Autism One is contemporaneous evidence infrastructure. The audited source set says what it says. The rest stays empty. Evidence state, coverage, the six layers you cannot subtract, the known money, and the audit clock are the same job: make legitimate care provable at the time it is delivered.
The board keeps observation year, publication year, and missingness on the same clock as the OIG wave and the Florida process. It does not invent a fraud percentage, does not sum Colorado, and does not promote identification into utilization.
Evidence state
Identified
1 in 31
CDC ADDM 2022, 16 communities. 32.2 per 1,000.
Known money
$10.1B
Medicaid/CHIP ABA paid, 2025, CMS T-MSIS. All beneficiaries.
Measured
10 / 51
States where treated lives are published. Not a census.
Fraud rate
Not collected
None exists; we do not publish one.
What exists for 51
Coverage is a property of each dataset, not a single statewide rate. Identification is 16 ADDM communities. Utilization unique recipients are 10 of 51 after two public searches. Filter the completion KPIs; missing cells stay missing.
| KPI | Phase | Filled | Missing |
|---|---|---|---|
| State demographicsdatasets/demographics.json · child_pop_3_17 Census Vintage 2025 PEP (states.json column still null) | Phase A | 51 / 51 | 0 |
| Workforcestates.json · workforce ∪ datasets/workforce.json BACB certificants 2026-08-22 (density is modeled, not counted) | Phase A | 51 / 51 | 0 |
| FMAPstates.json · fmap ∪ datasets/funding.json fmap_standard FY2027 (90 FR 54696) | Phase A | 51 / 51 | 0 |
| Managed carestates.json · managedCarePct ∪ datasets/payers.json CMS 2024 (AK/CT -- recorded as 0, not missing) | Phase A | 51 / 51 | 0 |
| Fee schedulesseed published 35 ∪ fee-schedules.json 36 with amounts (9 verified + 27 attributed). File-only = 36/51; union = 38/51. Do not treat the 13 as $0. | Phase A | 38 / 51 | 13 |
| Policydatasets/policy.json feeSchedulePointer 51/51. Many field values (diagnosis 16/51, caps, supervision) remain needed. | Phase B | 51 / 51 | 0 |
| Treated livesstates.json FL ∪ normalized/aba-utilization.json present=true (FL IN MD MA MN NV NY NC OH WY). Remainder searched, not zero. | Phase B | 10 / 51 | 41 |
| ABA spendstates.json FL ∪ normalized/aba-spending.json verified|attributed (AZ CO FL IN MD MA MN NE NV NC OH WI WY). NM $75.67M is modeled — not counted. Remainder searched, not zero. | Phase B | 13 / 51 | 38 |
| Claims prevalencedatasets/claims-prevalence.json CDC Medicaid layer CY2022 51/51 (PR + US national extra; not in the 51). Not NSCH. Not ADDM. Not ABA treated lives. | Phase B | 51 / 51 | 0 |
| Provider organizationsdatasets/provider-organizations.json nppes_aba_org_count 51/51 (21,739 org NPIs → 19,520 deduped). Orgs exist in every core jurisdiction. Not billed dollars. PE-7 providers.json untouched. | Phase C | 51 / 51 | 0 |
| Provider-level claimsP10–P99.9 / $/recipient / hours not_public. NPPES org census does not fill this KPI. providers.json is 7 PE traces, not a claims census. | Phase C | 0 / 51 | 51 |
Known flows, not a TAM
Most Appendix C / Figure 2–3c / 8a notes: 50 states + DC + PR + USVI + Guam (54). Footnote 38 on the 2025 $1.47B / 14.5% sentence names 50 states and DC only. Map FIPS filter is isMapRenderableFips (51). Do not subtract $1.47B from $10.1B to invent an ASD remainder. Do not subtract published cells to invent an ASD remainder.
