DSTU2

This page is part of the FHIR Specification (v1.0.2: DSTU 2). The current version which supercedes this version is 5.0.0. For a full list of available versions, see the Directory of published versions . Page versions: R5 R4B R4 R3 R2

2.9 Managing Push and Pull

FHIR Infrastructure Work GroupMaturity Level: N/ABallot Status: DSTU 2

The primary focus of FHIR is exchanging healthcare data between systems. Whenever one system has information that another needs or wants, one question that automatically arises is whether the source system will push the data, or whether the destination system will pull the data.

Push As new information becomes available, or whenever it deems appropriate, the source system sends the information to the destination. It expects the destination system to maintain and index the information on receipt. The source system must trust the destination system to manage access/security appropriately
Pull The source system maintains and indexes data. When the destination system wants or needs the information, it must retrieve it form the source system. Either the source or the destination (or both) manages access/security

Note that it is possible to mix these modes by adding an interface repository into the mix. For example, to connect a push-based source with a pull-based destination, simply have the source push information at the repository, and the destination can pull data from it as required. If the source is pull based, the interface repository can query it for new data on a regular basis, and when it finds any, it can pull it into the repository and/or push it to the destination.

Which combination of push, pull and interface repositories is appropriate for any given context depends on a host of factors around deployment architecture, network and business topology, security policies, etc. However, it is often constrained by the capabilities of the various systems and the standards used in the exchange.

The FHIR specification supports and enables both push and pull:

  • REST: The REST API can be used in either fashion - this is discussed in more depth below
  • Messages: FHIR defines message events for both push and pull
  • Services: Services may be defined to support either model
  • Documents: This question doesn't apply, because documents are bundles that are exchanged using either REST, messages, or services

2.9.1 Push and Pull on a RESTful interface

  • Push: The source is a client, and when new data is available, it uses create, update, and transaction to push data to the destination, which is a server
  • Pull: The source is a server, and the destination is a client. When the client wants to use the data, it uses a combination of search and read so the users can find the data they need
  • Push/Pull: The source is a server, and the destination is a client. On a regular basis, the destination queries the server for its history, and replicates the changes in its own copy of the database