Most Kafka “healthcare interoperability” examples are theoretical or focused on US/EU EHRs. I wanted to share a concrete case from a mid‑size hospital in Venezuela where we used Kafka as the event backbone between HIS, LIS and billing systems.
Context:
4–8 systems needed to know the same things in near real time: admissions, lab results, charges.
Point‑to‑point integrations had become a mess: custom scripts, duplicate records, fragile error handling.
The hospital knew it would keep adding systems (BI, SAP, new clinical modules) over the next few years.
What we ended up doing:
Defined topics by clinical domain (admissions, lab-results, orders, billing-events) instead of “one topic for everything”.
Partitioned by patient or encounter ID to preserve clinical event order per case.
Used replication factor 3 and min.insync.replicas=2 because losing a lab result is not just a business data issue.
Enabled idempotent + transactional producers (exactly-once semantics) for lab results and financial events, where duplicates are harmful.
The article goes into more detail on:
How we mapped HL7/FHIR messages to Kafka topics.
Trade‑offs in infra cost vs. resilience for a hospital context.
When Kafka was clearly the wrong tool (small clinics with only 2–3 systems and low event volume).
Link (in Spanish, but I’m happy to summarize or answer questions in English here):
Kafka in hospitals