EMR: How likely?

I often wonder if we will ever have a real electronic record (EMR) in the USA. I get pessimistic for a number of reassons, three of which I mention here. One reason is the Prisoner's Dilemma problem. Although there is a clear case for a universal EMR benefiting the public health, the benefit to those who must define, construct and maintain it is far from clear. Likewise, an EMR will involve expenditures by the implementers, expenditures of a magnitude as yet undefined.

A second concern has to do with the need to convert data from legacy systems and the cornucopia of headaches that will ensue as we try to convert. In addition to the costs and the difficulty of conversion, there is the risk of losing data. When businesses convert computer systems and lose data, dollars can be lost; when heatlhcare convert systems and lose data during conversion, lives can be lost. Most of us still value lives over dollars.

The difficulties in conversion stem from the indeterminate complexity of the problem domain. If you enjoy reading things I write in this blog, you most likely don't need to be told how complex the underlying data of biomedical research is. You already know there is no final answer to that question. Does biomedical research include biology? It seems logical, but if so, where does it stop in that domain? The same questions can be asked of every science except astronomy, mathematics, and geology. Biomedical research includes but is not limited to parts of physics, chemistry, psychology, sociology, and anthropology.

Further complicating the picture, data from human subjects are subject to an intricate web of regulations that differ from country to country and from context to context. Here in the USA, the HIPAA Security and Privacy Rules apply to any identifiable data; the so-called Common Rule for human subjects protection applies in all cases; the FDA Good Clinical Practice predicate rule applies in FDA-regulated clinical trials. Federal and state legislation like that arising from the stem cell controversy may apply. (GCP) and other Predicate Rules apply to data captured for FDA-monitored clinical trials. Institutional review boards (IRBs) can apply rules specific to the institution, which may include what amount to local customs but are largely driven by the applicable informed consent documents. These are often done study-by-study and are natural-language documents that include terminology with both legal and clinical implications.

If this is not enough complexity for you, add to these the whims of the sponsor, investigator, and biostatistician, who all have ideas of their own about what data must be captured and how they should be validated. And all this is before we even get close to the real problem, which is semantics. When humans communicate, they negotiate meaning; I say specimen, you hear biosample, if that is your term for some biological substance - tissue, serum, or whatever - obtained from a human or other carbon-based life form. When a computer says specimen, another computer hears specimen, and neither computer has any idea what the word means, so if there is no translation mechanism in place in which a human has associated the term specimen when used by computer system A with the term biosample within computer system B.

The problem is further complicated by the issue of granularity. Suppose computer system A has a data element called gender that uses the value M for male, F for female, U for unknown, and R for refused to answer. The field is required. Suppose computer system B has a field called sex that uses 0 for male, 1 for female, 2 for unknown, and 3 for refused to answer. The field is required. No problem translating: gender=sex, M=0, F=1, etc. But what if the field is not required in system A, and the absence of the value means (we must assume) that the field was inadvertently skipped? Should system B assume the value is unknown, or the respondent refused to answer? The correct answer is that in either case, you lose information going from A to B.

Now, starting with the orginal metadata for the two systems, suppose system A has a value H that means hermaphrodite, a value S that means gender reassignment by surgery - male to female, and a value T that means gender reassignment by surgery - female to male. Moving data from system B to system A is easy - 0=M, 1=F, 2=U, 3=R - but moving data from A to B is essentially impossible. What is the corresponding value for S? Is a surgically transgendered person his or her original gender, or his or her current gender? The correct answer is, of course, "it depends on the context". Genetically speaking, the Y chromosome is still there, but in a mental health context, the person's voluntary feminization is a vital part of the picture that needs to be recognized and addressed. Even adding new values to system B's domain for this data point does not necessarily help. All that will do is break system B unless it is modified to take the new values into account.

This may seem like a made-up scenario, and it is, but the problem is not. A great many standard vocabularies define gender; of these few have the same level of granularity. In fewer cases still is there agreement on eitherr the names or the encoding values even when the semantic granularity is the same.

An EMR will need to standardize a very large number of common data elements. The semantic problems I have described are not unique to gender. The standards process will go on for a long time, and per ONCHIT's official stance, the definition is expected to result in a certification process for EMRs. This process will act as a concrete requirements specification for conversion of legacy system, which will then lead to cost estimates.

Here, I would predict, the process will break down. Who will fund the conversion? Given the pace of scientific advances occurring around us, which is breathtakingly rapid and accelerating, who will maintain the EMR, and how? Will those who pay for it benefit from it, and if not, who will compel the required parties to fulfill an unfunded mandate?

In my darkest hours I wonder if this is simply some sort of dodge invented by people who will profit from the attempt - the contracts involved in setting all this up have been, are, and will be quite lucrative. Then I wonder how I can get in on the bidding.

My third concern is the dilution of effort through well-intended but doomed universal standards efforts. HL7 version 3 is one example; caBIG is another. I am old enough to remember the long, slow slide into irrelevance of the Esperanto language and the Technocracy movement, both of which proposed to replace near-chaotic systems of human communication and governance with rational, global replacements.

A better use of EMR funding would be on creating truly intelligent interchange systems that could negotiate meaning between local or regional EMRs. To the extent these EMRs leverage existing standards, the interchange system's job would be made easier, its crucial task would be to navigate ambiguous information and the need tomake decisions based on partial and imperfect knoweldge in a manner that emulates how humans handle them. The irrational, near-chaotic nature of human communication and governance won't go away in our lifetimes. Neither should it, necessarily; there is a lot to be said for intuition and creative guesswork. Maybe I'll get time to write about that soon, but for now I gotta go.

Comments

[ulo]

"I am old enough to remember the long, slow slide into irrelevance of the Esperanto language"

An Update on Esperanto:
http://www.uea.org/info/angle/an_ghisdatigo.html

Info:
http://www.2-2.se
http://www.esperanto.net/
http://en.wikipedia.org/wiki/Esperanto_in_popular_culture
http://www.amuzulo.net/quotes.html
http://esperanto-usa.org

Courses:
http://www.lernu.net
http://www.cursodeesperanto.com.br/
http://pacujo.net/esperanto/course/
http://www.institutoesperanto.com.ar/

Music in Esperanto!!
http://webjay.org/by/Fajro
http://www.vinilkosmo.com

News in esperanto:
http://raporto.info/
http://es.chinabroadcast.cn/gj/
http://www.liberafolio.org/
http://gxangalo.com/

Esperanto rulez!

;D

[Abdul Quddus]

I think that there are two views points to be considered when strongly advocating the use of EMR in practices. For a Doctor it needs to easy to use and save time so that more time can be spend on the patient. Currently we have many vendors with different types of EMR that are so hard to use that it simply puts them off. I think healthcare technology companies need to develop product after regular interaction with doctors to ensure that they provide just what is required. At binaryspectrum we have developed our healthcare solutions after spending countless number of hours with doctors to ensure that its work flow is kept simple and intuitive. This is then followed up with a period of Beta testing in real time environment before it is offered as a product in the market.