Behavioral health · Medicaid · Growing agencies
Catch issues before submission, not in the findings letter.
HealthThread checks every encounter against your state's Medicaid rules before anyone signs it, so the problem surfaces while it's still a note someone can fix — not a year later, in a letter from the state or an MCO.
- Built for Medicaid, configurable for any state — rules encoded per program
- Every claim traceable backward to the assessment that started it
- Any signed record reproducible exactly as it stood at signature
- 1 · Assessment ANSA · current
- 2 · Treatment plan Active plan · 3 goals
- 3 · Rendering clearance QMHP-CS · cleared for this code
- 4 · Session note Targeted case management · draft Billed service is not tied to an active goal Golden-thread rule · blocking · resolve in the note
- 5 · Timesheet entry Held
- 6 · Claim line Not built
The problem
Audit letters always refer to services you barely remember.
It comes as a records request, a findings letter, an MCO audit notice, or a demand for money already spent. Whatever the envelope says, the encounters are a year old or more. The clinician who wrote them may have left. Finding the documentation that would settle it becomes someone's full-time job for three weeks.
Recoupment is the worst version. It is rarely the expensive part.
- Staff pulled off billable work to assemble records by hand, one claim at a time — before anyone has decided you owe anything.
- A corrective action plan you now have to write, follow, and prove you followed.
- Extrapolation — a finding against a sample of twenty claims applied across hundreds.
- Prepayment review, where nothing gets paid until someone reads it first — and cash flow stops while they do.
- The quiet erosion of your standing with an MCO you depend on for referrals.
- And the thing nobody puts on paper: every clinician now writing defensively, for an auditor rather than for the next person treating the client.
Almost none of this is a story about bad clinical work. It's a story about disconnected systems, data that never lined up, and documentation judged against a rule nobody had the time to read.
Where it starts
- 01One encounter, keyed three times, in three systems. Schedule, timesheet, and note are entered separately, so they drift apart before anyone submits anything.
- 02Assessments and plans live as PDFs. You can't run a compliance check on a stored document, so the golden thread is a policy nobody can enforce.
- 03Clearance is recorded as of today, not as of the date of service. The most expensive findings turn out to be about the clinician, not the note.
- 04Conflicting time entries surface only when a reviewer finds them. By then the encounter is a year old and nobody remembers the day.
- 05Rework is what makes money slow, not the payer. A claim that needs a second submission was already paid for once, in staff time.
- 06Notes get written at nine at night, unpaid. For a contracted workforce, documentation time is time nobody is billing for.
- 07AI writes the note and nobody can tell who authored it. An auditor asking who attested that this service was rendered gets a signature with nothing behind it.
Our focus
Agencies carrying public-payer risk with a private-practice toolkit.
Practice software assumes one clinician, one payer, one set of rules. Commercial platforms assume negotiated contracts and predictable fee schedules, where the payer publishes its requirements once and rarely looks back. Neither is built for an agency running multiple programs, a supervised and often contracted workforce, and a public rulebook that changes without warning — so the rules stay in a manual nobody has time to track, validation happens too late to matter, and the record gets assembled by hand every time somebody asks.
HealthThread does one thing instead: it is built for this agency, and every decision in it reflects that.
The two-minute version
How HealthThread works, in six lines.
- 01One record behind every encounter. Schedule, timesheet, note, and claim all derive from the same signed encounter, so nothing is keyed twice.
- 02Assessments and plans live in the record, not in attachments. Which is what lets the golden thread be enforced by a rule instead of asserted by a policy.
- 03Clearance resolved as of the date of service. Enrollment, MCO credentialing, contracting, and reassignment of benefits, as they stood that day.
- 04Conflicting entries caught before billing. Two encounters in the same hour never reach a claim.
- 05The first submission is the only submission. Claims are built from a record that already passed its checks, so there is nothing to resubmit.
- 06Notes drafted from the session itself. Captured in the room, waiting when the clinician sits down, with findings already marked.
- 07AI drafts, the clinician authors, the record proves which. Nothing is signed on AI content alone, and the compliance checks are rules rather than model output.
Compliance-first
Compliance validation belongs before the signature, not after the denial.
A well-written note can still be unbillable — wrong credential for that code in that program, place of service that doesn't match, an authorization that lapsed on Thursday. HealthThread evaluates all of it while the note is still open, and anything that would cost you money holds the signature until it clears.
Eligibility is a four-part question
Evaluated as a combination, not as a code.
