Three codes, three time bands, and a scheduling convention that puts a great many practices on the wrong side of the line without anybody intending it. This is the most audited pattern in outpatient mental health and the most avoidable.
Individual psychotherapy is coded by session length: 90832 covers roughly 16 to 37 minutes, 90834 covers 38 to 52 minutes, and 90837 covers 53 minutes or more. A 50-minute session is 90834, not 90837. Billing 90837 for shorter sessions is a heavily audited upcoding pattern.
CPT describes these codes by a nominal duration — 30, 45 and 60 minutes — but payment is decided on the time actually documented, against ranges built around those midpoints:
A session under 16 minutes does not reach the threshold for the lowest code at all.
The bands are not guidance to be applied sensibly. They are thresholds, and there is no rounding convention that lets a 52-minute session become 90837.
The traditional therapy hour is fifty minutes. It is how practices schedule, how clinicians are trained, and what most patients expect.
Fifty minutes is 90834.
A practice that schedules 50-minute sessions and bills 90837 across the board is upcoding every single claim — not occasionally, not in edge cases, but structurally. And because the pattern is perfectly consistent, it is trivially visible in claims data. Psychotherapy coding in this code range has been named repeatedly as an area of audit focus.
The exposure is not a handful of adjusted claims. A practice billing 90837 by default across a year has a systematic overpayment, and the recovery is calculated across the whole period.
If the clinical work genuinely needs 60 minutes, schedule 60 minutes and document it. If it needs 50, schedule 50 and bill 90834. What cannot hold is a 50-minute calendar and a 60-minute claim.
The code asserts a duration, so the record has to establish one. Start and stop times are the cleanest form. A stated total session time is generally acceptable. What does not survive review is a note that describes the content of a session thoroughly and never says how long it lasted.
Behavioral health is delivered remotely more than almost any other specialty, so the modifier matters constantly rather than occasionally:
These are distinct and not interchangeable. A telephone session billed with modifier 95 misstates how the service was delivered, and payers that cover audio-video but not audio-only treat that difference as material.
Place of service requirements for telehealth vary by payer and have changed more than once. Confirm the current expectation per payer rather than carrying forward a configuration set during an earlier period.
Federal parity law requires that limitations applied to mental health and substance use disorder benefits be no more restrictive than those applied to comparable medical and surgical benefits. Visit limits, prior authorisation requirements and concurrent review that apply to therapy but not to comparable medical care are the kind of thing it is aimed at.
In practical billing terms, this means a denial citing a visit limit or an authorisation requirement is not automatically the end of the matter. It is worth asking whether the same limitation is applied to comparable medical and surgical benefits by that plan, and appealing on that basis where it is not.
This is genuinely specialised ground and the answer depends on the specific plan and its documented criteria. We would not represent otherwise. But a practice writing off parity-adjacent denials without ever asking the question is leaving something on the table.
Send three months of claims and your typical appointment lengths. We will tell you whether your code mix is defensible and where it is not. No charge, and if it is clean we will tell you that.
Request a free A/R review