ABA Billing and CPT Codes: The Complete Sourced Reference, and the Six New Codes That Arrive January 1, 2027

On January 1, 2027, the code set your entire revenue cycle runs on stops existing in its current form. The placeholders are gone. The AMA has published the real code numbers, and the ABA Coding Coalition has published a line-by-line crosswalk from the old set to the new one.

The short version: six new Category I codes (97148, 97149, 97159, 97160, 97173, and 97180), revisions to every code from 97151 through 97158, and the permanent deletion of 0362T and 0373T, replaced by Category I equivalents. One word change runs through the whole set: “protocol modification” is now “analysis.” And one code, 97155, stops doing two jobs and starts doing one.

When we first published this page in July 2026, none of that could legally be described. The descriptors sat under the AMA’s CPT confidentiality agreement, and the organizations that wrote the code change application were barred from detailing their own work. We said we would update the page when that lifted. This is that update.

What follows is built on the two documents that matter, both from the coalition that filed the application: the September 11, 2026 code conversion crosswalk and the supplemental guidance article on applying the new set. Not vendor summaries of them. Where the field’s own guidance and a federal press release disagree, which happens once below, we say so.

What changes on January 1, 2027

The primary sources are the 2027 CPT Code Conversion Crosswalk (ABA Coding Coalition, September 11, 2026), which maps every 2019 code to its 2027 successor with the personnel and time units for each, and the supplemental guidance article, which carries the approved descriptors, patient vignettes, clinical examples, and roughly fifty FAQs. The Coalition that produced them is the group that filed the application: APBA, Autism Speaks, the BACB, CASP, and the ABAI ABA Billing Codes Commission. The original Panel action is still public in the September 2025 CPT Editorial Panel Summary of Panel Actions, Tab 56, where these six codes first appeared as 97X1X through 97X6X.

97180

The one to understand first. Non-face-to-face work by the QHP becomes separately billable for the first time: reviewing data and session notes, deciding whether to change targets or protocols and making those changes, developing a discharge or transition plan, and training technicians on a specific patient’s protocols. In the 2019 set this work was not separately reportable at all.

97173

QHP treatment with analysis, face-to-face with one patient. This is the half of old 97155 that was direct QHP treatment. It now has its own code.

97155 narrows

97155 survives but stops being a dual-use code. In 2027 it describes direction of the technician with analysis, and nothing else. The Coalition states plainly that it “is no longer a dual-use code.”

97148 / +97149

Harmful behavior supporting assessment delivered by two technicians in a customized environment, replacing 0362T, plus an add-on for each additional technician to a maximum of five.

97159 / +97160

The treatment counterpart, replacing 0373T, with the same two-technician base and the same add-on structure and five-technician ceiling.

10 → 14 codes

The set grows from eight Category I codes plus two Category III strays to fourteen Category I codes. Nothing is left in Category III, which is the category payers most often treat as investigational and decline to price.

Every code in the 2027 set is still a 15-minute timed code. That did not change, and it remains the most common misconception in ABA billing.

The word change that drives everything else

Across every code a QHP performs, the term “protocol modification” has been replaced with “analysis.” That sounds cosmetic. It is the most consequential change in the set, because of what the Coalition says next: the analysis codes are reportable whether or not the analysis results in an actual change.

In the Coalition’s words, adjustments to protocols, targets, or goals “are not required for reporting codes that include analysis services (97180, 97155, 97173, 97156, 97157, and 97158).” The cognitive work is the service. A QHP who observes a session, evaluates the patient’s response against the protocol, and concludes that nothing needs to change has performed analysis and may report it, provided the work is documented in the session note.

If you have ever had a 97155 claim denied because the note did not show a protocol change, that is the denial this revision was written to answer. Analysis also may not be delivered by a technician under any circumstance, which is the other half of the definition.

97155 splits in two, and this is the part that moves money

Under the 2019 set, 97155 did two jobs: QHP treatment with protocol modification, and simultaneous direction of a technician. Clinics billed both through the same code. In 2027 those separate.

97155 becomes direction of the technician with analysis, face-to-face with the patient. 97173 becomes QHP-delivered treatment with analysis, one-to-one with the patient, no technician in the picture. The Coalition’s instruction for planning is direct: look at what your QHPs currently bill under 97155 and work out which half of it belongs under the narrowed 97155 and which belongs under 97173.

