These are binary, reproducible checks. If the evidence is present, the rule fires. If it is not, the rule fires. A qualified reviewer does not need a model to know a signature is missing.
Flagship intelligence product
AEGIS
You can automate much more than you might think.
Audit & Evidence Governance Intelligence System — turn raw facility records into an evidence-backed prediction of where external scrutiny is most likely to find weakness.
The customer question
Not a checklist. A prediction of scrutiny.
If a surveyor, MAC, TPE reviewer, RAC, OIG investigator, payer auditor, lender, or buyer walked through the door today, what would they find first — and how much could it cost us?
That is the difference between data analytics and audit intelligence. AEGIS is built to answer it with cited evidence, not a chatbot sitting on a pile of PDFs.
How analysis is divided
Three categories. AI only where it earns its keep.
Most “AI audit” products ask a language model to read a chart and opine. AEGIS splits the work so rules stay rules, clinical judgment is assisted, and statistical outliers become the investigation list.
AI becomes extremely useful when the question is consistency across the record, not a yes/no field. High-risk findings still require a qualified human before they become a validated finding.
- Does the clinical narrative support the MDS score?
- Does the resident’s condition appear consistent across notes?
- Does the care plan address the documented problem?
- Does the incident investigation actually identify a root cause?
- Does the documentation support the stated medical necessity?
- Does infection-control documentation demonstrate implementation — not merely policy existence?
Instead of asking “Is this wrong?”, the system asks “Is this unusual enough to investigate?” CMS publishes extensive nursing-home datasets — MDS-derived information, quality measures, staffing, ownership, and other facility-level data — which become an external benchmark layer.
So DRL could tell a facility: your RN staffing isn’t necessarily noncompliant. But your night-shift RN staffing is materially below your peer group, and 71% of your falls occur between 7 PM and 7 AM. That’s a much more valuable insight.
Or: antipsychotic utilization is not automatically evidence of noncompliance. However, 14 residents receiving antipsychotics have incomplete documentation supporting indication, monitoring, or gradual-dose-reduction considerations.
The upload process
Simple intake. Enormous evidence.
Potentially hundreds of thousands of pages. One facility record. Six steps from create to Audit Today.
- Step 1 Create facility
- Step 2 Upload evidence
- Step 3 Ingest everything
- Step 4 Evidence graph
- Step 5 Run audit engines
- Step 6 Audit Today
Step 1
Create the facility
Facility identity is the jurisdictional key. State loads the federal and state requirement pack before a single PDF is read. CCN ties the record to CMS public data. Beds, census, and audit period set the denominator for every rate and peer comparison that follows.
- FacilityLegal name and operating identity.
- CCNCMS Certification Number — the public-data join key.
- StateLoads federal + state requirement layers automatically.
- Beds / censusCapacity and occupancy for rate math and peer grouping.
- Audit periodThe window every date, MDS, claim, and incident is tested against.
Step 2
Upload evidence — drag and drop
The system does not require a single EMR. It requires the record as it actually exists.
- 📁 MDS
- 📁 EHR
- 📁 Nursing
- 📁 Therapy
- 📁 MAR
- 📁 TAR
- 📁 Physician
- 📁 Hospital
- 📁 Incidents
- 📁 Falls
- 📁 Infection
- 📁 Pharmacy
- 📁 Staffing
- 📁 Payroll
- 📁 PBJ
- 📁 Policies
- 📁 QAPI
- 📁 Survey
- 📁 Claims
- 📁 Remittance
Step 3
DRL ingests everything
OCR, extraction, entity resolution, and clinical concept identification — before any engine scores a finding.
- OCRs PDFs
- Extracts tables
- Identifies residents
- Identifies employees
- Identifies dates
- Identifies documents
- Identifies clinical concepts
- Identifies MDS items
- Identifies medications
- Identifies diagnoses
- Identifies incidents
- Identifies claims
- Identifies staffing
- Identifies policies
- Establishes relationships
Step 4 — critical
Build the facility evidence graph
The audit engines do not operate on folders. They operate on a graph: residents, employees, dates, documents, MDS items, medications, incidents, claims, and the edges between them.
That is the difference between searching a document dump and reconciling a facility.
- Resident 1847 has: MDS dated X · fall dated Y · medication Z · therapy · wound · hospital transfer · care plan · claim
- Then the engines run Every finding is a path through that graph — clickable back to the source note, date, and document.
Step 5
Run 15+ audit engines simultaneously
DRL CORE
Clinical · Financial · Regulatory — one evidence graph
Clinical
- MDS
- Falls
- Infection
- Medication
- Wounds
- Incidents
Financial
- PDPM
- Claims
- Denials
- Cost
- Opportunity
Regulatory
- Survey
- OIG
- MAC
- RAC
- TPE
- State
Step 6
Produce the DRL “Audit Today” result
Brutally useful. If an external audit started tomorrow, this is the first screen.
Illustrative facility report · not a live engagement
Audit Today
If an external audit started tomorrow — probability of significant finding: HIGH
- 17Estimated priority findings
- 3Critical
- 8High
- 6Moderate
- 23Residents requiring immediate review
- 41Claims requiring review
- 37MDS assessments requiring validation
- 9Staffing exceptions
- 14Incident investigations requiring review
- 11Policies / practices requiring remediation
The most important screen
Show me why
Every finding is clickable. A facility administrator should understand the conclusion immediately — from score, to resident, to the notes that contradict the MDS.
