Clinical
Assessments & Treatment Plans
Assessments & Treatment Plans is the clinical documentation module of Coralia, the practice management platform for Applied Behavior Analysis (ABA) agencies. The assessment is the treatment plan: a Board Certified Behavior Analyst (BCBA) builds it from live client data, signs it under enforced review gates, submits it to the payer, and records the approval as a claim-ready authorization.
The problem in real agencies
In ABA, the treatment plan is the funding document. A payer authorizes hours by Current Procedural Terminology (CPT) code — 97153 direct therapy, 97155 supervision, 97156 caregiver guidance — based on what a BCBA writes and signs, and the cycle repeats roughly every six months. Most agencies still assemble that document by hand: copying demographics from an intake packet, retyping Vineland-3 scores, hunting through the prior report for baselines. Every retyped number is a chance to contradict the chart a payer can audit.
The plan and the money live in separate systems. The payer often grants fewer hours than the plan requested — a step-down someone must catch, convert into authorization units, and hand to the scheduler before sessions run past the cap. When the signed plan, the authorization, and the calendar are disconnected tools, requested-versus-granted gaps surface months later as denied claims instead of on the day the determination arrives.
Then there is the version problem. A plan gets edited after the BCBA signed it, and nobody can say which text the payer actually approved. Parent consent lives in a filing cabinet, discovered missing during an audit. When a reviewer asks for the document behind a paid claim, the agency needs the exact signed version — not the current draft — and most tools keep only one copy that everyone has been editing.
How it works in Coralia
- 1
Readiness before writing
Creating an assessment starts a preparation workspace, not a blank page. A readiness dashboard runs up to 24 checks across five categories — client foundation, clinical data, observation data, clinical team, prior context — at three severities and computes a weighted 0–100 score. Required checks block generation: demographics complete, an active diagnosis on file, three completed 97151 observation sessions, hypothesized functions on maladaptive behavior targets. A BCBA can override and generate anyway; the override is stored as an audit record with who acknowledged it, when, which checks were skipped, and the score at generation.
- 2
One click generates a pre-filled document
Generation creates every section from a six-act template — Identification, History & Context, Evidence, Behaviors & Goals, The Plan, The Ask — and fills it from the client record. Ten of the nineteen section types auto-populate: demographics, diagnoses and medications, behavior targets with baselines and mastery criteria, an intervention matrix built from function mappings, antecedent-behavior-consequence (ABC) records, a document checklist. Populated sections carry a provenance chip such as 'From client record · 3 diagnoses', with a fix-at-source link when the source is empty — so corrections happen in the chart, not in a copy.
- 3
Write with cited evidence and AI drafts
An evidence rail sits beside the editor with the client's diagnoses, medications, instrument scores from the document's own Vineland-3 and BASC-3 tables, 97151 observation dates, authorization utilization, and per-section excerpts from the latest signed prior plan. Each item has a Cite button that inserts the fact at the cursor and records a traceability citation inside the section — citations never print in the payer document. AI drafting works per section; an optional per-agency pipeline, off by default, queues drafts, verifies each against a clinical rubric across 28 section keys, and finishes with a document-level coherence pass.
- 4
Signing is a legal gate, not a checkbox
Only an active BCBA or Board Certified Assistant Behavior Analyst (BCaBA) can sign — Registered Behavior Technicians (RBTs) and office staff are blocked at the server. Signing is refused while any section is unreviewed or any required section is blank (whitespace-only rich text counts as blank), and any content edit flips a reviewed section back to unreviewed, so nothing unread gets attested. The signature freezes an immutable, encrypted, content-hashed snapshot as a numbered version and locks the document. Changing a signed plan requires an amendment with a stated reason; re-signing produces the next version and marks the prior one superseded.
- 5
Submission and approval close the loop to billing
A plan cannot go to the payer without an active analyst signature and a signed parent or guardian treatment-plan consent on record; agencies can additionally require a current signed progress report before a reauthorization is submitted. When the determination arrives, recording approval is a dedicated workflow: it creates the authorization period with per-CPT-code units in one atomic step, links it to the assessment, infers renewal lineage from the prior authorization, and flags a step-down when any code was granted fewer units than requested. A payer can also pend a plan back for revision without a formal denial.
