Unique Selling Propositions (USPs)¶
This document states the main unique selling propositions of MedExpertMatch: what distinguishes it from manual processes, simple directories, or generic matching tools.
1. Match in Minutes, Not Days¶
USP: Specialist matching in minutes instead of days; target is to reduce consultation delays from days to hours.
Why it is unique: Combines vector similarity, graph relationships, and historical performance in a single scoring pipeline (Hybrid GraphRAG). Replaces ad-hoc "who do I know?" with a consistent, automated process that returns ranked specialists with rationales. No other single system in the scope combines these three signals with medical case analysis in one flow.
2. "Who Is Good at What" Becomes Visible¶
USP: Real expertise is visible: which specialists and facilities actually handle complex cases in specific domains ( e.g. by ICD-10 code), not only job titles or self-declared specialties.
Why it is unique: Network analytics on a graph of doctors, cases, conditions, and facilities. Queries like "top experts for I21.9 in the past 2 years" return data-driven rankings. Replaces implicit, anecdotal knowledge with transparent, measurable expertise mapping for routing, planning, and mentorship.
3. One AI Copilot: Analysis, Evidence, Recommendations, and Experts¶
USP: A single flow delivers structured case analysis, differential diagnosis, evidence retrieval (guidelines, PubMed), clinical recommendations, and matched colleagues to discuss with.
Why it is unique: One medical-domain copilot instead of separate tools. Case-analyzer, evidence-retriever, recommendation-engine, clinical-advisor, and doctor-matcher work together in one conversation or API flow. Specialists get analysis plus evidence plus recommendations plus expert matching without switching systems.
4. Three-Signal Scoring: Vector + Graph + History¶
USP: Matching score combines semantic similarity (40%), graph relationships (30%), and historical performance (30%) in a single, explainable formula.
Why it is unique: Hybrid GraphRAG with explicit weights tuned for medical matching. Not only keyword or only vector or only graph: all three with Apache AGE for doctor–case–condition relationships and PgVector for embeddings. Rationales support transparency and audit.
5. Urgent First, Not First-In-First-Out¶
USP: Consultation queue is ordered by clinical urgency (CRITICAL / HIGH / MEDIUM / LOW), so the sickest patients are seen first.
Why it is unique: AI classifies urgency and risk; queue is prioritized by clinical need, not arrival time. Reduces the risk of critical consults waiting behind routine ones and gives coordinators a single, urgency-based view.
6. Right Sub-Specialist, Not Just "Oncologist" or "Cardiologist"¶
USP: Second opinions and complex cases are matched to the appropriate sub-specialist by diagnosis and complexity, not only by broad specialty.
Why it is unique: Matching uses case analysis (ICD-10/SNOMED, complexity) and experience in that specific diagnosis. Reduces misrouting to a generic specialty when a sub-specialist is needed; can prioritize telehealth when relevant.
7. Facility Matches Case Complexity (Regional Routing)¶
USP: Regional routing scores facilities by case complexity, historical outcomes, capacity, and geography, and suggests lead specialists.
Why it is unique: Reduces facility–case mismatch in hierarchical systems. Not only "nearest" or "first available": routing is driven by capability and outcomes, with ranked facilities and explanations. Referrals and transfers become transparent and measurable.
8. Medical-Domain AI and Agent Skills¶
USP: MedGemma models and seven medical-specific agent skills (case-analyzer, doctor-matcher, evidence-retriever, recommendation-engine, clinical-advisor, network-analyzer, routing-planner) for end-to-end clinical workflows.
Why it is unique: Purpose-built for medical expert matching and decision support, not a generic chatbot. Skills are modular and documented; prompts and tools are medical-domain (ICD-10, guidelines, PubMed, risk, differential diagnosis).
9. Privacy-First and Deployment-Friendly¶
USP: Architecture supports local deployment and HIPAA-aware data handling; no PHI in logs or error messages; anonymization in code and test data.
Why it is unique: Designed for healthcare privacy from the start. Can run on-premises or in a controlled cloud; AI outputs include disclaimers; human-in-the-loop is explicit. Suited for environments where data cannot leave the organization.
10. FHIR and EMR-Ready¶
USP: Consumes FHIR Bundles (Patient, Condition, Observations, Encounter); supports EMR integration and portals; REST APIs for matching, queue, analytics, and routing.
Why it is unique: Fits into real clinical workflows via standard interoperability. Same logic supports direct text input (e.g. after OCR) and FHIR-based flows, so hospitals and regional systems can integrate without replacing existing EMRs.
Summary Table¶
| USP | One-line claim |
|---|---|
| 1 | Match in minutes, not days |
| 2 | "Who is good at what" made visible (graph + analytics) |
| 3 | One copilot: analysis + evidence + recommendations + experts |
| 4 | Three-signal scoring: vector + graph + history |
| 5 | Urgent first, not FIFO queue |
| 6 | Right sub-specialist, not just specialty |
| 7 | Facility matches case complexity (regional routing) |
| 8 | Medical-domain AI and agent skills |
| 9 | Privacy-first, local deployment, HIPAA-aware |
| 10 | FHIR and EMR-ready |
Last updated: 2026-02-08