Skip to content

Anyone know about American medical electronic records?

Earlier in the day I was reading something that suggested that the different vendors of the various medical records systems all use incompatible file formats.

Whoever let them do this needs to be shot, of course. But does anyone know this field in any detail? Sounds like the sort of thing where there would be an army of people writing conversion software. Is this so? If not, why not?

22 thoughts on “Anyone know about American medical electronic records?”

  1. The reason no one is writing conversion software yet is because pretty much the entire country is still in the implementation phase. The American Recovery and Reinvestment Act of 2009 (ARRA) created tiered subsidies for moving to electronic health records (EHR) tied to usage metrics and the subsidies still run out for the immediate future (Phase 2 runs through 2017, Phase 3 starts in 2018, I’m not sure how many years of money there are under Phase 3). Pretty much every health system in the country is still scrambling to comply with the next wave of usage metrics on their current system, the rules for which are not all written, the final 2015 rules were just released on 10/16/2015.

    Once the ARRA subsidies run out I suspect we will start to see the EHR companies begin to write that conversion software and really start competing with each other, but for right now everyone is riding the Federal Gravy Train (FGT) and every time a new requirement is announced that requires a software update or new feature the software companies are getting their cut by charging the healthcare providers a significant amount for it, which they can do because healthcare providers are fairly locked-in if they want to keep going after that those FGT funds.

  2. The NHS still has regulations that prevent A&E accessing your patient records from your GP. That’s not just down to incompetence, it’s part of the data protection rules.

  3. Why would different suppliers write to the same schema design ? Interoperability is about being available on demand, not having to be physically persistent.

  4. Meh. This is what data warehousing software is for. This is something that large multi nationals with a number of software packages have had to deal with for years.

    Move along folks, nothing to see here.

  5. There are a variety of tools available to perform data transformations such as would be required to allow interoperability among disparate health systems. Some of these tools enable data transformation ‘in flight’ (ie as an integral part of some greater transaction) and some only as part of some sort of batch data transfer method. For either type the data transformation requirements are largely described as data (or, more correctly, meta-data) so that the requirement for coding (writing new software) is minimised.

  6. I’d prefer a thousand formats. No one outside my doctor’s office should have any access to my medical records. Proprietary formats is an excellent barrier.

    7 Come, let Us go down and there confuse their language, that they may not understand one another’s speech.”

  7. Look at read codes. Take a look at the various DSMs from the US.

    The above two are examples of standards that can be a tad problematic to map between versions without even thinking about mapping sometimes textul data to the standard.

    Then consider the other data fields/values that can vary by various companies and bodies

    This is without even considering different doctors, different diagnoses. Or even multiple diagnoses.

    How on earth can you map between any these with any degree of accuracy?

  8. Whoever let them do this?? No-one let them do it. They risked their time and money to create intellectual property. Why should they make it compatible with competitors, unless the market forces them to. I don’t understand you, Tim.

  9. During software design the data design process will identify all the individual pieces of information that need to be captured and organise them into what seem like sensible groupings. Different data designers will come up with somewhat different groupings, for a variety of reasons, but the individual pieces of information will generally be quite similar for all systems addressing the same functional area. This is for the simple reason that that it is the functional area that defines what data must be captured.

    Modern medicine seems to be a pretty standardised thing, particularly in an individual country. All doctors use the same reference information for identifying what might be wrong with a patient and the same reference information for deciding what to do about it. That’s why it so easy to make a computerised ‘expert system’ which out performs a human doctor. The standardisation, though, would probably also help to ensure a degree of data consistency between different medical records keeping systems.

  10. Holy crap, I’ve just read Josephine’s second link:
    “She then e-mailed me my records as a PDF — without encryption, to an email address that wasn’t already on file, with no written or electronic patient authorization, and without verifying my identity beyond asking for my Social Security Number.”
    So it seems like medical records are exchangeable in the least-useful-to-bulk-use and most-useful-to-targeted-abuse fashion. Yay.

  11. I don’t know USA systems, but all the UK GP patient record systems I’ve worked with are able to import data from their rivals’ systems built in as standard, just like you can open an MSWord document in OpenOffice and open a OpenDocFormat file in MSWord.

    I would have thought it would be a sales killer to explicitly stop your product importing the data your user wants to get into it.

  12. It has now been 27 years since I worked in US healthcare. I was reasonably high up in the food chain of a very large CA healthcare organisation.

    One thing that struck me was just how many competing systems of codification the US health industry supports. There are literally multiple incompatible ways to codify every procedure, diagnosis, product and medical service – updated regularly in a never ending stream of data.