- 6
A state-machine pipeline from preparing to paid
Every assessment moves through twelve enforced statuses — preparing, scheduled, in progress, report writing, report review, submitted to payer, approved, billed, paid, plus canceled, on hold, and denied. Invalid transitions are rejected at the server, and each change writes a history row with who, when, and notes. The list view shows the pipeline as status tabs with per-status counts, search, and pagination, and the module tracks clients whose last assessment is more than five months old — ahead of the six-month Behavior Analyst Certification Board (BACB) reassessment cycle.
The specifics
Four assessment types — Initial Assessment & Treatment Plan, Reassessment & Treatment Plan, Supplemental, and Discharge — each with its own document-header title on the printed report.
Twelve lifecycle statuses from preparing to paid; invalid transitions are rejected server-side and every change is recorded with who, when, and notes.
Readiness runs up to 24 checks across 5 categories — including 3 completed 97151 observation sessions — with a weighted 0–100 score; required failures block generation.
19 section types; 10 auto-populate from the live client record and show a provenance chip linking back to the source.
16 assessment tool options, including VB-MAPP, ABLLS-R, Vineland-3, ADOS-2, PEAK, and QABF; the assessments-conducted checklist also accepts custom tools.
Payer unit math is exact: 4 fifteen-minute units per hour, authorization periods in whole weeks, 6 months = 26 weeks.
Signing freezes a numbered, content-hashed, encrypted version snapshot; signatures are append-only, and amendments require a stated reason before re-signing.
Submit-to-payer is blocked without an active analyst signature and a signed parent or guardian treatment-plan consent record.
Recording approval creates the authorization with per-CPT-code units, infers renewal lineage, and logs requested-versus-granted numbers when a code is stepped down.
97151 observation sessions link to the assessment — same client only — and unlinking is blocked once the observation has been billed.
Integrations
Authorization Tracking — recording payer approval mints the authorization period with per-CPT-code units, renewal lineage, and step-down detection · Consents & Signatures — submit-to-payer requires a signed parent or guardian treatment-plan consent; plan signatures are content-hashed and append-only · Progress Reports — the evidence rail reads authorization utilization, and an optional agency gate ties reauthorization submissions to a current signed report · Clinical Data Collection — behavior targets, baselines, mastery criteria, ABC records, and function mappings auto-populate plan sections
Access control
Access is gated by two permission codes — one to view assessments and one to manage them, covering every write. On top of the permission gate, each assessment resolves to its client and enforces client-level access, and only an active BCBA or BCaBA can conduct, author, sign, or amend a plan.
Frequently asked questions
Is the assessment the same as the treatment plan in Coralia?
Yes. In ABA practice the assessment report and the treatment plan are one payer-facing document, and Coralia models them that way: one record, four types (initial, reassessment, supplemental, discharge), each with its own printed title. The document is built from nineteen section types organized into six acts, signed by a BCBA or BCaBA, versioned on every signature, and linked directly to the authorization the payer grants.
Does AI write the treatment plan, and can a BCBA trust what it writes?
AI drafts; a BCBA signs. Sections can be drafted individually, and an optional per-agency pipeline (off by default) verifies each draft against a clinical rubric, revises it, and finishes with a document-level coherence pass. Any content change resets a section to unreviewed, and signing is refused while any section remains unreviewed — so nothing reaches the payer that an analyst has not read. An evidence rail with one-click Cite keeps every number traceable to its source record.
What prevents an incomplete or unsigned plan from reaching the payer?
Three server-enforced gates. Signing requires every section reviewed and every required section non-blank — whitespace-only text counts as blank, judged on the decrypted content. Submission requires an active BCBA or BCaBA signature plus a signed parent or guardian treatment-plan consent. Agencies can add a third gate for reauthorizations: no submission unless a current signed progress report covers the expiring authorization. Each gate is enforced on the server, not just in the interface.
What happens when the payer approves fewer hours than the plan requested?
Recording the determination is a dedicated approval workflow, not a status flip. It creates the authorization period with per-CPT-code unit rows in one atomic step, links it to the assessment, and infers renewal lineage from the prior authorization. If any code was granted fewer units than requested, the system detects the step-down and logs a partially-approved event with per-code requested-versus-granted numbers — so authorization tracking starts from the granted units, not the requested ones.
How is the clinical content of a treatment plan protected?
Section content, version snapshots, and comments are encrypted at rest as protected health information (PHI); Coralia is built for HIPAA compliance. Lifecycle writes are audit-logged, with signing and amending marked as sensitive events. Every request that touches an assessment passes an object-level check resolving the plan to its client, so a staff account cannot read another child's plan by changing a URL. A live demo with a fully synthetic agency is open at coralia.app/demo — no sign-up.