Mar 19 2009
This is a very big rock that I’m pushing up a very big hill, but I guess I try it one more time.
I just ran across this hearing announcement from the House Armed Services Committee:
“The Joint Military Personnel and Terrorism, Unconventional Threats and Capabilities Subcommittees will meet to receive testimony on Department of Defense Health Information Technology: AHTLA is “Intolerable,” Where Do We Go From Here?”
A little evolutionary history here: AHLTA is a $5 billion DoD-wide hospital information system replacement for the $1.6 billion Composite Health Care System (CHCS) system that I worked on at SAIC. (CHCS was an adaptation of the VA’s VistA system that I worked on while an employee of the Veteran’s Administration. VistA is an open source Electronic Medical Record system.
30 years ago, in December 1978, I attended a meeting with folks from the VA, DoD, Indian Health Service, and others about building a shared software infrastructure for medical records. (This is where I met Rob Kolodner, who has patiently stuck it out in federal service and the National Health Information Network.) It was an exciting time for me to think up a software architecture that would serve as a foundation for federal health information. I was big on the Whorf-Sapir hypothesis and how language can shape our thinking, so my idea was to design a speech community built around the language that I thought would communicate the needs of the community using it. Rather than trying to integrate all the “silos” I saw in the bureaucracies surrounding me, I sought to build tools that allowed folks (or sometimes forced them) to communicate. Out of this came the MailMan system, one of the first implementations of the SMTP mail protocol widely used today. I saw all of this as primarily a problem of communication; the data base was merely one aspect of that communication.
The system was wildly successful in the VA, winning numerous awards and becoming a topic in Philip Longman’s Best Care Anywhere, Why VA’s Health Care is Better Than Yours (see my video interview of him) I interfaced the VA Loma Linda’s system to the neighboring March Air Force Base hospital in 1983, which drew a lot of congressional attention. The DoD responded by hiring Arthur D. Little consultant (this is where I met John Glaser, who has since rehabilitated himself with Partners Health Care) to study the conversion, paying them more to study the installation than it cost to install it. This was my first introductions to beltway economics.
Congress looked at the situation and saw implementations of the VA system working at Fitzsimmons Army Medical Center and March Air Force Base and decreed that one of the DoD’s Composite Health Care System competitors had to bid an adaptation of the VA’s system. (The could have simply decreed that the DoD would deploy the VA system, but that would have been far too simple and would have disrupted the beltway economics food chain. The elephants had to be fed.)
I moved to SAIC from the VA to work on the conversion effort, and was stunned to see the DoD’s software development style. I was hit with a book of 9600 requirements to be met, some of which were throwbacks to the era of the punched card. We would have perfectly functional software that would do the same thing as the requirements, but had to modfiy things to meet the weirdly specific requirements. This introduced me to the DoD’s “requirements” relationship to the feed-the-beltway laws of economics.
SAIC won the award for CHCS for $1.01 billion, about 60% of our nearest competitor (McDonnell Douglas bid $1.7b). The DoD requirements, however, forced the system to fork from the VA’s version. These changes were forced by a brittle, unthinking requirements management system and DoD’s hyper-specific mindset.
DoD eventually decided to replace CHCS with a more vendor-friendly version, feeding the beltway with a $5 billion replacement. I took one look at the architecture and my heart sank… it was the highly centralized, single-point-of-failure design straight out of the 1960’s thinking of mainframe computers. It was based on proprietary, vendor-specific designs that we had so studiously avoided in VistA in order to keep an open source system. Two decades of evolutionary growth was simply being thrown away in order to appease the beltway bandits.
The bureaucracy ground on, and AHLTA was deployed. I immediately began hearing horror stories confirming my initial repulsion to the architecture. Hospital users received cryptic messages such as:
“AHLTA is experiencing latching issues at the Clinical Data Repository hosted at the DISA computing facility in Montgomery, Alabama. This issue has forced local instances of AHLTA to go into failover mode across the Enterprise. Failover mode means AHLTA will be performing in a degraded mode till such time as the cause of the latching issue is discovered and remediated. DISA is investigating the cause but have no anticipated recovery time at this moment.”
How anyone could design in which the “latching issue” in Alabama could shut down a worldwide complex of medical information systems is beyond me, but such is AHLTA, the bureaucracy, and the beltway bandits that support it.
The remediation of the “latching issue” is to never have designed such a system in the first place.
Here’s what I predict will happen at the hearing:
1. There will be lots of self-righteous indignation at the AHLTA system. People will be outraged at such a terrible state of affairs.
2. No one will mention that the architecture was self-evidently flawed from the outset. I predicted its state today with just a quick glance at the approach they were following. People will throw rocks at the system, not the management style and procedures that produced it.
3. They will probably change the name of the system, and maybe the program office managing it. Same software, same people, but a new and improved name.
4. The people at the hearing next week will likely end up some time in the future as employees of government contractors who will be providing services to the future son-of-AHLTA. The future job security will be based on proprietary, vendor-specific systems. They can comply with all the conflict of interest policies and be squeaky clean, but their post-government employment prospects are heightened by vendor-friendly solutions.
5. In 10 years, we will have another hearing of Congressional Indignation at the state of shared medical information technology in the federal government, not unlike the Sonny Montgomery hearings of decades past.
Here’s what I think needs to be done:
1. Focus on a personal health record approach, designing the system around the individual, not the bureaucracy treating the individual. This business of creating outrageously complex, expensive, and brittle enterprise systems is great for the beltway economy, bad for the patients and taxpayers.
2. Change the incentives. Federal managers who make their tasks more complicated requiring more workers earns greater job security and a promotion. Those who streamline their work, share resources with other agencies risk their jobs and their retirement. Congress can huff and puff about VA/DoD sharing all they want, but as long as this translates to loss of jobs to those responsible for implementing it, they will find themselves stonewalled. I was demoted for my work in the VA, because I was working laterally with my peers rather than building up a bureaucratic pyramid below me.
3. Blame DoD policies and management for the fiasco. The AHLTA system is just the symptom of a much deeper problem.
Leave a Reply
You must be logged in to post a comment.