Showing 14 / 14 published cells
| Published cell | Year | Value | Unit |
|---|---|---|---|
| Beneficiaries with ASD, any serviceFrom 2021 to 2025 the number of beneficiaries with an ASD diagnosis who received any Medicaid or CHIP service increased 67 percent, rising from 1.15 million to 1.92 million. | 2021 | 1.15 million | beneficiaries |
| Beneficiaries with ASD, any serviceFrom 2021 to 2025 the number of beneficiaries with an ASD diagnosis who received any Medicaid or CHIP service increased 67 percent, rising from 1.15 million to 1.92 million. | 2025 | 1.92 million; +67% | beneficiaries |
| ASD Medicaid beneficiary growth, 2021 to 2025the number of beneficiaries with an ASD diagnosis who received any Medicaid or CHIP service increased 67 percent, rising from 1.15 million to 1.92 million | 2021 to 2025 | 67 | percent increase |
| ABA recipient growth, 2021 to 2025there was an 189 percent increase in the number of children with an ASD diagnosis who received ABA services | 2021 to 2025 | an 189 percent increase; absolute ABA-recipient counts not printed | percent increase |
| ABA spend growth among children with ASD, 2021 to 2025a 421 percent increase in spending on ABA among children with an ASD diagnosis | 2021 to 2025 | 421 | percent increase |
| ABA spend per ASD child recipient, 2021 to 2025 | 2021 to 2025 | 1.80× (+80%); correction of the 2.76× / +176% misuse | percent increase |
| ABA share of ASD Medicaid beneficiaries, 2021 to 2025 | 2021 to 2025 | — | — |
| All ABA payments | 2021 | approximately $1.94 billion | USD |
| All ABA payments | 2025 | $10.1 billion; +421% | USD |
| All beneficiary ABA spend growth, 2021 to 2025the total amount paid by Medicaid and CHIP for ABA services for all beneficiaries increased from approximately $1.94 billion in 2021 to $10.1 billion in 2025, an increase of 421 percent | 2021 to 2025 | 421 | percent increase |
| ABA payments without an ASD diagnosis | 2021 | $77.6 million | USD |
| ABA payments without an ASD diagnosis | 2025 | $1.47 billion; 14.5 percent of 2025 ABA spending | USD |
| ABA payments without an ASD diagnosis, growth | 2021 to 2025 | 1,789 percent | percent increase |
| Mean weekly ABA hours, ASD | 2025 | +22% vs 2021; 2023 excluded from Figure 8a for an encounter-quantity outlier | hours per week |
Six layers. Six datasets.
A state-to-state identification gap is not fraud. Width is coverage. You cannot subtract these layers.
0/51 · 51 jurisdictions
Need
Who would meet criteria if everyone were assessed.
- Need. Who would meet criteria if everyone were assessed. (0/51; 51 jurisdictions)
- Identification. Who the system has found and recorded. (16/16; 16 ADDM communities — not 51 states)
- Claims. Who has an ASD code on a bill. (51/51; 51 jurisdictions)
- Utilization. Who actually received ABA. (10/51; 51 jurisdictions)
- Payment quality. Documentation and eligibility — not intent. (national only; national only)
- Intent. Adjudicated deception for payment. (none; none)
Ten facts, constraints, and open questions
Each item is an observation, an interpretation, and a boundary — not a claim that the case is closed.
1 · 2021–2025 · CMS
$1.94B → $10.1B (+421%)
Total Medicaid + CHIP payments for ABA, all beneficiaries, 2021–2025. CMS/T-MSIS Figure 3a.
Why it matters. The published national paid total is now in the tens of billions.
Boundary. Do not compare directly with ASD-child counts without matching universes. Page-6 421% is a different universe.
2 · FY2024 · CMS
74%
Insufficient documentation was the largest named driver of FY2024 Medicaid improper payments.
Why it matters. The largest named driver in that split is missing paperwork, not a proven stolen dollar.
Boundary. Medicaid-wide, not ABA-specific. Improper ≠ fraud. Insufficient documentation ≠ proof a service did not occur.
3 · 2024 · CMS OIG
100 / 100
Every sampled Indiana enrollee-month contained at least one improper or potentially improper ABA claim line.
Why it matters. Documentation and payment-support failures were pervasive in this audited sample.
Boundary. Not 100% fraud. Not 100% of services improper. Not a national prevalence estimate.
4 · 2021–2025 · CMS
+1,789%
ABA billed without an autism diagnosis in CMS-defined administrative data. CMS printed +1,789%. 14.5% of 2025 Medicaid/CHIP ABA payments. National; not Florida; not fraud.
Why it matters. A rapidly growing administrative category that deserves separate analysis.
Boundary. National category. Not Florida. Do not call this fraud.
5 · 2022 · CDC
Austin vs Laredo
Austin versus Laredo, ADDM 2022, per 1,000 eight-year-olds in those communities.