    Managing this alone is a nightmare, never mind managing file interchange formats (which are a relatively trivial exercise)

    At the time, I thought, “only in America”. I wonder if it has changed in the last quarter of a century, or has the situation worsened?

  13. Bloke in Costa Rica

    Data interchange founders on semantics rather than syntax. If the set of data is sufficient to requirements then the underlying representation doesn’t really matter. It could be ISO TC 215, XML, JSON, COBOL RECORDs or Morse code crocheted into a scarf. But if there is not a one-to-one correspondence between two datasets representing the same entity then lossage, fail and abject teeth-gnashing will likely be the order of the day. And once you’ve added in human factors such as data-entry peons making a pig’s breakfast of things, reconciling two records that nominally encapsulate the same thing in a semantically equivalent form can be impossible, even in principle. These are the sorts of things that fucked the NHS Spine and are strangling Obamacare. They’re ka-ching if you land the contract, though.

  14. Are you sure we’re not looking at this in an NHS-centric fashion? Meaning, all this talk of doctors interchanging information, e.g. A&E contacting your GP.

    I’m not completely sure how it works in France, but I’ve been going to various doctors and a hospital recently, and – due to the decentralised nature of the system here – it appears a patient is expected to keep control of his medical records and/or ask whatever doctor he has seen recently for copies of whatever he needs. I saw a GP here, and was referred to a specialist: the GP wrote a letter with a brief summary of the complaint and medicines I’d taken, and was left to get on with it. I saw the specialist, he read the letter, then asked me a load of questions and proceeded with his own recommendations.

    When I first went to the American Hospital in Neuilly I was surprised to find it less of a hospital and more a collection of individual clinics. I once saw a specialist there and she told me I needed an MRI. Expecting her to pass a case file onto the scanning department and lead me to it, she in turn wondered why I was just sitting there. Basically, it was my job to take my file down to the scanning department and organise it myself. Or simply go home and not bother. Up to me, you see.

    Most people on here say they want to move to a system made up of thousands of independent practitioners and not have to go through a GP in order to see a specialist…but now we want a centralised information system where each practitioner can requisition files from others, all in a standard format. I’m not sure the two are compatible.

  15. Dweeb
    October 27, 2015 at 12:57 am

    One thing that struck me was just how many competing systems of codification the US health industry supports.


    Why codify it at all? Computerization of medical records is a billion dollar solution to a non existent problem, compelled by government.

  16. I used to work for Epic Systems, an EMR company in Madison, WI, that had a rather large market share in the US (it claimed to have the largest, but I couldn’t stand working there long enough to know if this was true or just PR).

    The format for their database system predates relational databases, although they said that their system is actually faster than relational databases when you’re dealing with systems that support multiple hospitals over a large region. It essentially has a 40-year old architecture.

    Even when you’re dealing with databases engineered in the past few years, when the architects understand that they should design their data software for portability & migration, actually implementing this migration is a fair-sized undertaking. It’s quite likely to run into unexpected delays/problems that require restoring data from backups.

    On a theoretical level, it is technically possible to have two different data specifications where converting equivalent information between them requires a program that can solve the halting problem. Put more directly, there exist models where the isomorphism problem is not computable. In practice, of course, a database that tried to store data that encoded information that complex would be useless, but it does mean that you shouldn’t assume that converting data from one format to another will always be easy.

  17. diogenes, I do; and not very fondly.

    At the small consulting company I worked for many years ago, we were purchased by an American megacorp. Their corporate bullshit was all managed by Notes, including email.

    So they popped down to our small office (no more than a dozen people, including boss, boss’s secretary, test department as well as developers) and installed a 6′ rack with the most ridiculously over-specced machine we had ever seen. Gigabytes of memory, multi-processor, the works.

    At the time, the shared compile server we all used was a 90MHz Pentium.

    We christened it “Bloatus Notes”.

  18. Bloke in North Dorset


    That wasn’t ADL was it? They purchased us and we had to transition to Notes and what a useless pile of shite it turned out to be.

    The worst thing was that in the early days you had to be in the office to enter expenses and given that our work was always on client sites overseas this meant a trip to the office over the weekend usually weekly so the project manager could be on top of the case. That really went down well with the family when you were leaving home on Monday at 5 am and getting back Friday at 10pm.

  19. Not ADL.

    Not sure if it was a Notes limitation, or just nameless megacorp’s incompetence, but the system couldn’t cope with more than one person having the same name. Others had to add middle initials, or use diminutives, etc. Anyway, when they came to close our small office, my redundancy notice went to completely the wrong person 🙂 Never did get that P45…

Leave a Reply

Your email address will not be published. Required fields are marked *