Documentation cluster
Psychiatry notes carry constraints that general medical notes don't. The mental status exam has its own grammar. Suicide risk documentation has its own legal weight. Controlled-substance prescribing has its own audit trail. Mental-health E/M decision-making turns on factors most generalist templates ignore — chronic risk, longitudinal medication response, the difference between an acute decompensation and a chronic course flaring. Boilerplate doesn't survive a chart review. This page covers the note types psychiatrists actually write, the structural elements auditors and plaintiff attorneys actually look for, and the points in the workflow where Sigmund — sitting in, listening, drafting — earns his keep without overstepping the clinician's signature.
The MSE is the part of the note that signals to any reader — auditor, consultant, attorney — that a psychiatric evaluation actually happened. A copy-forwarded MSE that reads the same every visit is one of the most common reasons a note gets flagged on review. The exam is a structured cross-section of how the patient presented in the room: appearance, behavior, speech, mood, affect, thought process, thought content, perception, cognition, insight, and judgment. Most of those domains carry standard descriptors, but each visit should reflect what was actually observed, not what was observed three visits ago.
The minimum viable MSE for an outpatient psychiatry visit covers appearance, behavior, speech, mood, affect, thought process, thought content (with explicit suicidal and homicidal ideation status), insight, and judgment. Perception and cognition are situational — a stable patient on a maintenance visit may not warrant a formal MMSE, but a confused or acutely psychotic patient absolutely does. What auditors look for is internal consistency: the affect descriptor should match the mood descriptor or the discrepancy should be addressed. The thought-content section should explicitly document the absence of SI/HI rather than leaving the field blank. Cognition descriptors should be tied to observation, not asserted.
This is the section where passive AI capture is most useful. A scribe that listens during the visit can populate observable domains — speech rate and rhythm, thought process linearity, affect range — without the clinician retreating to the keyboard. The errors that get notes flagged are predictable: copy-forward without observation, conflicting affect and mood descriptors, and missing SI/HI status. A good capture system surfaces those at draft time so the clinician fixes them before signature.
The psychiatry progress note is a SOAP-variant. The subjective is denser than in general medicine — interval history, sleep, appetite, substance use, adherence, side effects, life stressors, response to last medication change. The objective is the MSE. The assessment is where psychiatric decision-making becomes visible, and it's the section payers and auditors read first. The plan ties what was assessed to what was changed.
Medical decision-making (MDM) for psychiatry, under the 2026 E/M conventions, turns on three axes: the number and complexity of problems addressed, the amount and complexity of data reviewed, and the risk of complications from management. A maintenance medication visit for stable, single-diagnosis depression sits in low-to-moderate complexity territory. A visit that addresses suicidality, a recent psychiatric hospitalization, or a controlled-substance dose change with side effects climbs to high complexity quickly. Time-based billing remains a viable alternative in 2026, and for psychiatry — where uninterrupted time with the patient is often the actual clinical work — it frequently produces a more honest level of service than MDM-based coding. Document total time, document what was done during that time, and the level follows.
The most common documentation failure here is a plan that doesn't trace back to the assessment. A medication increase needs a clinical "why" written next to it. An auditor reading the note should be able to follow a straight line from interval history to MSE to assessment to plan without filling in inferences. This is the structural property that Sigmund drafts toward by default.
Medication management notes are the highest-volume note type in outpatient psychiatry and the one most prone to documentation drift. The required elements are well-defined: target symptoms (what the medication is being used for, in this patient), response to date (with specifics — not "doing well"), side effects (asked, not assumed), adherence (asked, not assumed), and the plan with rationale. Every one of those elements has to appear, and the rationale has to be visible enough that someone other than the prescriber can reconstruct the clinical reasoning.
Common audit failures are predictable. "Tolerating well" with no symptom-level data. Side effects documented as "none" when the section was never asked. Dose changes without a clinical trigger written next to them. A medication continued for two years with no recurring documentation of why it's still indicated. Each of those, in isolation, is a minor issue; cumulatively, they are the pattern that opens a clinician to recoupment on a chart review.