That is a remapping project in your scheduling templates, your authorization requests, and your note templates, not a billing-software setting. A clinic that maps all of its old 97155 volume to the new 97155 will be describing QHP direct treatment as technician direction, and the authorization it requested will not match the service it delivered.

One useful anchor the Coalition supplies for utilization planning: CASP guidelines describe an average of one to two hours of QHP activity for every ten hours of technician-delivered treatment, counting 97155, 97173, and 97180 together. The Coalition is explicit that this is not a formula for any single code and does not set medical necessity. Authorization requests are still individualized.

97180: the work you have been doing for free

The new non-face-to-face code is the one most likely to change what a clinic collects, because it prices activity that the 2019 set simply did not recognize. Five activities qualify: reviewing and analyzing data and session notes on treatment targets; deciding whether to modify targets, goals, or protocols, and making those modifications; deciding whether additional assessment is needed and developing or modifying assessment protocols; developing a discharge or transition plan; and training technicians on a specific patient’s targets or revised protocols.

The reporting mechanics are unusual and worth getting right. Qualifying time is aggregated across a seven-day period (the guidance uses Sunday through Saturday), each date documented individually, then reported together on one claim line. The Coalition’s own worked example runs five hours of 97180 across four days and bills twenty units on the Friday. One practical consequence it spells out: patient cost-sharing then applies once, to that single billed date, rather than to each day the work happened.

What does not qualify matters just as much, because this is where recoupment will come from. Writing session notes is not billable under 97180. Neither is reviewing records for audit or compliance purposes, or any internal quality assurance. Neither is care coordination with other QHPs, which the Coalition says sits outside the adaptive behavior code set entirely. And administrative discharge work is excluded even though discharge planning is included: preparing payer notices, transferring records, writing a letter to another provider, or emailing a parent are not 97180.

Two clarifications that answer the questions clinics actually have. Training technicians is billable, and it is not limited to new hires or to new protocols: retraining an experienced technician on an existing protocol for a specific patient counts. And when a QHP trains technicians under 97180, only the QHP’s time is billable. The technicians’ time in that session is not separately reportable.

“Destructive” becomes “harmful,” and the door opens wider

The two-technician codes were the least-used part of the 2019 set, partly because they were Category III and partly because “destructive behavior” read narrowly. The 2027 set replaces that term with harmful behavior, and the Coalition defines it as any behavior that jeopardizes the patient’s health or safety or risks harm or medical consequences for the patient or others.

The examples it lists go well past aggression and self-injury: pica, elopement, feeding difficulties, sleep difficulties, rumination, bruxism, mouthing objects, inadequate fluid intake, skin picking, and resistance to dental or medical care. A clinic that read 0362T and 0373T as codes for severe aggression cases only should re-read the population it serves against that list.

The guardrails are real, though. The environment must be customized to the specific behavior, which can mean a padded room for severe head-banging or simply swapping a wooden bat for a foam one for a patient who uses objects aggressively. A QHP must be on site and immediately available to join the session, and “on site” means interruptible rather than in the room; where clinically appropriate that may be met by telehealth, subject to each payer’s policy and each state’s law. The add-on tops out at five technicians total. And the Coalition states directly that 97148 and 97149 “should not be reported solely because an additional technician happened to assist during a session” — the need has to be medically necessary and anticipated in a QHP-designed protocol.

One boundary case it answers cleanly: if harmful behavior erupts unexpectedly during a scheduled 97151 or 97152 session and a second technician steps in, that is not 97148, because the customized-environment element was never in place. But adding a third, fourth, or fifth technician mid-session to an actual 97148 is appropriate when the protocol anticipated it.

The 2027 code set, mapped from the one you use today

Summaries below are ours. The AMA owns the descriptors and reproducing them commercially generally requires a license, so what follows is the structural shape of each code in plain English: the number, who may perform it, the unit, and what replaced what. For verbatim descriptors, vignettes, and the full FAQ set, go to the Coalition’s supplemental guidance or the CPT 2027 Professional code book.

New in 2027.

