Psychiatry Medication Management Visits and AI Scribes: What Actually Gets Captured

A 20-minute medication management visit looks simple from the outside. It is not. You are covering medication response, side effects, adherence, current symptom severity, psychiatric history context, risk, functional impairment, and often an updated diagnostic impression — all while billing an E&M code that auditors will later match against your documentation. If your note does not support the level of complexity you actually managed, you have either under-billed or left yourself exposed.

AI scribes are increasingly used in psychiatry — but not all of them are built for it. General-purpose scribes handle therapy progress notes reasonably well. Medication management visits require something more specific: accurate medication name transcription, structured E&M documentation, controlled substance language that meets DEA and state standards, and monitoring parameters that vary by drug class. This post covers what a properly configured AI scribe captures in a med-management visit, and where you still need to verify.

E&M Coding and Documentation Requirements

Psychiatric medication management visits are typically billed as E&M codes — 99212 through 99215 for established patients, 99202 through 99205 for new patients — rather than the add-on psychotherapy codes (90833, 90836, 90838) that require separate documentation when combined with E&M services.

The 2021 AMA E&M guidelines shifted the basis for code selection to either total time or medical decision-making (MDM), removing the requirement to document all elements of history and physical examination for established patients. This is significant for psychiatry: you no longer need to document a full review of systems and physical exam to justify a 99214. The note must instead demonstrate one of two things:

MDM at the appropriate level: number and complexity of problems, amount of data reviewed, and risk of complications or morbidity
Total time documented: all time personally spent on that date of service, including review, documentation, and care coordination, not just face-to-face time

An AI scribe that captures the session but does not understand MDM-level documentation leaves the clinician to retrofit the note structure afterward — which defeats much of the efficiency gain. Scriber's psychiatry template explicitly prompts for problem complexity, data reviewed (labs, prior records, external consultation), and risk tier, and generates the MDM table structure that supports the appropriate code.

For a 99214 on MDM: at least one problem with moderate complexity, or two stable chronic conditions. For 99215: high-complexity MDM, which requires at least two of — two or more chronic illnesses with exacerbation, drug therapy requiring intensive monitoring, or decision to hospitalize.

Document the reasoning, not just the decision.

Controlled Substance Documentation

Controlled substance prescriptions — Schedule II through V — carry documentation requirements beyond what standard medication management notes require. Most state medical boards and the DEA specify that the record must reflect:

Indication: the diagnosis or condition being treated, linked explicitly to the medication
Dose and frequency: as prescribed, not as "discussed" or "as before"
Rationale for continued use: especially for Schedule II stimulants and opioids; documentation of ongoing benefit and periodic reassessment
PDMP check: most states require documentation that the Prescription Drug Monitoring Program was queried before prescribing a controlled substance; many require a note entry
Patient agreement: for Schedule II long-term prescriptions, documentation that the patient was counseled on risks, including diversion
Monitoring parameters: UDS frequency, pill counts, behavioral observations relevant to misuse risk

An AI scribe cannot run your PDMP query. What it can do is capture your verbal attestation — "I reviewed the PDMP today, no concerns" — and place it in the correct section of the note. If your workflow includes a verbal statement as part of the session summary, the scribe will capture it. If it does not, add it as a spoken tag before ending the recording.

For stimulant prescriptions specifically — Adderall, Vyvanse, Ritalin, and their generics — auditors look for documented ADHD severity, functional impairment justification, and evidence of response monitoring. A generated note that covers medication name, dose, therapeutic response, side effects, and a plan for next-visit reassessment will support the record; one that says only "continued on current regimen" will not.

Medication Accuracy: The Highest-Stakes Accuracy Problem in AI Transcription

Clinical terminology accuracy is the single most important capability for a psychiatry AI scribe, and it is where general-purpose engines show their limits.

Psychiatric medication names are phonetically similar in ways that create dangerous transcription errors:

Bupropion vs. buspirone (psychopharmacologically very different)
Lamotrigine vs. lorazepam (different drug classes, different routes)
Quetiapine, quietapine, and "kwet-eye-a-peen" (phonetic spellings of the same drug)
Clonazepam vs. clonidine (Schedule IV benzo vs. alpha-2 agonist)

Speechmatics and Azure Speech, both trained on medical vocabulary including psychiatric drug names, perform significantly better on this class of error than general-purpose engines. In production use, clinicians using Speechmatics for medication management visits report needing fewer medication-name corrections than those using consumer transcription engines.

The practical workflow: at the end of the session, read back the medication name and dose aloud as part of your summary — "I am continuing lithium 600mg twice daily" — rather than assuming the engine parsed it from a conversational exchange. Explicit restatement dramatically reduces medication transcription errors across all engines.

Verify every medication name in the generated note before signing. This takes 30–60 seconds for a medication management note and is not optional.

