For rehabilitation operations, patient access, authorization, revenue cycle, compliance, clinical documentation, and therapy leadership responsible for keeping PT, OT, and speech episodes administratively ready from intake through discharge.
Keep PT, OT, and Speech Therapy Benefits, Visits, and KX/ABN Decisions on Track
Turn scattered benefit responses, visit balances, authorization units, plan-of-care tasks, Medicare threshold signals, and payer replies into one source-linked work queue before a therapy episode drifts into an avoidable hold or billing dispute.
Before this workflow, staff often reconcile a portal, spreadsheet, schedule, therapy note, signature inbox, and claim history by hand. After adoption, the same evidence is versioned around explicit human decisions: verify, pursue, schedule, hold, notice-review, appeal, or close. The software organizes administrative evidence and exceptions; it does not prescribe care, judge progress, determine medical necessity, select modifiers, issue a notice, authorize treatment, or promise payment.
A therapy episode can look ready in one system and unresolved in another
Therapy access is not a single eligibility check or authorization number. Benefit design, visit definitions, authorization units, discipline, provider, setting, plan-of-care evidence, attendance, documentation, claims, and patient-liability paths can change at different times. A control tower must preserve those differences instead of hiding them behind a green status.
The benefit response does not answer the scheduling question
An eligibility response may mention a visit limit, referral requirement, coinsurance, or prior authorization, yet omit how PT, OT, and speech are grouped, whether the count is annual or episode-based, which settings share the limit, or which source is authoritative. A portal view may also differ from a plan document or payer representative response.
Operational consequence
Patient access either repeats research at every visit or passes an uncertain balance to scheduling. The result can be premature reassurance, unnecessary holds, inconsistent financial conversations, and rework when a later payer response changes the interpretation.
Authorized units, visits, and delivered services are treated as the same thing
An authorization can be expressed in visits, service units, procedure-specific units, discipline-specific quantities, or a date range. Scheduled, attended, documented, billed, reversed, and adjudicated activity are separate events. Cancellations and no-shows do not automatically consume payer-recognized authorization, while a completed note does not by itself prove billability or payment.
Operational consequence
A single decrementing spreadsheet can overstate or understate remaining capacity. Staff may schedule outside scope, stop care too early, miss a renewal window, or spend days reconstructing how an apparent balance was calculated.
Clinical documentation milestones live outside the authorization queue
Evaluation, plan of care, signature or certification, order or referral follow-up, progress report, recertification, functional outcome, treatment note, and discharge summary may sit in different modules or inboxes. Presence alone does not prove that a document belongs to the current discipline, episode, payer request, provider, or period.
Operational consequence
Authorization staff discover missing or mismatched evidence late, therapists receive urgent signature requests, billing cannot explain the source timeline, and leadership cannot distinguish a clinical decision from an administrative follow-up failure.
KX, GA/ABN, and therapy modifiers are reduced to automatic flags
Original Medicare keeps PT and speech-language pathology expenses together for the KX threshold and tracks OT separately. KX reflects an attestation supported by the medical record; GA communicates an applicable ABN path. GP, GO, GN, CQ, and CO also carry discipline or assistant context. None should be selected from a visit count alone.
Operational consequence
A generic modifier suggestion can create unsupported claims, an invalid liability conversation, or the wrong work queue. Clinicians, coders, compliance staff, and patient-access teams then have to reconstruct why the system made a high-risk recommendation.
Payer, setting, and liability changes arrive after the original plan
A payer switch, Medicare Advantage plan instruction, Medicaid managed-care assignment, commercial policy change, active home-health episode, workers’ compensation claim, automobile accident, provider change, discipline addition, missed visit, or discharge decision can alter the administrative path. The earlier approval remains evidence, not universal permission.
Operational consequence
Teams can pursue the wrong entity, reuse a notice that does not apply, schedule against stale scope, overlook coordination-of-benefits work, or leave a discharged episode open with unresolved units and claims.
A named agent team with visible decision boundaries
Each agent handles a defined part of the benefits, visits, authorization, and kx/abn control tower workflow. Confidence signals and human escalation remain visible rather than being hidden behind an automation label.
Benefit and Visit Baseline Agent
This AI agent assembles the payer, plan, line of business, therapy benefit, visit language, referral or order condition, network context, and effective dates into a source-linked baseline. It reports what each source actually says and never converts an incomplete eligibility response into a coverage promise.
Inputs
- Approved eligibility and benefit responses with source, retrieval time, and response scope
- Current plan documents, payer instructions, provider-contract references, and human call notes approved for use
- Requested PT, OT, or speech discipline, setting, provider context, and anticipated administrative episode
- Existing referral, order, authorization, coordination-of-benefits, and liability status references
Checks
- Match the payer, product, line of business, discipline, setting, network context, and effective period before comparing facts
- Keep visit limits, dollar or service thresholds, authorization triggers, referral requirements, and patient-cost fields distinct
- Identify whether PT, OT, and speech share a limit or have separate language rather than inferring a grouping
- Compare portal, electronic response, plan document, contract reference, and human verification for conflicts and staleness
- Treat an absent field, generic message, or failed transaction as unknown rather than no requirement
Outputs
- Versioned benefit and visit baseline with every statement linked to its source
- Conflict and missing-information queue assigned to patient access or payer follow-up
- Human-readable scheduling caveats without a coverage or payment guarantee
- Confidence method
- Confidence rises only when identity, discipline, setting, date, and benefit language agree across current authoritative or payer-issued sources. The score is reduced for generic messages, unverified call notes, missing scope, conflicting dates, or a source that cannot be reproduced.
- Low-confidence action
- The agent labels the affected field unknown, keeps the episode in a visible verification state, and asks for a specific source or human confirmation. It does not replace uncertainty with a default visit count or infer that authorization is unnecessary.
- Human escalation
- Patient-access, contracting, authorization, and payer-relations staff decide which benefit interpretation governs the operational conversation. They also control payer contact, financial explanation, network questions, and any schedule release based on the verified facts.
Authorization Scope and Unit Ledger Agent
This AI agent converts the payer’s actual authorization response into a versioned scope and reconciles it with planned, scheduled, attended, documented, billed, reversed, and adjudicated events. It makes the arithmetic explainable without deciding whether a service was clinically appropriate or payable.