- ProgramTargeted case managementeach program carries its own approved set
- CredentialQMHP-CSwith supervision requirements and status
- Service codeH0036with modifiers and unit rules
- Place of servicePOS 12 — homecommunity, school, office, telehealth
Findings surface in the note, while it can still be fixed
- PassCredential eligibilityLicensed to render this service, in this program, at this place of service.
- BlockGolden threadService traces to an active goal, plan to a current assessment, assessment in-window.
- BlockAuthorization statusA current authorization covers this date, this service, these units.
- FlagDaily service limitsUnits per clinician per day, across every program the client touches.
- FlagSupervision and co-signatureSupervised services route to the right supervisor first.
- BlockPart 2 handlingSUD records stay segmented; disclosure is governed by recorded consent.
Eligibility and authorization resolved at booking
Booked
The authorization on file is checked against the service being booked, so a gap surfaces while the appointment can still change.
Seen
In-person or telehealth from one record, with recording consent and audio-only attestation captured in the encounter.
Drafted
AI turns the capture into a structured draft in your program's note format, carrying the client's active goals.
Signed
The clinician edits, augments, and re-runs the check as often as they like; findings update as the note changes, and every edit is recorded as authorship.
Intake, assessment, and the plan
Intake, assessments, and plans are part of the record, not documents attached to it.
Nearly every behavioral health system treats intake, assessment, and treatment planning as paperwork — forms to complete, documents to store, a box that turns green once something is attached. The file looks finished. Nothing inside it can be checked, reported on, or connected to anything else.
HealthThread works the other way around. Assessments and treatment plans are part of your core clinical record rather than attachments to it. An assessment is a set of scored needs across defined domains, with the clinician's reasoning recorded wherever a rating calls for action. A treatment plan is a set of goals, each one traceable to a need the assessment actually identified. Intake establishes eligibility, consent, and authority to treat inside that same record — not in a packet that gets scanned and filed beside it.
It is the least visible decision in the product and the one everything else depends on — what lets every claim, every review, and every state report draw on the same information instead of a fresh interpretation of it.
Who can bill what, where
Provider clearance is resolved as of the date of service, not just today.
A perfectly documented service is worthless if the rendering provider wasn't enrolled that day, wasn't credentialed with that MCO, or hadn't completed reassignment of benefits. HealthThread tracks clearance as a state that changes over time — and knows what was true on the date of service, not just what's true today.
Clearance history — what we knew, and when
Credentialing is not a checkbox that gets ticked once. A clinician is enrolled with the state, credentialed separately with each MCO, contracted separately again, and linked to your billing entity through reassignment of benefits — each with its own effective date, each capable of lapsing quietly.
Most systems store the current answer. When an auditor asks whether a provider was cleared on a date fourteen months ago, the current answer is the wrong one.
HealthThread resolves clearance as of any date. Services rendered before an effective date don't reach a claim, expiring credentials surface before they lapse rather than after, and the record of what was known on the date of service survives intact.
Time, documentation, and the claim
One signed encounter produces the time entry, the note, and the claim.
A clinician logs the time in one place, documents the session in another, and a biller re-keys the visit a third time. Nobody is doing anything wrong — it's the same information entered three times, into systems never designed to agree.
Nothing is re-keyed, so nothing drifts
When the timesheet says ninety minutes and the note documents forty-five, one is wrong — and you have either paid for time you can't bill or billed time you can't support. Nobody committed fraud; two people entered the same visit on two different days. The money is gone either way, and you usually learn which way during a review.
Reconciliation stops being a monthly job
Matching a pay period against signed notes and claims by hand is a recurring cost that grows with every clinician you add. For contractor-heavy agencies it consumes days each cycle and produces nothing at all when the numbers agree.
Time entry
The hour on the timesheet is the hour in the note. Payroll draws from what the supervisor approved.
Clinical record
The signed note, tied to the plan goal it served and the assessment behind it.
Claim line
Units, modifiers, place of service, and rendering provider, from the record that already passed.
Three views of one artifact — not three copies that have to agree
Conflicting entries are caught before billing
Sometimes it has an ordinary explanation. Either way the bottom line doesn't move: the same clinician can't be in two places at once, and the same unit can't be billed twice under two programs. HealthThread compares every encounter against every other one on the same clinician, client, and day, and routes conflicts back to whoever wrote them and whoever reviews them. Most turn out to be ordinary — a session ran long, a timesheet entered from memory on Friday — and catching them early protects the clinician who did the work as much as the agency. Under a corrective action plan it runs as a standing control with its own audit trail.
Conflicts surface before the claim, not after the letter
AI and authorship
AI drafts the record. The clinician remains its author.
