Mandatory Interoperability

Picture the path of one lab result. It begins inside an IVD analyzer — a chemistry instrument, an immunoassay platform, a molecular panel. The analyzer emits its result in the dialects it was born speaking: HL7 v2.x and ASTM, usually over the IHE LAW (Laboratory Analytical Workflow) profile, which is itself built on HL7 v2. It does not speak FHIR, and it was never designed to.

From there the result flows to the LIMS — the laboratory information system that receives, validates and stores it. The LIMS is the lab's system of record, and internally it too lives in the v2/ASTM world. It then passes the validated result upward to the HIS, the hospital information system that holds the patient's encounter. And it is the HIS — not the analyzer, not the LIMS — that assembles the RDA: the FHIR R4 Bundle the IHCE expects.

The shape of the chain matters more than any single box in it. FHIR appears only at the very top. Everything below — the instrument, the lab system, most of the internal traffic — speaks the older dialects, and will keep speaking them for years. So the right question is not the intuitive one. It is not “does every machine speak FHIR?” They do not, and they do not need to. The real question is narrower and sharper: is there a single point in the chain where v2/ASTM gets translated into FHIR R4? That point is the whole game.

If FHIR lives only at the top and the instruments live only at the bottom, then somewhere in between a translation must occur. That translating layer has a name in health IT: middleware — an integration engine that converts and routes between formats. When it is present, the result climbs the chain, changes languages once, and arrives as a clean FHIR resource. When it is absent, the result reaches the top still speaking v2, and the RDA for that data cannot be assembled. The gap is not a missing cable. It is a missing interpreter.

Here the easy version of the story overreaches, and a serious analysis has to say so. The tempting claim is: “if the analyzer does not speak FHIR, the IPS cannot comply.” Stated that strongly, it is wrong today — and two nuances, both drawn straight from the RDA guide, correct it.

First, the detailed laboratory result is deferred. The RDA v1.0.0 does not model a lab result with a DiagnosticReport profile bound to LOINC. It models it as ProcedureResultRDA — a single Observation coded in CUPS, the Colombian procedure nomenclature, not in LOINC. The structured, analyzer-level feed of discrete results is contemplated for a later phase. So the near-term pressure of April 2026 falls on the encounter RDAs — outpatient, emergency, inpatient — not on a structured feed coming off the IVD instrument.

Second, the point of compliance is the HIS, not the analyzer. The HIS aggregates everything that reaches it — including lab results that arrive the old way, over HL7 v2 — and it is the HIS that builds the RDA Bundle. An IPS can comply at the HIS level even while the internal stretch from analyzer to LIS to HIS stays entirely in v2. The instrument does not have to learn FHIR for the hospital to meet the mandate this year.

There is a structural feature of the Colombian chain that a box-and-arrow diagram hides. Most IPS do not run a full laboratory. They draw the sample and send it to a shared reference laboratory that serves dozens — sometimes hundreds — of providers at once. That arrangement is efficient, but it creates coupling: the translation capability of one reference lab is silently inherited by every IPS that depends on it. Build the FHIR bridge there, and many providers comply at once; leave it unbuilt, and the same gap propagates across all of them simultaneously.

This is the first place in the study where network thinking earns its keep. Compliance is not a property each hospital holds alone; it is a property of a graph with hubs. A handful of reference laboratories sit at high centrality, and the system's readiness depends disproportionately on them. That same lens — supply structure as a network with concentrated, load-bearing nodes — is the engine of Report 2, where we turn it on the import market itself.

Before judging the gap, give the State its due: the foundation is genuinely built. The IHCE backbone is real and substantial. It runs on Azure Health Data Services and bundles the pieces a national exchange needs — a FHIR server as repository and API, the MPI / VIDA national patient identity, a terminology server carrying SNOMED CT, LOINC, CIE and CUPS, an RDA Manager and Viewer, and an API Gateway for authentication and audit. The IHCE projects on the order of 400 million FHIR documents a year across 9,000-plus provider organizations. This is not a sketch; it is a platform.

But notice exactly where the platform stops. The IHCE expects each provider to operate its own ETL layer — extracting from its HIS, RIS, LIS and pharmacy systems and mapping them into the Colombian FHIR profiles. That ETL is the middleware. The State built the highway and the destination; every provider still has to build its own on-ramp. The national backbone does not close the gap of Section 2 — it relocates it, cleanly, to the edge of each institution.

The on-ramp is not unoccupied. Three distinct layers of vendors are already working the translation problem, and they are easy to confuse until you separate them by what they actually mediate.

Read the table top to bottom and the pattern is clear. The HIS layer is crowded — nearly every local vendor now advertises an RDA module — but each one generates the summary from its own system; none is a neutral engine that mediates a third party's analyzers and labs into FHIR. The manufacturer layer (Roche navify and its peers) solves connectivity beautifully, but only for its own boxes, and on lock-in terms. The engine layer is the one that truly translates — Mirth Connect supports lab analyzers natively and is the most plausible workhorse for a device-to-FHIR bridge; InterSystems brought its “FHIR façade” pattern and a Colombia office; K2BHealth built Uruguay's national EHR and is openly marketing Resolution 1888. But this layer is thin and unevenly deployed, and the device-to-FHIR “last mile” is its thinnest stretch.

To see why that last mile is so thin, compare the two countries at the layer nobody photographs. Australia spent two decades building an intermediate layer that Colombia simply skipped: a mature secure clinical-messaging / HIE network — HealthLink (operating since 1993), Medical-Objects, Argus — carrying clinical traffic between more than 15,000 organizations and over 100 million messages a year. It replaced the fax for results, referrals and discharge summaries, and in doing so it already translated and routed between systems long before FHIR arrived.

Colombia jumped straight to a centralized FHIR hub without that middle layer ever existing at national scale. Historically, clinical exchange here moved on paper, fax, or CSV and Excel files. The consequence is structural, and it is the real content behind the maturity gap noted in Part I: the last mile from devices and labs into FHIR is, to a large degree, still unbuilt. Australia poured FHIR onto a network of interpreters that already worked. Colombia wrote the law first and is now discovering it must build the interpreters underneath it.

So is the gap an open field? No — and saying so would discredit the whole analysis, because anyone who knows the market would name the two players already standing on it. Assist (Bogotá) does precisely the engine angle, using IBM ACE v13 to move HL7 v2 to FHIR bidirectionally across HIS, RIS, LIS, pharmacy and laboratory with the national catalogs — a strong fit for large hospitals on an enterprise-consulting model. Clau (a LATAM SaaS platform) translates HL7 to FHIR and normalizes to SNOMED automatically across Colombia, Mexico and Chile, and attacks on price — roughly USD 500–2,500 a month against the USD 50,000–200,000 setups of the traditional model.

That is the bridge out of Report 1. Part I gave the mandate; Part II followed the data through the pipe and located the true bottleneck — not the analyzer that everyone fears, but the unfinished last mile and the deferred lab profile that almost nobody is watching. Report 2 now turns from the pipe to what flows through it: who actually supplies Colombia's diagnostic equipment, how that supply network is shaped, where it concentrates, and where — read as a graph rather than a list — its fragilities and its openings really lie.