Custom Integration Development in Seattle, Washington

Healthcare / Clinical Data Interoperability

A Pacific Northwest multi-hospital health system — operating 6 hospitals, 45 clinics, and 12 urgent care centers across Washington State with 8,200 employees and 1.4 million patient encounters annually — faced critical interoperability challenges after a series of acquisitions. The health system had grown through the acquisition of four independent medical groups over 7 years. Each acquired group brought its own technology stack: Hospital System A (the original system): Epic EHR, plus 35 ancillary systems (lab, radiology, pharmacy, scheduling). Hospital System B (acquired 5 years ago): Cerner EHR, plus 28 ancillary systems — partially migrated to Epic but with key departments still on Cerner. Hospital System C (acquired 3 years ago): Athenahealth for ambulatory clinics, custom-built inpatient system. Hospital System D (acquired 18 months ago): MEDITECH for inpatient, eClinicalWorks for ambulatory. The result: patient data was scattered across 4 different EHR platforms and 80+ ancillary systems. A patient who had been seen at Hospital A and then visited a clinic from the acquired System C was treated as a new patient — the clinic had no access to the patient's history from Hospital A. Clinical impacts: duplicate testing — physicians ordered tests that had already been performed at another system facility because results weren't available. Analysis showed 8,400 duplicate lab tests per year across the system ($1.76M in direct costs, plus patient inconvenience and unnecessary blood draws). Incomplete medication lists — when patients moved between facilities, their medication lists didn't follow them. Medication reconciliation was performed manually by nurses calling other facilities or relying on patients' own medication lists — a process that missed medications 23% of the time, creating drug interaction risks. Delayed referral processing — referrals between facilities required faxed records (because systems couldn't share data electronically). Average referral processing time was 8 days — during which the patient waited without care. Emergency department challenges — ED physicians at Hospital A couldn't access records from System C or D facilities. In emergencies, this meant: allergies unknown (relying on patient self-report, which is unreliable for unconscious or confused patients), prior imaging unavailable (leading to repeat CT scans and X-rays), and surgical history unknown (critical for anesthesia planning). The health system had attempted integration twice before — once with a custom HL7v2 interface engine and once with an enterprise integration platform (MuleSoft). Both projects were partially successful (connecting the two largest systems) but failed to achieve comprehensive interoperability across all four EHR platforms and their ancillary systems. The previous integration attempts cost $2.8M total and took 24 months before being abandoned.

We developed a healthcare interoperability platform over 16 weeks — building a FHIR-based integration layer that connected all four EHR platforms and their critical ancillary systems through a unified clinical data service. Architecture: FHIR-First Integration — rather than building point-to-point connections between all systems (which would have required 200+ individual interfaces), we built a central clinical data service with FHIR APIs: each EHR connected to the central service through a single bidirectional interface (4 interfaces instead of 12 point-to-point connections between EHR pairs), the central service maintained a master patient index (MPI) that linked patient identities across all systems — using probabilistic matching on name, date of birth, SSN, and address to link records with 99.2% accuracy, clinical data from all systems was accessible through standardized FHIR APIs — any application or system could query the central service for a patient's complete clinical data regardless of which system originated the data. EHR Connectors: Epic connector — the Epic system (Hospital A and partially migrated Hospital B) connected through Epic's FHIR R4 APIs: patient demographics, conditions, medications, allergies, lab results, imaging reports, encounters, and care plans were accessible through standard FHIR resources. Real-time ADT (admission/discharge/transfer) events were published through Epic's event notification service. Bi-directional: external clinical data was written back to Epic through FHIR APIs, appearing in the patient's chart as imported external records. Cerner connector — the remaining Cerner departments at Hospital B connected through Cerner's FHIR R4 APIs with supplemental HL7v2 feeds for data not yet exposed through FHIR. Athenahealth connector — System C's ambulatory clinics connected through Athenahealth's FHIR APIs with custom webhook integration for real-time event notification (Athenahealth's event system was less mature than Epic's and required custom development). MEDITECH/eClinicalWorks connector — System D's systems had limited FHIR support, requiring: HL7v2 ADT, ORM, and ORU message processing for real-time clinical data, database-level integration for historical data migration, and custom REST APIs built on top of the existing systems for ongoing data exchange. Master Patient Index (MPI): patient identity resolution — the MPI was the most critical component of the platform. It resolved patient identities across 4 systems that had independently assigned medical record numbers (MRNs) to the same patients: probabilistic matching algorithm (using name variants, date of birth, SSN partial, address, phone number, and insurance ID to calculate match probability), manual review workflow (match candidates between 85-95% probability were queued for manual review by Health Information Management (HIM) staff — matches above 95% were auto-linked, below 85% were kept separate), and ongoing identity maintenance (new patient registrations at any facility were automatically matched against the MPI — preventing new duplicate records from being created). Clinical Data Services: unified patient record — the platform provided a consolidated view of each patient's clinical data: medication list (all active and recent medications from all systems — reconciled to identify duplicates, conflicts, and potential interactions), allergy list (all documented allergies from all systems — with severity, reaction type, and documenting provider), problem list (all active diagnoses/conditions from all systems — with onset dates and managing providers), lab results (all lab results from all systems — with trending visualization for common lab values across time and facilities), imaging (all imaging studies with report text and study availability — enabling clinicians to view images from any facility's PACS through a unified viewer link), encounters (chronological history of all encounters across all facilities — providing a complete picture of the patient's interactions with the health system), and care plans (active care plans from any facility — enabling care coordination for patients managed by providers across multiple facilities). Clinical Notifications: real-time event routing — the platform routed clinical events between systems: ADT notifications (when a patient was admitted to any facility, their primary care provider at any other facility received a notification — enabling proactive care coordination), results delivery (lab and imaging results were delivered to the ordering provider regardless of which system they used — a provider at System C who ordered labs processed at System A's reference lab received results directly in their Athenahealth chart), referral routing (electronic referrals between facilities replaced faxed records — the referral included the patient's relevant clinical data from the sending provider's system, accessible to the receiving provider immediately), and care gap alerts (when a patient had overdue preventive care items (missed mammogram, overdue colonoscopy, lapsed vaccination) identified at any facility, the alert was surfaced to any provider seeing the patient at any facility).