Inputs
- Payer authorization request, acknowledgement, determination, amendment, and appeal response artifacts
- Authorized discipline, service category, quantity or units, date range, provider, location, and stated conditions
- Scheduling, attendance, cancellation, and no-show events from approved operational systems
- Therapy documentation status plus claim, correction, reversal, and remittance events when approved
Checks
- Preserve visits, timed units, untimed services, procedure-specific quantities, and discipline-specific allowances as different measures
- Match each event to the authorization version, discipline, provider or group, site, service category, and date window
- Keep scheduled, delivered, documented, billed, adjudicated, denied, reversed, disputed, and unknown states separate
- Detect duplicate imports, corrected claims, retroactive changes, partial approvals, and overlapping authorization versions
- Calculate a projected balance separately from a payer-confirmed balance and expose every source event behind the calculation
Outputs
- Explainable authorization ledger with authorized, allocated, observed, and unresolved quantities
- Projected exhaustion or expiration work queue with the triggering assumptions visible
- Discrepancy report for unmatched services, duplicate events, corrections, or scope conflicts
- Confidence method
- Confidence is based on exact scope matching, source acknowledgements, event uniqueness, complete version history, and reconciled quantities. It falls when units need interpretation, the payer changed scope, claims are pending, events conflict, or the authorization lacks a reproducible determination.
- Low-confidence action
- The ledger preserves the disputed amount as unknown and blocks automatic schedule release or renewal assumptions. The agent shows which events and conversion rules require review instead of forcing them into a remaining-units total.
- Human escalation
- Authorization specialists, therapy operations, qualified coders, billers, and payer representatives resolve unit definitions, corrections, retroactive decisions, exhausted scope, and discrepancies. A clinician decides the care plan; an authorized scheduler decides whether a slot can be offered.
Plan-of-Care Evidence Coordinator
This AI agent organizes the evaluation, plan of care, clinician-authored goals, order or referral, signature or certification, progress report, recertification, functional outcome, treatment-note, and discharge evidence for the verified episode. It tracks administrative presence and linkage, not clinical adequacy.
Inputs
- Authenticated therapy evaluation and current plan-of-care references for PT, OT, or speech
- Order, referral, signature, certification, recertification, and practitioner-response status
- Progress-report, functional-outcome, treatment-note, and discharge-summary references
- Payer documentation questions and approved human-maintained evidence checklists
Checks
- Match documents to the correct person, discipline, episode, author, service period, and authorization request
- Verify source, date, version, signature state, and supersession without judging the clinical content
- Distinguish an order or referral pursuit from plan-of-care certification and from payer authorization
- Identify a missing report, expired interval, conflicting version, late signature, or unresolved external-record request
- Block generated clinical rationale, goals, progress conclusions, functional scores, signatures, or attestations
Outputs
- Dated evidence timeline with present, missing, pending, superseded, and human-reviewed states
- Practitioner follow-up and document-retrieval tasks with owner, due context, and source request
- Source-linked administrative packet index for qualified authorization or audit review
- Confidence method
- Confidence reflects authenticated provenance, exact episode linkage, legible dates, clear version order, and confirmed signature state. It does not claim confidence in diagnosis, goals, progress, medical necessity, or clinical sufficiency because those judgments remain outside the agent’s authority.
- Low-confidence action
- The agent marks the evidence unresolved, asks a named owner for the missing source or linkage, and prevents the document from satisfying an administrative checkpoint. It never fills a gap with generated clinical language or a copied signature.
- Human escalation
- Treating therapists and authorized practitioners own clinical content, orders, plans, signatures, progress interpretation, and discharge decisions. Provider-relations, authorization, health-information, compliance, and legal staff resolve administrative pursuit, disclosure, timeliness, and policy questions.
Medicare Threshold and Modifier Review Agent
This AI agent prepares a calendar-year Original Medicare review of KX threshold status, PT and speech-language pathology combination, separate OT tracking, discipline modifiers, assistant-participation indicators, and GA/ABN evidence. It supplies traceable questions to qualified reviewers rather than placing a modifier on a claim.
Inputs
- Current official CMS therapy-services source and buyer-approved coding guidance
- Original Medicare claim-history and remittance references for the applicable calendar year
- Discipline, plan-of-care, assistant-participation, setting, and provider facts from authenticated systems
- ABN issuance, beneficiary-choice, delivery, and retention evidence when an authorized human has determined the path applies
Checks
- Keep PT and speech-language pathology threshold accumulation combined while tracking OT separately under the current CMS source
- Separate a threshold signal from the clinician-supported medical-necessity attestation represented by KX
- Separate KX from GA and confirm that an ABN question is not triggered solely because expenses reached a threshold
- Present GP, GO, GN, CQ, and CO facts for coding review without choosing a modifier from discipline name or staff title alone
- Flag home-health, institutional, Medicare Secondary Payer, corrected-claim, and other setting or payment contexts for qualified review
Outputs
- Source-dated threshold worksheet showing PT plus speech and OT paths separately
- KX, GA/ABN, discipline, assistant, and setting exception queue for coding and compliance
- Audit-ready explanation of source amounts, included claims, exclusions, corrections, and human decisions
- Confidence method
- Confidence depends on confirmed Original Medicare identity, current calendar-year CMS guidance, complete claim-event provenance, correct discipline grouping, and consistent coding facts. It falls for payer ambiguity, incomplete claims, corrected services, home-health overlap, uncertain assistant participation, or missing clinical support.
- Low-confidence action
- The agent withholds any modifier recommendation, labels the threshold view provisional, and routes the exact missing fact or source conflict to the accountable reviewer. It does not auto-create an ABN, infer liability, or treat a threshold as a therapy limit.
- Human escalation
- Treating clinicians determine and document medical necessity; qualified coders select claim modifiers; compliance and revenue-integrity staff control ABN and liability procedures; patient-access staff deliver approved communications; and the beneficiary or representative makes the applicable choice.
Payer Response and Appeal Orchestrator
This AI agent classifies payer acknowledgements, requests for information, partial approvals, denials, expirations, and appeal communications. It keeps the exact response attached to the exact request and prepares administrative work without inventing clinical rationale or treating transmission as approval.