Revised in 2027. All eight surviving codes were rewritten. Three of the revisions change what you can bill, not just how the descriptor reads.

One more thing the crosswalk settles. It lists the modifiers commonly used with these codes as HM, HN, HO, and HP, and defines them by education level: high school, four-year degree, master’s, and doctorate. If a vendor page has told you HN means home-based and HO means telehealth, the field’s own coding document disagrees. The QHP definition is equally worth pinning down, because payers vary: licensed behavior analyst, licensed assistant behavior analyst, BCBA-D, BCBA, BCaBA, or another credentialed professional whose scope, training, and competence include applied behavior analysis.

The 2026 code set, with official descriptors

Through December 31, 2026, this is still the set you bill. A note on what follows: the official CPT descriptors are copyrighted by the AMA, and reproducing them in full generally requires a license. So the summaries below are ours, not the AMA’s wording. The structural facts — which code, who performs it, the time unit, whether the QHP must be present — are facts, and they are verified against the ABA Coding Coalition’s official supplemental guidance. For the verbatim descriptors, go to that document or the CPT code book.

The thing to internalize: every code in this set is a 15-minute timed code. There are no per-hour codes, no per-day codes, and since 2019 no “first 30 minutes plus add-on” codes. That is the most common misconception in ABA billing.

Assessment codes.

Treatment codes.

Three rules that cause more denials than any coding error. Supervision without the patient present is not billable — that is an indirect service, and CPT has no stand-alone code for it. Ongoing treatment planning is bundled into the direct service codes. And indirect services have no CPT code at all unless your payer allows a HCPCS workaround. If your model assumes you can bill for team meetings or planning time, the code set disagrees.

On concurrent billing, the rule people get wrong: 97153 and 97155 can be billed at the same time, but not by the same person. The Coalition is explicit that a single QHP may not report both concurrently. It requires two people — technician delivering, QHP directing, both face-to-face with the patient. Also settled: 97154 and 97158 cannot be billed together, and 0373T cannot be billed with 97155 because directing the technician is already bundled into 0373T’s on-site requirement.

Two pieces of history worth knowing, because they explain why 2027 matters. This set only became permanent Category I on January 1, 2019. Before that, ABA billed entirely through temporary Category III codes (the 0359T–0374T family) introduced July 1, 2014 — codes payers could and did refuse to price. The 2019 transition deleted fourteen of those codes, converted untimed codes to 15-minute units, abolished the 30-minute-plus-add-on structure, and left 0362T and 0373T behind as Category III strays. The 2027 change finally cleans that up: the strays are deleted and everything lives in Category I.

And to answer the question that brings most people here: there were no CPT descriptor revisions in 2025 or 2026. The 2019 descriptors have been in force this entire time. Everything is happening on January 1, 2027.

Denials and prior authorization, by the numbers

Start with the thing nobody in this industry will tell you, because it is bad for business: there is no citable ABA-specific claim denial rate. It does not exist. Every “ABA denial rates run 10–30%” claim you have read traces to a vendor blog citing another vendor blog. The federal transparency data that produces real denial rates cannot be broken out by service — KFF, whose analysis is the gold standard, says so explicitly: because the data do not link denial reasons to the services denied, neither the share denied for a given reason nor the service most often denied can be calculated.

So anyone quoting you an ABA denial rate sourced to KFF is fabricating. What follows is real data that is either all-payer or behavioral-health-wide, labeled as such. It is less satisfying than a made-up number and considerably more useful.

The first federal audit of behavioral health denials landed six weeks ago, and it is damning.

100 of 100

Sampled denied behavioral health prior authorization requests that failed to meet all requirements, in an HHS Office of Inspector General audit of a Pennsylvania Medicaid managed care organization. A 0% compliance rate. The OIG notes this is its first audit ever to review behavioral health service denials — issued June 8, 2026. Report A-03-24-00204. One MCO, one state — do not generalize it to your payer. Do notice that the first time anyone federal looked, every single denial was out of compliance.

1 in 8

Prior authorization requests denied by Medicaid managed care organizations, across 115 MCOs covering 29.8 million enrollees. Twelve of those MCOs denied over 25% — twice the overall rate. HHS OIG, 2019 data.

