CRUSH: The AI Fraud Rule Now at OMB and the Hospice Billing Anomalies It Flags

4 days ago

6 min read

CMS sent its proposed CRUSH rule to the Office of Management and Budget on August 7, 2026, and the anomalies it is built to detect — unusually long stays, certification irregularities, geographic billing outliers — describe a large share of legitimate hospices as accurately as they describe fraudulent ones. CRUSH stands for Comprehensive Regulations to Uncover Suspicious Healthcare, and its stated purpose is to move Medicare from a "pay and chase" posture to pre-payment detection driven by AI and claims analytics. Hospice News reported on September 10, 2026 on what the models are actually surfacing. For hospice operators, the practical question is no longer whether an algorithm will read your claims. It is whether your documentation explains the pattern the algorithm will find. Here is what CRUSH looks at and how to prepare for a pre-payment review you will not be warned about.

Key Takeaways

CRUSH = Comprehensive Regulations to Uncover Suspicious Healthcare, a CMS program-integrity initiative that began as a Request for Information in early 2026.
CMS sent the proposed CRUSH rule to OMB on August 7, 2026, moving it from information-gathering into formal rulemaking.
The design shift is from "pay and chase" to pre-payment detection — stopping improper payments before they leave Medicare.
Models flag long lengths of stay, certification anomalies, geographic clustering, and unusual coding combinations.
The National Alliance for Care at Home asked CMS in April 2026 to revamp the CRUSH framework to reduce false positives on compliant providers.
CRUSH runs alongside a nationwide unannounced hospice site-visit project and the hospice enrollment moratorium effective May 13, 2026.
CMS reported preventing $1.6 billion in fraudulent Medicare laboratory payments (August 28, 2026) and announced a $3.4 billion DME supplier fraud action (September 8, 2026) — evidence the pre-payment posture is already operating.
MedPAC projects a 9% hospice fee-for-service margin for 2026 — a cushion that disappears fast under suspended payments.
The defense against an anomaly flag is contemporaneous clinical and financial documentation, not an appeal filed afterward.

What CRUSH Is and Where It Stands

From RFI to proposed rule

CMS opened CRUSH as a Request for Information in early 2026, seeking input on regulatory changes that would let the agency use advanced analytics and artificial intelligence to identify suspicious billing before payment rather than recover it afterward. The SBA Office of Advocacy circulated the RFI to small providers in March 2026. The National Alliance for Care at Home responded in April 2026, urging CMS to narrow the framework so that compliant providers are not swept up by pattern-matching.

Metric	Value	Source
Initiative name	Comprehensive Regulations to Uncover Suspicious Healthcare	CMS
Stage in early 2026	Request for Information	CMS / SBA Advocacy
Alliance comment urging revision	April 2026	Hospice News
Proposed rule received by OMB	August 7, 2026	OMB regulatory review docket
Enforcement model	Pre-payment detection ("detect and deploy")	CMS
Status	Proposed rule under OMB review — not final	OMB

Why hospice sits near the center of it

CMS has been explicit that hospice is a priority category. In a February 2026 report the agency said it had received numerous reports of hospice fraud, waste, and abuse in states that saw sharp increases in the number of Medicare-enrolled hospices. That same concern produced the nationwide unannounced site-visit project covering every Medicare-enrolled hospice, and the six-month nationwide enrollment moratorium effective May 13, 2026. CRUSH is the analytics layer under all of it. Even the CY 2027 Home Health PPS proposed rule fact sheet opens its enrollment section by describing CMS's "ongoing efforts to crush fraud, waste, and abuse."

The Anomalies the Models Look For

Hospice News's September 10, 2026 reporting describes CRUSH-style analytics examining utilization patterns, outliers such as unusually long stays, and anomalies in certifications. Existing CMS fraud-detection tools already flag suspicious claims before payment, analyze large volumes of claims and provider data, and identify outlier billing patterns, geographic anomalies, and unusual coding combinations.

