A Vision for Scalable High Quality Health Information Exchange - Part 2

This is the second of a three-part article looking at the area of interoperability and health information exchange (HIE) in the healthcare industry. In the first part, I intended to clearly articulate the key challenges and barriers to adoption faced by those looking to engage in HIE. Part 2 will examine an architectural approach to address those challenges and discuss some technology enablers to realize a vision for high quality HIE. In the last part, I will share with you a success story from industry where this approach was put to the test.

In the last post, I touched on some of the business and technical challenges that have led to slow growth of wide scale health information exchange and interoperability initiatives. Clearly this stuff is hard, as is to be expected with goals to improve the quality and cost of delivering healthcare. And I should point out that it's equally important to acknowledge the significant political, organizational and policy hurdles to HIE adoption. Ignoring those aspects can bring any new business initiatives to its knees, not just HIE. So let's put this out of scope for the time being. Doing so will allow us to focus on what I argue is the most important technical enabler for high quality health information exchange: getting the architecture right.

Systems integration is not a new phenomenon. For decades, disparate business and administrative systems have been modified, customized, or adapted to share information. The first architectural approach that gained wide scale adoption was adapter-oriented integration. In this model, each endpoint would be fitted with a custom software adapter responsible for translating data to and from other endpoints . Each endpoint in the architecture was required to do the same, and all adaptors where deployed by all network participants. This was fine for relatively small deployments. But it didn’t scale and was expensive to build and maintain, since point-to-point integration requires ongoing change management. The biggest limitation was that costs would rise exponentially as the number of network participants increase. For example, connecting 4 systems together requires 12 bi-directional interfaces, 10 systems called for 90 interfaces, and 100 systems required 9,900 interfaces!

Not long after, the Integration Broker (IB) approach gained broad acceptance with the emergence of EAI in the 1990s. But this just shifted the management of said adaptors to a central broker, and reduced the scale pain to linear from geometric i.e. total number of interfaces to write has been reduced. Any gains realized by the broker architecture tended to be offset by additional costs to maintain the integration solution. Another limitation of this approach is its reliance on being centrally deployed, which presents both network and compute scale challenges for large, distributed deployments. Compute performance could be mitigated by the development of optimized custom wire and data formats, but usually meant locking-in participants to a specific vendor package.

In the last few years, Service-Oriented Architecture (SOA) has re-emerged in a big way, heralded along with Web Services standards as the savior for systems integration. In a SOA, standard technologies and communication protocols enable a common framework for access to and distribution of information. Adoption of XML as the lingua franca for information sharing has leveled the playing field for how data is packaged for exchange between systems. Both of these, along with the core principles of separating access from implementation, loose-coupling and separation of concerns have enabled the distribution of the integration function more evenly across the network and its participants. So SOA and web services really has emerged as a best practice for complex data integration. This is what is commonly referred to as the SOA Network Model.

Bringing our topic back to healthcare, a service-oriented architecture is fundamental to the development and deployment of a successful HIE initiative. Leveraging the SOA Network Model for wide scale health information exchange can substantially reduce cost and complexity, addressing the challenges described in the previous post. This approach is founded on the creation of a standards-based information bus , on which participants agree to share information that conforms to a common, or "canonical" information model based on XML. For HIE between hospitals, IDNs, regional or national networks, this information is often shared as a set of messages based on the HL7v3, CDA/CCD, or ASTM CCR standards.

Participants are responsible for mapping their source data to the target "canonical" data model, and because all sources are delivering data onto the bus in a standard format, shared services can be offered to provide capabilities like semantic translation, record location, identity mapping, security policy, etc. This technique yields very high quality information exchange, as well as a one-to-one conversion cost between a network participant and the collective network services, which is much simpler and more cost effective than alternative integration models. The bus is highly distributed (often over a secure Internet channel) so that all end points can communicate without the need for a central mediator, and shared service endpoints can be virtualized i.e. deployed as location independent. As an added cost benefit, as new participants join the network there is no need to regression test all the other endpoints.

The net result is an architecture which is semantically precise, loosely coupled, and highly flexible to changes in number of participants, configuration and aggregate capability. But there's still one significant technical hurdle with an architecture like this (and even more so in IB deployments): highly verbose XML data formats and web services protocols can often hamper the performance and scalability of the architecture. Intel has addressed this in its new SOA healthcare interoperability platform, SOA Expressway for Healthcare. Intel SOA Expressway for Healthcare is a specific implementation of a new product category called a SOA "soft appliance", which delivers a breakthrough in simplicity, cost and scalability of implementing data interoperability. It natively delivers on the vision for this architecture and has the performance, scalability and security functionality needed to support an HIE deployment at scale.

In part 3 of this article, I will share with you some success stories from industry where SOA Expressway is enabling high-quality health information exchange via this very architecture.

For more complete information about compiler optimizations, see our Optimization Notice.