The EHR Game Has Yet to be Decided

Recently I've mentioned two Institute of Medicine (IOM) reports (“To Err is Human” and “Improving Diagnosis in Health Care”). If you’ve read them you know that, although a great deal is known about medicine, our knowledge is woefully, unavoidably incomplete. This, coupled with the fact that patients refuse to get sick in the ways the textbooks say they will, interferes with accurate diagnoses. You also know that even with best intentions, errors are difficult to eliminate. It will require more people and assistive technology if the error rate is to significantly reduced. In short, the expectations of both practitioners and the community at large are greater than what is possible today. We are, quite simply, limited by a lack of knowledge, too few people and not enough money. Our ‘eyes’ are bigger than our ‘stomachs’.

A long-time colleague of mine has, of late, begun to think about this subject and had an epiphany. He independently arrived at conclusions similar to those of the IOM. He contacted me excitedly with the idea that we could develop a ‘killer app’ that would address some of the problems. This was my response to him:

I have followed this field pretty closely, always from my skeptical and increasingly philosophical point of view. I have watched the software that Octo Barnett and Homer Warner created, and other stuff that mimicked theirs, as time has passed. The names applied to the apps have changed regularly but there has been little evolution and only rarely has a new species appeared on the scene. What today’s apps can do is severely constrained by their inheritance. That’s why all the big players sell ‘legacy systems’ - legacy is another word for inheritance. For all the reasons enumerated by Daniel Kahneman, people can’t bring themselves to abandon their inheritance: Greed, risk aversion, sunk cost fallacy, etc.

One might also say, with some justification, that it would be unwise to abandon their legacies and cash flow without an idea of what to do differently/better. It is my contention that the initial designs of these systems was ad hoc and pragmatic. The pioneers (understandably) built the first thing that came to mind that seemed to fall within the scope of what the available technology could support and was likely to perform some useful function immediately. The designs were not based on theory - there wasn’t any. The developers had administrators breathing down their necks and had to produce results. Taking the long view would have interfered. Octo says much the same thing in his undated piece, “History of the Development of Medical Information Systems at the Laboratory of Computer Science at Massachusetts General Hospital”.

In particular, the great failing of the early (and hence current) designs is that, rather than unifying the creation of records that succinctly describe the patient’s condition and treatment with the collection of the specific, discrete data elements needed for process-control and analysis, they treated the two as separate subjects. The majority of effort was concentrated on data collection (it’s what computers do best and easiest), while documentation was an afterthought forced to ride in the system’s rumble seat.

Now people are confronting the truth. The data that is being collected is divorced from its context. Its meaning is difficult to ascertain with certainty even immediately after collection and whatever meaning it might have will have a very short half-life.

Take a physician order as an example, the order-entry system creates the context necessary for the order to be processed. It is sent along with sufficient identifying information to the pharmacy that, in turn, know what to do with pharmacy orders. There is little certainty, however, that the doctor ordered the correct med. The system may (as does Cerner at the County) display every drug that can be ordered by any user of Cerner anywhere in the country. There is no indication which meds and which dosage forms are available at our pharmacy - you just have to ‘know’. That kind of unwritten context may work in the short-term but who, 10 years from now, will appreciate the subtlety of the challenge facing the doctor and imagine what mistakes might have been made to account for the otherwise inexplicable data than drug x was prescribed, and perhaps administered instead of drug y.

Kahane from Harvard is a prominent exponent of big data and how it’s going to make the future of medicine bright. All that is required to realize big data’s promise is to overcome about two dozen poorly understood technical hurdles, and a few sociological ones), something he believes likely in the near future.

I dream of the same goal. I just don’t believe it’s possible to realize that goal by continuing to follow the same path that has failed to fulfill these expectations in the past. It’s a bit like the search for the Northwest Passage. We had to wait for a new paradigm (global warming) to assert itself. Once the ice clears, and when medical records contain material that is equally meaningful to both people and machines, the prospects of reaching both destinations will improve greatly.

People regularly lament diagnosis and treatment errors and have ideas about what could/should be done. The merit of many of those ideas is almost self-evident. How to get there, not at just an isolated institution, but globally, is not evident at all. In my opinion it would be next to impossible today to build a generalized ‘killer app’ that would be game-changing and, even if it was possible, the likelihood of it being a success given the regulatory environment and the state of the market is low.

My hypothesis is that the current systems will ‘hit the wall’ in the not too distant future. The trigger-event will be the advent of some new technology (or regulation) that forces that many systems need to be replaced in a short span of time with ones (that don’t exist) that meet new/different requirements. What paradigm, what theory and principles will inform the design of these new systems? At the moment there isn’t one, except for what I’ve written about in this blog, and on that subject I’ve said about all I have say. It’s all there in the archive should anyone care to reread or study it.  

For this reason, I’m shifting my attention away from analyzing and discussing the already obsolete systems of today. Instead I will attempt to articulate a theoretical framework that might be useful when it comes time to develop that next generation of systems. That will take some time and a lot of effort, so it’s good bye for now. I’ve enjoyed the writing. I hoped you’ve found the articles informative or provocative.

Au Revoir, Sayonara, Hasta la vista, Arrivederci