Length of stay

Long stay is the most consequential flag because it is also the most legitimately variable metric in hospice. A dementia-heavy census produces long stays. So does a fraudulent one. An algorithm scoring you against national distributions cannot tell the difference; your certification and recertification documentation is what does.

Certification and recertification patterns

Face-to-face encounter timing, physician narrative quality, and the internal consistency between the narrative and the clinical record are all machine-readable. Templated narratives that repeat across patients are exactly the signal a model is trained to find.

Geography and growth rate

Rapid census growth in a market CMS has already identified as high-risk is a compounding flag, independent of anything in your individual claims.

Live-discharge and revocation rates

Discharge patterns are already embedded in the Special Focus Program methodology and the Hospice Special Focus / SSVI claims-based metrics. CRUSH extends the same logic to pre-payment.

Anomaly type	What triggers it	What defends it
Long length of stay	Stay distribution far above peer norms	Certification narratives, decline documentation
Certification irregularity	F2F timing gaps, templated narratives	Physician-specific, patient-specific narratives
Geographic clustering	Rapid growth in a flagged market	Referral-source documentation, admission criteria
Coding combinations	Unusual diagnosis/level-of-care mixes	Clinical rationale tied to the plan of care
Live-discharge rate	Elevated revocations or discharges	Discharge-reason documentation

What a Pre-Payment Posture Does to Cash Flow

This is the part hospice operators consistently under-model. A post-payment audit is an accounting problem: you were paid, you may have to pay some back, and you have time to respond. A pre-payment review is a liquidity problem: the money does not arrive at all, and the clock runs at whatever pace the contractor works.

Metric	Value	Source
Projected hospice FFS margin, 2026	9%	MedPAC, March 2026
Fraudulent lab payments prevented	$1.6 billion	CMS press release, Aug. 28, 2026
DME supplier fraud scheme announced	$3.4 billion	CMS press release, Sept. 8, 2026
Hospice enrollment moratorium effective	May 13, 2026	Federal Register

A 9% margin means roughly 33 days of operating cost sitting between a hospice and breakeven in a normal year. Hold 40% of a mid-size hospice's claims for 90 days and that cushion is gone — not because the claims were wrong, but because they were queued. Agencies that carry a documented 13-week cash forecast and a defined borrowing base survive an anomaly flag. Agencies that run on the assumption that Medicare pays in 14 days do not.

Reconciling your claims data against your general ledger monthly — rather than at cost-report time — is what makes that forecast real. That reconciliation is ordinary bookkeeping discipline, and it is the cheapest insurance available against a payment hold.

A Preparation Checklist for the Next Two Quarters

Pull your own outlier report before CMS does. Rank your patients by length of stay and identify the top decile. Confirm each has current, patient-specific certification documentation.
Audit ten recertification narratives at random. If two read alike, fix the process, not the two.
Reconcile face-to-face encounter dates against certification periods for the last two quarters.
Document your referral mix. Geographic and source concentration is a flag you can explain in advance.
Build a 13-week cash forecast with a scenario where 30% of Medicare receipts are delayed 60 days.
Tie your claims data to your cost report ledger monthly. Discrepancies you find in September are correctable; discrepancies a contractor finds in an ADR are not.
Track your SSVI inputs. The claims-based metrics that drive the hospice indicator overlap heavily with CRUSH anomaly categories.
Watch the Federal Register. Once OMB clears the rule, the comment window is the industry's last structured input.

The cost report is where all of this eventually reconciles, and a hospice cost report that already agrees with your claims data is the strongest single piece of evidence that your billing pattern reflects your patient population.

The Bottom Line

CRUSH is not hypothetical and it is not a hospice-only rule — it is a Medicare-wide shift to pre-payment analytics, and hospice is a named priority category inside it. The proposed rule is at OMB now. The models look for exactly the patterns that distinguish a dementia-heavy, rural, or rapidly growing hospice from the national average, which means being flagged is not the same as being wrong, but proving it requires documentation you either created at the time or did not. If you want help stress-testing your census documentation and building the cash forecast that goes with it, schedule a free consultation.