Healthcare interoperability case study
Reducing Epic API Costs by 90% Without Breaking Healthcare Workflows
When Epic shifted vendor APIs from contract-based pricing to usage-based pricing, API efficiency stopped being an abstract engineering concern and became an operational business problem overnight.
Starting point
~$1.3M/month projected Epic API costs
Outcome
~$134K/month after optimization
Constraint
Production workflows had to remain reliable
Monthly API cost
From projected exposure to sustainable utilization
Approximate monthly run-rate
~$1.3M
Projected monthly cost under usage-based pricing
~$134K
Reduced monthly cost after workflow-aware optimization
Production workflow
Epic APIs
FHIR + proprietary endpoints
The pricing change exposed the system
The initial problem looked financial: projected Epic API costs were on track to reach roughly $1.3M per month. But the deeper issue was operational. A pricing model change made previously hidden API utilization visible, and that visibility showed how small inefficiencies could become expensive at production scale.
This was not a situation where one code change could solve the problem. The integrations already supported live customer workflows. The work required understanding why calls were happening, which workflows depended on them, which endpoints were actually necessary, and which assumptions had accumulated over time.
Why this was difficult
Healthcare interoperability problems rarely stay inside one department. This effort touched engineering, implementation, support, finance, customer operations, Epic configuration, and production monitoring at the same time.
- Production integrations were already live and supporting real patient engagement workflows.
- Some fixes required customer-side Epic web service configuration changes, not just internal engineering updates.
- Scheduling, messaging, refill, and synchronization workflows could not lose reliability while API behavior changed underneath them.
- Finance, support, implementation, engineering, customer operations, and Epic-facing teams all needed the same operating picture.
That changed the shape of the work. Engineering changes mattered, but they were only part of the path. Customer environments needed to be reviewed. Epic web service configuration changes had to be planned. Testing had to account for the actual healthcare workflows customers relied on every day.
Reliable healthcare integrations are socio-technical systems. The API call is technical; the workflow it supports is operational.
Finding the leaky pipes
The process felt less like rebuilding an integration platform from scratch and more like tracing and repairing hidden leaks inside a large operational plumbing system. The goal was not to make the system look different. The goal was to reduce waste while preserving the behavior customers depended on.
The work started with mapping expensive usage back to real workflows. Some calls were tied to necessary production behavior. Others were artifacts of polling intervals, synchronization assumptions, duplicate checks, or endpoint choices that had been reasonable under one pricing model but unsustainable under another.
- Audited expensive proprietary Epic endpoint usage and mapped it back to the workflows driving each call pattern.
- Replaced usage with lower-cost or free FHIR alternatives where the workflow and customer environment allowed it.
- Introduced safer caching and synchronization windows so data stayed useful without unnecessary repeated calls.
- Reduced polling patterns that had become operational habits rather than workflow requirements.
- Monitored utilization after each change to confirm savings without introducing hidden reliability regressions.
Optimization lifecycle
01
Audit
02
Analyze
03
Optimize
04
Validate
05
Roll out
06
Monitor
The human side of integration engineering
The most important work happened in the space between systems and people. Every optimization had to be translated into an implementation plan: which customers were affected, what configuration needed to change, how the change would be tested, when it could be rolled out, and how the team would know whether it was safe.
That required communication more than theatrics. Customer-facing teams needed clear explanations. Implementation teams needed sequencing. Support needed to understand expected behavior. Engineering needed feedback from production validation. Finance needed visibility into whether the run-rate was actually improving.
My role sat across those boundaries: helping connect technical findings to workflow impact, coordinating implementation realities, validating production behavior, and keeping attention on the operational outcome rather than the elegance of any single technical change.
Results
The effort reduced projected Epic API costs from roughly $1.3M per month to about $134K per month, an approximately 90% reduction. Just as importantly, the reduction was achieved while maintaining production workflow reliability and improving visibility into ongoing API utilization.
The outcome was not only a lower bill. It was a more observable operating model: clearer ownership of API usage, better understanding of which workflows drove cost, and a stronger connection between integration design and business sustainability.
What I took away
API architecture decisions have operational consequences. A workflow that makes sense in isolation can become expensive when repeated across customers, environments, and production schedules. Interoperability systems need continuous operational analysis because the economic and technical assumptions around them can change.
One of the most surprising outcomes was realizing how much optimization was possible without sacrificing workflow reliability or customer experience. In many cases, the largest improvements came not from rebuilding systems entirely, but from understanding how operational assumptions, workflow design, and API usage patterns interacted at scale.
That is the work I find most meaningful in healthcare technology: making the system understandable enough that teams can improve it without breaking the people and workflows depending on it.
Cross-functional system
Connect
Want to talk Epic integrations, healthcare interoperability economics, or production workflow reliability?