HealthThread drafts across all clinical content — progress notes, assessments, treatment plans, intake — and tracks, item by item, what the model produced and what the clinician did with it. Nothing reaches a signature on AI content alone.
In Medicaid, a signature is an attestation that the service was rendered as documented. That attestation has to belong to a person. So every element of clinical content carries its origin — drafted by the system, or written by the clinician — and, where a clinician has changed a draft, the record keeps both facts rather than replacing one with the other. The origin is never overwritten by the review.
Content that is still AI-drafted and unreviewed cannot be signed. It surfaces as a blocking finding in the same queue as a missing plan goal or an expired credential, and it resolves the same way: a clinician reads it, changes what needs changing, and takes authorship of what remains. Some items are never drafted at all — on suicide and self-harm, the model may surface what it heard in the session, but the judgment stays with the clinician and no value is assigned on their behalf.
This matters most where the thread runs. Because drafting reaches assessments and plans, not only notes, a record could in principle be synthetic end to end and still pass a structural check — an AI note, tied to an AI goal, on an AI plan. Tracking authorship at every link is what makes that impossible here.
The compliance rules engine is deterministic, written against state-specific regulations, and it validates AI-drafted and clinician-written content the same way. Nothing gets a lighter check because a person typed it, and nothing gets a stricter one because the model drafted it. What makes a service billable follows from the rules and the record, not from the drafting layer.
Authorship is tracked per item, not per document
When you're audited
You answer with the record as it stood, not as it looks now.
You'll still get audited. That part we can't change. What changes is what the day costs you when the request arrives.
-
Select the claims or the date range in question
No one goes looking for which clinician saw which client under which program. The system already knows.
-
Every supporting artifact comes with each claim
Claim, timesheet entry, session note, treatment plan, assessment — assembled as a single packet rather than gathered from six systems and a cabinet.
-
Each document appears as it did the day it was signed
Rendered under the rules, reference data, and terminology in force on that date. A record signed in 2026 reads as a 2026 record when it's produced in 2029.
-
A reviewer can follow any claim backward on their own
From the claim line to the assessment that justified the plan that authorized the service — minimizing the time your clinical staff spend supporting audit reviews.
It stops being a scramble and starts being a query.
And the money moves
The first submission is the only submission.
What stretches the calendar is the loop — a note returned, a timesheet held because one entry is wrong, a claim denied for something visible the moment the session ended. Close each loop where it opens and the calendar shortens on its own.
Encounter
Captured by voice, on the device already in hand.
Note signed
Checks run while the note is open; blocking findings resolve first.
Review
Supervisors can review by exception. Clean entries move; flagged ones come back.
Timesheet
One disputed entry rolls forward while the rest gets paid.
Claim
Built complete from the checked record, then handed to billing with nothing re-entered.
Supervisors review the exceptions, not the pile.
When every note needs a signature, review is a bottleneck that grows with your census. When only flagged ones need a decision, it stops growing — and a single disputed entry rolls forward instead of holding an entire payroll run.
Review by exception, not by volume
What documentation costs
Notes get written when the session ends, not in the middle of the night.
In an agency running contract clinicians, the visit is billable and the charting afterward is not. Forty minutes at the kitchen table appears on nobody's budget, which is exactly why it never gets managed.
Clinicians stop absorbing unpaid documentation time
A contractor absorbing an hour of charting a night is paid for the visit and not the record of it. Some leave over it — and replacing a clinician here isn't a hiring problem, it's a credentialing one. Enrollment, CVO, contracting, and reassignment of benefits all happen again before the new person bills a single unit.
Same-day notes are stronger notes
A session documented four days later is written from memory, and memory writes short. Thin notes are what a reviewer finds: no linkage to the plan, a service described too generally to support the code. Closing that gap is a compliance intervention, not only a quality-of-life one.
So the work moves back into the day it belongs to — captured where the session happens, drafted before the clinician sits down, with findings appearing while they still remember the room. That has to hold up wherever they actually work.
In the home
Speak the session into the phone in the driveway. No signal needed; it syncs when there is.
In a school
Capture between students with the roster loaded. Group and individual units counted across the day.
In the office
Ambient capture with client consent, the active plan and last assessment alongside the draft.
Telehealth
Video beside the chart, with audio-only and video recorded distinctly — the billing rules treat them differently.
Records and privacy
Records stay defensible, private, and portable.
- 42 CFR Part 2 segmentationSubstance use records separated at the data layer, disclosure governed by recorded consent.
