EHR and practice management integration designed around the failure path first: what happens when a message does not arrive, a system is down, an identity does not match, or a transaction half completes at two in the morning.
Connecting two systems on the happy path is a week of work. Everything that determines whether the integration is safe to run for five years happens in the other paths, and that is where most healthcare integration is thinnest. CaliberFocus builds interfaces with reconciliation, replay, alerting and ownership designed in, and helps organizations reduce an interface estate that most cannot fully enumerate.
An interface that fails loudly is an incident. One that keeps running and quietly drops a small percentage of messages is a clinical risk nobody is looking for.
Every interface was justified at the time. A department bought a system, a vendor needed a feed, a project needed data moved by a date. Each was built, tested against the happy path, and handed to an operations team that inherited it without documentation. Ten years on, the organization runs an estate it cannot fully enumerate, cannot safely change, and cannot retire because nobody is certain who still consumes what.
The estate grows and never shrinks
Interfaces are built as projects and inherited as operations.
Built for the happy path
Testing covers the message that arrives correctly. Production is defined by the ones that do not.
Partial success is the normal failure
Order sent and result never returned. Charge posted and claim never created.
Silent failure beats loud failure to production
A stopped interface raises an alert. One dropping a subset quietly may not.
Vendor interface economics shape the architecture
Per-interface and per-connection costs are design constraints.
Staff are the fallback integration layer
Re-keying, searching two systems, uploads and spreadsheets hide integration cost in headcount.
If staff routinely move information between two systems by hand, you have an integration problem.
Even when both systems are working exactly as designed.
Even when both systems are working exactly as designed.
Step 1
Inventory the estate
What exists, what it does, who consumes it, and when it last carried traffic.
Step 2
Define the business transaction
What event is occurring, what must be true afterwards, and what happens if it does not.
Step 3
Choose the method per transaction
Real time messaging, API, batch, event or file based on need and cost.
Step 4
Design failure modes first
Retry, idempotency, ordering, dead-letter handling, reconciliation and alerts.
Step 5
Write the data contract
Fields, identifiers, code sets, mappings, transformations, validation and errors.
Step 6
Map against real data
Vendor implementations diverge from the standard.
Step 7
Build reconciliation alongside delivery
Sent and received are two different counts.
Step 8
Test failures deliberately
Inject downtime, malformed messages, duplicates, out-of-order delivery and identity mismatches.
Step 9
Deploy with monitoring and ownership
Named owner, alerting and runbook defined at go-live.
Step 10
Operate and retire
Review interfaces on a cycle and switch off anything superseded or unused.
Sent is not received, and received is not processed.
Reconciliation that counts both ends and reports the difference is one of the highest-return controls in healthcare integration.
Reconciliation that counts both ends and reports the difference is one of the highest-return controls in healthcare integration.
Connect
EHR and Practice Management Integration
Bidirectional integration using methods platforms actually support.
Interface Engine and Middleware
Design, build and migration on your existing integration engine.
API and Event Integration
FHIR and vendor APIs, webhooks and event streams where justified.
Third-Party and Departmental Systems
Laboratory, imaging, pharmacy, scheduling, contact centre, patient engagement, ERP and workforce applications.
Transform and Validate
Data Mapping and Terminology Translation
Mappings maintained as versioned assets with documented owners.
Identity Resolution at the Interface
Patient, provider and encounter matching before a write.
Validation and Business Rules
Structural, semantic and business validation before acceptance.
Error Handling and Replay
Retry policy, idempotency, dead-letter queues and replay tooling.
Operate
Reconciliation
Automated comparison of what was sent, received and processed
Monitoring and Alerting
Volume, latency, error rate and running-but-failing conditions.
Version and Change Control
Specifications and mappings under source control with rollback.
Estate Management and Retirement
Maintained inventory with owner, consumers, last traffic and business purpose.
What CaliberFocus does, and does not do?
We are not reselling an integration platform and we build on the engine you already run. We will also tell you when the answer is not an interface at all.
We are not reselling an integration platform and we build on the engine you already run. We will also tell you when the answer is not an interface at all.
| Integration | Direction | How It Typically Fails |
|---|---|---|
| Patient Demographics and Registration | Bidirectional | Updates applied to the wrong record after a weak identity match, or a merge in one system never propagating to the other. |
| Scheduling and Appointments | Bidirectional | Cancellations and reschedules arriving out of order. |
| ADT and Patient Movement | Inbound | Event sequence processed out of order. |
| Orders | Outbound | Order accepted at the boundary and never acted on downstream. |
| Results and Reports | Inbound | A subset silently rejected on a code or format the mapping never covered. |
| Charges and Billing | Outbound | Charges posted with no corresponding claim, or duplicated on retry. |
| Claims, Remittance and Eligibility | Bidirectional | Rejections at the clearinghouse never routed back to a work queue. |
| Prior Authorization | Bidirectional | Status updates never reach scheduling or clinical workflow. |
| Referrals and Transitions | Bidirectional | Referral sent, result never returned, and no reconciliation to reveal the open loop. |
Write-Back Is Where the Risk Concentrates.