This is the note type where evidence-grounding earns its keep. A dose increase from sertraline 100 to 150 mg should be accompanied by a documented reason — partial response on PHQ-9, residual anhedonia, side-effect-limited at higher doses. An AI scribe that drafts the dose change without surfacing the rationale is producing a note that ages badly. The evidence-grounding architecture that drives Sigmund's recommendations is the same architecture that ties each plan-line to its clinical justification at draft time.
Suicide risk documentation is the single section of a psychiatry chart most likely to be subpoenaed. When a patient dies by suicide, the plaintiff's expert reads the risk section first — not to find a checkbox, but to find reasoning. A note that documents a C-SSRS score of 0 and nothing else, on a patient with active depression and recent stressors, is a note that will not hold up. A note that documents chronic and acute risk separately, names the modifiers driving the current acuity, identifies protective factors, and ties the disposition to the risk level — that note will.
The required structural elements are: chronic baseline risk (demographic, diagnostic, historical), acute risk modifiers (current ideation, plan, intent, recent attempt, recent loss or stressor, intoxication, access to means), protective factors (treatment engagement, social connection, reasons for living, future orientation), and the disposition rationale — explicitly why outpatient management, hospitalization, increased contact, or means restriction is the right call given what was just documented. C-SSRS and SAFE-T language are useful scaffolding, but they are not a substitute for the reasoning. A template alone doesn't fix the malpractice exposure; the visible clinical reasoning does.
Clinicians at Integrative Psychiatry Manhattan have built their risk documentation conventions around exactly this principle — structured scaffolding plus visible reasoning — and the lesson generalizes. The malpractice exposure is real, and templates that flatten the section into a score collapse the part of the note that actually defends the clinician.
Controlled-substance prescribing carries its own documentation layer on top of the standard progress note. For every scheduled medication — stimulants, benzodiazepines, z-drugs, controlled mood stabilizers — the chart should contain the diagnostic justification (the indication, not just the ICD code), the response to the current dose, side effects, a screen for misuse or diversion concerns, and a documented review of the state prescription monitoring database. In New York, that means the I-STOP/PMP check at each controlled-substance prescription, documented in the chart with date and finding. The check itself takes seconds. The documentation of the check is what makes the prescription defensible.
The chronic-stimulant-use audit trap is the most common failure mode. A patient stable on 30 mg of mixed amphetamine salts for six years, with notes that say "continue current regimen" and nothing else, is a chart that will not survive a payer audit or a DEA inquiry. Each controlled-substance visit needs a fresh documented justification: target symptoms still present, functional benefit specific to this patient, side effects monitored, alternatives considered or contraindicated. A dose change without cited rationale is the documentation equivalent of an unsigned check.
This is the second note type — alongside medication management — where evidence-grounding matters most. Stimulant dose changes age into the record. Two years later, in a peer review or a board inquiry, the rationale either exists on the page or it doesn't. The clinical perspective that informs Dr. Ryan Sultan's research on ADHD pharmacoepidemiology — including the longitudinal stimulant data published from the Sultan Lab for Mental Health Informatics — sits directly on top of this documentation problem.
Sigmund is a psychiatry-specific scribe. He sits in during the visit, listens, and drafts the note structurally, by section, with the documentation conventions above built into the draft. The signature line stays empty until the clinician signs it.
The full product picture lives on the Sigmund homepage. The architecture that ties every line to its evidence is on the evidence-grounding page. Related cluster reading: clinical decision support in psychiatry and how Sigmund compares to generic ambient AI scribes.
Sigmund is investigational and intended to assist — not replace — clinical judgment. Research and development conducted at the Sultan Lab for Mental Health Informatics, Columbia University Irving Medical Center, and the New York State Psychiatric Institute.
