Jan 21 2013
I just noticed that the VA VistA Electronic Health Record System was named as one of The 10 oldest, significant open-source programs in the company of Linux, Python, Perl, and other software. Back in the early days of VistA, the term “Open Source” didn’t exist – we called it “public domain.”
What makes VistA unique in this list is that it is more than just code, it is a longitudinal data base extending back over 30 years across 172 hospitals and millions of patients. This is an irreplaceable resource – VistA has petabytes of information about clinical conditions, activities, locations, and demographics from a period when most other hospitals were just dreaming about putting their records online. This information is mapped by a “data dictionary” – a “road map” to the data base that defines the semantics (what it means) of the data, not just the syntax (how it is formatted). It is also structured around a network model of information (the connectors between the dots), rather than just a hierarchical model (think of how the Dewey Decimal System tries to form a hierarchy of the books in a library).
This data is not complete, perfectly formatted, nor coordinated with precision. Clinical data is not research-grade clinical trials data, and the patient population is that of (mostly male) veterans. But it is a treasure trove of information, waiting for future data mining. And what we consider “junk” information today (say, appointment schedules from 30 years ago), may hold great value in future research (say, how epidemics spread).
I hope the VA has the wisdom to retain this information in its original form.
At the 2012 OSEHRA summit meeting, I happened to be on a panel with medical researcher Leonard D’Avolio who commented on how easy it was to extract research data from VistA. It was quite a touching moment for me, to have helped set in motion something that captured data that would be available 33 years later.
To quote my friend Doc Searls: “When young you think life is a sprint. When older you see it’s a marathon. And when mature you see it’s a relay race.”
The question, just who are we handing off the baton to? Is the medical information we’re dealing with today going to be available 33 years from now?
I see a number of necessary conditions for this to happen:
1. We need a curator. Some organization needs to take on the role of “librarian” to maintain the archives. The VA is a likely candidate for this, but even so, I think they might need a bit of prodding to keep them focused on their responsibility. And maybe if Chuck Hagel (one of the Fathers of VistA) is confirmed as Secretary of Defense, he’ll recall his roots as one of the fathers of VistA and bring DoD medical information along, as well. Roger Baker, Assistant Secretary for Information and Technology for the Department of Veterans Affairs, is moving things along so well that he earned a VIP membership card in the Underground Railroad.
2. We need to manage the data at a “meta” level. The VistA Data Dictionary is being updated to more modern semantic web concepts, for example through Conor Dowling’s Semantic VistA. The PCAST report called for greater use of Metadata, apparently unaware that they were recommending an approach already used by VistA for 1/3 century, which they cited as a case study in a successful large-scale system. This is a bit like Monsieur Jourdain in Moliere’s play Bourgeois Gentleman, who suddenly realized that he had been speaking prose all his life, but didn’t know it.
3. We need to continue the tradition of Open Source. There is some movement in this direction with the OSEHRA effort, but it’s funding is but a tiny fraction of what it needs to be.
4. We need to fund the effort. Some folks think that Open Source means “free labor” – that the VA can just declare VistA to be an open source system and programmers will flock to it, offering volunteer time to do the VA’s bidding. This isn’t going to work. VA needs to fund these efforts, and release the work as open source.
5. We need to think of what we are building as an “information space” rather than just an “integrated electronic health record.” If someone said that they were going to improve our legal system by systematically storing and arranging lawyer’s documents, most of us would be a little skeptical. But there is a huge chasm between our goal of improving our health care and the current task of automating the medical record. As we’ve seen with Twitter, Facebook, Google, Linked In, and Wikipedia, there are huge benefits to focusing on “connecting the dots” in large scale systems, rather than just focusing on specific dots. Here’s a webinar I gave at Kitware, a leading open-source company.
All in all, though, it’s pretty cool to see VistA placed in this company.