Free Markets, Free People

Another software disaster in the making?

Over at The Corner, there’s a discussion going on about medical records, prompted by this sentence in the “stimulus” plan:

“Computerizing every American’s health record in five years, reducing medical errors and saving billions of dollars in health care costs.”

There are the usual (and valid) privacy concerns. But Iain Murray goes further, and wonders:

I’m not sure why insurance companies haven’t insisted on it, but my guess (and I stress it is a guess) would be some regulations related to privacy, which was the source of the AMA and ACLU’s opposition in the past.

From someone sitting inside the world of healthcare software, perhaps I can enlighten things a bit. Privacy is only one of a host of challenges. A bigger obstacle is the difficulty involved. Creating the software that manages patient data electronically is, to put it bluntly, beyond the capabilities of almost all software developers. It’s really, really hard.

Here are some of the challenges in creating such software:

1.  The data is very complex. It’s not just numbers and text; it includes all kinds of media, which needs to be interpreted and annotated.

2.  The data evolves rapidly over time. New tests are constantly being created.

3.  How the data is interpreted varies rapidly over time. Today’s rule might be “you need a prostate exam if you’re over 50 and blah, blah”, but a cheaper, less invasive test next year might mean it changes to “you need a prostate exam if you’re over 40, period.”

4.  The users are very difficult to please. Doctors are the most difficult users I’ve worked with in an entire career of software development. They won’t sit still for two weeks of training. If it’s too hard to use, they just won’t use it. They’ll keep using paper. (Given the responsibilities with people’s lives they have, that’s understandable.)

5.  There are laws (HIPPA) concerning privacy that are difficult to design for. The rules are not prescriptive, so you don’t really know if you have satisfied the law until some auditor tells you whether you have.

6.  Existing systems are very fragmented, and typically include only a small minority of information such as prescriptions. But that data must still be brought in. So transitioning to a new system is very, very hard because all kinds of weird data must be imported. That transition has to be right; errors introduced during transition would be a huge legal risk.

I’m not sure it was even possible to satisfy all these constraints with technology until the very recent past. We now have much better technology for user interfaces, and better technology for transporting records around. But it still takes extreme architectural and design skills to create a system that can incorporate entirely new types of data and rules by clinicians without the involvement of a programmer.

That means in particular that this can’t be a big government job. The IRS and their four billion dollar debacle showed the problems government has with creating large systems. I simply don’t think any government effort could attract and keep the talent needed for this task.

Even private entities in the healthcare world have trouble with complex systems. HCA Healthcare attempted to write a next-generation patient accounting system, and wrote off some $130 million, and I’m pretty sure the actual amount of money spent (on a system that was thrown away) was much higher than that. 

It’s easy for liberals, and even some “compassionate conservatives” to see the opportunity for saving lives and saving money, and just want to pass a law to make it happen. I don’t think I have to tell the people who frequent this site why that’s a bad idea. We could end up wasting tons of money.

But there’s a potential outcome that’s even worse. If an inadequate, buggy, brute force, low tech system were rushed into being, and its use was mandated, that would block the adoption of an innovative, more modern, better system that could be developed later. We would effectively be frozen into using that system, just as the air-traffic control system was frozen on old, obsolete technology for decades.

To sum things up, there is enormous opportunity to improve healthcare by applying technology to clinical patient data. But it’s a huge challenge too. And the more government tries to push it before it’s ready, or to command it into being the more likely that the potential won’t be reached.

Tweet about this on TwitterShare on FacebookShare on Google+Share on TumblrShare on StumbleUponShare on RedditPin on PinterestEmail this to someone

16 Responses to Another software disaster in the making?

  • Well it sounds like the Doctors are the problem…well they receive substantial Federal reimbursements…I propose we MANDATE they use the sytem AND that no doctor make more than the POTUS!  It’s only fair.

    I might add you dense righties will NOT stop this stimulus package, not when 500 MILLION Americans are losing their jbs per month!

  • I don’t think it’s doable in 5 years as a whole, but certain record types could well be transferred to/from a central repository in far less time. 

    I mean, if my state BMV can block my DL renewal because of a 19 year old speeding ticket I failed to pay, they can certainly put my weight, height, blood pressure, temp, blood sugar levels, cholesterol, etc. into a central database.

    Now, of what benefit that is to my health, I don’t know.  My current dr already does that.  When I moved 1 1/2 ago, I took all my records, and my new dr’s office had them entered in (probably by someone in India.)  My insurance company can tell from my Rx’s what conditions I have, and sends me info from time to time to remind me how to stay healthy w/ my conditions.

    But, beyond any supposed efficiency, what business of the government what my dr and insurance provider are doing.  They’re probably far more efficient, and inovative now, then if they had to follow even more mandates.

  • Amen Billy.  “All we want is a system that….” Hah! I watched NHIC Texas throw 2 years (sorry, more than that, but I was there for 2 while it was going on, supporting the system that was marked for death) of effort into updating their systems using a (an alleged….) 4GL that would miraculously solve the problem of involving programmers in adding or updating rules for procedures etc, and that was before anyone really wanted to start blending all the various media into a common repository (or at least making hooks to all the related media from a centralized location). It’s nice to say “all we want”, but it seldom works that way. Heh, I’m pondering the equally easy “all you need to do…” mandate that old data is transmogrified into newer formats (either on the fly or on the database) to enable the record keeping/retrieval mechanisms. Heh, and that’s just the electronic stuff, what about the stuff that’s sitting in a series of pendaflex folders….oh lord…..

