
CLINICAL MEMORY
Longitudinal Patient Memory
Find 40+ prompt packs at templates.scribing.io to draft psychiatric notes in seconds and trend PHQ-9 and C-SSRS scores across visits.
Specialty Clinical Playbook
Defensible 90833/90836 documentation with discrete MSE, C-SSRS risk formulation, and G2211 capture. Book your audit at https://cal.com/merryai/demo.
Specialty Architecture
Engineered to mirror the pacing, diagnostic frameworks, and documentation requirements of Adult & Geriatric Psychiatry.

CLINICAL MEMORY
Find 40+ prompt packs at templates.scribing.io to draft psychiatric notes in seconds and trend PHQ-9 and C-SSRS scores across visits.

CONTEXT RETRIEVAL
Retrieve prior MSE domains, medication reconciliation history, and risk trajectories in one structured draft without tab bouncing.

WORKFLOW INTELLIGENCE
Capture longitudinal complexity add-on revenue for continuing psychiatric focal-point care. Claim your 15-Minute Workflow Audit today.

SPECIALTY-AWARE REASONING
Diarize clinician, patient, and caregiver dialogue and map findings into discrete E/M versus time-based psychotherapy note sections.
Point-of-Care Flow
Zero IT friction, zero complex API setup, and human-verified attestation on every note.