- HIPAA architecture and BAAA Business Associate Agreement with every agency. Encryption in transit and at rest.
- Minimum-necessary accessAdmin staff see the status of a clinical record without seeing its contents.
- Your records leave with youComplete structured export, during the relationship and at the end of it.
Where we are
Working software, in market validation.
We're refining HealthThread alongside a small number of agencies in the Texas Medicaid market before we widen access. Taking part is a working relationship, not a waiting list, and it comes with three things.
Your operations evaluated first
We start with your programs, your credential mix, and the findings you've actually received — not a generic walkthrough. We can then provide a tangible demonstration of how HealthThread will handle your specific use cases before you commit to moving forward.
Direct influence on what we build
Your needs and priorities shape what gets built next. Our initial stakeholders provide real-world feedback, and we refine specific rules, workflows, and reports as needed to ensure your needs are met.
Hands-on transition support
Hands-on support moving your practice data to HealthThread when you're ready — not a migration guide and a login.
Your rate, locked for two years
Agencies who join during market validation keep the published rate for twenty-four months from signing, whatever happens to list price after that.
Questions
Questions agencies ask before they switch.
Which states do you support?
Texas is our initial market. Texas Medicaid support is built in, against the Texas provider manuals, and we're prioritizing additional states based on customer demand.
How is this different from the EHR we already have?
Most systems record what happened and then report on what's missing. HealthThread evaluates each encounter against your state's Medicaid rules while it's being written, and holds the signature on anything that would make the service unbillable. The difference shows up as fewer denials and less rework.
What happens when the system blocks a note it shouldn't?
HealthThread allows you to override compliance checks, but every override is tracked and visible for audit purposes. Override access is typically granted to a clinical supervisor or compliance lead, not the clinician who wrote the note. An override requires a reason, is recorded against the encounter, and travels with the audit packet alongside the finding it cleared. Overrides are also reportable, so a rule that gets overridden constantly can be reviewed and revised if necessary. Only findings that would make a service unbillable prevent signature.
We're already under a corrective action plan. Does that change anything?
It usually makes the case stronger rather than weaker. A corrective action plan commits you to a documented process and to demonstrating it works. Rules enforced by the system, with an audit trail showing they were enforced, is a more durable answer than retraining staff and hoping. Bring the plan to the demo and we'll look at which of its commitments the software can carry.
Do you handle 42 CFR Part 2 programs?
Yes, and as a structural matter rather than a policy note. Records originating in a Part 2 program are segmented at the data layer, and disclosure is governed by recorded consent rather than by staff discretion.
Do you support CANS and ANSA?
Yes — CANS, ANSA and TX-eCANS are first-class structured rating instruments within HealthThread, including support for reassessment window tracking and results linked to the treatment plan and the golden thread. Support for other standard screeners such as PHQ-9, GAD-7, and the SUD screens is planned.
Most of our clinicians are contractors. How does that work?
Contract clinicians are first-class users with credential tracking, supervision requirements, and supervisor sign-off enforced by the same rules as employees. Timesheets are reviewed entry by entry: clean entries are approved and paid while a disputed entry returns to the clinician and rolls into the next period, so one problem doesn't hold up an entire pay run.
We use an outside billing company. Does that still work?
Yes. HealthThread's job is to produce a checked, claim-ready record; your billing partner can keep doing what they do with it. What's different is that we provide encounters that already carry the code, units, modifiers, place of service, rendering provider, and supporting documentation, instead of requiring your billing team to reconstruct claim details from disparate artifacts.
The handoff sits at submission. HealthThread builds the checked claim record; your billing partner or clearinghouse submits it. If you're weighing whether to keep that relationship, we'll tell you plainly what we handle today and what we don't.
What does it cost?
$60 per clinician per month, with non-clinical staff — administrators, billing, scheduling, executives — included at no charge. No volume tiers, no per-claim or per-note fees. Paying annually up front takes a month off the year. Full detail is on the pricing page; implementation timing depends on how many programs you run and what has to be migrated, and we offer a free scoping assessment to size it accurately.
What happens to our records if we leave?
You get a complete structured export of your clinical and billing record, including signed documents in their rendered form. Your records are yours.
Request a demo
See the entire golden thread, end to end.
We'll walk one client's record from intake through assessment, treatment plan, service delivery, supervisor review, claim, and audit response — showing where each check fires, what it holds, and how every artifact stays tied to the one before it.
In forty-five minutes you'll see how much risk and how many administrative hours come off the table when clinicians, supervisors, and billing all work from one record — instead of three systems reconciled afterward.