Click the finding. Illustrative walkthrough of a documentation / MDS inconsistency.
37 assessments identified
Section GG: transfer = 3
Open supporting evidence to see the notes that do not match the coded score.
Supporting evidence
- Nursing note 06/12: “Requires extensive assistance.”
- Therapy note 06/13: “Max assist ×2.”
- CNA note 06/13: “Mechanical lift.”
- Care plan: “Hoyer lift.”
- Fall report 06/14: “Attempted transfer.”
DRL conclusion
High-confidence documentation / MDS inconsistency. That is something a facility administrator can understand immediately.
Where we are careful
AI does not declare you out of compliance.
AEGIS does not autonomously conclude: “You are out of compliance.”
It concludes: “DRL identified evidence consistent with a potential compliance deficiency requiring human validation.”
Once a qualified reviewer confirms it, the status becomes Validated Finding. That is an auditable AI-assisted workflow — not a black-box model making regulatory determinations.
- Confidence: 96%Model certainty on the evidence path — not a legal conclusion.
- Reviewer status: PendingHeld until a qualified human signs the finding.
- Validated FindingThe only status that should travel into a memo, IC deck, or survey response.
Often more valuable than bad evidence
The missing-evidence engine
Suppose the facility uploads MDS, nursing, care plans, and MAR — but not therapy, hospital records, or physician notes. The system should not say “no problem.”
Evidence gap
DRL cannot fully validate 23 findings because required corroborating evidence is missing.
Request these documents
- 23 physician records
- 17 therapy records
- 8 hospital discharge summaries
- 11 incident investigations
That is extremely powerful for diligence. Incomplete packets stop being silent. They become a worklist.
Risk-based review
It can also tell you what not to audit.
Instead of manually reviewing 500 charts, AEGIS stratifies the census so human time starts where exposure is densest. That is how an enormous audit becomes manageable.
- 500Residents in the census
- 92Low-risk
- 247Moderate
- 119High
- 42Critical — recommended human review starts here
Jurisdictional layer
The 50-state layer fits directly into this.
When the facility is created with State = Utah, the system automatically loads federal requirements, the CMS survey framework, Utah nursing facility requirements, Utah Medicaid requirements, Utah reporting requirements, and Utah-specific licensing requirements.
If another facility is Texas, it loads Texas. The exact same uploaded evidence is interpreted through the appropriate jurisdictional layer.
- FederalRequirements + CMS survey framework, always on.
- State packLicensing, Medicaid, reporting, and state-specific survey overlays.
- Same evidence, different lensA finding in Utah is not automatically a finding in Texas — and the reverse.
Lenders and investors
Diligence Intelligence for a portfolio — not a star rating.
A buyer uploads the diligence package for a 12-facility SNF portfolio. AEGIS returns what they are actually buying — and why Facility A is not just a bad survey score.
Illustrative portfolio report
Portfolio Audit Intelligence
12 facilities · 1,842 residents · annual revenue as provided in the packet
- 61Portfolio risk / 100
- 12Facilities
- 1,842Residents
| Facility | Risk | Revenue risk | Survey risk | Clinical risk |
|---|---|---|---|---|
| Facility A | 82 | High | High | High |
| Facility B | 76 | High | Medium | High |
| Facility C | 43 | Low | Low | Medium |
| Facility D | 68 | Medium | High | Medium |
Facility A isn’t just a bad survey score. Here’s why it matters economically. That is Diligence Intelligence.
Two products, one engine
Facility self-audit. Transaction diligence.
Same evidence graph. Different buyer question. That is where DRL Holdings becomes extremely differentiated.
Mode 1
Facility self-audit
Output: “How exposed are we today?”
- SNF administrator
- DON
- MDS director
- Compliance officer
- Operator
- Regional clinical leader
Mode 2
Transaction diligence
Output: “What are we actually buying?”
- Private equity
- Lender
- Healthcare investor
- Auditor
- Acquisition group
- Consultant / management company
Built into the existing stack
Not a new island. A new engine.
The conceptual pieces already exist across the portfolio. AEGIS’s job is to turn raw facility records into an evidence-backed prediction of where external scrutiny is most likely to find weakness.
One architectural decision, made now: do not build this as “upload PDFs → ask ChatGPT questions.” That is too fragile.
Documents → structured evidence → normalized entities → rules → AI analysis → cross-document reconciliation → risk scoring → human validation → report.
That is the difference between a chatbot and a healthcare audit intelligence platform.
- RevOptix1 Revenue intelligence
- PDPM Audit Optimization Group Clinical / reimbursement audit
- DRL Holdings Enterprise diligence intelligence
- Updox / operational platform Data ingestion + workflow
- ARGUS External market intelligence
- ORION Commercial / opportunity intelligence
- AEGIS Audit & Evidence Governance Intelligence System
Technical precedent
CMS already described the operating pattern.
CMS’s own Medical Review Management System describes uploading MDS records and medical-chart documentation, storing it, assigning reviewers, documenting findings at the data-element level, calculating inter-rater reliability, tracking audit progress, and generating facility-level and individual audit reports.
What AEGIS adds is the evidence graph, the three-layer analysis, the missing-evidence engine, the 50-state overlay, and the diligence mode — so the output is not a simple automated compliance checklist.
It is the flagship intelligence product of DRL Holdings: an auditable prediction of where scrutiny lands first, and what it could cost.