Interoperability is an operating cost, not a project

Most health information exchanges are budgeted as projects: a build phase, a launch, a ribbon, and then a quiet slide into irrelevance. The pattern is so consistent that I now read any HIE business case by looking for one line first — what is funded after go-live, every year, indefinitely. If that line is missing or thin, the exchange is already failing; the date is just not on the calendar yet.

The exchange is the cheap part

Moving a message from facility A to the registry is a solved engineering problem. HL7 FHIR gives you the shapes, the tooling is mature, and a competent team can stand up a working exchange in months. This is exactly why programmes are seduced by it: the visible, demonstrable part is the part that finishes.

What does not finish is everything underneath the transport. Patient identity matching degrades the moment a new facility onboards with a different naming convention. Terminology drifts the moment a laboratory updates its test catalogue. Every connection you add is not an asset — it is a maintenance liability that someone must own forever.

The three permanent functions

• Identity stewardship. Someone investigates the duplicate record queue, tunes the matching thresholds, and adjudicates the cases the algorithm cannot. This is a person, every week, not a feature.
• Terminology governance. Code systems version, mappings break, and new facilities arrive with spreadsheets instead of code lists. A terminology service without a human operator rots within a year.
• Exception handling. A percentage of every message flow fails, always. The question is whether failures queue visibly for someone whose job is to clear them, or disappear into a log nobody reads.

What I do differently now

On the state-wide exchange I operate today, the operational functions were costed before the platform was. Each of the three has a named owner, a weekly review, and a line in the recurrent budget that survives leadership changes. The technology stack matters far less than whether those three names still have jobs in year three.

So the question to ask of any interoperability proposal is not 'which standard' or 'which vendor'. It is: show me the organogram for year three, and show me the budget line that pays for it. If the answer is a project manager and a spreadsheet, you are looking at a demonstration, not an exchange.