95% / 18%

Share of appealed Medicare Advantage prior-auth denials that were overturned, versus the share of denials that were actually appealed. HHS OIG, June 2026. This is skilled nursing facility admissions, not ABA, but it is the clearest evidence anywhere that denials are frequently wrong and rarely challenged.

Now the appeal math, which is the part most clinics get wrong. Across ACA marketplace plans, fewer than 1% of denied claims are ever appealed, and when they are, insurers uphold their own denial 66% of the time, meaning roughly a third get overturned. Only 40% of consumers even know they have the right to appeal. KFF, 2024 data.

19%

In-network claim denial rate across HealthCare.gov issuers in 2024, from 451 million in-network claims across 157 insurers. Out-of-network runs 37%. Note the scope: marketplace plans only — not employer plans, not Medicaid, not Medicare. The famous “insurers deny 1 in 5 claims” line comes from here, and the qualifier gets dropped constantly.

3% – 36%

The spread in in-network denial rates between individual insurers in the same year. A twelve-fold difference. Which payer you contract with matters more than almost anything you do operationally.

36%

Share of in-network denials whose stated reason was “other” — no reason disclosed. The single largest category. Medical necessity, which is what everyone builds their appeals strategy around, accounts for 5%.

Prior authorization is the tax underneath all of it. The AMA’s 2025 physician survey (n=1,000, fielded December 2025) is the best measurement of the burden, and Medicaid (where most ABA volume lives) sits near the top.

40 / 13 hrs

Prior authorizations per physician per week, and physician-plus-staff hours per week spent on them. 40% of practices have staff working exclusively on prior auth. AMA, 2025.

59%

Physicians who do not appeal adverse PA decisions because they do not believe an appeal will succeed; another 52% cite insufficient staff or time. Set that next to the OIG’s 95%-overturn finding. The belief that appeals do not work is itself the most expensive thing in the revenue cycle.

26%

Physicians reporting that prior authorization led to a serious adverse event for a patient; 8% report it contributed to disability, permanent damage, or death. Note this is “more than 1 in 4” — if you have seen “1 in 3,” that is the stale 2021 figure.

95% / 79%

Physicians reporting PA delays necessary care, and physicians reporting PA can at least sometimes lead to treatment abandonment. Read that second one carefully: it is a share of physicians reporting, not a share of patients who abandon treatment. It is mangled into the latter constantly.

Finally, the one genuinely ABA-specific, actionable billing finding in the peer-reviewed literature, and the real answer to “what is my per-day unit limit.” It is not a payer policy. It is the Medically Unlikely Edit, and we pulled the current values straight from CMS’s practitioner MUE table effective July 1, 2026.

8 units

Medicare’s MUE for 97151 — two hours per day. The tightest limit in the set, and the one the ABA Coding Coalition is actively lobbying CMS to raise.

32 / 24 units

Medicare’s MUE for 97153 (8 hours) and 97155 (6 hours). For the rest: 97152, 97156, 97157, 97158 and 0362T are 16 units; 97154 is 18; 0373T is 24.

Not a benefit limit

CMS defines an MUE as the maximum units reported on the vast majority of appropriately reported claims — not a cap on what is medically necessary. Kornack, Unumb, and Williams, in Behavior Analysis in Practice, document payers misusing MUEs as hard limits and denying ABA on that basis.

Medicaid MUEs differ from Medicare MUEs, and Medicaid’s are the clinically appropriate ones for ABA. The typical ABA patient is a child, not a Medicare beneficiary. Kornack et al. reported Medicaid’s 97151 MUE at 32 units against Medicare’s 8 (as of 2023; we could not re-verify the current Medicaid value, because the CMS dataset the Coalition links to is dead). Yet commercial payers routinely adopt the stricter Medicare edit. If your 97151 claims are getting chopped at two hours, that is a concrete, checkable thing to raise with your payer.

That paper is also, as far as we can find, essentially the only peer-reviewed literature on ABA insurance denials. A field billing billions of dollars a year has one study. That is not a gap in our research; it is a gap in the evidence base, and it is why the vendor blogs have had such an easy time filling it.