Why it matters. Two communities in the same state, same year, are not interchangeable identification rates.
Boundary. ADDM identified prevalence reflects both the children identified and the surveillance/diagnostic systems capable of identifying them. It is not underlying biological prevalence.
6 · 2026 · Published FFS
$12.26 / 15 min
Florida published FFS 97153 $12.26 / 15 min. Four jurisdictions in the audited comparable published-FFS set have lower rates.
Why it matters. This weakens the narrow hypothesis that Florida’s aggregate spending is primarily explained by a uniquely high published FFS unit price.
Boundary. Does not speak to MCO contracted rates, intensity, duration, service mix, or provider mix.
7 · Publication coverage · Audited extract
10 / 51
Ten of 51 jurisdictions currently have a treated-lives (or near-recipient) value in the audited dataset. Filled: FL · IN · MA · MD · MN · NC · NV · NY · OH · WY.
Why it matters. This is publication coverage in the audited dataset — not a market census.
Boundary. Publication coverage is uneven; missing jurisdictions are unknown, not zero. California empty ≠ 0. Not a national rate. Not Florida’s peer set. Not one comparable series (MN EIDBI, NV do-not-sum, MD 2021 testimony).
8 · Category · Competitors
Mixed
Social-robot evidence remains mixed. Separately: Moxie/Embodied shut down. RoboKind moved away from hardware-first delivery and reported that virtual avatars were preferred over hardware for engagement and progress.
Why it matters. Dedicated hardware must demonstrate incremental value over tablets or avatars.
Boundary. The category has not been shown to have failed. The study numbers stay attributed until the meta-analysis is independently reopened.
9 · Gaze study · Matrix
Female validity not established
Female validity is not established in the cited study (male association; no significant female association; small; traits not diagnosis).
Why it matters. Do not use gaze as a clinical or classification feature unless sex-stratified external validation establishes fit-for-purpose performance.
Boundary. “Do not ship” is a product decision, not a finding of the paper.
10 · 2000–2024 · Denmark
6.5% / 3.9%
Nationwide Danish register, cumulative ASD incidence by age 18: 6.5% male / 3.9% female across 2.8M children (2000–2024).
Why it matters. Denmark complicates the claim that high measured autism prevalence is uniquely a product of U.S. ABA reimbursement incentives.
Boundary. Not a causal test of funding incentives. Cumulative incidence to 18 is not comparable to ADDM prevalence at 8. Different diagnostic system, healthcare, registry, and incentives.
State explorer
A state is a bundle of different datasets, not one prevalence number. Empty treated-lives cells were searched twice and are not zero. Colorado $77.8 / $42.6 / $112.5 are never summed.
Showing 51 / 51 jurisdictions
| State | Workforce | FMAP | 97153 | Treated lives | OIG |
|---|---|---|---|---|---|
| AL Alabama | 3180 | 72.6% | $10 | — | No published HHS OIG ABA report |
| AK Alaska | 327 | 52.4% | $22.63 | — | No published HHS OIG ABA report |
| AZ Arizona | 8944 | 64.3% | $17.91 | — | No published HHS OIG ABA report |
| AR Arkansas | 1825 | 69.2% | — | — | No published HHS OIG ABA report |
| CA California | 37952 | 50.0% | $19.39 | — | No published HHS OIG ABA report |
| CO Colorado | 10065 | 50.0% | $17.2 | — | Published Feb 2026 |
| CT Connecticut | 3908 | 50.0% | $14 | — | No published HHS OIG ABA report |
| DE Delaware | 687 | 59.4% | $15.68 | — | No published HHS OIG ABA report |
