Friday, 4:40 pm
A sixty-page referral arrives by fax.
Someone has to read it against the acuity limits, the payer mix, the isolation beds and the diagnoses this building cannot take — before the discharge planner calls the next facility on the list.
Orenda reads the documents your facility already produces and returns a structured, cited answer — a fit score, a PDPM review, a survey-ready chart, a drafted note — with every finding quoted to the page it came from. Your team verifies in seconds. Your team decides.
“Stage 3 sacral pressure injury, 4.2 × 3.1 cm. Wound VAC, change M/W/F.”
Built around the standards and systems your building already runs on
The pattern repeats all month. In every case the work that eats the hours is not judgment — it is finding the one sentence on page thirty-four that settles the matter.
Friday, 4:40 pm
Someone has to read it against the acuity limits, the payer mix, the isolation beds and the diagnoses this building cannot take — before the discharge planner calls the next facility on the list.
Day 5 of the window
Items to verify, a comorbidity documented in a consult note nobody coded, Section GG scores to support. Reimbursement rides on what gets found before the window shuts.
Monday, 8:05 am
Falls, psychotropics, weight loss, pressure injuries — every chart has to answer for itself. The binder says what to audit. Nobody has the hours to pull it.
Orenda does the retrieval — and shows you the sentence. The recommendation never travels without its quote, its page and its section.
Admissions, MDS, nursing, compliance and the business office each get their own workspace — all reading the same record, so adding a tool never forks your data or your workflow.
Admissions
Drop a referral packet. Get a 0–100 fit score, an accept, review or decline recommendation, risk factors with severity, staffing impact and a qualifying-stay check — scored against your building’s criteria, not a generic rubric.
Case mix
A PDPM register across the census and MDS review while the assessment window is still open.
Survey
Readiness across the building, and a working surface for answering a 2567 with the evidence attached.
Clinical QA
The audit instruments your QA binder already specifies, run against the chart instead of a clipboard.
Business office
Receivables and documentation requests tracked to the resident and the claim, reconciled before billing.
Data
Referral packets drive the admission decision. Every tool after admission reads the resident’s PointClickCare chart, synced into one record — and a result goes stale when the chart changes, not when the calendar says so.
SBARs, progress notes, care-plan changes and IDT notes are drafted from the chart. A nurse edits them in place and enters them into PointClickCare under their own name.
Nothing auto-posts. No draft reaches a resident’s chart without a clinician putting it there.
Edits save as they type. No forms, no extra attestations — edit, then copy.
Gaps stay visible. Anything the chart could not answer is left as a placeholder, and copy is held until it is filled.
A reviewer confirms the quote rather than taking a summary on faith — and every answer comes back in the same shape, so a decision can be filed, compared across months and audited a year later.
Each assessment returns the same fields every time. Nothing is parsed out of free text.
Scores, deadlines, F-tags and disclaimers are recomputed after the model responds — never taken from it.
A referral packet is untrusted input, fenced off as quoted data — and nothing written in it can move a server-owned field.
Each result is stamped with the record it read, so a change in the chart marks it stale the moment it lands.
Acuity ceiling, accepted payers, isolation beds, excluded diagnoses. Stored as a versioned record, so you can see which rules produced last quarter's decisions.
The upload goes from the browser directly to encrypted storage. Protected health information never passes through an application server.
Scanned pages go through OCR first; identifiers are stripped; the output is forced into a fixed schema and checked on every run.
Every finding carries its quote, page and section. A reviewer verifies the source, and the access is written to the audit log.
Every architectural decision assumes real patient data, even where only synthetic data is being processed. Your IT reviewer will find the controls where they expect them.
Documents upload from the browser straight to encrypted storage through a presigned, size-limited policy. The API never holds the bytes.
Names, MRNs and dates of birth are stripped before anything reaches the model. Clinical tools see age, sex and admission date.
Who, when, what and from where — on every read of a resident's record, not just every write. No update, no delete.
Every row carries its facility and every query is scoped to the caller's. A role never widens the boundary.
Nursing, MDS, business office, therapy, compliance, admissions and IT each see their own workspace — enforced on the server, per route.
Model inference runs on AWS Bedrock under a signed business associate agreement, alongside storage and the database.
Connectors for PointClickCare ship in the platform and stay dormant until your facility supplies credentials.
Recommended first
US Core 3.1.1 and 6.1.0, from published documentation. Reaches progress, therapy and consult notes, vitals, weights, labs, orders, medications, immunisations, the care plan, coverage, encounters and the document list.
Deeper surface
PointClickCare documents a partner programme covering the medication administration record, CNA and ADL charting, and the MDS item sets. We have not yet verified it against a live tenant, so we describe it as their published surface rather than something we read today.
By design
There is no bulk link and no name matching. A wrong link writes one resident’s chart into another’s record and nothing downstream would catch it — so a person confirms each one.
Something else? Bring it to the walkthrough — we will answer it with the product open.
No. The platform drafts; a nurse reads, edits and enters the note into PointClickCare themselves. Nothing reaches a resident's chart without a clinician putting it there, and a draft with an unfilled placeholder cannot be copied until it is completed.
The model reads and extracts; the platform scores. The fit score, deadlines, F-tags and disclaimers are recomputed from structured fields after the model has answered, so a score can always be traced to its parts and an instruction hidden in a document cannot move it.
Not for admissions: the fit assessment reads the referral packet you upload. The tools that work after admission — PDPM, survey readiness, clinical QA, drafted notes — read the resident's chart, so they need the resident linked to PointClickCare.
Inference runs on AWS Bedrock under our BAA, and identifiers are removed before a prompt is built. Your residents' records are processed to answer your question, not collected to train anything.
Each building keeps its own criteria, staff and data. One account can work across the buildings it has been granted, with every access recorded against the building and the authority it was made under.