ONESTEP SCRIBE

The three psychotherapy add-on codes: choosing between 90833, 90836 and 90838

By Selam Koch, PMHNP-BC, founder of OneStep Scribe and practicing psychiatric nurse practitioner. Published August 2026.

90833 vs 90836 vs 90838: picking the tier - OneStep Scribe

Three codes, one variable. 90833, 90836 and 90838 describe the same service, psychotherapy delivered alongside a separate evaluation and management visit, at three lengths. What moves you between them is how many minutes of psychotherapy you actually did.

That sounds like the easiest decision in psychiatric billing. In practice it is reliably botched, because the minutes clinicians count are the wrong minutes: the whole appointment, the record review before the patient walked in, the note written at nine that night. None of that belongs in the add-on.

This guide is about the choice. For a line by line annotated note on the most common pairing, read the 90833 note example, and for whether a 99214 and a 90833 can sit on the same claim, see billing 99214 with 90833. Here we decide which tier the visit earned.

The family at a glance

All three are reported only in addition to a qualifying evaluation and management service, performed and documented by the same clinician in the same encounter.

CodePsychotherapy time reportedReference time published with the code
9083316 to 37 minutes30 minutes
9083638 to 52 minutes45 minutes
9083853 minutes or more60 minutes

CPT is a registered trademark of the American Medical Association. Descriptors here are paraphrased, and the current code set carries the official language.

Coding conventions change with each annual code set, and payers and state Medicaid programs publish their own policies on who may report these codes and what documentation they expect. What follows is the general framework, not your contracts. Verify against your own payer policies and compliance resource before changing how you bill.

The bands in the second column are published with the codes rather than derived here. Two things set them. Psychotherapy of less than 16 minutes is not separately reported, which fixes the floor. Each upper boundary then sits halfway between adjacent reference times: halfway between 30 and 45 is 37.5, so 90836 begins at 38, and halfway between 45 and 60 is 52.5, so 90838 begins at 53. The arithmetic is only a way to remember the published ranges.

Fourteen minutes of supportive work is real clinical work, but with no add-on to report it folds into the E/M and appears nowhere else on the claim.

The word "add-on" is doing real work here

These three codes cannot stand alone. Several consequences follow, and each shows up in denials:

Not every E/M code qualifies as the primary service. The add-ons attach to office and other outpatient visits, hospital inpatient and observation care, consultations, nursing facility services, and home or residence services. They do not attach to emergency department E/M codes, and prolonged service time add-ons are out as well, since the E/M cannot be leveled by time when an add-on is reported. Check the pairing list in your code set before building a claim outside the office.

90785 for interactive complexity is the add-on most often reported alongside this pair, when the specific complicating factors are present and documented. Other same-date services, such as brief screening instrument administration and scoring, follow their own rules and payer policies.

How the add-on time is counted

Add-on time is psychotherapy time only: time in direct, real-time contact with the patient delivering psychotherapy, in person or over an approved interactive telehealth connection, separate and distinct from the work that constitutes the E/M service. It is not the length of the appointment slot.

Counts toward the add-onDoes not count
Direct, real-time psychotherapy using a recognized therapeutic approachHistory taking for the medical evaluation, review of systems, medication history
Structured intervention: cognitive restructuring, exposure planning, behavioral activation, skills coaching, psychodynamic workMental status examination performed as part of the E/M
Processing the patient's response to that intervention in sessionDiscussion of medication risks, benefits and side effects, and consent
Assigning and reviewing between session work that is part of the therapyReviewing labs, outside records or prior notes
Note writing, ordering, prescribing, care coordination, calls after the visit
Time the patient spent waiting, on forms or screeners

Two things follow. First, the medication conversation is E/M work. A twenty minute visit in which fifteen went to a cross taper discussion is a good visit, and it does not contain fifteen minutes of psychotherapy.

Second, screening instruments do not buy add-on time. The patient's own time completing a PHQ-9 or an AUDIT-C is neither E/M nor psychotherapy time. Your review and interpretation of the results is E/M work, not add-on time, and whether administration and scoring is separately reportable at all depends on payer policy. Unit limits per date are common, and separately reported scoring time cannot also count toward E/M total time.

Time is counted the same way in a video visit as in the office: what changes is the place of service, the modifier and the technology attestation, not the arithmetic. What varies is coverage: whether a plan pays a psychotherapy add-on delivered by video, and whether an audio-only encounter qualifies, is a payer-by-payer and state-by-state question that has changed more than once. Check the policy before the visit, not after the denial. Mechanics are in telehealth psychiatry documentation.

