What a 95% clean-claim rate actually requires

Mental health billing carries denial patterns most specialties never see, and psychiatry billing in particular concentrates them. HFMA sets 95% as the benchmark for a high-performing billing operation. Most independent and group practices run 75–85%. That gap is worth real money: at $150 average reimbursement and 3,000 monthly claims, each point of clean-claim rate is roughly $4,500 a month in delayed or reworked cash.

The instinct is to blame billing. In our experience running a multistate psychiatry practice, that is almost never where the problem is.

Denials are usually documentation failures that surface at the claim

A claim is the last place an error becomes visible, not the first place it occurred. Walk the common behavioral-health denials backwards and most originate before the biller ever opens the encounter:

• Diagnosis specificity the note does not support — documentation describes a general anxiety presentation; the submitted code is a specific F41 subtype.
• Add-on code without the primary service documented — 90833 billed alongside an E/M where psychotherapy time is not separately evidenced.
• Modifier 25 without a separately identifiable service — one of the most audited constructs in outpatient psychiatry.
• Eligibility accurate at scheduling and stale at service — plan changes, terminated coverage, or a behavioral-health carve-out routing to a different payer entity.
• Rendering provider not credentialed with that payer, in that state, on that date of service.

None of these are billing errors. They are readiness, documentation and eligibility errors that a biller inherits and cannot fix without going back to the clinician — which is why they become denials and A/R days rather than clean claims.

The four checks that move the number

Reaching 95% is mostly a matter of moving verification upstream of submission. In order of impact:

1. Verify eligibility close to the date of service, not at booking

A 270/271 check run at scheduling describes coverage as it existed when the appointment was made. For a practice booking three weeks out, that is three weeks of drift. Re-checking within a day or two of the visit catches terminations, plan changes and carve-out routing while the front desk can still act on them.

2. Check code combinations against NCCI before submission

CMS publishes Procedure-to-Procedure and Medically Unlikely Edits quarterly. These are deterministic: a pair either conflicts or it does not. Running them pre-submission turns a class of denial into an exception someone resolves in seconds. In behavioral health, E/M plus psychotherapy add-on combinations account for a disproportionate share.

3. Tie provider readiness to the claim, not to a spreadsheet

A provider can be licensed in a state, listed on a payer roster, and still not be billable for a specific service on a specific date — because supervision requirements apply, because the roster entry post-dates the encounter, or because the contract state scope does not cover where the patient was located. If readiness lives in a credentialing spreadsheet and claims live in the billing system, nothing connects them until a denial arrives.

4. Make corrections stick

This is the difference between a clean-claim rate that spikes and falls back, and one that holds. When a denial is worked, the correction has to be captured so the next claim can use it — as a rule, a check, or a required field. Otherwise the same error recurs next month with a different patient, and the team re-solves a problem it already solved.

What 95% does not require

It does not require replacing your EHR, and it does not require AI that submits claims autonomously. Every check above is deterministic — rules, edits and eligibility responses, not model judgment. Automation runs those checks on every claim without a human remembering to, and routes exceptions to whoever can resolve them. Clinical and billing decisions stay with people.

For a view of your own gap, the fastest diagnostic is three numbers: first-pass acceptance rate, denial rate by reason code, and days in A/R. The reason-code distribution usually reveals within an hour which of the four checks is missing.