Documentation in adult and geriatric psychiatry carries a structural burden that few other specialties share: a single same-day encounter often contains two clinically distinct services that must be defended independently. A psychiatric evaluation and management (E/M) visit reported alongside psychotherapy add-on codes 90833 or 90836 must demonstrate a fully supported E/M service—anchored by a discrete mental status examination—plus a separate, time-based psychotherapy service with its own modality, interventions, and minute count. Merry AI is engineered to keep these two data streams parallel and never blended.
The compliance risk in this specialty is not merely under-documentation; it is over-attribution. An ambient scribe that infers suicidal ideation from a diagnosis history, or implies subtle psychosis where none was verbalized, manufactures audit and liability exposure. Our anchor discipline is absolute: document only what was observed, measured, or verbalized, and clearly separate patient report, standardized instrument results, and clinician interpretation.
Consider the geriatric encounter specifically, where reliability is frequently constrained by neurocognitive decline. The system flags when collateral information from a caregiver is the source of a finding, and prompts the clinician to state reliability limits explicitly rather than allowing the draft to present caregiver interpretation as patient-reported fact.
| Diagnostic Framework | Required Clinical Data Points | Billing Evidence |
|---|---|---|
| Psychiatric E/M (99213–99215) | Chief complaint, HPI, discrete MSE domains, medication reconciliation, MDM | E/M selected by elements not time; Modifier 25 appended |
| Psychotherapy 90833 | Modality, interventions, therapeutic focus, patient response, 16–37 min | Start/stop time distinct from E/M activities |
| Psychotherapy 90836 | Modality, interventions, patient response, ≥38 min face-to-face | Time log excluding history, exam, MDM |
| C-SSRS Risk Stratification | Items endorsed, assigned risk level, protective/dynamic factors | Formulation tied to instrument criteria in MDM |
| PHQ-9 Severity | Total score, item 9 response, administration date | Score justifies treatment intensity decisions |
| G2211 Complexity | Continuing focal-point care documentation | Longitudinal complexity add-on per CMS |
The mental status examination is an E/M exam element, not psychotherapy content, and payer guidance consistently expects it to live in the objective section of the note, distinctly separated from the psychotherapy narrative. Merry AI routes observed and verbalized findings into discrete MSE domains: appearance and behavior, speech, mood and affect, thought process, thought content, perceptions, and cognition. Each domain is populated only from what the clinician actually observed or the patient actually stated.
Thought content documentation demands particular restraint. When a patient denies suicidal ideation, the note records “no suicidal ideation reported” rather than a hedged inference. When hallucinations are denied, the draft states “no hallucinations reported or observed” and does not imply latent psychosis on the strength of a prior diagnosis. This mirrors the AMA and CMS expectation that the psychiatric exam be a factual record, not a speculative one.
Cognitive status carries outsized weight in the geriatric MSE, where orientation, attention, memory, insight, and judgment inform capacity and safety decisions. The system captures evidence relevant to dementia, delirium, or neurocognitive disorders only when those domains were explicitly assessed, and it annotates the limits of reliability when advanced impairment constrains the examination. Collateral sources are labeled as collateral, never merged into the patient's own verbalized history.
Medication review anchors the medical decision-making component in psychiatric E/M, and the draft documents reconciliation, adherence, side effects, and adjustments as a distinct block. Treatment decisions are then linked explicitly to the risk formulation—for example, an intensified antidepressant dose and weekly follow-up justified by a documented PHQ-9 of 21 and C-SSRS endorsement of active ideation with plan.
Validated instruments carry the risk narrative in modern psychiatric practice, and Merry AI treats the PHQ-9 and Columbia-Suicide Severity Rating Scale (C-SSRS) as source-of-truth data rather than raw material for interpretation. The PHQ-9 total, the item 9 response, and the administration date are recorded verbatim, and any discrepancy between the score and the clinical presentation is surfaced for the clinician to reconcile.
The C-SSRS demands item-level fidelity, so the note records precisely which items the patient endorsed—“wish to be dead,” “active ideation with plan”—and assigns the risk level that the instrument's own framework supports. The system refuses to generate a bare “moderate suicide risk” statement unless it is tethered to the endorsed items and the clinician's documented formulation, preserving a transparent, non-speculative chain of reasoning.
Protective and dynamic factors round out the formulation, and for cognitively impaired patients the draft explicitly names the reliability constraints and the collateral sources consulted. When a higher-risk decision such as urgent hospitalization is made, the note enumerates the exact data points—prior attempts, absent protective factors, instrument endorsements—so the decision reads as evidence-grounded rather than intuitive. Practices seeking the underlying guideline structure can reference the AMA CPT resources at https://www.ama-assn.org/practice-management/cpt.
The two-service split is where claims are won or lost. Because 90833 and 90836 are add-on codes that cannot stand alone, the note must render two distinct, individually necessary services on the same date. Merry AI selects the E/M level from history, discrete MSE, and MDM elements—never from time—while reserving time exclusively for the psychotherapy code selection. Time spent on history, examination, and decision-making is affirmatively excluded from the psychotherapy minute log.
The psychotherapy section stands on its own, capturing the modality (CBT, interpersonal therapy, supportive therapy, behavioral activation, grief work), the specific interventions delivered, the therapeutic focus, the patient response, and the exact start and stop times. A hard stop fires when psychotherapy time falls below sixteen minutes for a claimed 90833, or below thirty-eight minutes for a claimed 90836, protecting the practice from a time-unsupported add-on. Explore ready-to-deploy split-note templates and Access Specialty Prompts at templates.scribing.io.
Modifier 25 is the connective tissue of the same-day claim, signaling a significant and separately identifiable E/M service furnished alongside psychotherapy. Merry AI appends the modifier to the E/M code, bundles the appropriate add-on, and confirms that a single psychiatric diagnosis can lawfully support both services without a fabricated second diagnosis. The compliance layer verifies that a discrete MSE is present before any psychiatric E/M code is permitted to finalize.
Longitudinal complexity is frequently left on the table in psychiatry, where continuing focal-point management of serious conditions is the norm. The system recognizes G2211-eligible encounters and surfaces the complexity add-on consistent with CMS guidance, converting the ongoing therapeutic relationship into properly captured, defensible revenue. Every recommendation remains subject to human attestation before the note is signed.
The result is an audit-resistant record that satisfies the E/M element requirement, the time-based psychotherapy requirement, and the risk-documentation standard simultaneously—without a single hallucinated clinical conclusion. To see this workflow validated against your own payer mix and EHR, Schedule a 15-Minute Specialty Workflow Audit.

Bind LVEF% to NYHA class, defend high-complexity MDM, and preserve GDMT titration history in every note. Book your audit at https://cal.com/merryai/demo.