The Marrakesh Express
French, Berber, and Arabic; sounds like HL7!

The Marrakesh Express song performed in 1969 By Crosby, Stills, Nash, and Young at Woodstock, was written by Graham Nash during his train ride from Casablanca to Marrakesh. Initially, he was traveling first class, but soon he found it very boring, and moved back to third class “where it was happening.” He changed the lyrics to “Ducks, and pigs, and chickens call, animal carpet wall-to-wall.” I still remember that song vividly as I saw the Woodstock movie a few times, which was a life-changing event for me then.

When I got the opportunity recently to visit Marrakesh in Morocco, I was really intrigued and wanted to experience that same magic. I did not take that infamous train, but my wife and I rented a car and drove a bit south along the coast and eventually toward Marrakesh. We saw a similar landscape, with goats on top of vans, which were packed with people coming from and going to local markets and little places to eat the “chicken tajine,” a mix of meat, veggies, and fruits prepared in an earthen pot on top of a charcoal fire.
Navigating through Morocco by car was not that hard. We noticed, however, that road signs were in three different languages. I easily recognized French and Arabic, but there was a third alphabet that was unrecognizable, so I had to google it to find out it was the Berber language. I have seen signs with multiple languages in other countries, for example, in Singapore, there are four: Chinese, Malay, Tamil, and English; but some countries only use the signs in the local language depending on where you are, and the direction you are traveling. I will never forget that I missed an exit when traveling from Paris back to Amsterdam seeing a sign for Anvers—not realizing that it was the French version of Antwerp. Having multiple languages in certain countries is a fact of life, and instead of forcing everyone to speak and read one language, it is just not possible—or realistic.

The same applies to the different “languages” we have in healthcare imaging and IT. The CIO of a large hospital asked me once why everyone doesn’t “speaks HL7” because it would make his life so much easier. I had to explain that each communication standard has a specific purpose. For example, converting binary encoded pixels representing density in an X-ray is hard and cumbersome to represent in an HL7 standard because HL7 is optimized to exchange textual information. But even if everyone spoke HL7, it would not help because there are probably as many HL7 dialects as there are deployments, due to local and vendor-specific semi-proprietary extensions.

The HL7 standard by itself had an interesting evolution, starting with V2 in the late 80s which has become the de-facto standard for exchanging patient demographic information, orders, and results from diagnostic procedures such as imaging, lab, and much more. HL7 V2 “glues” the many computer systems together and can easily consist of over 100 systems in a typical hospital, with tens of thousands of transactions daily. These operations typically flow through an Interface engine that deals with many different dialects, versions, and extensions.

There have been at least three attempts to replace V2 with a more modern standard. The first attempt created HL7 V3 which replaced the ancient “pipe” encoding with XML, but that died gracefully after it choked the hospital infrastructure because of its verbosity. In addition, it also was universally hated by implementers due to a lack of tooling to develop their interfaces. A spin-off of V3 was the Clinical Document Architecture (CDA) which is still used quite a bit to exchange documents between EMRs of different institutions. A typical CDA is a hospital discharge form that can be ingested by a physician’s EHR to update medications and other clinical artifacts. However, most EMRs archive these documents “as-is” so that a specialist would have to electronically flip through many documents to find what happened at an ER episode. The third attempt to replace HL7 is somewhat promising, it uses the FHIR (Fast Healthcare Interoperability Resources) standard based on web technologies; however, it struggles with immaturity and version control as most of it is still being debated.

The fact of life is that we need to deal with multiple “languages” for the foreseeable future and need to be able to connect these systems with middleware that can map an HL7 order to a DICOM worklist while accessing an FHIR resource for a patient MPI, while mapping and knowing the many different dialects. And a logical extension to being a “broker” in the middle of these different interfaces is that it allows for orchestrating the workflow of an institution as well.
Back to Marrakesh, yes it was magical! I recommend it, but navigating the country and ordering food was challenging. Apparently, when you get older, your short-term memory capacity shrinks, but the long-term memory should stay; however, I find that not much of my five years of college French was retained. But to see that a country can apparently function very well speaking and writing different languages, each with a completely different alphabet was encouraging. We should be able to do the same in healthcare; HL7 V2 is not DICOM, and looking at a JSON-encoded FHIR message is quite different than interpreting a DICOM header. But nevertheless, it all seems to work . . . kind of.

Herman Oosterwijk is a trainer/consultant and guest writer for Dicom Systems.