A result is not a document

Most systems treat a laboratory result as a document to file. That decision is made once, early, and everything downstream inherits it: you cannot trend a PDF, you cannot threshold it, you cannot code from it reliably, and you cannot release part of it to one party and withhold the rest from another. Metaclinic treats a result as a set of coded observations with a release state attached to each one.

What arrives

Diagnostic data reaches us in four shapes, and each one has a different failure mode.

Discrete observations

LOINC-coded, with a value, a UCUM unit, a reference interval, and an abnormal flag. This is the only shape that supports computation. Everything we do that looks like intelligence depends on getting this right at intake.

Narrative reports

Gross, microscopic, final diagnosis, comment. Anatomic pathology lives here. The text is the deliverable, but the synoptic elements inside it are structured data that a registry and a coder both need.

Referenced objects

Whole-slide images, gross photographs, immunostain captures, and PDFs from laboratories that will not send discrete values. Large, and moved by reference.

Interface artifacts

Acknowledgements, control segments, and the message envelope. Not clinical, but the only evidence available when a claim is denied for a result nobody can find.

Normalization is where the value is

Two laboratories will report the same analyte under different local codes, different units, and different reference intervals, and both will be correct. Normalization means mapping local codes to LOINC, converting units to a canonical UCUM expression without discarding the original, preserving the performing laboratory's reference interval rather than substituting our own, and keeping the local code alongside the mapped one forever.

OBX|1|NM|2160-0^Creatinine^LN||1.9|mg/dL|0.60-1.30|H|||F|||20260810143000
OBX|2|NM|33914-3^eGFR^LN||38|mL/min/1.73m2|>60|L|||F

Versions, corrections, and the record that has to hold

Preliminary becomes final. Final becomes corrected. An amended pathology report changes a diagnosis after a claim has been paid and after a patient has read the original. A system that overwrites cannot answer the only question that matters in a dispute, which is what was known, by whom, on what date.

• Every version is retained and addressable. Nothing is overwritten in place.
• A correction generates its own release event to every party that received the prior version, and the notification records whether it was opened.
• Amended anatomic pathology reports trigger a coding review, because the final diagnosis can move the surgical pathology level in either direction.
• Retractions are marked as entered in error rather than deleted, and remain visible to the roles entitled to see the history.

Send-outs and reflexes

A specimen leaves for a reference laboratory and a result comes back days later with a different accession number, a different patient identifier format, and no memory of the case it came from. The result is clinically important, frequently expensive, and structurally easy to lose. Reconciling it back to the originating accession is unglamorous and it is most of the work.

Reflex testing has the inverse problem. The reflex fires correctly, produces a billable service, and then that service has to be defended against a payer policy that may or may not consider the trigger value sufficient. The trigger value has to be retained next to the reflex result, not inferred later.

What we do not do