Bottom line. Medicare's digital mental health treatment-device codes create a credible reimbursement pathway, but only for a narrow product-and-care configuration. CMS began paying in 2025 for specified FDA-authorized devices used incident to ongoing professional behavioral health services, and in 2026 extended the policy to qualifying ADHD therapy devices.[1][2] Physician investors should not value this as reimbursement for mental health apps generally. They should verify the exact FDA classification and indication, who buys and furnishes the device, how it enters a behavioral health treatment plan, whether practitioners complete and document management work, what claims actually collect, and whether clinical and operating evidence survives real-world use. The moat is the integrated care-and-payment system, not the existence of three billing codes.

Key takeaways

  • Eligibility begins with the authorized device and its labeled use. A wellness app, coaching tool, or generic chatbot does not become reimbursable because it addresses mental health.
  • The billing pathway is clinician-centered. The device must augment ongoing behavioral health treatment under a plan of care rather than operate as a detached direct-to-consumer subscription.
  • G0552 and G0553-G0554 carry different economic signals. The device-supply component remains contractor-priced, while the management services retain national pricing in 2026.[2]
  • FDA authorization is necessary but insufficient for an investment case. Clinical effectiveness, engagement, safety operations, claims yield, cash timing, retention, and contribution margin need separate proof.
  • The physician practice, not just the patient, is a core user. Ordering, onboarding, monitoring, escalation, documentation, billing, and follow-up must fit limited staff capacity.
  • Generative AI can expand utility and risk at the same time. Scope, failure modes, human oversight, change control, and postmarket monitoring should be explicit before investors credit an AI roadmap.
  • Contractor payment variation, patient cost sharing, device acquisition terms, denial rework, and clinician time can turn apparent reimbursement into weak unit economics.

What Medicare actually pays for in 2026

CMS established payment for digital mental health treatment devices in the 2025 Physician Fee Schedule. The policy covered devices cleared through 510(k) or granted de novo authorization and classified under the specified computerized behavioral therapy regulation, when furnished incident to professional behavioral health services and used with ongoing care under a behavioral health treatment plan.[1] This language defines a linked chain: authorized treatment device, eligible patient and indication, billing practitioner, device supply, clinical plan, professional management, and compliant claim. Break one link and the revenue thesis may fail even if the software works.

The covered object is a treatment device, not a category label

FDA's product classification for computerized behavioral therapy for depressive disorders describes a Class II software-based mobile application that provides computerized behavioral therapy to treat depressive disorders and uses product code SAP under 21 CFR 882.5801.[3] That is materially different from an app that tracks mood, sends educational content, facilitates communication, or promotes general wellness. Diligence should capture the device's decision summary or clearance, product code, regulation number, prescription status, intended user, age range, indication, contraindications, warnings, course length, and required clinician involvement. Marketing language should reconcile word for word with the authorized use.

The 2026 expansion is real but still bounded

For 2026, CMS expanded G0552-G0554 payment policy to devices classified under 21 CFR 882.5803 for ADHD therapy. CMS describes that classification as software intended to provide therapy for ADHD or its symptoms as an adjunct to clinician-supervised treatment. The agency also states that the practitioner must incur the device cost, furnish it incident to professional behavioral health services under the practitioner's plan of care, and use it according to the FDA classification.[2] This adds a new eligible pathway; it does not erase the indication, supervision, or incident-to constraints.

Pricing design shapes the commercial model

CMS states that G0552—the device supply plus initial education and onboarding per course of treatment—continues to be contractor-priced, while G0553 and G0554 continue to have national pricing.[2] Investors should therefore avoid a single national reimbursement assumption. Rebuild revenue from the actual Medicare Administrative Contractor jurisdiction, allowed amount, place of service, patient responsibility, claim acceptance, payment lag, and contractual split between the startup and practitioner. Then test whether the practice incurs the device cost in the manner CMS requires and whether collected revenue covers onboarding, support, clinical management, denials, and bad debt.

Why an authorized product can still be a poor investment

Medical-device status establishes a regulatory perimeter; it does not supply distribution, workflow adoption, payment certainty, or durable outcomes. FDA's January 2026 clinical decision support guidance distinguishes non-device CDS functions from software that remains a device, including patient- or caregiver-facing functions that meet the device definition.[4] FDA's policy navigator separately notes that low-risk coaching or prompting without specific treatment suggestions may fall within enforcement discretion.[5] Those categories can sit next to each other in one product. The company needs a function-level map showing which module is the authorized treatment device, which modules are non-device or enforcement-discretion functions, and how commercial claims and data flows preserve those boundaries.