  • Billy – the other big road block to EMR’s is just plain old data storage.  Think of the gigabytes and bachigobytes and holyfreakingbatshitobytes of data that would need to be stored.  Never mind access control…

    • Oh, I haven’t even gotten started on the problems. Yours is a big one on the list. Another bit one is handling separate specialties. Every specialty needs it’s own data, and it’s own views into the overall data. An overarching system would have to satisfy all specialties, which means a pluggable design with a lot of dynamic, on-the-fly generation of user interface.

      These people who don’t work in this field have no bloody idea what they’re asking for. It can be done, but it will take money and some very fine talent. If it’s done via some sort of government-driven program, it will fail. I can’t see any other possible outcome.

  • The Air Traffic Control system upgrade was a debacle.  The IRS computer system upgrades were a debacle.  The FBI’s computer system was a debacle.

    What could possibly go wrong?

  • There is a cottage industry in correcting the data generated by doctors involved in clinical trials.  You’ll often get records where, due to unit confusion, a patient was 3 feet tall and weighed 324 pounds…or was 16 feet tall and weighed 70 pounds!

    How is a “national automated database” going to correct every error entered by every doctor, every day?

  • Excellent summary of the vast legion of problems with EHR software. As a physician who has written my own EHR over 15+ years, I can testify (and you no doubt already know) that the list a problems is far vaster than your brief summary. I spoke of many of these in an older post on my blog <a href=””>here</a>.

    It is quite simply nonsense that “computerization of health records” will save billions in health care dollars. If the solution were easy (or even challenging but doable) it would have been done years ago, as it has in most other industries.

    • You’ve gotten a good idea of some of the obstacles. As you’ve seen, it’s hard but possible to create a solution that’s highly customized for one specialty and a small set of users.

      I’ve seen even some commercial systems that do that. One here in Nashville handles physical therapists very well, mostly because it automates Medicare paperwork (elderly patients are the ones who most commonly need physical therapy) and thereby saves the doctor 1 to 2 hours per day. They have installations in all 50 states. But nothing they’ve done will transfer out of physical therapy. (I’ve looked at their systems.)

      • mostly because it automates Medicare paperwork (elderly patients are the ones who most commonly need physical therapy) and thereby saves the doctor 1 to 2 hours per day.

        Here is an area where the government OUGHT to provide a standard computerized solution.  Why doesn’t the government get it’s own house in order before trying to force a solution on private industry?

  • In connection with my application for VA benefits, I recently received my own medical records (military and civilian) for the period 1966 through 2009.  Since I have had constant care and treatment (including numerous surgical procedures) for heart disease and high blood pressure since 1966, and for diabetes since 1991, these records were voluminous.  I have had to review them, cull out duplications, compile and organize them, and annotate them so someone can make sense of them in connection with my claim.  This has taken me on average about 10 hours per week for the past 8 months.  While I intend to scan these docs and convert them to PDF format, the resulting product will not be searchable. Admittedly my records are more complicated than most.  Let’s assume I am in the top 1% for complexity. Multiply my situation by 3 million and you have a huge project.  And that does nothing for the remaining 287 million potential patients.  We haven’t even scratched the surface. 

  • Liberals are so cute when they save the planet.

  • The Veterans Administration medical information system is close to 100% computerized. All patient information, records, appointments, medication information, tests,  (including videos, spreadsheets, narratives and graphic information like X-rays and CAT scans),  and other information are stored in the VA IT system and accessed and entered through desktop computers by care givers very quickly and easily regardless of geography. Patients also have some access into the system from their home computers, for instance, VA patients can renew prescriptions on their home desktop computer easily and the medication arrives by mail in about a week. The wheel doesn’t have to be invented, it already exists. Whether we want or need a single national health care information system is another matter.

    • That system is nowhere near to what’s needed for a universal patient management system. First, the VA can force their doctors to use it, but based on screen shots of the UI, I’d be willing to bet long odds you couldn’t force the entire medical profession into it.

      Plus, it’s optimized to treat the problems that arise in a restricted patient set. I rather doubt it’s ready for OB/GYNs.

      There’s no indication that is has a rules-based to facilitate data handling and addition of new data. The words “rule” or “dynamic” do not occur in that article. (Granted, it’s a short overview, but I know how hard such system are to create, so I’d be surprised if it were well-designed in that area.)

      Then there’s the accounting aspect. Most VA care is paid for by the VA, so the system is unified. An overall system would need much more complex interface to payment and pricing mechanisms.

      It’s a client-server design, according to the wikipedia link, which means it does not have the geographical distribution needed. Client-server designs typically differ quite a bit from highly distributed systems. Transactions and concurrency, to name two considerations, are typically done quite differently in those two environments.
      Look, that system is open source, and has been around for several years. If it was that all-fired great, the healthcare industry would be falling over themselves to implement it. They’re not.

      • Of course there have been women in the armed forces for quite a while now so there a more and more women veterans and they need and receive OB-GYN care.

        The VA seamlessly accesses my medical records no matter which facility I visit all over the state on NY. I can access the system for my personal records from any desktop computer using security procedures similar to what you use to access your banking or investment websites.

        • I think you ignored the most important point. Allow me to state it again:

          Look, that system is open source, and has been around for several years. If it was that all-fired great, the healthcare industry would be falling over themselves to implement it. They’re not.

          I’m sure it works fine in the VA restricted setting, and I’m guessing it’s better than what most doctors have today. That does not necessarily mean it would work well in a wider setting, or that it would evolve well as technology and medicine evolve.