ITI White Paper: Publish/Subscribe Infrastructure

Introduction

Event driven information exchange patterns dominate the data interchange in most healthcare settings. For example, most HL7 version 2.x interfaces send messages based on events within the sender's system. Most current IHE profiles assume either static, out-of-band determination of the senders and receivers of event driven information exchange, or describe query-response patterns. There is a need for a profiled dynamic infrastructure for event-driven information exchange patterns within IHE. This white paper describes such a framework based on the publish/subscribe data exchange model, and the methodology to apply the framework to IHE profiles.

Publish/subscribe patterns of data exchange are conceptually simple, and involve a limited numbers of actors and transactions. The transaction allow for automating the determination of information consumers based on events or content "topics". For example, if an IHE profile describes information content where a diagnosis is present and coded using a particular coding system, a subscriber can request to receive any transactions where one from a particular set of diagnosis codes is present. If the subscription is accepted, the system which keeps track of information recipients will start sending the transactions which match the described criteria to the subscriber.

The above example demonstrates two important issues which need to be addressed by profilers and implementers of publish/subscribe interactions. The first one is that the implementation of transactions and actors are dependent on the information exchange environment. Prescribing a specific technology for publish/subscribe is not a practical option - it is unlikely that, for example, IHE actors which exchange information using DICOM transaction will be inclined to add Web Services based transactions for enabling publish/subscribe. To address this issue this paper describes the publish/subscribe actors and their transactions in an abstract manner, and present the way to make these abstract constructs into regular IHE actors and transactions.

The second issue that needs to be addressed is the use of subscription topics. Based on the information exchange environment, topics can be described in various ways. This paper presents various approaches to specifying topics, and some discussion on how these approaches can be used in IHE profiles.

The implementation of publish/subscribe in healthcare environments also needs to take into account the need for security and privacy of the exchanged information. The methodology to include publish/subscribe transactions in IHE profiles contains a discussion on the security implications arising in such situations.

As a proof of concept, Appendix A of the paper demonstrates the use of publish/subscribe as part of the XDS profiles.