Worth knowing where the limits do not come from: no major commercial payer publishes per-day or per-week CPT unit caps for ABA. We went through Aetna’s medical necessity guide, Cigna/Evernorth policy EN0499, Optum’s ABA FAQ, and Anthem’s state provider guides. The unit-limit tables circulating on billing-vendor blogs, attributed to these payers, are not in these payers’ documents. What the payers actually publish is supervision ratios — Optum expects about 2 hours of QHP direction per 10 hours of technician time, Aetna and Cigna 1 to 2 hours per 10. That is the real constraint on your staffing model, and it is nowhere near as widely discussed as the fake unit caps.

The HHS OIG audit series on Medicaid ABA payments

Everything above is about money payers keep from you. This section is about money the federal government is trying to take back, and it is the most ABA-specific data in this article.

HHS OIG is running an audit series on Medicaid ABA payments, announced in January 2022 and running through FY2027. Four states are published so far. In all four, the finding is the same sentence: all 100 sampled enrollee-months included payments for one or more claim lines that were improper or potentially improper. One hundred out of one hundred. Four times.

$56M

Improper fee-for-service Medicaid ABA payments in Indiana (2019–2020), with a $39.4M federal refund recommended. In 97 of 100 sampled months, services were rendered by staff without appropriate credentials.

$18.5M

Wisconsin (2021–2022). Top two causes: session notes that did not support the CPT code billed, and session notes that did not support the number of units billed.

$45.6M

Maine (2019–2023), $28.7M federal refund recommended.

$77.8M

Colorado (2022–2023), issued February 2026. Documentation failures concentrated in 97155, plus uncredentialed behavior technicians. Colorado’s FFS Medicaid ABA spending had grown from $60.1M in 2019 to $163.5M in 2023.

Read the root causes rather than the dollar figures, because the root causes are a checklist for your own documentation. Session notes that do not justify the code billed. Session notes that do not justify the units billed. Staff without the credentials the code requires. Missing signatures. No documentation of protocol changes. Nothing exotic — it is the ordinary stuff, at scale, in every state examined.

Two fairness notes. The OIG’s framing is contested: the ABA Coding Coalition met with OIG in September 2025 to address what it called misunderstandings evident in the Indiana and Wisconsin reports, and OIG agreed to present at APBA and CASP conferences. And be careful with the Colorado number — some coverage reported “$285.2 million,” which aggregates improper and potentially-improper federal-share figures. OIG’s own headline is “at least $77.8 million.” Use OIG’s.

Billing statistics that are wrong, stale, or mangled

Same disease as the workforce numbers, different organ. These are all in current circulation.

Two numbers we went looking for and could not responsibly publish, which we mention because their absence is itself information. There is no citable figure for the share of a therapy clinic’s revenue lost to denials and underpayment — every version traces to vendor marketing. And the MGMA benchmarks everyone quotes for days in A/R and clean claim rate (“35 days,” “top quartile under 5%”) could not be verified against MGMA’s actual paywalled report; every instance we found was a billing company restating them without attribution. We would rather leave a hole than fill it with something we cannot stand behind.

What this means for your clinic

Billing is where a rate cut becomes real. When reimbursement per unit falls, and it is falling in the states with the most volume, as we covered in A Rate Cut Is Not a Margin Problem — the two levers left are collecting more of what you have already earned, and serving more families. The first is a billing problem, and largely a tooling one: our comparison of ABA practice management software looks at how each platform handles authorization tracking and claim scrubbing. The second is a demand problem, and it is the one most clinics leave untouched.

Citing this page

Suggested citation: ABA Clinic Marketing, “ABA Billing and CPT Codes in 2026: The Complete Sourced Reference,” July 16, 2026, abaclinicmarketing.com/blog/aba-billing-cpt-codes-2026. Code descriptors belong to the AMA and are reproduced here for reference; cite the ABA Coding Coalition for the code set itself.

Updated October 16, 2026 with the published 2027 code set, mapped from the 2019 codes using the ABA Coding Coalition’s September 11, 2026 crosswalk and its supplemental guidance article. The confidentiality period that made the July 2026 version of this page necessarily vague has ended. Two things we are still watching: the CMS physician fee schedule final rule, which settles telehealth status for the six new codes, and the payer policies and fee schedules that will attach prices to them. If you spot an error, tell us and we will fix it and say so.