Integrations & interface development
Getting data out of a healthcare system is rarely the hard part. Getting it out reliably, in the right format, without losing a record on a Tuesday night when nobody is watching — that is what we build.
The systems already in your building
If it holds data you need and exposes any way to reach it, it can be integrated. These are the sources we work with most.
Every shape your data shows up in
Most organisations need several of these at once — a FHIR API here, a nightly CSV there, and an X12 835 that has to reconcile against both.
| Standard | What we handle | Transport |
|---|---|---|
| HL7 v2.x | ADT, ORU, ORM/ORD, SIU, DFT, MDM, ACK handling, custom Z-segments | MLLP, TCP, file drop |
| FHIR R4 | Patient, Encounter, Observation, Coverage, Claim and related resources; search and pagination; bulk export | REST over HTTPS |
| C-CDA / CDA R2 | CCD, referral notes, discharge summaries, transitions of care; parsing and generation | API, SFTP, HIE gateway |
| EDI X12 | 837P/837I claims, 835 remittances, 270/271 eligibility, 276/277 status, 834 enrolment, 999/TA1 acknowledgements | SFTP, AS2, clearinghouse API |
| REST APIs | Vendor APIs, OAuth 2.0 and token auth, pagination, rate limiting, webhooks; plus custom APIs we build for you | HTTPS |
| SOAP APIs | WSDL-based services, WS-Security, legacy vendor endpoints | HTTPS |
| CSV & flat files | Delimited, fixed-width, multi-record and header/trailer files; encoding and delimiter quirks included | SFTP, FTPS, shared storage |
| Staging tables | Direct reads from staging schemas, replicas or read-only views when a system offers no interface at all | SQL over VPN or private link |
Built to survive the night shift
An interface that works on demo day is easy. One that keeps working through a vendor upgrade, a malformed message and a two-hour network outage takes deliberate engineering.
- Mapping and transformation. Field-level maps agreed with you, including the code sets and edge cases specific to your specialty.
- Validation before load. Structural and business rules applied on the way in, with failures quarantined for review rather than dropped.
- Retries and replay. Transient failures retry automatically; anything that still fails can be re-processed once the cause is fixed.
- Monitoring and alerting. Feed-level health checks that tell you a file did not arrive — before the people downstream notice.
- Documented handover. Interface specs, mappings and runbooks written down, so you are never dependent on one person's memory.
One interface at a time, proven before the next
Specify
We agree the source, the trigger, the format, the fields and what a failure should do.
Build
The interface is developed against sample data, with mapping and validation rules reviewed by you.
Test
Parallel runs against real volumes, reconciled record-for-record before anything is trusted.
Run
Go-live with monitoring, alerting and a support path when a vendor changes something.
Integration questions, answered
Usually not. Between vendor APIs, scheduled exports, HL7 feeds, report extracts and direct staging-table access, there is almost always a route to the data. Part of discovery is finding which one your contract and your vendor will actually support — and telling you honestly if the answer really is no.
Not necessarily. If you already run one, we will build inside it. If you do not, we can deliver the same result with a lighter-weight pipeline on your own infrastructure — and we will tell you when the volume genuinely justifies buying a platform.
Yes. Bidirectional interfaces are common — results back into the EHR, eligibility responses into the PM system, status updates into a portal. Writing into a clinical system needs more care and testing than reading from one, and we scope it accordingly.
Encrypted transport throughout — TLS for APIs, SFTP or FTPS for files, VPN or private links for database connections. Credentials are stored in a secrets store rather than in scripts, access is least-privilege, and every transfer is logged for audit.
You do. The code, the mappings, the documentation and the credentials are yours, and they run on your infrastructure unless you ask us to host. We would like you to stay because the work is good, not because leaving is difficult.
Which two systems should be talking first?
Bring us the pair that causes the most manual work today. We will scope that one interface end-to-end and show you what the rest would look like.