2026 AI-suite announcement and a billing-capability product page; reviewed August 2026. Source copies, URLs and verbatim product-scope excerpts retained in Itirra's research file.

Related reading: The Cost of a Fast Yes · The 20-Minute Referral Gamble · The Prior Authorization API Is Still "Coming Soon."