Inputs
- Portal, API, electronic transaction, fax, mail, and call-response artifacts approved for the workflow
- Submitted packet index, transmission receipt, payer reference, and human approval record
- Determination scope, reason text, additional-information request, and appeal or reconsideration instructions
- Current payer, plan, program, contract, jurisdiction, and effective-date context
Checks
- Distinguish a technical acknowledgement from an authorization or coverage determination
- Compare approved and requested discipline, units, visits, services, providers, sites, and date ranges
- Detect a partial approval, changed scope, missing-information request, denial, duplicate response, or superseding decision
- Preserve payer-stated reason text and deadlines without paraphrasing them into a clinical conclusion
- Require human approval before submission, disclosure, peer-to-peer coordination, reconsideration, or appeal
Outputs
- Versioned response classification linked to the request and transmission evidence
- Additional-information, clarification, renewal, peer-review, or appeal work item with accountable owner
- Human-reviewed administrative draft and packet index without generated clinical assertions
- Confidence method
- Confidence is based on an authenticated payer source, exact request-response match, complete pages or payload, legible scope, and reproducible timestamps. It is reduced for portal truncation, ambiguous reason text, contradictory messages, missing attachments, or a response that cannot be linked to the active request.
- Low-confidence action
- The case remains in a response-review state. The agent asks for the complete payer artifact or qualified interpretation and does not change the authorization ledger, release scheduling, calculate liability, or start an appeal deadline from an uncertain message.
- Human escalation
- Authorization leaders interpret the administrative response; clinicians own peer-to-peer and clinical rationale; compliance, legal, and payer-contract experts address rights and deadlines; and authorized personnel approve every submission, disclosure, reconsideration, or appeal.
Scheduling, No-Show, and Discharge Coordinator
This AI agent turns the verified benefit, authorization, documentation, and clinical-status signals into scheduling or follow-up work. It helps recover missed appointments and close completed episodes while preserving the distinction between an operational slot, an authorized service, and a clinician-directed plan.
Inputs
- Human-approved schedule-release status and authorization scope
- Appointment, attendance, cancellation, no-show, reschedule, and wait-list events
- Current discipline, provider, site, visit cadence reference, clinical hold, and discharge status
- Approved outreach preferences, communication templates, and minimum-necessary contact fields
Checks
- Confirm that a proposed appointment fits the approved discipline, provider or group, site, date window, and visible quantity assumptions
- Keep cancellations and no-shows separate from attended, documented, billed, and payer-recognized utilization
- Detect a plan-of-care, authorization, payer, home-health, liability, or clinical hold before offering a slot
- Prevent outreach when consent, channel, language, contact, privacy, or legal status is unresolved
- Reconcile discharge, remaining authorization, open claims, document tasks, equipment or referral follow-up, and appeal status before closure
Outputs
- Schedule-ready, verify-first, and hold queues with the controlling reason visible
- Approved no-show recovery and rescheduling tasks without clinical or financial promises
- Discharge-closure checklist with open administrative exceptions assigned
- Confidence method
- Confidence reflects current human release, exact authorization scope, reconciled event states, active consent and channel rules, and absence of unresolved holds. It falls when the schedule source is stale, a visit balance is projected, outreach permission is uncertain, or closure dependencies conflict.
- Low-confidence action
- The agent does not contact the person or offer a slot. It creates a narrowly scoped verification task and shows the scheduler which source, approval, or communication rule is missing before the workflow can continue.
- Human escalation
- Schedulers approve appointments and outreach, therapists control clinical frequency and discharge, patient-access staff control financial communication, and privacy or legal owners decide disputed contact and liability paths. The agent never ends care or represents that a future visit is covered.
The operating sequence, evidence by evidence
The sequence separates agent work from accountable human checkpoints so teams can see what moves forward, what stays pending, and why.
Patient access and authorization
Establish the episode and payer path
Start with the requested discipline and an authenticated payer, plan, line of business, setting, provider, and effective period. Link but do not merge PT, OT, and speech work. Record whether the case is Original Medicare, Medicare Advantage, Medicaid, commercial, workers’ compensation, automobile or other liability, or still unresolved.
Agent actions
- Normalize approved source references without copying sensitive values into marketing or analytics
- Classify the administrative payer path and flag secondary-payer or home-health questions
- Open separate discipline and episode records when source scopes differ
Evidence produced
- Source-linked episode identity and payer-path record
- Unresolved payer, setting, provider, discipline, or effective-date queue
- Named human owner for the first release decision
Human checkpoint: Patient access or authorization staff confirm the payer and episode identity. Liability, home-health, coordination-of-benefits, or ambiguous line-of-business cases remain pending until qualified staff establish the correct path.
Benefits and payer follow-up
Verify benefits, visit language, and authorization triggers
Compare the current eligibility response with plan, contract, portal, and verified payer sources. Preserve how the source defines visits, services, disciplines, periods, and authorization conditions. Build a patient-access explanation only from reviewed facts and label every missing field as unknown.
Agent actions
- Assemble benefit, network, referral, cost-share, and visit facts by source
- Identify shared versus separate discipline language and conflicting limits
- Route specific questions for payer or contract verification
Evidence produced
- Reviewed benefit-and-visit baseline
- Source conflict log and follow-up record
- Approved scheduling caveats and financial-communication inputs
Human checkpoint: Patient-access, contracting, or payer-relations staff approve the interpretation used operationally. They do not describe an eligibility or authorization response as a guarantee of coverage, final cost, or payment.
Therapy documentation and authorization
Build the documentation and authorization record
Organize the evaluation, plan of care, order or referral, signature or certification, progress report, recertification, functional outcome, and other payer-requested evidence. Map explicit source facts to the request while keeping all clinical judgments and attestations with the qualified clinician.
Agent actions
- Version documents and verify episode, discipline, author, date, and signature state
- Create targeted practitioner or records follow-up tasks for missing evidence
- Assemble a source-linked administrative packet index and preflight exceptions
Evidence produced
- Dated plan-of-care and documentation timeline
- Missing, conflicting, late, or superseded evidence list
- Human-approved authorization packet and transmission record
Human checkpoint: The treating therapist or other authorized practitioner approves clinical content and signatures. Authorization, privacy, compliance, and records staff approve the packet, minimum-necessary disclosure, and submission channel.
Authorization and therapy operations
Reconcile authorization scope with scheduling and utilization
Translate the payer’s response exactly as issued, then reconcile it with planned, scheduled, attended, documented, billed, corrected, and adjudicated activity. Projected usage is useful for planning but remains visibly different from payer-confirmed utilization and claim payment.
