Debunking the Health IT Banking Myth

Shawn Vincent, Vice President of R&D at MD Practice Software LP wrote a thoughtful article in Technology for Doctors titled “Interoperability: Banking did it… how hard can it be?” In the article, Vincent explores a common myth: the comparison that is frequently made between healthcare data sharing and the banking industry.

He states, “There are many reasons for HIT’s lag when compared with other industries. Politics and costs play a big role here, but we’re not going to talk about these just now. Instead, let’s concentrate on the technical problem itself. ‘Just add standards’ isn’t as easy as it sounds!”

He also points out that the complexity of healthcare information is significantly greater than information managed within the banking industry. Bank account elements have simple structures and attributes and in fact have to be simple in order for the banks to “function as businesses”. Whereas healthcare diseases and concepts are not restricted in the same ways as debits and credits. The relationships between diseases can be varied and multiple. For example, Diabetes Mellitus can be referred to as NIDDM, DM, or Adult onset Diabetes (and many others). A clinician would know that all of these terms mean exactly the same thing. But it is much more difficult for a software application to correctly categorize these different names as the same disease. This is where standards become extremely important. All of the diseases, definitions, descriptions, and associations need to be codified and then correctly linked in order to create logical relationships. SNOMED CT (Systematized Nomenclature of Medicine – Clinical Terms) is a global standard designed to be used in clinical systems. It evolved over a 30-year period from work done in both the UK and U.S. and now comprises more than a million clinical terms. However, SNOMED CT is not widely deployed in clinical systems and — although it is a goal of EMR programs to have all systems using standardized coding systems and vocabularies — it is extremely complex to integrate.

As Vincent explains, combining historical data with new data (standardizing the information) and the continued use of old (legacy) systems frequently requires modifications to the standard. This in turn results in the standard becoming proprietary to that system; as a result, it can no longer be used as a universal standard. Sound complicated? It is!

The bottom line is that the scale necessary to get all clinical systems to speak to one another using essentially the same language is of a significantly greater magnitude than that in banking and finance. When you hear someone in the future draw parallels between Healthcare IT and banking, remember some of these concepts.

Comments

[Gerald Tevaarwerk]

Right on! Now "lets stop dithering" and put the tax payers money made available for electronic medical record keeping into development of the standardization essential for interoperability, rather than for deployment of inadequate systems that give this industry a bad reputation.

Gerald Tevaarwerk, Chair, Information Technology & Communication Committee of the British Columbia Society of Specialist Physicians and Surgeons.

[Dr Zen]

Sorry, but as a software professional who's helped implement real-time air traffic control systems, credit card payment systems, real-time satellite communication systems, etc... I respectfully disagree with this article.

I quit after working 1 month as an Architect on a health care project as I could not deal with the stupidity involved.

From a technical point of view, the challenges are relatively trivial in comparison.

The major reason for the poor state of progress is the dreadful level of bureaucracy, politics, and FUD spread by those who cannot do, but wish to protect their own empires (... this includes doctors)

The only thing needed is competition and the abolishment of Provincial EMR Programs (nothing but a barrier to entry to protect an inefficient oligopoly and a self-serving bureacracy).

Open standards would then develop naturally, efficiently, and organically in quick fashion.

[Shawn Vincent]

@Dr Zen:

I don't disagree with you! :-) The "bureaucracy, politics, and FUD spread by those who cannot do" is very real -- I just didn't focus on those parts of the problem in my article.

I guess I don't explain my point of view well enough in the original article. What I'm arguing against is the creation and forced adoption of big bang universal health care standards. This has been proven again and again to fail (not just in healthcare, although i would argue that the data complexity makes health care particularly problematic).

The much better solution is what you talk about: agree on some low level (secure messaging) standards, and let the industry exchange data. I think (although I cannot prove) that ad hoc standards that solve real world problems would quickly evolve.

This goal is actually being stymied right now, unfortunately. Mostly the people who fight against it think that the current big bang standardization efforts will save the industry, if we just wait long enough. I'm just pointing out that maybe instead of waiting for this (which I think will never happen), we should instead focus on solving the simpler problems (connectivity, security) and let the data standards evolve naturally.

Thanks!

[Sherif]

Do you think that Reaching to common Universal Language is possible in EMR/EHR? How? Where can I found these Dictionary of Medical Terms? thanks

[Alan Brookstone]

Sherif, can you explain what you mean by Universal Common Language. Is this related to coding of healthcare information?

[Shawn Vincent]

@Sherif, good question.

Many people have tried to create a universal standard for healthcare data.

HL7v3 (http://hl7book.net/index.php?title=HL7_version_3) attempts to model all possible medical actions using XML, by analyzing many many clinical use cases with working groups filled with experts.

SNOMED CT (http://www.ihtsdo.org/snomed-ct/) has attempted to build the "ultimate" coding system, going so far as to allow "post-coordinated" terms, where a single code "expression" can say things like "broken lower left leg" without having to have an infinite number of codes.

I love these standards: they are excellent to read to get a perspective on the complexity of the data modelling exercise that we are running into.

The challenges that I worry about are caused by these complexities. Because the standards do a decent job of modelling a good chunk of reality, they're *really* *really * complicated, subtle, and hard to understand. There are very few people I've met who really understand these standards. I've studied them extensively, and while I can speak intelligently about them, and have worked with them in production, I'd by no means consider myself an expert.

The failure comes when you have something this complex that has to be implemented by many real people independently.

So it's fine to sit here and rant about how bad things are, but: how to solve the problem? Is there a path we could follow that would yield real, useful specs?

I have an idea that I think may work. They keys are: simplicity, incremental improvement, and asynchronous bilateral cutover. :-)

First, you need to create a layer where messages containing arbitrary information can be shared securely between all healthcare providers. Secure email, basically. Get rid of the fax machines. This is being done already in production in the US with the Direct Project: http://directproject.org/

Next, let clinical users send messages with attachments. These attachments can contain anything: word documents, scanned files, etc. This is valuable right off the bat. Maybe create a two-way fax gateway to talk to the users that aren't on the system yet.

Next, let people attach their EMRs into it and start sending data around. Folks using one vendor could use proprietary protocols, there could be ad hoc standards or common simple standards to share data between vendors.

Left to it's own devices, this may result in some useful ad hoc standards arising after a few years (and in the meantime, lots of real world problems getting solved in pragmatic ways).

To encourage the rise of common standards, I'd propose one last step: to promote an envelope standard like MIME (Multi Internet Mime Extensions) or CDA (Clinical Document Architecture) without mandating anything about the contents. The idea is that when you send a message, you send it in multiple redundant formats: one human readable, and others with various machine readable formats. This allows new formats to be created and sent, and vendors and users to upgrade their software to take advantage of the new data. This is what is known as "asynchronous bilateral cutover", as described by Wes Richel: http://blogs.gartner.com/wes_rishel/2012/04/13/the-biggest-healthcare-interop-issue-frozen-interface-syndrome/

This approach (I claim) will result in one or more useful standards. More importantly, it would provide an ecosystem where lots of useful problems could be solved in the meantime.

And heck, we could even send HL7v3 and SNOMED CT messages over this infrastructure along with all the other messages flying around! :-)

-Shawn Vincent.
--
VP R&D, MD Practice Software LP