Select the E/M level by medical decision making, never by time

E/M levels in most categories can generally be chosen either by total time on the date of the encounter or by medical decision making. The moment you report a psychotherapy add-on, that choice collapses to one option.

The logic is simple. The minutes you assign to psychotherapy are already accounted for by the add-on, and you cannot spend them again on a time based E/M level. So the E/M level rests on medical decision making: the problems addressed and their complexity, the data reviewed and analyzed, and the risk from the management options considered.

That makes the therapy tier and the E/M level independent. A 90838 does not push you to a 99215, and a 16 minute 90833 does not hold you down to a 99213. I have made both errors myself, in the same month. Two poorly controlled chronic illnesses and a medication change carry the same medical decision making whether the psychotherapy ran eighteen minutes or fifty five.

The costliest mistake in this family is recording the appointment length as the psychotherapy time. A note that says "50 minute session" for a fifty minute slot that also held a medication review, a risk assessment and a refill discussion does not support 90836. It shows fifty minutes of combined work with no boundary between the services, and that boundary is what a reviewer commonly asks about.

What the note has to show

Two services means two footprints, and the note should show where one ends and the other begins. On the E/M side, the reader should see the problems addressed, the data considered, the risk weighed and the decision made, with the reasoning visible rather than asserted.

On the psychotherapy side, four elements carry the weight:

  1. Time. Total psychotherapy minutes, stated explicitly. Start and stop times are stronger, and some payer policies expect them.
  2. Modality. Name the approach. "Supportive therapy provided" is thin. "CBT, cognitive restructuring of catastrophic appraisals" is not.
  3. Target and content. What the intervention aimed at, tied to a goal in the treatment plan. Measurable goals make this paragraph easier to write.
  4. Response. What the patient did with it in the room, and what comes next.

FICTIONAL EXAMPLE NOTE, EXCERPT ONLY

Details below are invented for illustration and do not describe a real patient.

CHIEF COMPLAINT

Follow up for recurrent depression and panic attacks. "The mornings are still the worst part."

HISTORY OF PRESENT ILLNESS

Marcus is a 42 year old man with recurrent major depressive disorder and panic disorder, seen at four week interval. Sertraline 100 mg daily, adherent. Sleep improved to six hours nightly. Two panic attacks since last visit, both on the highway, down from five. Denies suicidal ideation, plan or intent. No new medical problems or medications from other prescribers. Mild delayed ejaculation, tolerable, no change wanted for this reason.

ASSESSMENT

Recurrent major depressive disorder, partial response, improving. Panic disorder with situational avoidance of highway driving, now the primary functional limitation. Sexual side effect mild, patient declines intervention. A sertraline increase would target residual panic frequency at the cost of a side effect he is willing to live with, and the four week trajectory argues for holding the dose.

PLAN

Continue sertraline 100 mg daily. Dose increase deferred given ongoing improvement and the side effect profile. Discussed increasing to 150 mg next visit if panic frequency plateaus, including the likelihood of worsening the sexual side effect. Return in four weeks. Reviewed when to call the office sooner, and reviewed use of the 988 Suicide and Crisis Lifeline between visits.

PSYCHOTHERAPY

Psychotherapy time 41 minutes, from 10:14 to 10:55, separate and distinct from the evaluation and management service documented above. These minutes were not counted toward that service.

CBT with an exposure focus. Built a driving exposure hierarchy with Marcus, from low traffic surface streets through a two exit highway segment. Identified his central catastrophic appraisal, that a panic attack while driving means losing control of the car, and tested it against his history of eleven attacks with no loss of control. He resisted the reframe at first, then generated the counterevidence himself. Screened for sedation, syncope and driving safety before assigning in vivo work, and Marcus agreed to the plan and to stopping if symptoms escalate. Assigned two surface street exposures with a brief log of anticipated and actual anxiety. Target goal: independent highway driving to work by the twelve week review.

Why this matters. The note draws the line itself. The psychotherapy carries a start, stop and total plus an explicit statement that those minutes were not counted twice, and the therapy paragraph names a modality, a target, the response in the room and a next step. At 41 minutes the example sits in the 90836 band. It names no E/M level on purpose: that depends on the whole record, and the support here would come from the documented deliberation about the dose increase and the side effect trade-off, not from a phrase reciting decision making elements. The exposure work is illustrative: what is safe to assign, and what state rules govern driving fitness, is a clinical judgment, not something to lift from an example.

Why 90838 draws more scrutiny

Reporting 90838 asserts at least 53 minutes of psychotherapy on top of a separately documented evaluation and management service. That puts the encounter well past an hour, and anyone reviewing a set of claims can do the arithmetic.

