FHIR R4 Data Platform for 20,000+ Skilled Nursing Facilities
Architecting the interoperability foundation that connects America's post-acute care network — from HL7 v2 legacy interfaces through FHIR R4, to agentic AI quality analysis.
The Challenge
The Challenge: 20,000 Facilities, 50+ Different Data Systems, Zero Interoperability
Octdaily's vision was ambitious: a single quality monitoring platform that could observe care quality patterns across every Skilled Nursing Facility in the United States and use AI to help each facility improve. The technical reality was daunting. US post-acute care is served by over 50 different EMR systems — with PointClickCare, MatrixCare, and American HealthTech dominating, but dozens of regional and custom systems also in use. Most of these systems communicated via HL7 v2 interfaces designed in the 1990s, some had no API whatsoever, and none spoke FHIR R4 natively. Each facility had its own patient identifier scheme, local diagnosis codes, and custom clinical workflows. Aggregating quality data across this landscape — in real time, at the scale of 20,000+ facilities — required a fundamentally new approach to healthcare interoperability.
The Solution
Solution: A FHIR R4 Data Warehouse with Multi-EHR Integration and Agentic AI
I architected a three-layer technical platform: a multi-EHR integration layer, a FHIR R4 data warehouse, and an agentic AI analysis system.
Multi-EHR Integration Layer
Built custom integration adapters for PointClickCare (REST API and HL7 v2), MatrixCare, American HealthTech, and legacy custom EMRs. Each adapter handled vendor-specific authentication, data model quirks, rate limiting, and error recovery. The PointClickCare adapter — connecting the majority of facilities — processes over 1 million ADT events daily using a combination of PCC's REST API and HL7 v2 feeds routed through Azure Service Bus.
FHIR R4 Data Warehouse
Designed and built a FHIR R4 data warehouse on Azure Health Data Services. All incoming clinical data — regardless of source EMR or format — is transformed to FHIR R4 resources conforming to US Core profiles before storage. The warehouse stores Patient, Encounter, Condition, Observation, MedicationRequest, and Procedure resources for every connected facility, queryable via the FHIR API or Azure Synapse Analytics for population-level queries.
Agentic Medical Director AI
Built using Claude 3 Opus (for complex clinical reasoning) and GPT-4o (for structured data extraction). The AI agent autonomously analyses each facility's quality metrics across CMS Five Star domains, retrieves relevant clinical guidelines via RAG, identifies root causes for quality measure underperformance, and generates prioritised QAPI action plans. The agent runs continuously, processing each connected facility's data on a rolling 30-day window.
Technology Stack
Tools & Technologies Used
Related Services
Services Used in This Project
Have a similar project?
Let's discuss your requirements. Book a free 30-minute consultation.
Book Free Consultation