The platform deployed in phases — Epic-Cerner integration first (highest volume, most clinical impact), then Athenahealth and MEDITECH/eCW over the following 8 weeks. Master Patient Index Results: patient records linked: 1.4 million total patients identified across all systems — 186,000 patients (13.3%) had records in 2+ systems that were previously unlinked. Auto-match accuracy: 99.2% (validated by HIM team review of 2,000 randomly sampled auto-linked records). New duplicate prevention: the MPI prevented an estimated 12,000 new duplicate records in the first year by matching new registrations against the unified index in real-time. Clinical Impact (12 months): duplicate testing reduction: duplicate lab tests reduced from 8,400 to 1,200 per year (86% reduction — $1.34M in direct savings). The remaining duplicates were primarily cases where the provider had clinical justification for repeating a test despite the prior result being available. Medication reconciliation accuracy: medication list completeness improved from 77% to 96% across facilities — meaning providers had accurate medication information for 96% of patients (vs. 77% previously). The improvement was driven by automatic medication aggregation from all systems. Referral processing time: reduced from 8 days to 1.2 days average (85% reduction) — electronic referrals with attached clinical data eliminated the faxing, record retrieval, and manual review that caused the previous delay. ED access to external records: ED physicians accessed cross-system patient records for 14,200 patient encounters in the first year. In a survey of ED physicians: 94% reported that cross-system data access changed their clinical decision-making at least once per shift, 68% reported preventing unnecessary imaging by accessing prior studies from other facilities, and 42% reported identifying medication allergies or interactions that the patient hadn't reported. Patient Safety: medication interaction detection — the unified medication list enabled system-wide drug interaction checking that identified 340 potential interactions per month that were invisible when medication data was siloed in separate systems. 89 of these (26%) were classified as clinically significant — requiring medication changes that prevented potential adverse drug events. Allergy documentation — 23,000 allergy records were unified across systems, identifying 4,600 patients who had allergies documented in one system but not in others — closing documentation gaps that represented patient safety risks. Financial Impact (annual): duplicate testing reduction: $1.34M. ED imaging avoidance (reduced repeat imaging): $820,000. Referral efficiency (faster specialist access, reduced fax and mail costs): $480,000. Care gap closure (increased preventive care compliance leading to quality bonus payments): $360,000. HIM staff efficiency (automated MPI reducing manual record matching): $240,000. Total annual savings: $3.24M. Platform Development Cost: $680,000 (16 weeks). Annual operating cost: $192,000 (infrastructure, EHR connector maintenance, MPI operations). ROI: 4.8:1 in the first year, improving as the platform matures and additional use cases are added.