Side-Effect Monitoring and Psychiatric Lab Work

Medication management visits regularly include monitoring that needs to be documented: lithium levels, metabolic panels for atypical antipsychotics, thyroid function, LFTs for valproate, QTc intervals for antipsychotics with cardiac risk. The documentation of monitoring serves two purposes — clinical continuity and billing justification.

Under the MDM framework, ordering or reviewing labs is a data point that increases the complexity level. If you reviewed a lab result in preparation for the visit, say so explicitly in the session: "I reviewed the lithium level from last Tuesday, it was 0.7, which is in range." The AI scribe will capture that statement and the note will reflect the data review.

For metabolic monitoring of second-generation antipsychotics — weight, blood pressure, glucose, lipid panel — the note should document the values if obtained and the clinical interpretation. "Weight stable at prior measurement; no metabolic concerns at this time" is better than leaving the section blank. Blank sections in an atypical antipsychotic follow-up note raise questions at review.

Tardive dyskinesia screening (AIMS exam) should be documented at the frequency your practice protocol specifies — typically every 6 months for patients on dopamine-blocking agents. If you performed an AIMS screen, verbalize the finding; the AI scribe captures it. If the screen was not performed this visit, the note can reflect that it is deferred to the next scheduled visit with a date.

Follow-Up Note Structure for Psychiatry

Medication management follow-up notes differ from therapy progress notes structurally. They are closer to primary care E&M visits than to behavioral health progress notes — shorter, denser, and organized around medication rather than therapeutic process.

A well-structured follow-up note for psychiatry covers:

Current medications and doses — every psychiatric medication, dose, frequency, and changes from last visit. Not "continued on current regimen" but a full list.

Medication response — patient's self-report of efficacy, with symptom severity language that corresponds to the diagnosis (PHQ-9 score, sleep duration, mood rating, concentration, energy — whichever is clinically relevant).

Side effects — active inquiry, not just patient-reported. Document what you asked and what the patient reported. Absence of side effects is documentation, not silence.

Adherence — especially for medications with significant consequences for missed doses (SSRIs, lithium, antipsychotics). "Patient reports taking medication as prescribed" is defensible. So is "patient endorsed difficulty with morning dose, discussed split dosing."

Objective assessment — MSE abbreviated to clinically relevant findings. For medication management, this typically covers appearance, affect, mood per patient, thought process, insight, and judgment. Not every element is required when the presentation is stable, but risk assessment is always required.

Plan — changes made, refills authorized, follow-up interval, and any referrals or coordination (PCP, therapist, labs).

Scriber's psychiatry template generates these sections from the session transcript. The follow-up interval is captured from your verbal closing: "I want to see you back in four weeks." The plan section consolidates medication changes and next steps spoken during the session.

What AI Scribes Cannot Do in a Med-Management Workflow

Being clear about this matters.

An AI scribe will not calculate your MDM level for you — that is a billing decision that requires clinical judgment about complexity and risk. The note can provide the evidence; the code selection is yours.

It will not run your PDMP query, flag a drug interaction, or check your patient's insurance formulary. It does not have access to your EHR in real time. It generates a note from what was said in the session.

It will not catch a medication name error caused by mumbling or background noise — which is why explicit restatement of medications at the end of the session is the correct workflow, not a workaround.

It will not apply state-specific controlled substance documentation requirements unless those requirements have been built into the template. If you practice in a state with specific language requirements for Schedule II prescriptions — and several states do — build that language into your custom template.

The scribe is a documentation accelerator, not a compliance system. The clinical and billing judgment remains yours.

Workflow for Psychiatry Medication Management Visits

A practical sequence that works in a 15–20 minute visit:

Open the session recording at the start of the visit, not after.
State the patient's diagnosis and the medications being reviewed — once, clearly — at the opening. ("Today we're following up on the patient's bipolar I disorder and reviewing lithium and lamotrigine.")
Conduct the visit normally. The scribe captures the full exchange.
Before closing, do a verbal medication summary: name, dose, changes, and refill interval.
State your plan: "I am going to continue the current regimen, labs due in six weeks, follow up in four weeks."
End the recording.
Review the generated note: verify medication names, check MDM documentation, confirm risk statement is present.
Sign.

Most clinicians doing this workflow report review time of 3–5 minutes per medication management note. At 8–10 med management visits per half-day session, that is 24–50 minutes of documentation time instead of 60–90 minutes of typing.

For more on psychiatry-specific templates and workflows, see Scriber for Psychiatry and the E&M coding documentation guide. For a side-by-side comparison of AI scribe engines on clinical terminology accuracy, see transcription engine comparison.

Reduce documentation time on medication management visits without sacrificing E&M support. Start a free trial at Scriber or review pricing for psychiatry practices.