Writing into a clinical system is a patient safety surface and deserves stronger identity confidence thresholds, validation, auditability and reversibility.
Writing into a clinical system is a patient safety surface and deserves stronger identity confidence thresholds, validation, auditability and reversibility.
| Method | Use It For | What to Design Around |
|---|---|---|
| HL7 v2 Messaging | ADT, orders, results, scheduling | Ordering, duplicates and acknowledgement handling. |
| FHIR and Vendor APIs | Discrete data retrieval, application integration, workflow launch, near-current reads | Vendor coverage and throughput limits. |
| X12 EDI | Eligibility, claims, remittance, claim status, authorization | Trading-partner variation and rejection routing. |
| CDA and Document Exchange | Transitions of care, external records, referral packets | Document-level payloads still need parsing. |
| Files and Secure Transfer | Legacy systems, departmental applications, bulk movement | Fragile ownership and manual dependency. |
| Events and Streaming | Operational workflows needing sub-minute reaction | Complexity justified only where latency matters. |
| Controlled UI Automation | Systems with no interface and no realistic prospect of one | Last resort with maintenance burden, documented as technical debt. |
Engine Over Point-to-Point.
Bringing accumulated point-to-point connections back inside the integration engine usually produces high operational return.
Bringing accumulated point-to-point connections back inside the integration engine usually produces high operational return.
Structural
Is the message well formed and complete?
Semantic
Do values conform to expected code sets, ranges and formats?
Identity
Does this resolve with sufficient confidence to act?
Business rule
Does this make sense in context?
Transient
Retry with backoff.
Malformed
Reject at the boundary, alert the sender, do not silently repair.
Unmapped value
Quarantine and alert.
Identity uncertain
Route to human review. Never write to a clinical record on an uncertain match.
Business rule violation
Route to a work queue with the reason attached.
Downstream unavailable
Queue and replay, with a defined depth limit and alert.
Idempotency, Ordering and Replay
A replay should not create a duplicate charge or second order, and operations should be able to reprocess a defined window without engineering involvement.
A replay should not create a duplicate charge or second order, and operations should be able to reprocess a defined window without engineering involvement.
Monitoring, in three layers
-
Technical health
availability, volume, error rate, latency, queue depth, retries. -
Data health
missing fields, mapping exceptions, identity failures, duplicates. -
Workflow health
end-to-end completion and reconciliation variance.
Security
Encryption, managed credentials, environment separation, least privilege and PHI minimization.
Governance and ownership
Five named owners per interface: technical operation, data definition, operational workflow, exception resolution and vendor coordination.
Resilience
Documented behaviour during outages, queue limits, replay tooling and manual fallback where needed.
Ask what your interfaces did last night.
A healthy estate can answer immediately: volumes against expected, errors by type, anything queued, anything reconciling short.
A healthy estate can answer immediately: volumes against expected, errors by type, anything queued, anything reconciling short.
| Category | What We Measure | Why It Matters |
|---|---|---|
| Estate Size | Interfaces in production, point-to-point connections consolidated, interfaces retired | The count should be falling. |
| Detection | Share of failures found by monitoring versus users, and time to detection | Separates a managed estate from a lucky one. |
| Reconciliation | Interfaces with active reconciliation, and variance detected and resolved | Surfaces silent failure. |
| Recovery | Time to recover, share recoverable by operations without engineering | Determines whether an incident takes minutes or a weekend. |
| Stability Through Change | Interfaces broken per vendor upgrade, and breakage found in test versus production | The recurring cost most organizations never measure. |
| Manual Work Removed | Re-entry, uploads, portal work and spreadsheet reconciliation eliminated | Connects integration spend to operational value. |
| Ownership | Interfaces with a named owner, documentation and a current specification | Predicts what happens when the person who built it leaves. |
Retirement is harder than building.
Budget for a shadow period where an interface is disabled but recoverable. The estate does not shrink any other way.
Budget for a shadow period where an interface is disabled but recoverable. The estate does not shrink any other way.
We will inventory what is actually running, identify which interfaces have no reconciliation and no owner, test a sample against real production data to find the silent failures, and give you a prioritized remediation and retirement plan. In most assessments a meaningful share of the estate turns out to be unused, undocumented, or failing quietly.
Bring us a specialty or clinical setting where clinicians are spending too much time creating notes. We will assess where ambient documentation fits, what must remain clinician controlled, how it should integrate with your EHR, and how to measure whether it is actually reducing burden.
One conversation with people who have run these deployments, and a written readiness view you can use with or without us.
- AI Agents and Workflow Automation
- Voice and Conversational AI
- Document AI and Intelligent Processing
- Generative AI and Enterprise Copilots
- AI Strategy and Governance
- HCC and Risk Adjustment Analytics
Security & Compliance
