We're upgrading how our provider search interprets patient queries and maps clinical terms. Here's everything your team needs to know — including answers to the most common compliance questions.
WHAT'S BEING UPDATED
Two components are changing
We are modernizing two components of our provider search product used by healthcare clients to power patient-facing "find a doctor" search experiences. Neither change affects how providers are presented or makes clinical decisions — the system routes search queries only.
NLP Engine Upgrade A new multi-model AI layer interprets patient search queries — understanding plain-language terms like "can't sleep" or "bloated belly" to route to the right specialties. | Clinical Knowledge Graph Replaces our flat taxonomy with an intelligent graph of clinical relationships, so searches for complex conditions like "ankylosing spondylitis" still find relevant results. | No Clinical Decisions Made The system routes search queries only. It does not diagnose conditions, recommend treatments, or change how providers are presented to patients. |
DATA HANDLING & PRIVACY
Your data stays protected
About patient data
Patient search queries are processed in real-time and are not stored beyond normal application logging. No patient or client data is used to train AI models. The clinical taxonomy powering this update is RevSpring's own proprietary data (the Kyruus Clinical Library), not sourced from any client or patient records.
Key data handling facts:
- Patient search queries are handled entirely within RevSpring's internally hosted NLP service — not passed to external systems.
- A limited vector embedding process uses Amazon Web Services (AWS) — applied only to RevSpring's own clinical terminology data, never patient or client data. A Business Associate Agreement (BAA) is in place with AWS covering this processing arrangement; this does not require a separate BAA between your organization and AWS
- Client provider data remains within the existing data environment and is not passed to any new external system as part of this project.
- All production models will be hosted and managed on RevSpring's own infrastructure.
- Patient search queries do not constitute Protected Health Information (PHI) under HIPAA. Queries are free-text strings entered by patients and are not linked to patient identifiers, medical records, or treatment data within this system.
ROLLOUT TIMELINE
What to expect and when
Q2 2026 Preparation Regression testing with real-world queries. Validation in staging environments. | Q2 2026 NLP Upgrade Goes Live Applies automatically to all clients. No action required. No migration needed. | Q3 2026 Knowledge Graph Launch Enabled by default for all clients. No data migration or client-side changes. |
Rollout notes:
- NLP Upgrade: Applies automatically to all clients in Q2. No action, configuration, or migration required.
- Knowledge Graph: Rolling out in Q3 and iterating ongoing to continuously improve the experience.
CLINICAL REVIEW & VALIDATION
How we validate accuracy
- The Kyruus Clinical Library (KCL), the foundation for the Knowledge Graph, has been clinically curated by RevSpring's internal clinical team over many years.
- The Knowledge Graph restructures and enriches this existing data — it does not introduce new unvalidated clinical content.
- A regression testing suite uses real-world search queries drawn from analytics, internal testing, and customer feedback to evaluate accuracy before and after deployment.
- The system is designed with guardrails to remain strictly within care navigation — it does not make clinical recommendations or diagnoses.
FREQUENTLY ASKED QUESTIONS
Common questions from clients
Will this change the search experience for our patients?
The search bar and results interface remain the same. Behind the scenes, the AI upgrade helps the system better understand what patients type — especially colloquial or lay terms — so they see more relevant results. Patients are less likely to encounter zero-result searches.
Does the NLP upgrade require any action from us?
No action is required. The NLP upgrade applies automatically to all clients in Q2. There is no configuration, data migration, or client-side change needed.
Is patient data used to train any AI models?
No. Patient search queries are processed in real-time and are not used to train AI models. The AI models used in this project are trained on RevSpring's own clinical taxonomy data (KCL), not on any patient or client records.
Does any patient data leave RevSpring's infrastructure?
No. Patient search queries are handled entirely within RevSpring's internally hosted NLP service. A separate, limited process uses AWS for generating vector embeddings — but this applies only to RevSpring's own clinical terminology data, not to patient or client data. A BAA is in place with AWS.
Is this a material change requiring compliance review?
This change affects the underlying search intelligence layer only — not clinical workflows, provider presentation, patient data handling, or your organization’s data environment. RevSpring does not consider this a change to the scope of processing under existing agreements. Clients with specific regulatory obligations (for example, under state AI transparency laws) should assess applicability with their own counsel.