In my experience, and in the payer policies I have read, the patterns that attract questions are:

None of that makes 90838 improper. Long sessions happen, especially in trauma focused work, early phase treatment and complex cases. It does mean the note should look like an hour of work: distinct content, a named modality, a documented response, a plan that has moved. If the same note would be equally true of a 25 minute session, the tier is exposed.

The longest sessions also tend to be the ones where risk is on the table. If risk is present, the note should show what you assessed, what you concluded and what safety information you gave, including the 988 Suicide and Crisis Lifeline.

A decision path you can run in thirty seconds

  1. Was there a separately identifiable E/M service? If no, stop. You are in the 90832, 90834 or 90837 family.
  2. Time the psychotherapy directly. Note when it started and when it stopped. Without that, subtracting the E/M work from the appointment gives a ceiling rather than a total, and anything that was neither E/M nor structured psychotherapy comes out too.
  3. Is what remains at least 16 minutes of psychotherapy? If no, report the E/M alone, selected by medical decision making or by total time on the date of the encounter, counting only activities that fall within the E/M time definition.
  4. Place the minutes in a band. 16 to 37 is 90833. 38 to 52 is 90836. 53 or more is 90838.
  5. Select the E/M level by medical decision making, independently, and write the minutes into the note before you close it.

The habit that makes step five painless is writing the start and stop time in the room rather than reconstructing later. Reconstructed minutes drift upward. Mine did, until I started writing the stop time before the patient left.

When the standalone family is the right answer instead

If the encounter was psychotherapy and only psychotherapy, the add-on family is the wrong tool no matter how long the session ran. That is what 90832, 90834 and 90837 exist for, and their bands describe the entire service rather than a piece of it.

The most common version I run into is a prescriber who does therapy for the full session, glances at the medication list, changes nothing, and bills an E/M with a 90836. A glance at an unchanged medication list will generally not, on its own, support a separately identifiable evaluation and management service. The question is whether medically necessary medical work was done and documented, not whether a medication list was opened.

The reverse error is also common: a bifurcated visit reported as a single standalone therapy code, understating the medical work. Both come from not deciding at the top of the visit what kind of visit it is.

Frequently asked questions

What if the psychotherapy ran 15 minutes?

No add-on. The work folds into the E/M service, and the E/M level may then be selected by medical decision making or by total time on the date of the encounter, whichever the record supports. Be careful what goes into that total: whether psychotherapy minutes count toward E/M total time when no add-on is reported is not settled and payer policies differ, and a note describing psychotherapy while the E/M is leveled by time is a pattern some reviewers look at closely. Counting only E/M activities is the conservative approach.

Do I have to document start and stop times, or is a total enough?

A stated total is the minimum, and it should be a number, not a range and not "approximately 45 minutes." Start and stop times are stronger, and some payer policies ask for them. Since time is all that distinguishes these three codes, I write both.

Can I use these add-ons with an intake visit?

Not with 90791 or 90792. Those diagnostic evaluation codes are their own service and the add-ons do not attach to them. If your intake was billed as an E/M service instead, and you delivered separate and distinct psychotherapy in the same encounter, the add-on rules apply normally. See 90791 vs 90792 for how the intake codes differ.

What happens if a claim comes back denied?

Read the denial reason before rewriting anything. Add-on denials usually trace to a handful of causes: no E/M line on the claim, an incompatible primary code, a missing modifier 25 where the plan requires one even though the code set does not, frequency or utilization edits aimed at the longer tiers, a modifier or place of service problem on a telehealth claim, or a plan policy about who may report psychotherapy. Documentation quality drives the outcome of a review rather than an automated edit, and some plans run prepayment documentation review on the longer tiers.

Related guides

OneStep Scribe drafts this documentation automatically.

OneStep Scribe is an AI scribe built for psychiatric prescribers. It listens to the visit, in person or over telehealth, and drafts the complete note from what occurred in the visit, including the telehealth details and a billing section, for your review and signature. Every account is NPI-verified.

Start a 14-day free trial

This article is educational and reflects one clinician's understanding of coding rules at the time of writing. It is not legal, billing, or payer-specific advice, and it does not guarantee reimbursement or audit outcomes. Coding rules, payer policies, and state and federal laws change; always verify current requirements with your payers, coding resources, and compliance advisors, and consult your licensing board(s), current DEA guidance, and a healthcare attorney for licensure, consent, and controlled-substance questions. CPT is a registered trademark of the American Medical Association.