HHS public Medicaid ABA data compared with CMSHow much ABA spending appears in the HHS public file?
For eight ABA billing codes in 2024, the February 2026 HHS public extract sums to $2.117 billion, compared with $8.103 billion in the August 2026 CMS data book.
Who this counts: National Medicaid and CHIP paid amounts, including fee-for-service and managed care, for CPT 97151 through 97158; February 2026 HHS public release compared with the August 2026 CMS data book.
At a glance
- The public-file amount is numerically 26.1% of the CMS total. That ratio does not measure missing claims or fraud.
- Both sources include fee-for-service, managed care and CHIP. This is not a comparison of fee-for-service with all Medicaid.
- The public file suppresses small rows and uses line-level payments. The sources also come from different releases.
Quote this figure
For CPT codes 97151 through 97158 in 2024, Autism Stats calculated $2.117 billion in the February 2026 HHS public extract versus $8.103 billion in the August 2026 CMS data book; the difference is not a fraud estimate.
The two releases produce different ABA totals
HHS public-file amount divided by CMS data-book amount (%)
Same eight CPT codes, 97151 to 97158. February 2026 HHS release versus August 2026 CMS workbook. This numerical ratio is not an estimate of missing claims, fraud or completeness.
Source: HHS Medicaid Provider Spending by HCPCS; CMS Autism Spectrum Disorder and ABA Data Book.
$2.117 billion in one release, $8.103 billion in another
For calendar year 2024, the HHS extract contains $2,116,900,673.48 in paid amounts for CPT codes 97151 through 97158. CMS Table 3 reports $8,103,311,330 for ABA therapy. Table 7 identifies the same eight codes.
Dividing those two amounts gives 26.1%. The four-year table shows the same calculation for 2021 through 2024. This comparison uses two different releases, not a matched set of individual claims.
Why the public file is not a national spending total
The HHS release groups claim lines by billing provider, servicing provider, procedure code and month. A row is published only when it contains at least 12 claim lines and 12 unique patients. Summing the surviving rows cannot restore the suppressed amounts.
HHS also uses line-level payments and warns that those amounts can differ from claim-header payments in T-MSIS. Both releases include fee-for-service, managed care and CHIP data, so the difference should not be described as managed care being absent from the public file.
The HHS methodology labels 2024 data preliminary and notes state reporting differences. The CMS workbook was released six months later. These facts limit the comparison; they do not tell us how much of the difference each factor explains.
A smaller sum is not evidence of Medicaid fraud
The roughly $5.99 billion arithmetic difference does not identify improper payments, hidden providers or fraudulent claims. Establishing those findings requires claim-level evidence and an investigation or audit.
Use the public file to examine the provider, code and month combinations it actually publishes. Use the CMS national series when citing the CMS-reported ABA total. Neither source measures whether therapy helped a child.
Do not add patient counts across rows
A person can appear in several provider, procedure and month combinations. Adding the patient-count column would count some people repeatedly. Claim-line counts are also not the same as the number of whole claims.
This analysis sums paid amounts only. It does not estimate unique patients or reconstruct suppressed records.
What this cannot tell you
- The HHS extract is the February 9, 2026 release saved on October 12. Its documentation was rechecked on November 2; the full extract was not downloaded again.
- Matching eight procedure codes and calendar years does not make the underlying claim populations, revisions or payment fields identical.
- The ratio is an arithmetic comparison between releases. It is not a measured percentage of all claims captured or money missing.
- The available aggregates cannot separate the effects of suppression, payment-field differences, later revisions or state reporting.
- The analysis excludes 0362T and 0373T, which are outside the eight-code scope used for this comparison. It does not cover every possible autism service.
How this report was built
Autism Stats queried its saved HHS ABA extract for exactly CPT 97151, 97152, 97153, 97154, 97155, 97156, 97157 and 97158. For each calendar year 2021 through 2024, we summed TOTAL_PAID after casting each value to a two-decimal fixed-point number. We compared those sums with PaidTotal in CMS Table 3 and checked the eight-code scope against Table 7. Ratio (%) = HHS sum / CMS total x 100, rounded to two decimal places in the CSV. The August 2026 CMS workbook was downloaded again on November 2 and matched the saved workbook byte for byte. This is a secondary analysis of published aggregates, not a claims audit.
Source retrieval dates are listed below. This report was checked against those source versions on November 2, 2026. Live source pages may change after that date.
Read the publication methodSources and downloads
| HHS Medicaid Provider Spending by HCPCS | Open the sourceSaved eight-code subset; payment methodology and suppression documentation rechecked November 2, 2026. Data period 2021 to 2024 subset of January 2018 to December 2024 release. Released February 9, 2026. Retrieved 2026-10-12. |
|---|---|
| CMS Autism Spectrum Disorder and ABA Data Book | Open the sourceTable 3 national PaidTotal; Table 7 eight-code ABA scope. Data period 2021 to 2024 subset of 2021 to 2025 workbook. Released August 2026. Retrieved 2026-11-02. |
Cite this report
Autism Stats. Public Medicaid ABA spending data compared with CMS totals. Version 1.0, November 2, 2026. https://autismstats.com/statistics/medicaid-aba-public-data-comparison
When you quote a figure, include the population and collection year. Link to the original source as well when your publication requires it.