Guide · Clinical documentation
Documentation burden is well evidenced in medicine, thinner in behavioral health, and rarely framed as what it also is: a fidelity problem. Notes written from memory lose the session’s physiological record.
SeroState is an evidence layer delivered to EHR and telehealth platforms by API — not an app a clinician installs.
The strongest documentation-burden data comes from time-motion studies outside behavioral health. Direct observation of 57 ambulatory physicians across 430 hours found 49.2% of the office day on EHR and desk work against 27.0% in direct clinical face time — nearly two hours of documentation for every hour with a patient, plus one to two hours nightly (Sinsky et al., 2016)[1].
That last row matters. In clinicians treating PTSD, higher administrative hours were among the baseline predictors of burnout at twelve months[5], and a systematic review of psychotherapist burnout points to work-related conditions rather than personal deficiency[2]. Among 5,341 VHA therapists, institutional support for evidence-based practice was associated with less burnout independent of workload[6].
Widely circulated figures putting behavioral-health documentation at a precise share of the working week generally trace to vendor surveys rather than peer review. We cite the observational literature and mark the gap instead.
Pick a failure mode. Each is a different reason the record diverges from the encounter.
Notes written hours after a session rest on reconstruction. Sequence blurs, detail drops, and inference fills the space. In behavioral health the losses are clinically loaded: autonomic arousal during a difficult disclosure is not recoverable from memory.
The measurement was available during the session. It simply was not recorded.
Transcription captures what was said. It does not capture what the body did while saying it. A patient says “I’m fine” and the transcript is accurate — and the physiological response in that same window is simply absent.
Accurate words, incomplete record. See the segment below.
When the record rests on recall, the defensive response is to write more. Length grows, review time grows, and objective corroboration does not — because narrative cannot supply what was never measured.
Coding and billing decisions always remain with the licensed provider. See reimbursement readiness.
Illustrative. A transcript preserves the sentence; the flag, the trace behind it, and a confidence score arrive in the post-session payload.
SeroState calls this divergence the say/body gap: a reviewable moment where words and physiology disagree. The reference definition →
Notably absent from that list: a claim that any tool writes the note for the clinician. How the platform delivers evidence →
Session evidence is the measured physiological and vocal record of a consented encounter, time-aligned to the conversation and returned after the session as structured evidence: flags, the evidence segments behind them, and confidence and signal-quality scores. Heart rate is the primary signal, vocal markers come from session audio, and heart-rate variability and electrodermal activity join the timeline when the session wearable supplies them.
It does not write the note. The clinician still authors the impression — now with measured corroboration attached to it rather than recall alone. Evidence reaches the chart through the EHR or telehealth platform via the QOAX API, and the organisation holds the PHI. Security architecture →
We publish no accuracy or time-savings numbers. Validation is blinded comparison against expert clinical judgment, reported by subgroup, pre-registered and published before any claim. The scientific foundation →
No. Scribes transcribe and summarise what was said. Session evidence is the physiological and vocal record of the encounter, time-aligned to the conversation and delivered as structured FHIR R4 evidence. It does not author the clinical note.
We do not publish time-savings figures. SeroState adds objective evidence to the record; any effect on workflow depends on how the platform implements it. We pre-register and publish before making outcome claims.
No. Every output is a reviewable flag with the trace and a confidence score behind it. The clinician authors the impression, can dismiss any flag, and always decides.
Nothing. Capture is passive during the consented encounter and all processing happens post-session, so nothing competes with therapeutic attention in the room.
Through their EHR or telehealth platform. SeroState is an API-delivered evidence layer for platforms, health systems, research groups, and trial sponsors — not a standalone app a clinician installs.
Session Physiology, defined → AI scribe vs session evidence → Talk to us about your platform →
Time-motion evidence [1] is drawn from family medicine, internal medicine, cardiology, and orthopedics — not behavioral health. It is cited for the scale of documentation burden in medicine generally, not as a behavioral-health measurement.