Written and maintained by CASRAI Editorial Board
Last updated
Fabric Health (often referred to as Fabric, formerly Florence) is an enterprise care enablement, conversational AI, and asynchronous virtual care platform used by health systems, payers, and clinical enterprises to automate digital front-door patient intake, symptom triage, and clinical documentation. By combining the conversational AI symptom navigation of GYANT, the adaptive asynchronous clinical interview engine of Zipnosis, and the virtual care infrastructure of MeMD, Fabric operates alongside specialized AI symptom checkers such as Buoy Health and Sensely to route patients to appropriate acuity levels and write structured History of Present Illness (HPI) data directly into the Electronic Health Record (EHR).
Informational Notice: This article evaluates clinical informatics architectures, triage algorithms, and health IT integration workflows for healthcare administrators and researchers. It does not provide individual medical advice or clinical diagnosis.
What Is Fabric Health?
Fabric Health is an integrated clinical workflow and virtual care suite that unifies conversational AI patient navigation (formerly GYANT), asynchronous adaptive clinical interviews (formerly Zipnosis), and in-clinic/ED intake automation into a single EHR-integrated platform.
Health systems deploy Fabric to address two operational bottlenecks: patient navigation at the “digital front door” (deciding whether a symptomatic patient requires an emergency department visit, urgent care, primary care, or asynchronous virtual treatment) and clinician documentation time during intake. Its core technical modules include:
- Conversational AI & Acuity Triage (GYANT Engine): Uses natural language processing (NLP) and deterministic clinical safety pathways to interview patients in free text or structured prompts, screen for red-flag emergent symptoms, verify insurance/demographics, and book appointments into matching provider slots.
- Asynchronous Adaptive Interviews (Zipnosis Architecture): Guides patients with common, low-acuity conditions through dynamic branching questionnaires based on evidence-based clinical guidelines. The engine compiles the patient’s responses into a structured SOAP-formatted clinical summary and presents staged order sets to a reviewing clinician, who can verify the assessment and sign a treatment plan asynchronously.
- In-Clinic & Emergency Department Digital Intake: Collects chief complaints, review of systems (ROS), medication reconciliation, and social determinants of health (SDoH) via SMS/web links prior to or upon arrival, writing structured data back to Epic, Oracle Health (Cerner), or athenahealth via bidirectional HL7 FHIR R4 and HL7 v2 EHR integration.
Comparing Fabric Health (GYANT), Buoy Health, and Sensely
When clinical informatics teams evaluate patient-facing conversational AI and symptom checking platforms, they typically compare Fabric (GYANT) against Buoy Health and Sensely across algorithmic design, care delivery scope, and clinical protocol provenance:
| Dimension | Fabric Health (integrating GYANT & Zipnosis) | Buoy Health (Buoy Symptom Checker) | Sensely |
|---|---|---|---|
| Primary Architecture | Unified conversational AI front door (GYANT) + asynchronous clinical interviews (Zipnosis) + ED/clinic intake | Probabilistic Bayesian clinical interviewing engine and employer/payer benefits navigation | Avatar-based and text/voice conversational AI mapped to standardized clinical triage protocols |
| Clinical Reasoning Model | Deterministic clinical decision trees + NLP intent recognition and guideline-based branching | Dynamic question selection computed against statistical models compiled from peer-reviewed medical literature | Rule-based triage pathways utilizing Schmitt-Thompson telephone triage protocols, Mayo Clinic, and NHS content |
| Care Delivery Handoff | Direct handoff into built-in asynchronous virtual visits, synchronous audio-video telehealth, or EHR scheduling | Acuity stratification and white-label routing to in-network health system, payer, or employer benefits | Multichannel triage (Web, mobile app, WhatsApp, SMS) routing to nurse triage lines, telehealth, or clinics |
| EHR Write-Back | Writes discrete HPI, ROS, allergies, and staged asynchronous notes into Epic, Oracle Health, and athenahealth | Passes structured triage summaries and referral context to partner portals and scheduling APIs | Integrates via HL7 FHIR APIs to populate patient intake summaries and appointment bookings |
Buoy Health (Buoy Symptom Checker)
Developed by emergency medicine clinicians and data scientists originating at the Harvard Innovation Lab, the Buoy symptom checker replaces static symptom lookup lists with an interactive interview that adapts each subsequent question in real time based on epidemiology, age, biological sex, risk factors, and co-occurring symptoms. Rather than acting as an EHR documentation platform, Buoy focuses on front-end diagnostic triage accuracy and benefits navigation for health plans, self-insured employers, and hospital portals—helping users distinguish self-care conditions from urgent or emergent care needs.