Clinical workflow is the real distribution channel

A reimbursement-enabled product asks a behavioral health practice to identify eligible patients, prescribe or order correctly, educate and onboard them, incorporate the device into a treatment plan, review relevant progress, intervene when appropriate, document work, submit claims, handle cost-sharing questions, and manage the end of a course. Each step competes with other clinical tasks. Ask for time-and-motion evidence by role and site. A product that shifts ten minutes from the clinician to three disconnected staff queues may improve neither capacity nor margin.

Safety operations have to work outside the trial

Digital mental health products encounter symptom worsening, disengagement, self-harm signals, medication changes, comorbidity, access barriers, and use outside intended conditions. The FDA convened its Digital Health Advisory Committee in 2025 to discuss benefits, risks, premarket evidence, and postmarket monitoring for generative AI-enabled mental health devices.[6] Investors should inspect escalation thresholds, response ownership, coverage hours, user messaging, clinician notification, false-positive burden, adverse-event assessment, complaint handling, outage plans, and evidence that the workflow performs across patient subgroups and care settings.

What an investment-grade demonstration should prove

A credible demonstration starts with a real eligible patient profile and ends with clinical follow-up and a reconciled claim. Use one standard case, one patient who disengages, one safety escalation, one claim denial, and one use request that falls outside the authorized indication. Require source records and timestamps rather than a narrated prototype. The goal is to see whether regulatory scope, care delivery, payment, and evidence remain connected when the happy path breaks.

Workflow stage Evidence that earns credit Investor interpretation
Eligibility and order The system verifies diagnosis, age, payer, device indication, practitioner role, treatment plan, and course status before fulfillment. Prevents a broad app funnel from being mistaken for a reimbursable patient cohort.
Supply and onboarding The practitioner incurs the device cost; access activation, education, consent, support, and start date are traceable to the course. Tests the G0552 operating model and exposes fulfillment or support labor.
Treatment use Engagement, module completion, symptoms, adverse signals, and deviations are visible with indication-appropriate thresholds. Shows whether product use creates clinically actionable information rather than vanity engagement.
Clinical management Practitioner review, patient interaction, decisions, escalation, and documentation are captured without duplicate charting. Connects management services to actual professional work and practice capacity.
Claim and collection Submitted claims reconcile to eligibility, documentation, allowed amount, patient responsibility, denials, appeals, and cash received. Converts theoretical coverage into auditable revenue and contribution margin.

The physician investor's seven-part diligence framework

1. Reconcile every product function to regulatory scope

Build a matrix of software functions, intended users, clinical claims, FDA status, product codes, labeling, data inputs, model versions, and planned changes. Compare the marketed product, current production build, billing description, sales demo, and authorized device. Flag features that diagnose, recommend treatment, generate therapeutic content, change contraindication handling, or expand to new ages and conditions. Require a regulatory decision process for releases rather than a historical clearance stored in the data room.

2. Define the reimbursable cohort from the bottom up

Start with active Medicare patients who match the authorized indication and the company's contracted care setting. Apply diagnosis, age, benefit, practitioner, plan-of-care, prior-course, technology access, language, clinical appropriateness, and patient-choice criteria. Reconcile the remaining cohort with sales projections. A prevalence statistic is not an addressable market when much of the population falls outside labeling, lacks an aligned practitioner, declines digital treatment, or cannot complete onboarding.

3. Audit care delivery and safety ownership

Trace responsibilities among the manufacturer, prescribing or billing practitioner, employed clinicians, support vendor, and patient. Review protocols for symptom deterioration, suicidality, misuse, adverse events, technical failures, and loss to follow-up. Test after-hours and cross-state cases. The FDA's 2025 executive summary emphasizes total-product-life-cycle evidence, postmarket monitoring, performance criteria, labeling, and risk controls for evolving digital mental health devices.[8] Contract language and staffing should match the operational safety story.

4. Separate efficacy, effectiveness, and commercial evidence

Premarket evidence may support the authorized indication under defined study conditions. Investors still need effectiveness in the target Medicare workflow. Request patient flow from eligibility through activation, meaningful use, course completion, clinical follow-up, and outcome measurement. Segment results by site, clinician, age, baseline severity, comorbidity, access needs, and engagement. Distinguish missing data from lack of improvement. Avoid uncontrolled pre-post results presented as causal proof when attrition or concurrent treatment could explain the change.

