Caution 1: NPPES is self-declared
- Selected by the provider, rarely updated after a promotion [3]
- The client's own certification register is the final word
Home / Case studies / ABA
Case study · Multi-state ABA · Compliance diagnostic
Billed honestly, configured imperfectly: a 370,000-line ABA compliance sweep
Short answer: we tested every claim line, not a sample, against a payer audit's opening questions. The clinics passed; the configuration did not; our biggest first-pass number was wrong until checked against the federal registry.
369,979 lines · 2022 to 2026
$2.82M to $72K · after NPPES match
Lines tested
369,979
Period
2022 to 2026
Archetype
Multi-state ABA
Front end
100% units match
Modifier exposure
$72K floor
Status
Diagnostic
0lines
Billing more units than the session lasted, 369,979 lines
100%
Of lines carry a linked authorization; 99.1% inside the window
$72K from $2.82M
Modifier exposure after NPPES matching, versus the first pass
4,683lines
CO-16 denials, $630K contract value, 23% collected
The payer audit's questions, in its order
A multi-state ABA and pediatric therapy nonprofit; its collection problem is in the first-pass yield case study. Every figure is diagnostic.
- 1Units vs time15-minute units against the session note
- 2Authorization linkedAn authorization ID on every line
- 3Inside the windowService date within the authorized dates
- 4Modifier vs credentialWho delivered, per roster and NPPES
- 5Denial clustersCO-16 templates, CO-97 bundling
- 6Remit routingCash posted where the claim lives
Result
Pass 100.0%
Pass 100%
Pass 99.1%
Review $72K
Fix $630K + $126K
Fix $566K + $203K
Owner
Clinics
Clinics
Clinics
Roster, fee table
Templates, EDI
ERA and EFT enrollment
PassReviewFix, configuration
Front-end integrity on all claim lines
100.0%
Units billed match time workedNo line billed more units than the session lasted
100%
Lines with a linked authorizationAn authorization ID on all 369,979 lines
99.1%
Service inside the authorized window0.9% exceptions listed for appeal or accept
Takeaway
The front end was honest. Every exposure the sweep found lived in configuration: fee tables, claim templates and enrollment, not in the therapy room.
The number that shrank: $2.82M to $72K
ABA codes carry a credential modifier; some plans pay by credential. The first pass inferred credentials from billing patterns; before presenting, we matched the roster to NPPES taxonomy [2] [3].
Modifier exposure, before and after credential matching
Contract value where a technician appeared billed at a higher credential
$2.82MFirst passinferred, 100%
$72KAfter NPPES matchabove registered credential, 2.6%
Root cause: configuration
1
Modifier attached to the fee-table code
It followed the code, not the person.
2
Roster held no credential field
Nothing drove it per provider.
3
Fix: modifier from the provider roster
Matched to NPPES, then the certification register.
Caution 2: payment is not compliance
- A plan can pay a wrong modifier today and recoup it after a documentation audit
- We state the exposure; we do not size a recoupment we cannot verify from published payer rules
Two denial families, on a contract-value basis
Denied contract value and how much of it was recovered
CO-16 missing information4,683 lines, one template defect per Medicaid payer
$630K denied23% collected
CO-97 bundling99% from one Blue plan on 97153
$126K denied12% recovered
0$200K$400K$630K
Contract value deniedRecovered
Orphaned remittances across two group NPIs
Where the claim livedWhere the cash postedFix
Therapy NPIone tax ID, two group NPIs
Landed inThe ABA instance: $566K of therapy-coded remits
FixSeparate ERA and EFT enrollment per group NPI at the clearinghouse
ABA NPItwo system instances
Landed inThe therapy instance: $203K of ABA-coded remits
ResultCash received, posted where nobody was looking
Seven tests, three statuses
| Test | Result | Status |
|---|---|---|
| Units billed vs time worked | 100.0% match; zero lines over session time | Pass |
| Authorization linked to every line | 100% of lines carry an authorization ID | Pass |
| Service inside the authorized window | 99.1% of lines | Pass, exceptions listed |
| Credential modifier vs the person who delivered | First pass $2.82M at risk; $72K premium after NPPES matching | Review |
| Claim templates per Medicaid payer (CO-16) | 4,683 lines, $630K contract value, 23% collected | Fix |
| Bundling edits (CO-97) | $126K, 12% recovered, 99% from one Blue plan on 97153 | Fix |
| Remittance routing across two NPIs | $566K and $203K posted in the wrong system | Fix |
Five habits, now a standing control
All lines, not a sample
Every claim line in the export, so a payer cannot find a population we never looked at.
Credential match before alarm
Rosters matched to NPPES taxonomy, then to the client's certification register, before any exposure is presented.
Templates to enrollment and EDI
CO-16 clusters fixed as one template defect per payer at the source, not reworked line by line.
ERA and EFT per NPI
Separate remittance enrollment per group NPI at the clearinghouse, so cash lands where the claim lives.
Adversarial review
We argue the client's likely objections against our own findings and keep only those that survive.
Stated, not sized
- $72KThe $72K premium is a floor, not the exposureIf a credential cannot be evidenced from the client's register, a payer can recoup the whole claim, not the difference. The certification check closes this.
- MODNo payer modifier mandate verified from a provider manualThe exposure is stated; the recoupment is not sized.
- 0.9%Lines outside the authorized windowAppeal-or-accept per line; the authorization record holds only a total unit cap, so weekly or daily sub-limits cannot be tested.
- 51.7%Authorization utilization 51.7%2,460,664 units authorized, 1,271,475 worked over 20 months; 24.5% of unused units sit behind clients who had stopped attending. Capacity, not revenue.
Frequently asked questions
What does an ABA billing compliance sweep test?
It runs the questions a payer audit opens with across every claim line: do billed units match the time in the session note, does every line carry a linked authorization, does the service date fall inside the authorized window, does the credential modifier match the person who delivered the service, and do denial codes cluster on a template or a payer. Here it covered all 369,979 lines rather than a sample.
Why did the modifier exposure fall from $2.82M to $72K?
The first pass inferred each provider's credential from their billing pattern and flagged $2.82M of contract value. Matching the roster against the NPPES registry, where providers record a behavior analyst or behavior technician taxonomy, showed the premium actually billed above the registered credential was $72K. The modifier was attached to the fee-table code instead of the provider roster.
Is a paid claim with a wrong modifier safe?
No. Plans adjudicate on automated edits and audit documentation later, so a claim paid at a higher credential can be recouped and the finding extrapolated. We state the exposure and ask the client to evidence credentials from its certification register, because NPPES taxonomy is self-selected by the provider.
What is an orphaned remittance?
A payment that posts into a billing system other than the one that holds the claim. Here one tax ID had two group NPIs and two practice-management instances; $566K of therapy-coded remits posted in the ABA instance and $203K the other way. The fix is separate ERA and EFT enrollment for each group NPI at the clearinghouse.
- 45 CFR 162.410, Implementation specifications: health care providers must obtain an NPI and use it on standard transactions (eCFR)
- CMS, NPPES Data Dissemination: the downloadable NPI file and its reference files
- NUCC Health Care Provider Taxonomy: taxonomy codes are self-selected by the provider
Checked October 27, 2026. Engagement figures are from ASP-RCM records and the client's own exports; the client is described by type only.
Run the sweep on your own ABA book
Every claim line tested for units versus time, authorization linkage, modifier-to-credential consistency and template denials.