| DC District of Columbia | 359 | 70.0% | — | — | No published HHS OIG ABA report |
| FL Florida | 67290 | 57.2% | $12.26 | 60000 | No published HHS OIG ABA report |
| GA Georgia | 17138 | 66.4% | — | — | No published HHS OIG ABA report |
| HI Hawaii | 3028 | 59.7% | — | — | No published HHS OIG ABA report |
| ID Idaho | 955 | 66.9% | — | — | No published HHS OIG ABA report |
| IL Illinois | 10997 | 51.8% | $13 | — | No published HHS OIG ABA report |
| IN Indiana | 13341 | 64.7% | — | — | Published Dec 2024 |
| IA Iowa | 2004 | 62.7% | — | — | No published HHS OIG ABA report |
| KS Kansas | 1580 | 60.7% | $16.25 | — | No published HHS OIG ABA report |
| KY Kentucky | 3402 | 71.4% | — | — | No published HHS OIG ABA report |
| LA Louisiana | 1801 | 67.8% | $12.5 | — | No published HHS OIG ABA report |
| ME Maine | 449 | 61.3% | — | — | Published 2025 |
| MD Maryland | 7561 | 50.0% | $19.17 | — | No published HHS OIG ABA report |
| MA Massachusetts | 8105 | 50.0% | $16.37 | — | No published HHS OIG ABA report |
| MI Michigan | 6849 | 65.3% | $16.5 | — | No published HHS OIG ABA report |
| MN Minnesota | 2522 | 50.7% | — | — | No published HHS OIG ABA report |
| MS Mississippi | 1112 | 76.9% | $15.08 | — | No published HHS OIG ABA report |
| MO Missouri | 5157 | 64.4% | — | — | No published HHS OIG ABA report |
| MT Montana | 354 | 61.5% | $11.36 | — | No published HHS OIG ABA report |
| NE Nebraska | 2388 | 55.9% | $18.7 | — | No published HHS OIG ABA report |
| NV Nevada | 4154 | 59.8% | $30.1 | — | No published HHS OIG ABA report |
| NH New Hampshire | 1716 | 50.0% | $17.79 | — | No published HHS OIG ABA report |
| NJ New Jersey | 9592 | 50.0% | $15 | — | No published HHS OIG ABA report |
| NM New Mexico | 1658 | 71.7% | — | — | No published HHS OIG ABA report |
| NY New York | 5959 | 50.0% | $14.45 | — | No published HHS OIG ABA report |
| NC North Carolina | 14300 | 64.6% | $20.81 | — | No published HHS OIG ABA report |
| ND North Dakota | 346 | 51.0% | $10.6 | — | No published HHS OIG ABA report |
| OH Ohio | 6853 | 64.8% | — | — | No published HHS OIG ABA report |
| OK Oklahoma | 3617 | 66.5% | $17.35 | — | No published HHS OIG ABA report |
| OR Oregon | 999 | 57.8% | $14.7 | — | No published HHS OIG ABA report |
| PA Pennsylvania | 8108 | 56.1% | $12.73 | — | No published HHS OIG ABA report |
| RI Rhode Island | 1148 | 57.5% | — | — | No published HHS OIG ABA report |
| SC South Carolina | 4363 | 69.5% | $14.88 | — | No published HHS OIG ABA report |
| SD South Dakota | 452 | 51.0% | $21.73 | — | No published HHS OIG ABA report |
| TN Tennessee | 7227 | 64.2% | — (No statewide schedule) | — | No published HHS OIG ABA report |
| TX Texas | 36428 | 59.8% | $14.5 | — | No published HHS OIG ABA report |
| UT Utah | 5572 | 62.5% | $19.67 | — | No published HHS OIG ABA report |
| VT Vermont | 387 | 59.0% | $15 | — | No published HHS OIG ABA report |
| VA Virginia | 7358 | 50.4% | $15 | — | No published HHS OIG ABA report |
| WA Washington | 2929 | 50.0% | $12.4 | — | No published HHS OIG ABA report |
| WV West Virginia | 672 | 74.2% | $9.9 | — | No published HHS OIG ABA report |
| WI Wisconsin | 2777 | 60.7% | — | — | Published Jul 2025 |
| WY Wyoming | 158 | 50.0% | $12.3 | — | No published HHS OIG ABA report |
Identification · CDC ADDM 2022
Research has not demonstrated that living in certain communities puts children at greater risk for developing ASD. Overall 1 in 31 (32.2 per 1,000; male 49.2; female 14.3; ratio 3.4). ADDM is 16 communities, not 51 states.
