The note expires when the claim pays. The data compounds for the life of the client.
The new PraxisNotes is data-first. Every fact starts in the assessment, is captured as structured rows in each session, is validated against the plan, and compounds into trends a note archive could never give. The note is a rendering of that data — the data is the asset.
required at every 6-month reauthorization — Florida Medicaid.
quantitative currency rule for continued authorization — Cigna.
today's PraxisNotes stores a baseline number and nothing plots it.
The data journey
One spine runs from the first assessment to the billed claim. Each stop below is a structured record, not a document in a drawer — and everything after this section walks the map.
Vineland-3 · BASC-3 · CDE · referral
DOB · diagnosis · guardians · insurance
behaviors · programs · goals · mastery
CPT · units granted · date window
voice · tap · captured · asked
value · type · source · plan target
per-behavior · caseload · org
reauth · QA · billing
via the extension · no retyping
The five stops in coral are where PraxisNotes lives; the rest it feeds. Scroll the row on a phone.
It starts with the assessment
A client is not a name. It's a spine of structured fields — every one of which a payer, a reauthorization, or a claim will eventually ask for. Intake captures them as data, not as a PDF nobody can query.
DOB, preferred language, and guardians as linked people — never sharing the child's phone or email identity. Relationships are records, not free text.
ICD-10 to the highest specificity, plus co-occurring conditions — the code that must support medical necessity on every claim line.
Payer, plan, member ID — and authorizations as first-class rows: auth number, CPT codes, units granted, date window.
Vineland-3 and BASC-3 scores — the exact instruments Florida requires in an authorization packet — held as scored values, not attachments.
Each with an operational definition, a measurement type — frequency, percentage, or duration — and a baseline that is the first point on every chart.
Replacement programs with mastery criteria, and goals tied to the plan — so a session can only record against something the plan actually authorizes.
The current app captures behaviors with baselines in a five-step wizard — good clinical instincts. But there's no DOB, no insurance, no diagnosis, no authorization field anywhere, and the baseline it collects is never charted. The data model exists; the data architecture doesn't. See the honest audit →
Paste or upload the assessment report. The agent extracts the structured profile — diagnosis, scores, payer, behaviors, baselines — and a human confirms every field before it saves. Today's chat assistant already builds clients from a pasted treatment plan; we make that the way every client begins, with review over data entry.
Nothing is invented and nothing is silently trusted — the agent proposes, the human approves, and the audit trail remembers who confirmed what.
An authorization exceeded or a date outside the window is the top preventable denial — CARC 197/198. The ledger is live, so overbooking against the auth never reaches a claim.
Every session adds rows
The four ways in — captured, tapped, spoken, asked — all land in the same measurement rows, each validated against the plan. One data model, four ways in → The note renders from these rows; it never invents them.
We store a confirmed zero distinctly from not measured. Conflating the two is the exact bug that bites Office Puzzle users today — data reads as though a session was never held. Two states, never one default.
Tap, voice at 4:47pm, an answered escalation, an edit before signing — the provenance travels with the number. Every point on every chart is auditable down to the voice note that produced it.
See the trajectory
Rows over weeks become a picture. This is the client's Data tab — the thing a note archive can never render, and the exact graphs Florida wants in the reauthorization packet.
This exact graph is what Florida wants in the reauth packet — it renders itself from the rows, no chart-building.
A flat line here is the signal Aetna documents: no improvement means the BCBA must modify the protocol — and record that they did.
Reauthorization is per-behavior table + graph + a 60-day data-currency check, one click — the shape Florida and Cigna require. The same rows feed QA (data-vs-note consistency) and the billing packet. One source, three destinations, zero re-charting.
Validated, always
A trend is only trustworthy if every point is anchored. The golden thread runs from the plan to the graph, and survives the plan changing underneath it.
A measurement can only record against a behavior or program that lives in the client's plan. No orphan data on any chart.
Trends survive plan revisions: each measurement references the plan version it was recorded under, so a Feb goal change never rewrites Jan history.
A behavior the plan doesn't have becomes a plan-update suggestion, never a silent drop — the BCBA decides whether to add it.
Every chart drills down to the tap, the voice note, or the answered escalation that produced the point. Defensible to the source.
The same data feeds Office Puzzle
Billing runs through Office Puzzle today, and its Event Modal expects the session data too — data-collection values, the note, signatures, and claim fields, all on one event. Today that's a double-entry tax: capture here, retype there — or capture only in Office Puzzle and lose the analytics entirely.
no transcription drift
EVV is never auto-filled — visit verification must originate in the EVV flow, a
permanent design invariant. Office Puzzle's Imported Data tab takes only historical DATE;VALUE;TYPE
numbers — fine for backfilling old data, useless for the live loop. The extension is the live path.
Extensible by schema
Each CPT code's capture is a versioned schema, not code. 97153 carries ABC and program data today; 97155 protocol modification, 97156 caregiver training, and 97151 assessment writing come next. A new code is a new schema paired with a new rule pack — no redeploys.
And the spine isn't ABA-specific. Assessment → plan → structured sessions → trends is the shape of speech and OT too — the same brain, pointed at a different discipline.
See the data model →New schemas ship as data, versioned and reviewable — the same pattern the plan and payer rules already use.
The data outlives the note. It's the clinic's compounding asset — and the client's story, told in evidence.
A note is spent the moment the claim pays. The rows behind it become baselines, trends, reauth packets, and proof — for the whole life of the client.
Meet the team that works this data →