5. Validate the real-world evidence system

FDA identifies electronic health records, claims, registries, device-generated data, and patient-generated data as potential real-world data sources, while emphasizing relevance, reliability, and accurate device identification.[7] For diligence, inspect the data dictionary, device and version identifiers, denominator logic, outcome definitions, missingness, linkage rates, provenance, quality controls, and analysis plan. A dashboard can look precise while combining non-equivalent product versions, unverified self-report, and selectively observed follow-up.

6. Rebuild claims yield and contribution margin

Obtain claim-level data from submission through adjudication and cash. Calculate clean-claim rate, denial categories, appeal yield, days to payment, allowed amount, patient responsibility, write-offs, and refunds by contractor and customer. Add the device acquisition cost, onboarding, practice training, clinical management, support, safety review, billing labor, integration, and revenue share. Model low activation, incomplete courses, denied G0552 supply claims, unbillable management time, and patient nonpayment. The relevant metric is contribution margin per eligible started course, not gross billed charges.

7. Align contracts, roadmap, and financing

Review who purchases the device, who bills, who bears denial and recoupment risk, how patient responsibility is handled, which clinical services are included, and what happens when labeling or payment policy changes. Inspect minimum commitments, refunds, data rights, security duties, service levels, implementation acceptance, regulatory cooperation, insurance, indemnities, and termination assistance. The financing plan should fund evidence generation, regulatory maintenance, integrations, and a slower claims ramp rather than assuming that policy immediately converts into cash.

Investment committee scorecard

Dimension Evidence that earns credit Reserve or term response
Regulatory and billing fit The production device, indication, practitioner workflow, plan of care, supply transaction, and claim documentation reconcile. Exclude unsupported populations and condition financing on closure of material scope gaps.
Clinical and safety value Indication-matched outcomes, completion, subgroup performance, escalation quality, and adverse-event operations are reproducible. Use evidence milestones and board reporting for safety, attrition, and postmarket signals.
Payment realization Actual claims show allowed amounts, denial reasons, cash timing, patient responsibility, and contractor variation. Haircut theoretical revenue and reserve for denials, recoupment, and billing-cycle working capital.
Operating scalability Onboarding, clinical work, support, safety review, billing, and integrations decline per completed course as volume grows. Milestone expansion capital to contribution-margin and practice-retention cohorts.
Change resilience Release governance, regulatory review, monitoring, customer communication, and rollback cover product and AI changes. Require reserved compliance capacity and consent rights for material product-scope shifts.

Red flags that should change price or terms

  • Management calls the product Medicare reimbursable but cannot identify the exact FDA classification, indication, authorized version, and eligible billing workflow.
  • The addressable market begins with mental health prevalence and never applies labeling, Medicare, practitioner, treatment-plan, activation, or completion filters.
  • Forecasts assign one national G0552 payment rate despite contractor pricing and no paid-claims evidence by jurisdiction.
  • The device is sold as a detached patient subscription while the model assumes incident-to billing and ongoing practitioner-supervised behavioral health care.
  • Clinical outcomes exclude non-starters and dropouts, mix product versions, or report symptom change without a defined cohort and follow-up denominator.
  • Safety escalation depends on customer clinicians, but contracts, integrations, hours of coverage, and response tests do not establish who acts or when.
  • Generative AI or new indications appear in the roadmap without a function-level regulatory assessment, validation plan, monitoring criteria, and rollback design.
  • Gross margin excludes onboarding, clinical management, claim rework, support, integration, postmarket surveillance, or patient-responsibility collection costs.

Frequently asked questions

Does Medicare reimburse every mental health app through G0552-G0554?

No. The policy is tied to specified FDA-authorized digital mental health treatment-device classifications and to use within ongoing clinician-supervised behavioral health care under a treatment plan. Wellness, coaching, communication, screening, and unregulated self-help functions are not automatically eligible. Investors should match the exact cleared or authorized device, indication, labeling, billing practitioner, and service workflow before assigning reimbursement value.

What changed for Medicare digital mental health devices in 2026?

CMS expanded payment policy for G0552, G0553, and G0554 to include qualifying FDA-authorized software classified under 21 CFR 882.5803 for ADHD therapy. CMS kept contractor pricing for G0552 and national pricing for G0553 and G0554. The expansion increases eligible scope, but the billing practitioner must still incur the device cost, furnish it incident to professional behavioral health services, and use it according to the FDA-authorized indication.