| Community | Per 1,000 | Male | Female | Ratio | Record access |
|---|---|---|---|---|---|
| California (San Diego) | 53.1 | 80.1 | 23.1 | 3.5 | all |
| Pennsylvania (subn Phila.) | 47.4 | 72.4 | 21.6 | 3.4 | all |
| Wisconsin | 38.4 | 57.5 | 17.9 | 3.2 | all |
| Minnesota | 35.5 | 53.6 | 17.4 | 3.1 | health+ed |
| New Jersey | 34 | 52.9 | 14.3 | 3.7 | health+ed+ei |
| Tennessee | 34 | 51.6 | 15.3 | 3.4 | health+ed |
| Georgia (Atlanta) | 32.6 | 51.1 | 13.7 | 3.7 | health+ed |
| Missouri | 32.1 | 47.8 | 15.7 | 3.1 | health+ed |
| Arizona | 31.3 | 48.3 | 13.7 | 3.5 | health+ed |
| Arkansas | 29.8 | 48.1 | 10.8 | 4.4 | health+ed |
| Utah | 27 | 39.9 | 12.3 | 3.2 | all |
| Puerto Rico | 26.4 | 39.8 | 12.5 | 3.2 | all |
| Maryland | 26.3 | 41.3 | 10.7 | 3.9 | health+ed+ei |
| Texas (Austin) | 19.5 | 28.5 | 9.6 | 3 | health+ed |
| Indiana | 18.3 | 27.5 | 8.7 | 3.2 | health+ed |
| Texas (Laredo) | 9.7 | 15.9 | 3 | 5.3 | health+ed |
The audit wave is already public
HHS OIG announced an 8-state Medicaid ABA series. 4 published, 4 unnamed. Documentation pressure is the CMS FY2024 split — 74% insufficient documentation, Medicaid-wide, not ABA-specific. Improper payment is not fraud. Florida's Task Force recommendations are due 31 Dec 2026.
Published audits
| State | Published | Improper $M | Federal refund $M | Potentially improper $M | Sample |
|---|---|---|---|---|---|
| Indiana | 2024-12 | 56 | — | — | 100 of 100 claims deficient |
| Maine | 2025 | 45.6 | — | — | — |
| Wisconsin | 2025-07 | 18.5 | — | 94.3 | 100 of 100 enrollee-months · No statewide postpayment review since program inception 2016 |
| Colorado | 2026-02 | 77.8 | 42.6 | 112.5 | 96 recipients / 100 months · $289.5M reviewed, 1M+ claims; spend $60.1M (2019) to $163.5M (2023); ~6,600 certified techs plus est. 1,500-2,000 uncertified |
Deficiency categories
- Notes do not support the CPT code, units, or dates — most-cited in every audit
- Staff lacking credentials or documented supervision
- Missing signatures / wrong rendering provider
- Recreational time, meals, naps billed as clinical
- Excessive units; overlapping service times
- Prior-auth not verifying diagnosis/referral
Written eligibility gates
Florida is not the only state whose written Medicaid rules appear to permit behavioral treatment without confirmed ASD. Our current 51-jurisdiction review identifies eight provisionally similar policy pathways; their legal and operational equivalence is now being audited.
Florida is not the only jurisdiction whose written Medicaid policy allows behavioral treatment without confirmed ASD.Supported, narrow
Verified written non-ASD paths include CA BHT, AZ BA, CO PBT, LA ABA, PA IBHS ABA, ID CHIS, WI behavioral treatment, KY MH/SUD BH ABA, and MN/WA/OR/NY named-class benefits.
Eight jurisdictions have rules that can reasonably be described as function/diagnosis-neutral.Not supported as one bucket
Provisional D=8 does not survive audit. Mechanisms split D1/D2/D3/D5; KY left D; ID object changed.
Those jurisdictions provide operationally comparable ABA benefits.Not supported
Taxonomy does not measure operational equivalence. Service definitions differ.
They exhibit comparable non-ASD utilization.Not supported
observedUtilization is a separate dataset. Not inferred from class.
They exhibit comparable spending.Not supported
Spending is a separate empirical object.
Provisional D split
- Split D1: AZ, CA
- Split D2: FL, ID
- Split D3: PA, RI, VA
- Split D5: CO, LA
Written gate audit status
- Verified, direct: 38
- Verified with interpretation: 5
- Conflicting authority: 3
- Stale source: 2
- Unresolved: 2
- MCO / state mismatch: 1
Hypothesis board
Verdicts were re-audited 2026-08-22. “Not supported” is not falsification. The stated mechanism can fail while the parent remains untested.
Geographic variance is measurement, not biologyConfirmed
Austin 19.5 vs Laredo 9.7 per 1,000 — same state, same Medicaid program, 235 miles. CDC states it directly.
The 22-year rise is identification, not incidenceLargely confirmed
DSM-5 consolidation (2013), criteria change, screening expansion.
Some genuine variation in population risk remains plausible at the margin.
