All insights

Clinical documentation

Four things generic RCM tools get wrong in psychiatry

Behavioral-health denials cluster around four constructs most specialties barely use — and every one of them is catchable before submission.

Generic revenue-cycle tools get psychiatry wrong in predictable ways. The errors are not random — they cluster around four constructs that behavioral health uses constantly and most specialties barely touch.

If your denials concentrate in these four, the problem is not your billing team. It is that the checks running against your claims were built for a different specialty.

1. E/M plus the psychotherapy add-on

A psychiatric visit is frequently two services in one encounter: an evaluation and management service, plus psychotherapy delivered alongside it. The add-on codes (90833, 90836, 90838) attach to the E/M rather than standing alone.

Two things go wrong. First, the documentation has to evidence the psychotherapy separately from the E/M work — time spent, modality, and content distinct from the medical management. A single blended narrative supports one service, not two. Second, the add-on must pair with a permitted primary code; the wrong pairing is a deterministic edit failure that never should have reached the payer.

2. Modifier 25

Modifier 25 asserts that a significant, separately identifiable E/M service occurred on the same day as another procedure. In behavioral health it appears constantly — and it is among the most audited modifiers in outpatient medicine.

The failure mode is habitual application: the modifier becomes something the billing system appends by default rather than something the documentation earns. That works until an audit samples the charts, at which point the recoupment covers years, not months. A pre-submission check that asks whether the note actually supports a separately identifiable service is worth more than any appeal process.

3. Incident-to and supervision

This is the construct where multistate groups get hurt most, and it is genuinely subtle.

Incident-to billing lets services delivered by a supervised clinician bill under the supervising provider, under specific conditions. Three things make it dangerous. Rules vary by state and payer. The conditions are factual, not inferential — you cannot conclude incident-to applies merely from credentials, NPI type or employment relationship. And crucially, when incident-to does apply, the claim reports the supervisor as rendering provider while the record must still show who actually furnished the service.

Collapsing those two facts into one field is the error. The claim needs the supervisor in CMS-1500 Item 24J; the clinical record needs the performing clinician. Systems that store a single “provider” per encounter cannot represent this correctly, and produce either a wrong claim or an unauditable record.

4. Telehealth across state lines

Behavioral health went remote faster than almost any specialty, and the rules did not follow uniformly. Three separate questions have to resolve for the same visit:

  • Licensure — determined by where the patient is located, not the provider.
  • Payer telehealth terms — audio-only allowance, rate parity, permitted place-of-service codes, all of which vary by contract.
  • Prescribing authority — particularly for controlled substances, governed by state law that changes with some regularity across all 51 US jurisdictions.

A patient who travels changes the answer to all three mid-treatment. Most systems never ask the question.

A diagnostic worth running. Pull your last quarter of denials and sort by reason code. If a meaningful share trace to add-on pairing, modifier 25, rendering-provider mismatch, or place of service, your checks are generic. Those four are deterministic — every one can be caught before submission rather than appealed after.

Why this is a systems problem

Each of these is knowable in advance. None requires judgment at claim time: the code pair either conflicts or it does not; the state either permits the prescription or it does not; the supervision agreement either exists for that state on that date or it does not.

What they require is that the system holds the right facts — performing clinician and supervising provider as distinct fields, patient location as explicit context rather than inferred from an address, contract terms as structured data rather than PDF prose, and reference data current to the quarter.

Practices do not fail these checks because their staff lack knowledge. They fail because the knowledge lives in someone’s head and the claim goes out at 4:45pm on a Friday.

Talk to us

Tell us where revenue is leaking.

A few questions about your provider count, states, EHR and claim volume — enough that the first call is about your numbers.