What evidence should a physician investor request beyond FDA authorization?

Request indication-matched clinical results, engagement and completion by cohort, safety escalation performance, clinician intervention data, treatment-plan integration, claims acceptance, net collections, patient cost-sharing effects, and renewal outcomes. FDA authorization answers a regulatory question; it does not prove that a provider can deploy the product reliably, obtain payment, improve care, or earn an attractive contribution margin.

Why is contractor pricing for G0552 important to valuation?

Contractor pricing can produce local payment uncertainty for the device-supply and onboarding component. A startup should show actual allowed amounts, denial patterns, appeal experience, payment timing, practitioner acquisition cost, and device transfer pricing by geography. A national addressable-market calculation that multiplies an assumed uniform G0552 rate by all eligible patients can materially overstate revenue and margin.

How should investors treat generative AI features in a digital mental health device?

Treat them as a separate risk and evidence layer. Review whether the feature is within the authorized device and labeling, how outputs are constrained, how self-harm and other urgent signals are handled, what changes require regulatory review, and how errors, hallucinations, drift, and subgroup performance are monitored. Do not assume a cleared non-generative product automatically supports an added conversational AI function.

Conclusion

Medicare's DMHT policy is best understood as a narrow operating pathway, not a broad digital-health subsidy. It rewards a specific combination of authorized treatment software, clinician-supervised behavioral health care, device supply, professional management, documentation, and successful claims execution. That combination can support a valuable company when it produces better access or outcomes and attractive practice economics. It can also expose weak distribution, safety ownership, evidence, and margins that an FDA logo and billing-code slide obscure. Physician investors should give credit only for patients who fit the indication, workflows that function under ordinary clinical pressure, outcomes with credible denominators, cash that has actually collected, and a release system capable of controlling future product and AI changes.

References

  1. Centers for Medicare & Medicaid Services. Calendar Year 2025 Medicare Physician Fee Schedule Final Rule Fact Sheet. November 1, 2024. Accessed August 7, 2026. https://www.cms.gov/newsroom/fact-sheets/calendar-year-cy-2025-medicare-physician-fee-schedule-final-rule
  2. Centers for Medicare & Medicaid Services. MM14315: Medicare Physician Fee Schedule Final Rule Summary, Calendar Year 2026. December 2025. Accessed August 7, 2026. https://www.cms.gov/files/document/mm14315-medicare-physician-fee-schedule-final-rule-summary-cy-2026.pdf
  3. U.S. Food and Drug Administration. Product Classification: Computerized Behavioral Therapy Device for Depressive Disorders, product code SAP, 21 CFR 882.5801. Updated July 27, 2026. Accessed August 7, 2026. https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfPCD/classification.cfm?ID=4174
  4. U.S. Food and Drug Administration. Clinical Decision Support Software: Guidance for Industry and FDA Staff. January 2026. Accessed August 7, 2026. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/clinical-decision-support-software
  5. U.S. Food and Drug Administration. Digital Health Policy Navigator, Step 7: Device Software Functions and Mobile Medical Applications. Current through August 7, 2026. Accessed August 7, 2026. https://www.fda.gov/medical-devices/digital-health-center-excellence/step-7-does-device-software-functions-dsf-and-mobile-medical-applications-mma-guidance-apply
  6. U.S. Food and Drug Administration. FDA Digital Health Advisory Committee: 2025 Meeting on Generative AI-Enabled Digital Mental Health Medical Devices. Updated 2026. Accessed August 7, 2026. https://www.fda.gov/medical-devices/digital-health-center-excellence/fda-digital-health-advisory-committee
  7. U.S. Food and Drug Administration. CDRH and Real-World Evidence. Current through August 7, 2026. Accessed August 7, 2026. https://www.fda.gov/medical-devices/science-and-research-medical-devices/cdrh-and-real-world-evidence
  8. U.S. Food and Drug Administration. Digital Health Advisory Committee Executive Summary: Generative AI-Enabled Digital Mental Health Medical Devices. November 6, 2025. Accessed August 7, 2026. https://www.fda.gov/media/189391/download

Editorial disclaimer: This article is for educational purposes only and does not constitute medical, legal, tax, accounting, regulatory, reimbursement, privacy, cybersecurity, or investment advice. Requirements, guidance, payment policies, contracts, and company circumstances are fact-specific and can change. Readers should consult qualified professionals and verify current primary sources before acting. Evidence reviewed through August 7, 2026.