Sex ratio is a surveillance-quality indicatorNot supported
Austin has the lowest ratio in the network (3.0) at the second-lowest prevalence; removing Laredo collapses the tercile trend to 3.10, below the high-prevalence group.
16 sites, no pre-specified test, and CDC reports significant ratio heterogeneity (p<0.01). Supports 'not usable as a cross-site proxy', not general failure.
Florida's spend reflects loose billing and fraudMechanism falsified
Published 97153 rate $12.26/unit, 11th percentile — falsifies the high-unit-price mechanism only.
Silent on volume, duration, authorization, eligibility breadth, coding, improper-payment exposure, and provider conduct. High volume is compatible with loose billing, not exonerating.
Provider density tracks reimbursement ratesNot supported
Largest workforce (67,290) on a near-lowest published rate; Nevada pays 2.5x with a fraction of the market.
One counterexample against published FFS, uncontrolled for population, contracted MCO rates (unknown; FL is 89-99% managed care), commercial mix, cost of living, entry timing. Rejects 'published FFS alone explains density' — not the relationship.
The dominant failure is documentation, not fraudConfirmed
CMS FY2024: 74% of Medicaid improper payments ($23.4B of $31.8B) are insufficient documentation; 15.6% ineligibility or lack of medical necessity.
Medicaid-wide, not ABA-specific.
Published fee schedules are not what providers are actually paidConceded
Florida runs 89-99% managed care; Tennessee publishes no statewide schedule.
Qualifies H4a and H5 without overturning either.
US prevalence is inflated by funding incentives; true rate ~1.5-2%Falsified, strong form
Denmark — universal healthcare, no ABA-funding incentive — reports cumulative incidence to age 18 of 6.5% male / 3.9% female across 2.8M children (2000-2024).
Cumulative-incidence-to-18 is not point-prevalence-at-8 and runs higher by construction. Falsifies the ~1.5-2% claim; does not establish that no share of US identification is incentive-related.
A five-dataset crosswalk beats any single sourceAccepted
IDEA Part B child count is public, per-state, and measures a fifth distinct thing: who a school classified.
Adopted as method. Not an empirical finding.
Provider behaviour is driven by working conditions, not only ratesSupported
RBT annual turnover ~65%; replacement 90-200% of salary.
Attributed, national, not state-resolved. Context — do not convert to dollars.
Instrument evidence matrix
Evidence grade reflects replication status, not intuition. Female validity is a separate column; gaze stays not established.
0 = no usable evidence · 1 = preliminary or contested · 2 = emerging, small samples or single lab · 3 = replicated across independent samples
| Modality | L1 | L2 | L3 | Female validity |
|---|---|---|---|---|
| Trial outcomeHuman-entered: correct/incorrect/prompted, prompt level, reinforcer. No inference, no training data, no bias surface. The measurement basis of ABA itself. | 3 · Fully trackable | 3 · Fully trackable | 3 · Fully trackable | Neutral |
| Response latencyTime from SD to response, from timestamps. A clock, not a classifier. Most valuable at L3 where other signals thin out. | 3 · Strong | 3 · Strong | 3 · Strong | Neutral |
| Turn-taking timingDiarization; gap and overlap duration only, not content. Not analysed by sex in any work found. Plausibly neutral, unverified. | 2 · Good | 2 · Moderate | 1 · Limited | Untested |
| Motor stereotypyVideo pose or accelerometer. Males show significantly more stereotyped behaviour. A stereotypy-centred product is tuned to boys. | 1 · Sparse events | 3 · Best case | 2 · Detectable, undocumented | Skewed |
| Gross motor / kinematicsGait, postural sway, rhythmicity. Best sex-balanced study (18 of 37 female) still lacked power to analyse sex. | 2 · Group-level only | 2 · Group-level only | 0 · Unknown | Not analysed |
| Gaze / eye regionFixation to eyes and faces. Single study, n=48, university-age adults, autistic traits not diagnosed autism; no pre-specified validation criterion. Cited male rho −0.45 to −0.47; no significant correlation in females. Female validity is not established. Do not ship gaze (product constraint: no established female validity, '725 patent proximity, impracticality above Level 1). Legacy femaleValidity=fails is SUPERSEDED. See matrix.corrections.json. | 1 · Evidence in males | 1 · Weak | 0 · Impractical | Not established |
| Speech prosodyPitch range, rhythm, intensity contour. Small samples, inconsistent cross-linguistically. BIPA voiceprint exposure. | 2 · Promising | 1 · Limited | 0 · Not applicable | Untested |
| Facial affectExpression classification. EU AI Act bans emotion inference in education. Faceprints regulated under amended COPPA. Structurally blind to camouflaging. | 1 · Do not ship | 1 · Contested | 0 · No evidence | Prohibited |
| Autonomic (HR/HRV/EDA)Bracelet telemetry. Arousal is measurable; arousal-to-distress mapping is the weak link. Legal to ship well before meaningful. | 2 · Indicative | 2 · Indicative | 1 · Noisy | Untested |
| Session structureStart/stop, environment, who present, materials. Also the generalization matrix. Inference-free and reimbursed. | 3 · Fully trackable | 3 · Fully trackable | 3 · Fully trackable | Neutral |
Social-robot category, not ABA software
None of these captures billable ABA documentation. That is the whole opening. Category evidence: 40 studies, 17 RCTs, 346 participants.
