HAPI vs GCP Healthcare API: Discrepancies in FHIR Server Selection

Details

This article reports on an empirical experiment comparing two major FHIR server implementations: HAPI FHIR and Google Cloud Healthcare API. The author loaded 1,000 Synthea patient records into both platforms and ran identical queries. The results showed that every single query diverged in at least one measurable field, indicating fundamental differences in how the servers interpret or process data. A major finding related to data ingestion was the stricter validation profile of GCP Healthcare API. For instance, GCP rejected resources like Claim or Questionnaire if they deviated from specific canonical formats (e.g., handling duplicate URLs or missing nested elements), while HAPI FHIR silently accepted them. Furthermore, the article highlighted critical operational differences: GCP imposes a hard limit of 4,500 entries per transaction bundle and enforces write quotas per minute. In contrast, HAPI FHIR demonstrated no such limits, allowing larger bundles to be processed without immediate failure. The author warns that relying solely on one server's sandbox environment can hide bugs that only surface when migrating to a stricter production profile. This is particularly critical for longitudinal patient records (chronic care), where the most clinically valuable data might fail to import or process correctly due to these vendor-specific limitations.