Improve FHIR Search Performance with Subject References

[alexanderkiel]

The performance of FHIR Search queries that include a search parameter of type token and a subject or patient search parameter should be improved.

The queries have the following form:

[base]/[type]?[token-param]=[system]|[code]&[subject/patient]=Patient/[id-0],Patient/[id-1],...

were:

• type – a resource type that is part of the patient compartment like Condition or Observation
• token-param – a search param of type token like code
• system, code – the value of the token search param that has to include a system URL
• subject/patient – either a subject or a patient search parameter
• id-0, id-1 – one or more patient ID's

Other additional search parameters are allowed. The order of the search parameters doesn't matter.

In case all requirements hold, the performance optimisation will use the patient compartment indices in order to process the query. Using the patient compartment indices ensures that the combination of token search parameter and patient is directly accessed via a single index.

Performance is documented here. TLDR It's super fast.