Sensely
Sensely pairs conversational AI (supporting text, speech recognition, and an optional interactive digital clinician avatar) with established nursing triage standards, most notably the Schmitt-Thompson Clinical Content telephone triage guidelines as well as NHS and Mayo Clinic clinical pathways. Because Schmitt-Thompson protocols are already the standard of care in hospital nurse call centers across the United States, Sensely allows health systems to maintain consistent acuity disposition rules between human telephone triage nurses and automated digital self-triage channels.
How Is Fabric Health Used in Clinical Research and Healthcare?
In healthcare operations and clinical research informatics, Fabric Health and peer symptom checkers are deployed to standardize pre-visit acuity stratification, reduce emergency department overcrowding, capture structured patient-reported symptom histories, and prescreen participants for decentralized clinical trials.
Health services researchers and clinical AI governance committees evaluate conversational triage platforms against rigorous empirical benchmarks and regulatory frameworks:
- Benchmarking Diagnostic & Triage Accuracy (Semigran et al., BMJ 2015): In the foundational audit study published in The BMJ (Semigran et al., 2015;351:h3480), researchers evaluated 23 online symptom checkers across 45 standardized clinical vignettes spanning emergent, non-emergent, and self-care presentations. Across all systems, the correct diagnosis appeared first in 34% of evaluations (and within the top 20 in 58%), while appropriate triage advice was achieved in 57% of evaluations. Crucially, triage accuracy was 80% for emergent vignettes but only 33% for self-care vignettes—demonstrating a systematic risk-averse “over-triage” bias that modern conversational engines must calibrate against to avoid shifting unnecessary volume to emergency departments.
- Asynchronous Protocol Efficiency & Guideline Adherence: In health systems utilizing Fabric’s Zipnosis-derived asynchronous care workflows for low-acuity conditions (such as uncomplicated urinary tract infections, upper respiratory infections, or dermatologic flare-ups), structured branching questionnaires enforce 100% completion of mandatory clinical red-flag exclusions before a provider can sign an order—supporting antibiotic stewardship and clinical quality auditing.
- Decentralized Trial Prescreening & ePRO Intake: Under ICH E6(R3) Annex 2 decentralized clinical trial (DCT) workflows, conversational intake agents are used to prescreen symptomatic patients against protocol inclusion/exclusion criteria at the point of digital care seeking and to capture structured baseline symptom severity using HL7 FHIR
QuestionnaireResponseresources. - AI Governance & Regulatory Oversight: Because patient-facing symptom checkers directly influence care-seeking urgency, hospital oversight committees subject them to healthcare AI governance reviews aligned with FDA Clinical Decision Support (CDS) guidance (FD&C Act Section 520(o)(1)(E)) and ONC HTI-1 transparency criteria (45 CFR § 170.315(b)(11)) when integrated into certified EHR technology.
Selection Checklist for Health System & Research Informatics Teams
- Validate Acuity Disposition Sensitivity & Specificity: Require vendor validation data across standardized clinical vignettes (emergent, urgent, primary care, and self-care) verifying zero missed high-acuity red flags alongside acceptable false-positive emergency routing rates.
- Verify Clinical Protocol Provenance: Confirm whether triage logic is deterministic and auditable (e.g., Schmitt-Thompson protocols or health-system-approved clinical pathways) versus opaque generative outputs, ensuring generative LLMs are restricted to conversational phrasing rather than unconstrained medical triage decisions.
- Test Bidirectional EHR Write-Back: Confirm whether the platform writes discrete clinical elements (chief complaint, HPI narrative, ROS, allergies) directly into clinician note templates via HL7 FHIR R4 (`QuestionnaireResponse`, `Observation`, `DocumentReference`) rather than depositing an unindexed PDF.
- Audit HIPAA BAA & Tracking Pixel Compliance: Because symptom checkers operate on public-facing health system domains, verify that a HIPAA Business Associate Agreement (BAA) covers all pre-login symptom chats and that third-party advertising/analytics tracking pixels are strictly excluded per HHS OCR online tracking guidance.