Agent actions
- Create or update the versioned authorization scope and unit ledger
- Forecast exhaustion and expiration using disclosed assumptions
- Route mismatches, no-shows, corrections, added disciplines, and provider or site changes
Evidence produced
- Explainable authorized-versus-observed ledger
- Schedule-ready, verify-first, and hold status with reason
- Renewal, amendment, or discrepancy work queue
Human checkpoint: Authorization staff approve the scope interpretation, schedulers approve appointment actions, clinicians control the care plan, and coding or billing staff resolve claim events. A projected balance cannot independently release or stop care.
Coding, compliance, and clinical leadership
Run Medicare threshold, modifier, and notice review when applicable
For confirmed Original Medicare cases, compare the current CMS calendar-year source with traceable claim history, keeping PT plus speech-language pathology and OT in their correct threshold paths. Present KX, GA/ABN, GP, GO, GN, CQ, CO, setting, and home-health questions to the people who hold the relevant authority.
Agent actions
- Refresh the official annual threshold source and calculate an explainable provisional position
- Separate medical-necessity support, modifier selection, notice applicability, beneficiary choice, and claim submission
- Hold conflicting or incomplete source, documentation, assistant, and setting facts for review
Evidence produced
- Calendar-year threshold worksheet and source version
- Coding, clinical-support, ABN, and liability exception queue
- Human approval and override chronology for each resulting action
Human checkpoint: Clinicians own medical-necessity support; coders own modifier selection; compliance and revenue integrity own ABN procedure; patient-access staff communicate the approved notice; and the beneficiary or representative makes the applicable choice. No agent performs those acts.
Revenue cycle and therapy operations
Close the response, appeal, no-show, and discharge loop
Classify each payer response against the active request, maintain additional-information and appeal work, and recover operational follow-up after missed visits. At discharge, reconcile open authorization, documentation, claim, liability, and communication tasks so closure does not erase unresolved evidence.
Agent actions
- Distinguish acknowledgements, approvals, partial approvals, denials, and information requests
- Prepare human-reviewed follow-up, reconsideration, appeal, or rescheduling tasks
- Reconcile the final episode chronology and retain approved records under buyer policy
Evidence produced
- Payer response and appeal timeline
- No-show recovery and patient-contact record under approved rules
- Discharge closure record with remaining exceptions and owners
Human checkpoint: Authorized staff approve every payer submission, appeal, outreach, financial statement, and closure. Treating clinicians make discharge decisions, and unresolved claims or liability work stays open under the organization’s retention and escalation policy.
Separate payer paths instead of one universal rule
Coverage, notice, authorization, and documentation requirements vary by program, plan, jurisdiction, service, and date. These paths show where teams must verify current authoritative instructions.
medicare
Original Medicare: threshold awareness is not a therapy cap or an automatic ABN
CMS states that the former outpatient therapy caps were repealed while KX thresholds remain. In the reviewed 2026 source, the threshold is $2,480 for PT and speech-language pathology combined and $2,480 separately for OT. KX confirms that services above the threshold are medically necessary as supported by the record; it is not a visit authorization or proof of payment. CMS separately explains GA when an applicable ABN was issued for therapy expected to be noncovered as not reasonable and necessary.
- Refresh the annual CMS threshold source and preserve PT plus speech-language pathology together while keeping OT separate.
- Do not generate KX from accumulated expense alone; qualified clinicians and coders verify the record, service, claim, and current guidance.
- Do not issue an ABN solely because the threshold was reached. Use the current Original Medicare instructions and an authorized notice process for a defined expected-noncoverage reason.
- Track evaluation and plan of care, certification or recertification when applicable, progress reports, treatment notes, and discharge documentation as separate evidence states.
- Review GP, GO, GN, CQ, CO, setting, home-health, corrected-claim, and other facts with qualified coding and compliance staff rather than inferring them from a discipline label.
Human handoff: Therapists and authorized practitioners own clinical content and medical-necessity support. Medicare-trained coders, billers, compliance personnel, and revenue-integrity staff verify thresholds, modifiers, coverage sources, ABN applicability, claim treatment, and appeals; patient-access staff deliver only approved notices and explanations.
Sources for this path: Centers for Medicare & Medicaid Services, Centers for Medicare & Medicaid Services, Centers for Medicare & Medicaid Services, Centers for Medicare & Medicaid Services, Medicare.gov
medicare advantage
Medicare Advantage: use the member’s plan determination and utilization-management path
A Medicare Advantage case remains a health-plan case. CMS defines an organization determination to include plan decisions about authorization or payment, enrollee cost, and quantity limits. The team must use the current plan’s benefit, network, prior-authorization, continuation, notice, and appeal instructions; an Original Medicare fee-for-service ABN workflow is not a universal substitute.
- Preserve the exact plan, product, provider, site, discipline, service, authorized quantity, duration, and determination version.
- Separate Original Medicare coverage context from the plan’s operational submission, determination, and appeal process.
- Treat an authorization as scoped permission, not a guarantee that every later claim condition will be met or paid.
- CMS-0057-F applies to defined impacted payers and phased requirements; it does not prove that a particular plan, therapy request, EHR, portal, or API is connected or ready.
Human handoff: Plan-trained authorization and revenue-cycle staff interpret the organization determination and appeal route. Clinicians own medical-necessity statements and peer review, while compliance, contracting, legal, and patient-access leaders control notices, financial communication, and disputed plan requirements.
Sources for this path: Centers for Medicare & Medicaid Services, Centers for Medicare & Medicaid Services, Centers for Medicare & Medicaid Services
medicaid
Medicaid: verify the state, delivery system, age, benefit, and current plan rule
Medicaid.gov explains that states establish and administer their programs and determine the type, amount, duration, and scope of services within federal guidelines; physical and occupational therapy are examples of optional benefits. Therapy coverage, visit limits, prior authorization, practitioner requirements, managed-care rules, appeals, and coordination of benefits therefore require the exact current state and plan source.
- Identify state Medicaid fee-for-service versus the assigned managed-care entity before applying an authorization workflow.
- Verify the specific PT, OT, or speech benefit, age and eligibility category, setting, provider qualification, referral or order rule, quantity, period, and continuation process.
