After many years working on both the VA VistA and DoD CHCS software architectures, I came to the conclusion that we were hitting a wall of complexity that could not be handled with the technologies that we were using. In the mid 1990's, I started wandering around looking for alternatives. I found two very interesting candidates.
My first discovery was a programmer at CERN in Geneva who was working on a thing called the World Wide Web. There were 150 web sites already linked, and I'd get emails, breathlessly announcing the latest addition to the web.
One of the things I wanted to do with CHCS and VistA (and IHS' RPMS) was to create a "Universal Namespace" - giving a name to every data object handled by any system. The idea, never implemented, was to prepend the domain name to the FileMan metadata, so that any information could be linked to any other information. I had some fuzzy thoughts about using FileGrams to do this, and saw the need for some kind of reference language to link it all together. When I saw Tim's design of the web, I realized that this was what I was trying to do, but done in a really smart way. He had a certain genius of knowing what had to be right from the start (the design of the URL) and what needed to be "good enough" (HTML, HTTP). He had a very practical vision of how to make it all scale. He allowed the web to be "broken" - the "404 Not Found error" - in contrast to prior hypertext systems (Doug Engelbart, Ted Nelson) who required bi-directional referential integrity. This vision was not appreciated by all: his first paper describing the Web was rejected by peer reviewers as being "not scalable" and leading to a "bowl of spaghetti." Here is a video interview I did with Mark Frisse, one of those reviewers.
The web is arguably one of the most complex and globally transformational invention in the history of technology. But Tim did not try to create Google, Facebook, or Amazon. Nor did he create a Dewey Decimal system to try to structure the web, nor did he assign experts to standardize travel, auctions, or book sales. He started with simple initial conditions and let the system evolve - something that is very hard for many to understand. A complex system requires a complex command and control system, they would say... the more complex the system, the more complex the command and control hierarchy.
A complex information system would require a complex "librarian" function to categorize, standardize, normalize, and edit information, they would say. But as we have seen with Google, this is not the case. We can find things with Google that would have been unimaginably complex in the old days of reference librarians and Dewey Decimal systems.
Today, it is trivial to look up a book and paste the URL in an email or Tweet. There is no need for a standards group to define a meaningful use "book exchange" standard - it is simply an intrinsic part of the information space we call the web.
I have long thought that health IT needs to adopt a similar "information space" architecture - a large-scale, fine-grained network that links all of our information, all the time, from anywhere. Concurrent with this universality, we also need an equally powerful privacy and security model that can deal at the fine-grained level of individual objects. The simplicity and the regularity of this model - putting the patient at the center of the health care universe - offers a dramatic improvement in our ability to manage the complexity of health IT.
I held a workshop at the World Wide Web Consortium offices at MIT Apr 20 (minutes of the meeting coming soon), and decided to continue the discussion at the SemTech conference in San Francisco June 3.
The President’s Council of Advisors on Science and Technology (PCAST) identified the need for a universal healthcare exchange language as a key enabler in addressing this problem by improving healthcare data portability. Many familiar with Semantic Web technology have recognized that RDF / Linked Data would be an excellent candidate to meet this need, for both technical and strategic reasons. Although RDF is not yet well known in conventional healthcare IT, it has been beneficially used in a wide variety of applications over the past ten years -- including medical and biotech applications -- and would exceed all of the requirements outlined in the PCAST report.
RDF offers a practical evolutionary pathway to semantic interoperability. It enables information to be readily linked and exchanged with full semantic fidelity while leveraging existing IT infrastructure investments. Being schema-flexible, RDF allows multiple evolving data models and vocabularies to peacefully co-exist in the same instance data, without loss of semantic fidelity. This enables standardized data models and vocabularies to be used whenever possible, while permitting legacy or specialized models and vocabularies to be semantically linked and used when necessary. It also enables a limitless variety of related information to be semantically linked to patient data, such as genomic, geographic and drug interaction data, enabling more effective treatment, and greater knowledge discovery. Other reasons for adopting RDF as a universal healthcare exchange language include: (a) its ability to make information self-describing with precise semantics; (b) its support for automated inference; and (c) its foundation in open standards.
I think that this is an exciting path forward for VA/DoD sharing, and the logical next technological step for the metadata models begun with VistA and CHCS...