Test data whose codes are still real
The quiet failure in healthcare test data is not a missing field. It is a code that used to be right. Code books are reissued every year, the test file is not, and one day your synthetic feed is carrying values your production system would refuse. MediFlow v12 closes that gap by checking every code it emits against the published release, in the build, before anything ships.
How test data rots
Nothing breaks loudly. A diagnosis code is subdivided in a new fiscal year: the old three-or-four character code stops being billable and becomes a header for the more specific codes beneath it. Low back pain and unspecified asthma both went that way. A drug identifier is retired and replaced by another. A vaccine code is deprecated when the product changes. In each case the code still looks like a code. It has the right shape, it passes a regular expression, and it fails only where it matters, in the system under test.
What gets checked
| Code system | The test applied |
|---|---|
| ICD-10-CM | Exists in the fiscal-year release and is billable, not a category header |
| LOINC | Exists in the release and is active — never a check-digit calculation, which proves nothing about whether a code is real |
| RxNorm | Current, not retired and not suppressed; a retired identifier names its replacement |
| CVX | A real CDC vaccine code. Inactive is allowed: a dose keeps the code of its day |
Enforced by the build, not by good intentions
A style guide saying "use current codes" decays the moment someone is in a hurry. The check runs as a test: the published releases sit beside the generator, every emitted code is scanned out of the source and checked, and a bad code fails the build. It cannot be argued with and it cannot be forgotten, which is the only way a rule like this survives contact with a deadline.
Where a standard has no code
Sometimes there genuinely is no standard code for a concept — a sedation scale, a triage level, a composite panel. The wrong answer is to pick a plausible-looking code that means something else. MediFlow sends a local code, explicitly marked as local, which is correct behaviour in HL7 and keeps the distinction visible to whatever maps it on the far side. A mapping team can see what it is being asked to map; it is not silently handed a real code with the wrong meaning.
Why a buyer should care
- A clean run means something. If the feed cannot carry an invalid code, a rejection is your system's behaviour, not the test data's.
- Mapping coverage is measurable. Orphan concepts are real gaps in the receiving system, not artefacts of stale inputs.
- Billing paths are testable. Claims built from billable diagnoses can be followed end to end instead of dying at the first edit.
- Nobody debugs the fixture. The most expensive hour in integration testing is the one spent chasing a defect that turns out to live in the test data.
Tell us what you need to trust
Describe the workflow, interface or release you are validating. We define the synthetic patients, the interface profile, the expected outcomes and the evidence you get back. Reply within one business day.