- Keep state-plan, waiver, managed-care contract, provider manual, portal, and determination sources versioned rather than copying a rule from another state.
- Do not reuse Original Medicare KX accumulation, discipline-modifier, or ABN logic unless the applicable Medicaid authority expressly requires the same element.
- Apply CMS-0057-F only within its actual impacted-payer, transaction, and compliance-date scope and retain a tested manual process.
Human handoff: State- and plan-trained authorization, clinical, billing, compliance, and legal staff verify the current benefit and appeal authority. Qualified clinicians control therapy content, and patient-access staff use only state- or plan-approved communications for coverage and potential liability.
Sources for this path: Medicaid.gov, Centers for Medicare & Medicaid Services
commercial
Commercial plans: contract and product details control the benefit and notice path
There is no universal commercial therapy visit limit, authorization rule, modifier convention, or ABN. HealthCare.gov defines preauthorization and warns that it is not a promise the plan will cover the cost. The workflow must preserve the member’s exact product, network, contract, policy, authorization response, cost context, state requirements, and effective date.
- Verify whether PT, OT, and speech share visits, have separate limits, use habilitative or rehabilitative categories, or require different review vendors.
- Reconcile plan, provider-contract, portal, electronic-response, and payer-representative evidence and route conflicts to contracting or payer relations.
- Use the payer’s own authorization, denial, appeal, estimate, consent, and financial-liability process rather than applying an Original Medicare ABN by analogy.
- Do not claim CMS-0057-F covers every commercial plan; CMS states that its commercial scope is limited to Qualified Health Plans on the Federally Facilitated Exchanges.
- Keep benefit verification, authorization, medical necessity, clinical ordering, coding, patient estimate, and claim adjudication as separate decisions.
Human handoff: Contracting, authorization, patient-access, billing, compliance, and legal owners decide which plan and contract sources govern. Clinicians own therapy decisions and clinical rationale; approved staff control financial estimates, notices, appeals, and any decision to schedule while a payer question remains open.
Sources for this path: HealthCare.gov, Centers for Medicare & Medicaid Services, Centers for Medicare & Medicaid Services
workers comp auto liability
Workers’ compensation, automobile, and liability: establish who may pay first
A work-related injury, automobile accident, no-fault policy, liability claim, settlement, or other recovery context can change the responsible entity and the authorization channel. CMS explains that workers’ compensation, no-fault, or liability insurance may pay before Medicare for related services. That federal context does not replace state law, carrier rules, legal review, or the facts of the individual claim.
- Identify the carrier or administrator, accepted body part or condition, approved provider and site, therapy discipline, authorized scope, date range, claim status, and contact channel.
- Keep carrier authorization, utilization review, state forms, attorney or adjuster communication, lien or recovery issues, and health-plan benefits in distinct records.
- Do not infer accepted liability from a referral, claim number, scheduled visit, prior payment, or patient statement.
- If Medicare is involved, route Medicare Secondary Payer and possible conditional-payment questions to trained staff before billing or liability communication.
- Do not use an Original Medicare ABN to replace a carrier, state, court, settlement, or legal process.
Human handoff: Workers’ compensation and liability specialists, carrier representatives, qualified billers, compliance personnel, and legal counsel establish the operative payer and state-law path. Therapists control care, authorized schedulers control access, and high-risk settlement, recovery, or patient-liability questions remain with qualified humans.
Sources for this path: Centers for Medicare & Medicaid Services, Centers for Medicare & Medicaid Services
Make every release, hold, modifier question, and notice review explainable
A reliable therapy workflow is a governed operating model, not a collection of autonomous shortcuts. Each source, calculation, agent action, confidence signal, human decision, override, and downstream acknowledgement needs an owner and a reproducible history. High-risk ambiguity stops the workflow instead of being smoothed into an answer.
Source and effective-date control
Assign every CMS page, payer rule, plan document, contract reference, portal instruction, call note, authorization, and internal checklist a source owner, scope, retrieved date, effective period, version, review date, and approver. Expired or conflicting authority creates a maintenance hold.
Explicit state and arithmetic control
Keep requested, submitted, acknowledged, approved, partially approved, denied, scheduled, attended, documented, billed, adjudicated, reversed, appealed, discharged, and unknown states separate. Every visit, unit, and threshold calculation exposes its included events, exclusions, conversions, and assumptions.
Role-based human approval
Name who may approve a benefit interpretation, packet, clinical statement, signature, schedule release, modifier, ABN procedure, patient communication, appeal, claim action, and closure. The workflow cannot expand its own authority because confidence is high or a deadline is near.
Minimum-necessary data and access
Configure access, purpose, field, display, export, disclosure, and retention around the buyer’s legal and privacy analysis. HHS describes minimum necessary as context-dependent and subject to exceptions; the software cannot replace organization-specific policies, safeguards, or qualified review.
Validation, monitoring, and rollback
Test source matching, false releases, false holds, ledger discrepancies, missed renewals, stale guidance, document linkage, response classification, override quality, interface failures, and downtime recovery by risk. Keep a manual fallback and a tested rollback for every released slice.
Non-PHI measurement boundary
GA4 measurement is limited to approved route, page-family, specialty, workflow, content-cluster, engagement, and CTA context. Search Console review uses aggregate page and non-sensitive query performance. Neither channel receives patient, payer, authorization, visit, document, claim, schedule, portal, or free-text values.
- Human authority
- Qualified humans retain every clinical, coverage, coding, authorization, scheduling, financial, privacy, legal, compliance, and liability decision. Agents may organize evidence, compare explicit facts, calculate disclosed planning values, and prepare work; they cannot diagnose, prescribe, establish medical necessity, sign, attest, select a modifier, issue an ABN, determine coverage, appeal without approval, release care, submit a claim, or promise an outcome.
- Audit trail
- Retain the approved source and version; the exact input references used; agent output and confidence; calculations and assumptions; missing or conflicting evidence; human reviewer, decision, reason, and timestamp; disclosure and transmission record; downstream acknowledgement; override; correction; and final reconciliation under the buyer’s retention and access policy. Never put credentials or secrets in the log.
- Data boundary
- Use synthetic, non-identifying fixtures for configuration and testing before any approved minimum-necessary production use. Do not place PHI, patient or order identifiers, payer-portal credentials, API keys, attachments, free text, or operational case values in page source, marketing forms, analytics, Search Console exports, screenshots, sales demonstrations, SEO evidence, or support logs. Unexpected sensitive search queries follow the buyer’s privacy and incident process.