RoboKindPivoted off hardware
Milo humanoid, now virtual avatars
~400 school placements; DoD $6.4M; Indiana $1.35M
Data showed that Virtual Avatars were preferred over hardware for boosting student engagement and progress.
Movia RoboticsActive
Robot-Assisted Instruction on Kebbi/iPal/NAO/Pepper
~300 robots since 2010 to ~40 schools, ~80 families
LuxAIActive
QTrobot + curriculum, session reporting, SDK
WEF/CES awards; research-lab presence
No ABA-specific features
RobotLABActive
Reseller: NAO V6, Pepper, ARES autism packs
claims 1,000+ schools and clinics
Embodied (Moxie)Defunct Dec 2024
Consumer AI companion
consumer
Devices bricked within days, no refunds
ABA practice software
EHRs and documentation tools are a separate competitive set from social robots. Do not collapse them.
Showing 18 / 18 ABA software products
| Product | Class |
|---|---|
| CentralReach, LLC (Florida) / CentralReach Holdings, LLC (Roper acquiree)CentralReach autism and IDD care software / EMR operating system | EHR / practice management |
| Rethink Behavioral Health (RethinkBH), a division of RethinkFirstRethinkBH | EHR / practice management |
| Motivity / Motivity Systems (founder launched Motivity Systems)Motivity | EHR / practice management |
| Passage HealthPassage | EHR / practice management |
| DataFinch Technologies, Inc. (App Store seller); product now branded Ensora Data CollectionEnsora Data Collection (formerly Catalyst by DataFinch) | Clinical documentation |
| Theralytics, LP (trade name Theralytics)Theralytics | EHR / practice management |
| Hi Rasmus (trade name)Hi Rasmus | Clinical documentation |
| Artemis ABA (trade name)Artemis ABA | EHR / practice management |
| Aloha Practice Management IncAlohaABA | EHR / practice management |
| Hex Matrix (company); ABA Matrix (product)ABA Matrix | EHR / practice management |
| Lumary (trade name)Lumary ABA | EHR / practice management |
| TherapyPM (trade name)TherapyPM | EHR / practice management |
| Portia (trade name / PortiaPro)Portia ABA Software | EHR / practice management |
| Noteable (trade name)Noteable | EHR / practice management |
| Raven Health (trade name)Raven | EHR / practice management |
| Praxis Notes (trade name)Praxis Notes | AI notes |
| AccuPoint (product); landing hosted on Therapy Brands / Ensora infrastructureAccuPoint 2.0 | EHR / practice management |
| Calmanac (product of HumaneBITS; now inside Motivity)Calmanac | EHR / practice management |
Patent working brief
Program Brief §06 — working brief for counsel, not an FTO opinion. Filter by risk. This is not an FTO opinion.
