Why Two Hospitals' Systems Don't Just Talk

There is no single EHR every hospital runs. More than a dozen major vendors compete for hospital contracts in the US alone, and large health systems, small clinics, and specialist practices each pick from that list, and nothing forces those vendors to store data the same way underneath. Two hospitals five miles apart can be running completely unrelated software, built by different companies, on different underlying database structures.

That would be a minor inconvenience if patients only ever saw one hospital in their life. They do not. A referral to a specialist, a transfer after an emergency, a second opinion — all of them require one system to somehow hand data to a different system that was never designed to receive it.

Compare the two mock records below. Same patient, same diagnosis, same date — coded under two different standards, formatted with two different date conventions, and filed under two differently named sections of the chart.

The mismatch above is worth naming precisely, because “interoperability” actually splits into two separate problems that get solved in two separate ways.

Solve syntax without semantics and you get a system that opens the file successfully and still cannot tell you whether the patient has diabetes — which is exactly what the two mock records above demonstrate. The message technically arrives. Nothing in it means anything to the system receiving it. Fixing syntax is a plumbing problem, and it is the easier of the two. Fixing semantics means every vendor agreeing on the exact same code for the exact same condition, which is a coordination problem at the scale of an entire industry.

Not every failure to exchange a record is an accident. Sometimes a vendor or a health system has a real financial reason to make data exchange harder than it needs to be — a practice with a name in US health policy: information blocking.

You could imagine solving this case by case: write a one-off translator between these two specific hospitals' systems. That approach breaks the moment a third hospital joins, because now you need a translator for every pair — three systems need three translators, ten systems need forty-five. Goodwill and one-off integrations do not scale; a shared standard that every vendor agrees to speak does.

What actually moved the needle historically was not goodwill, and it was not engineering elegance either — it was regulation forcing every vendor to adopt the same deadline at once, the same lever that eventually curbs information blocking above. Left purely voluntary, a shared standard is a coordination problem nobody wants to move first on. Made mandatory, it stops being a competition and starts being table stakes.