Connect the evidence chain without claiming a universal integration
The control tower can use approved read, task, acknowledgement, and write-back patterns, but every connection must be validated in the buyer’s actual environment. Vendor permission, identity matching, field meaning, latency, downtime, reconciliation, retention, and cost are part of the implementation decision—not assumptions hidden behind an integration logo.
Therapy EHR or clinical documentation system
Information in scope
Approved references to evaluations, plans of care, signatures, orders or referrals, progress reports, recertifications, treatment-note status, functional outcomes, and discharge summaries can support an administrative evidence timeline.
Boundary
The workflow does not author or alter clinical findings, goals, scores, plans, signatures, progress conclusions, frequency, or discharge decisions. Exact fields, permissions, acknowledgements, and write-backs require vendor and buyer validation.
Practice-management and scheduling system
Information in scope
Episode, discipline, provider, location, appointment, attendance, cancellation, no-show, reschedule, wait-list, and human-approved hold or release states can feed operational queues.
Boundary
An appointment is not proof of authorization, medical necessity, documentation, billing, or payment. The agent cannot schedule, cancel, discharge, or contact a person without the buyer’s approved authority, communication rules, and acknowledgement controls.
Eligibility, clearinghouse, and benefit sources
Information in scope
Electronic eligibility responses, payer identifiers, plan context, benefit fields, referral or authorization indicators, and transaction status can be attached to the source-linked baseline.
Boundary
A returned transaction can be incomplete, delayed, generic, or inconsistent with plan and contract sources. No clearinghouse reach, real-time accuracy, payer coverage, or benefit guarantee is assumed; failed and absent fields remain unknown.
Payer portals and authorization channels
Information in scope
Approved request requirements, packet references, submissions, receipts, status checks, determinations, additional-information requests, partial approvals, denials, and appeal instructions can enter the chronology.
Boundary
Portal automation, credential use, scraping, API access, submission authority, and response completeness must be contractually and technically permitted. A receipt is not an approval, and no page claim implies every payer or portal is connected.
Billing, claims, and remittance systems
Information in scope
Approved claim lines, corrections, reversals, remittance events, payer responses, and calendar-year therapy expense references can support ledger reconciliation and Medicare threshold review.
Boundary
The workflow does not choose codes or modifiers, release claims, post payments, calculate final patient liability, or infer payer-recognized utilization without configured sources and human approval. Billing and coding systems remain systems of record.
Document, fax, task, and communication services
Information in scope
Document provenance, inbound and outbound transmission status, practitioner pursuit, staff tasks, approved templates, delivery acknowledgements, and retention metadata can close administrative handoffs.
Boundary
Content, recipient, purpose, consent, minimum-necessary scope, accessibility, language, delivery, signature, and retention must be governed by the buyer. Marketing forms and analytics are never an operational document or communication channel.
Model recoverable administrative capacity with buyer-owned inputs
Use a transparent cases × minutes saved × loaded labor rate formula. Replace every illustrative input with a measured baseline from the selected queue, include review and exception time, and compare the resulting labor capacity with subscription, usage, interface, third-party, validation, maintenance, and change-management costs. Capacity is not cash savings unless the buyer documents how it will be used.
Monthly benefit and authorization cases
600 cases
Illustrative planning assumption only. Count unique cases in the bounded pilot and define whether a case includes initial verification, renewal, amendment, or appeal before using a real baseline.
Administrative minutes saved per case
12 minutes
Illustrative difference between measured before-and-after active labor, net of human review, payer follow-up, exceptions, correction, and reconciliation. Do not substitute elapsed turnaround time for labor time.
Loaded administrative labor rate
38 USD per hour
Illustrative blended rate only. The buyer should use its approved wage, benefit, payroll, management, and overhead method and document which roles are included.
Formula
600 cases × 12 minutes saved per case × ($38 loaded labor rate ÷ 60 minutes) = $4,560 of illustrative monthly labor capacity.
Illustrative result
$4,560 in modeled monthly labor capacity before software subscription, usage, interfaces, third-party services, internal validation, governance, maintenance, and change-management costs. This is not revenue, cash savings, or a forecast.
Illustrative planning model—not a customer result, guarantee, or quote.
Illustrative workflow examples
These examples explain process behavior. They are not customer stories, measured outcomes, clinical advice, or promises of coverage.
Illustrative example
A commercial visit limit and authorization-unit balance do not agree
A hypothetical, non-identifying PT episode has a benefit response that describes a plan-year visit limit and a separate payer determination that authorizes procedure-specific units within a date window. The schedule shows attended, cancelled, and future appointments, while several claims remain unadjudicated. This example does not state what the plan covers or whether any service should occur.
- The Benefit and Visit Baseline Agent preserves the visit language, scope, date, and source instead of turning it into a universal remaining-visit number.
- The Authorization Scope and Unit Ledger Agent records the authorized units separately and links scheduled, attended, documented, billed, and pending events without assuming that a cancellation used authorization.
- A conflict between the projected unit balance and the portal’s visit display creates a payer-follow-up task; neither number silently overwrites the other.
- Authorization staff obtain and interpret the current payer response, then approve the operational scope or keep the case on hold. The scheduler acts only on that human decision.
Illustrative outcome: The illustrative outcome is an explainable verified, schedule-ready, or hold decision with separate visit and unit evidence. It is not a coverage promise, payment prediction, clinical recommendation, or claimed customer result.
Illustrative example
An Original Medicare threshold signal reaches review without an automatic KX or ABN
A hypothetical, non-identifying episode includes PT and speech-language pathology services in the same calendar year, a separate OT episode, a pending plan-of-care signature task, and claim history that may include a correction. The team wants to know whether the next administrative step is KX review, ABN review, documentation pursuit, claim reconciliation, or no action. This example makes no medical-necessity or liability determination.
- The Medicare Threshold and Modifier Review Agent refreshes the current official CMS source and keeps PT plus speech-language pathology accumulation separate from OT.
- The agent exposes included claim events, the possible correction, and the provisional threshold position instead of counting scheduled visits as incurred expense.
- The Plan-of-Care Evidence Coordinator routes the signature question to the authorized practitioner without treating a pending signature as proof that services are noncovered.