| Patent | Risk | Design around |
|---|---|---|
| US 12,380,988Behavioral-health engagement platformGUI generation for child/therapist/parent displays | Med high | Restructure around asynchronous role-scoped surfaces rather than synchronized session screens. Element-by-element claim chart required. |
| US 2021/0043106Learning platform for persons with autismguardian/therapist dashboards, skills to modules to progress | Medium | No skill-tree / module-progression dashboard. Targets with mastery criteria and generalization probes instead. |
| US 10,176,725ASD intervention system incl. robotrobot providing/assisting ABA, gaze sensing | Medium | ABA logic server-side and clinician-attributed. No gaze-coordinate storage. |
| US 2018/0117479Voice-enabled connected smart toybutton to SIP voice session | Medium | Wake-word or scheduled trigger, WebRTC/MQTT, partial on-device NLU. |
| US 10,864,453Mobile robot for child developmentexpressive face, therapy/game modes | Low med | Stationary housing. No device-run therapy or game modes. |
| US 11,185,659Digitally-interactive plush therapeutic apparatus | Low med | No tactile biofeedback loops or therapeutic vibration patterns. |
| US 10,792,581Robotic therapeutic learning toy (train) | Low | Different form factor entirely. |
| US 10,885,719Methods and systems for treating autismVR/AR emotional regulation | Low | Different modality. |
| WO 2001/069830Networked interactive toy system | Prior art | Defensive. Cite against broad connected-toy networking claims. |
| US 2025/0032945LLM-based interactive dollpending | Watchlist | Quarterly docket watch. Cite as related art in our own filing. |
Why now
The audit wave and the Florida clock are contemporaneous with this board. Filter by kind; critical dates stay labeled.
2022-01 · Federal
OIG announces 8-state Medicaid ABA audit series
2024-12 · Audit
Indiana audit published — $56.0M, 100/100 claims deficient
2024-12 · Market
Embodied/Moxie folds; cloud robots bricked, no refunds
2025 · Audit
Maine audit published — $45.6M
2025-07 · Audit
Wisconsin audit published — $18.5M / $94.3M potentially
2025-02-01 · Policy
Florida BA carved into SMMC MMA — plan claims for enrollees; FFS remainder continues
2026-01 · Regulatory
FDA revised general wellness guidance
2026-01-01 · Regulatory
USCDI v3 required for ONC-certified EHRs
2026-02 · Audit
Colorado audit published — $77.8M improper / $42.6M refund
2026-02-16 · Regulatory
42 CFR Part 2 single-consent rule effective
2026-04-22 · Regulatory · critical
Amended COPPA fully enforceable — biometrics now PI
2026-08-12 · Policy
Florida ABA Task Force first meeting
2026-12-31 · Policy · critical
Florida ABA Task Force recommendations due to Governor + Legislature
2027-07-01 · Policy
Florida Task Force authority expires
2027 · Federal · critical
OIG audit series completes — 4 unnamed states publish
Method
A number without a source is a bug. A conclusion without a boundary is a bug. Missing cells stay empty.
- No fraud percentage, no “prevents fraud,” no vaccine content.
- Colorado figures are never summed.
- No imputation. Hatched is not a low value.
- Identification is not utilization. Utilization is not payment quality. Payment quality is not intent.
Sources
Each figure keeps the observation year printed by its source.
- CDC Autism and Developmental Disabilities Monitoring (ADDM) Network, MMWR Surveillance Summaries, 2022. Identified prevalence in 16 communities, not 51 states.
- HRSA Maternal and Child Health Bureau and US Census Bureau, National Survey of Children's Health Indicator 2.8, 2023 to 2024. Parent reported current ASD, ages 3 to 17.
- CMS State Medicaid and CHIP ABA Toolkit, T-MSIS national paid totals, August 2026. All beneficiary ABA payments, $1.94B (2021) to $10.1B (2025).
- HHS Office of Inspector General, Medicaid ABA audit series, 2024 to 2025. Published reports for Indiana, Maine, Wisconsin, and Colorado.
- CMS Payment Error Rate Measurement, FY2024 Medicaid improper payment split. Documentation share is Medicaid wide, not ABA specific.
- Behavior Analyst Certification Board, region specific certificant counts, 2026 snapshot. BCBA, BCaBA, and RBT certificants by location, not Medicaid enrolled providers.
- BACB and Lightcast, US employment demand for behavior analysts, 2010 to 2025. Job posting growth, not treated lives.
- State Medicaid published FFS 97153 schedules. Statewide technician rates where a public fee schedule exists. Not MCO contracted rates.
- CDC Medicaid claims coded ASD prevalence, continuously enrolled ages 3 to 17, calendar year 2022. A claim code is not ABA receipt.
- Census Bureau Vintage 2025 Population Estimates Program, civilian ages 3 to 17. Denominator for workforce density only.
- State Medicaid unique recipient extracts. Treated lives are published for 10 of 51 jurisdictions. Empty is not zero.