Patient device
BYOD
Personal Apple Watch via HealthKit. One-time consent, then continuous heart rate streams during the session.
- Best for established patients on RPM
- Patient keeps their own wearable
- Lowest clinic logistics
Session Evidence Platform
After the session, voice and continuous heart rate become a QOAX Clinical Session Report — structured evidence your platform can ingest by API.
Patient-side
Two ways in
Intelligence layer only. You pick who owns the watch.
Clinic device
Clinic pool
Checkout at intake, wear during the session, return after. Same evidence quality without requiring a personal watch.
- Best for intakes, acute visits, equity access
- Clinic owns the hardware pool
- Typical bridge: 2–3 sessions, then BYOD
Either path hits the same POST /v1/session endpoint — FHIR R4 out, with signal quality on every response.
Fusion protocol
Four streams, one asset
01 — Semantic
Voice NLP
Diarization, BH narrative, prosody — the language layer.
02 — Vocal biomarker
Voice Physiology
Prosodic and acoustic markers time-aligned to conversational moments.
03 — TIMESTAMP
Biological Timestamp Engine
Continuous HR and motion (HRV where sampling permits), locked to session timestamps.
04 — FUSION
Evidence Asset
Fused into one FHIR R4 report — voice plus wearable when present.
Deliverable
QOAX Clinical Session Report
Redacted wearable-present sample; clinician review required. Audio-only sessions still produce documentation — wearable unlocks full corroboration.
QOAX Clinical Session Report
1. Executive summary: clinical note
Client [redacted] presented for follow-up regarding occupational stress. The session moved from frustration and feeling stuck toward identifying a concrete workplace resolution path. No safety or crisis indicators were present. Recommend follow-up to assess implementation of the agreed strategy.
2. Clinician flags
- No safety, self-harm, or crisis indicators in session.
- Say/body gap at 00:18–00:22 during supervisor topic (HR spike; verbal affect flat).
- Presenting issue is situational stress; no co-occurring diagnoses discussed.
- Strong self-directed problem-solving. Reinforce in follow-up.
3. Session statistics
- Words
- 1,680
- Active speaking
- 460s
- Silence (>2s)
- 16 events · 8.7%
- Speakers
- 2
- Signal quality
- 0.91
- Arousal score
- 0.78
4. Affect timeline
30-second windows · green positive · red negative · grey neutral
5. Topics & clinical analysis
Workplace workflow conflict anger
Uncertainty markers detected in self-advocacy with supervisors. Session arc: stuck → brainstorming → reduced frustration. Physiological corroboration present for the mid-session arousal peak.
Sample is illustrative and redacted. Not a clinical claim about accuracy.
Full feature comparison → Documentation & billing economics →
Fusion Engine
Composite Arousal Scorer & Evidence Confidence
High Confidence
Strong physiological-conversational alignment. Clear objective support for the clinical narrative.
Medium Confidence
Partial alignment. Documentation supported; flagged for clinician review before charting.
Low Confidence
Weak or noisy signal. Insufficient for standalone corroboration. Surfaced for audit context, not as proof.
Incongruence Flag
Physiological signal contradicts self-report, a clinically significant pattern for risk assessment.
Policy context
Built for where healthcare is heading
The WHO’s Global Strategy on Digital Health asks platforms to turn data and biological evidence into informed clinical action, and WHO guidance on ethics and governance of AI for health warns against opaque systems clinicians cannot question. QOAX is built on the right side of that line: no app-usage patterns, no keystroke proxies. Only involuntary physiological signals a clinician can see and question, fused into one interoperable FHIR R4 stream.
Further Reading
Feature comparison vs. AI scribe category: on About.
Compare FeaturesHIPAA, 42 CFR Part 2, and Medicare audit alignment. full posture on Security.
Explore SecuritySee It in Your Stack.
Explore API integration scope or partnership economics with our team.