- Qualified clinicians, coders, compliance staff, and revenue integrity reviewers decide medical-necessity support, modifier selection, ABN applicability, claim treatment, and any patient communication.
Illustrative outcome: The illustrative outcome is a source-dated human review with distinct KX, GA/ABN, documentation, and correction paths. Reaching a threshold never becomes an automatic therapy stop, modifier, liability transfer, or assertion that Medicare will pay.
Adopt one payer path and one accountable queue before expanding
A credible buying decision starts with a bounded workflow, not a promise to automate every therapy episode. Choose one discipline mix, payer path, location, and queue; map the current work; validate actual connections; run silently; and release only the actions that named humans can supervise. Evaluate source maintenance, exception workload, total cost, manual fallback, and rollback before expansion.
Define the first operating slice and baseline
- Select one payer or plan, location, therapy discipline mix, episode type, and benefit or authorization queue
- Map the before-state from intake through benefit verification, documentation, submission, response, scheduling, claim reconciliation, and discharge
- Define visits, units, active labor, elapsed time, exception, false-release, false-hold, renewal, and rework measures
- Name clinical, authorization, patient-access, scheduling, coding, billing, compliance, privacy, legal, security, and executive decision owners
Exit criteria: The sponsor approves scope, definitions, baseline method, authoritative sources, human decision rights, excluded actions, success and stop criteria, total-cost categories, and a manual process that works if the software is unavailable.
Configure sources, states, controls, and connections
- Inventory every payer, policy, benefit, authorization, plan-of-care, schedule, claim, notice, and liability source with version ownership
- Configure separate requested, submitted, approved, scheduled, attended, documented, billed, adjudicated, corrected, appealed, and unknown states
- Validate identity match, permissions, minimum-necessary fields, credentials, acknowledgements, latency, downtime, reconciliation, retention, and write-back behavior
- Create confidence thresholds, human approval gates, exception routes, monitoring, source-review cadence, rollback, and credential-rotation procedures
Exit criteria: Cross-functional owners approve the source register, field mapping, interface behavior, access model, decision matrix, audit events, security and privacy controls, downtime method, rollback, maintenance workload, and all separate third-party or internal costs.
Run synthetic tests and a silent comparison
- Start with synthetic, non-identifying fixtures covering routine, conflicting, corrected, partial-approval, no-show, home-health, liability, and appeal paths
- Run the agent team silently on approved minimum-necessary production cases without submitting, scheduling, contacting, selecting modifiers, issuing notices, changing records, or releasing claims
- Compare benefit interpretation, document linkage, ledger math, threshold grouping, response classification, holds, and explanations with qualified staff
- Measure false releases, false holds, missed renewals, ledger discrepancies, source staleness, override reasons, review time, and interface failures by risk
Exit criteria: Named reviewers accept source fidelity, accuracy by risk, explanation quality, privacy, security, interface reconciliation, manual fallback, and error handling. High-risk defects are corrected and re-tested rather than averaged into an overall score.
Release a human-in-the-loop pilot and govern expansion
- Allow named staff to approve only the configured low-risk tasks, packet states, ledger updates, and queue changes
- Keep clinical statements, signatures, schedule release, modifiers, ABNs, patient communication, appeals, and claim actions under explicit human authority
- Review accuracy, overrides, adoption, exception volume, source drift, downtime, labor capacity, subscription and usage charges, and third-party costs on an agreed cadence
- Expand by one payer, discipline, location, task, or connection at a time and rehearse rollback before each new slice
Exit criteria: The operating council accepts the pilot evidence and net value, users can explain and override every released action, source and control owners remain staffed, manual fallback remains viable, and no expansion depends on unverified payer, portal, vendor, or outcome claims.
Authoritative sources used for coverage context
Sources support the cited coverage and administrative context. Teams must confirm the current policy, contract, jurisdiction, and effective date for each real case.
Therapy Services
Centers for Medicare & Medicaid Services · government · reviewed
Current official CMS therapy-services page describing the repeal of outpatient therapy caps, the 2026 KX threshold amounts, combined PT and speech-language pathology tracking versus separate OT tracking, and current therapy-assistant modifier context. The annual source must be refreshed for future years.
Medicare Benefit Policy Manual, Chapter 15
Centers for Medicare & Medicaid Services · government · reviewed
Current official Medicare manual chapter covering outpatient rehabilitation therapy policy, plan-of-care and certification context, and expected documentation such as evaluations, progress reports, treatment notes, and discharge notes. Qualified reviewers must use the current section and applicable facts.
FFS ABN
Centers for Medicare & Medicaid Services · government · reviewed
Current official Form CMS-R-131 resources and instructions for defined Original Medicare fee-for-service expected-noncoverage situations. The source supports keeping an ABN distinct from Medicare Advantage, Medicaid, commercial, or liability notices and from a generic financial waiver.
Advance Beneficiary Notice of Noncoverage Tutorial
Centers for Medicare & Medicaid Services · government · reviewed
Official CMS tutorial distinguishing the GA modifier used when an applicable ABN was issued under payer policy for therapy expected to be noncovered as not reasonable and necessary from KX, which reflects medically necessary services justified in the record.
Organization Determinations
Centers for Medicare & Medicaid Services · government · reviewed
Official CMS explanation that a Medicare health-plan organization determination includes decisions about authorization or payment, enrollee cost, and limits on quantities of items or services. It supports a plan-specific Medicare Advantage determination and appeal path.
Benefits
Medicaid.gov · government · reviewed
Current federal overview explaining that states establish and administer Medicaid programs and determine the type, amount, duration, and scope of services within broad federal guidelines, with physical and occupational therapy identified as examples of optional benefits.
Medicare Secondary Payer
Centers for Medicare & Medicaid Services · government · reviewed
Current official CMS overview of situations in which another entity may pay before Medicare, including workers’ compensation, no-fault, and liability insurance for related services. It is not a case-specific coverage, settlement, billing, or legal determination.
Home Health Services Coverage
Medicare.gov · government · reviewed
Current official beneficiary guidance explaining that physical therapy, occupational therapy, and speech-language pathology can be included in the Medicare home-health benefit when applicable conditions are met. It supports checking setting and benefit context rather than assuming an overlap result.
