Post-session physiological evidence for behavioral health — biological ground truth, as an API.
We fuse wearable and vocal biomarkers into FHIR R4 session evidence via the QOAX API.
One obsession: post-session physiological warnings of deterioration in the record.
When the two devastating earthquakes hit Hatay, deterioration went unseen by the experts. Not because they were inexperienced: they were affected themselves, there was no time for supervision, and their determination to help meant they would not leave the disaster zone. The longer they worked without anyone watching over the work, the worse the outcomes became. Support stopped landing, and most survivors never built trust in the psychologists trying to help them.
Everyone was doing the right job. But because there was no objective measure of the care being given, the few who actually felt better were simply the ones lucky enough to build a bond with their expert. Chance is not a clinical standard.
Later I found the same gap at the top of the system. The World Health Organization recommends that mental health services be part of medical care, and that care for noncommunicable diseases integrate mental health as well. Both are the right calls, and neither comes with an objective measure of whether the care is working. There was no tool to make that possible for the people in healthcare, in mental care, or in the therapy room.
These are the reasons we left our professional careers and jumped into the ocean: to make this technology available where it touches lives, and to support experts in making their work stronger. That is why we are building SeroState.
A 2026 Nature Medicine paper audited a related blind spot at scale: nine frontier AI chatbots, 810 simulated conversations with vulnerable users. Concerning behavior was widespread — reduced in newer models, but still present — and highest when supportive replies reinforced the user’s vulnerability (Weilnhammer et al., 2026).
That is the gap we are building for: physiological grounding a clinician can review or dismiss — not an AI alone with a vulnerable person.
SeroState (serostate.com) is a behavioral health technology company. It produces post-session physiological evidence (wearable biosignal and vocal biomarkers fused into audit-ready FHIR R4 records) and delivers it via the QOAX REST API to EHR and RPM platforms.
AI scribe platforms focus on transcription and note generation. SeroState fuses voice with wearable physiology into confidence-scored evidence. It supports reimbursement documentation and audit defense; it does not replace the scribe.
Those tools converse with patients without an external signal of help or harm. Weilnhammer et al. (Nature Medicine, 2026) audited nine frontier chatbots across 810 simulated conversations: concerning behavior was widespread — reduced in newer models, but still present — and highest when supportive replies reinforced the user’s vulnerability.
SeroState does not replace the clinician with an AI conversation — we add physiological evidence for the clinician to review.
Behavioral health platforms, RPM vendors, and EHR integrators that need objective physiological evidence alongside session documentation, embedded in existing stacks via the QOAX API. Also payers, TPAs, and program integrity teams evaluating documentation quality — see program integrity.
This compares SeroState against AI scribe platforms. For a comparison against passive wearable/wellness tracking, see Platform →
| Capability | AI Scribe Category | SeroState |
|---|---|---|
| QOAX session report | ||
| Executive summary & structured clinical note | Included | Companion, not the product |
| Full diarized transcript appendix | Included | Platform-dependent companion |
| Clinician flags (risk, safety, follow-up) | Not included | Included |
| Voice NLP | ||
| Speaker diarization & session timing | Included | Included |
| Behavioral-health clinical narrative | Included | Included |
| Clinical workflow | ||
| Structured clinical note support (evidence-backed) | Limited | Included |
| Supervisor-ready evidence package for review | Not included | Included |
| FHIR R4 artifacts for EHR chart ingestion | Not included | Included |
| Session intelligence | ||
| Session topics & affect-relevant clinical analysis | Not included | Included |
| Arousal timeline & significant pause mapping | Not included | Included |
| Session statistics, speaker analysis & arousal score | Limited | Included |
| Evidence & documentation support | ||
| Wearable heart-rate arousal fusion | Not included | Included |
| HRV / EDA fusion | Not included | When present |
| Documentation support for coding corroboration (clinician decides) | Not included | Included |
| Pre-submission documentation checklist support | Not included | Included |
| Per-code criteria mapping support (e.g. 90785, 90839) | Not included | Included |
| Post-session REST API & FHIR R4 delivery | Not included | Included |
Audio-only = documentation mode. Wearable unlocks audit-defensible coding corroboration. Clinician sign-off required.
Redacted sample report, fusion engine, arousal scoring, and evidence confidence.
Published literature per signal, our validation method and honest limitations.
Publish, serve, specialize — and the same session engine for pain, rehab, and trauma counseling.
Endpoints, response shape, and integration model.
Leakage zone vs. evidence zone: Documentation gap recovery.
HIPAA, 42 CFR Part 2, and audit-trail architecture.
Documentation integrity, not claims scoring — a second source of session evidence.
Integration scope and partnership economics for RPM and EHR platforms.