Minimum Necessary Requirement
U.S. Department of Health and Human Services · government · reviewed
Official HHS guidance describing the HIPAA Privacy Rule minimum-necessary standard, organization-specific policies, and stated exceptions. It supports buyer-controlled role, purpose, access, and disclosure design rather than an invented universal software limit.
Physical therapy, occupational therapy, and speech therapy workflow FAQs
Open a question to review the answer. The disclosure controls use native browser behavior and remain available by keyboard without page JavaScript.
What does physical therapy prior authorization software mean on this page?
It means a governed administrative orchestration layer for PT, OT, and speech benefits, visit language, authorization scope, documentation tasks, payer responses, Medicare threshold questions, scheduling holds, appeals, and discharge closure. It is not an EHR, payer, utilization-review authority, clinical decision-maker, coding engine, or promise that every portal is connected.
Does this replace our therapy EHR, practice-management, scheduling, or billing system?
No replacement is assumed. The workflow can read approved references, coordinate tasks, reconcile events, and return agreed statuses while the existing systems remain authoritative for their fields. Every connector, permission, acknowledgement, write-back, downtime path, reconciliation rule, retention term, vendor dependency, and cost must be validated in the buyer’s environment.
How does the workflow distinguish a visit limit from authorized units?
It preserves each source’s exact measure and scope. Visit limits, timed units, untimed services, procedure-specific units, date ranges, and discipline groupings stay separate. Scheduled, attended, documented, billed, adjudicated, reversed, and unknown events also remain distinct. Qualified staff decide conversions and which payer response governs; the system exposes the arithmetic and assumptions.
Can the agent team determine whether therapy remains medically necessary?
No. Agents may locate clinician-authored support, verify source and version, map explicit facts to a payer question, and show what is missing. They cannot diagnose, set goals, assess function, interpret progress, prescribe frequency, decide maintenance or rehabilitation status, establish medical necessity, sign a plan, discharge a person, or write clinical rationale. Qualified clinicians retain those decisions.
What is the difference between KX and GA or an ABN for therapy?
CMS describes KX as confirmation that services at or above the applicable therapy threshold are medically necessary as justified by the record. CMS describes GA for an applicable ABN issued under payer policy when therapy is expected to be noncovered as not reasonable and necessary. Reaching a threshold does not itself justify KX, require an ABN, stop therapy, transfer liability, or guarantee payment; trained humans review the current source and facts.
Does the system combine PT, OT, and speech for Medicare KX tracking?
Not as one total. The current CMS source keeps physical therapy and speech-language pathology together for the annual KX threshold and tracks occupational therapy separately. The configured review uses the current calendar-year source and traceable claim events. It does not count scheduled visits as incurred expense or carry a past year’s amount forward without review.
Can the workflow add GP, GO, GN, CQ, CO, KX, or GA to a claim?
Not in this operating model. It can gather explicit discipline, assistant-participation, documentation, threshold, notice, setting, and claim facts and route a review. A qualified coder selects modifiers using current authoritative guidance and the complete record. Human approval, claim-system controls, acknowledgement, reconciliation, correction, and rollback remain required.
How are plan-of-care signatures, orders, progress reports, and recertification handled?
The workflow versions the evaluation, plan of care, order or referral, signature or certification state, progress report, recertification, functional outcome, and discharge evidence for the correct discipline and episode. It can pursue an authenticated missing item, but it cannot decide clinical sufficiency, create a signature, invent rationale, or assume one payer’s timing rule applies to another.
What happens when home health, workers’ compensation, or automobile liability may overlap?
The case moves to a distinct setting or liability review rather than receiving a generic eligibility or ABN answer. Staff verify the active benefit, carrier, accepted scope, responsible payer, state or plan rule, and any Medicare Secondary Payer context. Qualified operations, billing, compliance, carrier, and legal owners decide the path; the agent preserves evidence and blocks unsupported release.
Are Original Medicare, Medicare Advantage, Medicaid, and commercial therapy rules interchangeable?
No. Original Medicare uses current CMS fee-for-service therapy, modifier, documentation, and ABN sources; Medicare Advantage uses the member’s plan determination and appeal process; Medicaid varies by state and delivery system within federal requirements; commercial coverage varies by product, contract, network, policy, and law. Each configured rule retains payer, program, plan, service, setting, jurisdiction, version, and date.
Can the agents submit authorizations, contact patients, or appeal without review?
No autonomous authority is assumed. Agents can prepare a source-linked packet, draft an administrative task, classify a response, and calculate a provisional ledger. Named people approve disclosure, submission, outreach, schedule release, clinical statements, peer review, reconsideration, and appeal. A transmission receipt never becomes approval, and low-confidence work remains visibly pending.
How should a therapy organization validate accuracy before adoption?
Use synthetic fixtures first, then a bounded silent comparison and human-in-the-loop pilot. Measure wrong-source matches, benefit conflicts, unit discrepancies, missed renewals, stale plans, false releases, false holds, modifier-routing errors, response misclassification, interface failures, overrides, and review time by risk. Require cross-functional sign-off, manual fallback, explainable correction, and tested rollback before expansion.
What therapy workflow information belongs in GA4 or Search Console?
Only approved non-PHI route, page-family, specialty, workflow, content-cluster, engagement, and CTA context belongs in GA4. Search Console evaluation uses aggregate page and non-sensitive query clicks, impressions, CTR, and average position. Patient, payer, authorization, visit, clinical, document, claim, schedule, portal, identifier, and free-text values stay out; unexpected sensitive queries follow the privacy process.
What does implementation cost, and is the therapy authorization software free?
No, the software is not free. $0 implementation fee. $0 customization charges. Those terms apply to the agreed implementation and customization scope. Software subscription and usage charges are separate and still apply. Payer, clearinghouse, portal, interface-vendor, third-party data, licensing, migration, validation, source maintenance, internal change-management, travel, and out-of-scope work may carry separate costs stated in the order form.
Bring one difficult therapy authorization queue to a working session
Choose one payer path, location, PT, OT, or speech discipline mix, and non-PHI process map. We will identify the benefit and visit sources, authorization ledger, plan-of-care handoffs, Medicare or liability boundaries, human decisions, integration assumptions, validation measures, total-cost questions, manual fallback, and the transparent cases × minutes × loaded-rate model. Do not submit patient, payer, authorization, visit, clinical, document, claim, schedule, portal, credential, identifier, or free-text values through the marketing form.