Wednesday, June 9, 2010

Do EHRs Kill People?

Back in the times when EHRs were just EMRs, they had a very simple and humble mission. The software was supposed to help providers of health care services better manage their business. EMRs were supposed to help physicians adhere to CMS documentation rules, automate patient flow management and get rid of all the mountains of paper floating around a typical medical office or hospital. It was assumed that EMR software will increase reimbursement rates, streamline workflow and even make the doctor more efficient. After all, every other industry that switched to computerized business management realized bottom line improvements.

Along the way, bolder statements started appearing, mainly from EMR vendors trying to sell their wares. EMRs could also reduce medical errors. The most common argument was for the benefits of replacing the notoriously illegible physician hand writing. Prescription errors would be reduced if only pharmacists and nurses could get a nice legible script. Then came the frequently misplaced paper charts. If the chart resides in the computer, it cannot be misplaced, it is always available to all and it is complete. All the information you need right at your fingertips, regardless of your physical location. It could save lives or at the very least, it could save time. The EMR was nothing more than an electronic chart. One vendor went so far as to create a computerized image of a yellow manila folder with tabbed pockets for various items in the electronic chart.

Nobody thought the electronic chart needed to be regulated by the FDA any more than the paper chart was. After all, the EMR was not providing medical care; it was just a more effective place to record administered care. Or was it? There is a funny thing about computers. They have a mind of their own, a mind created by programmers, a mind which makes them interactive. A paper chart is passive. If you want to see all lab results in a paper chart, you have to decide where to look and actively flip the pages. If you missed one page, it’s your omission, not the chart’s mistake. If you want to see all lab results in an EMR, you click a button and the software does all the paging returning a convenient list for you to look at. If the software missed a page, it’s the software’s fault, not yours. The software is controlling what you see and how you see it. This small fact renders the electronic chart a full partner in delivering health care; it is now a medical instrument.

And then the EMR grew up and became an EHR. The EHR performs tasks for you, like calculating dosage for medications or just simple BMI. The more advanced EHRs presume to give you advice on what to order for a certain condition, or what not to order based on what it knows about your patient. There are EHRs now, and there will be more in the future, which communicate with other software and automatically, bring in medical data and place it in the chart. This sophisticated software makes decisions regarding patient identities and about schedules for preventive care and disease management. Computers are not infallible. Their mistakes are called “bugs” or “software glitches” and just like the nurse or the medical records clerk sometimes placed the wrong piece of paper in the chart, EHRs can, and do, corrupt medical records. Incorrect, incomplete and indecipherable medical records can lead to injury and even death. But does it really happen?

Do EHRs actually kill people? 

The Huffington Post has been investigating this exact question. Between January 2008 and February 2010, the Huffington Post identified 237 reports in the voluntary incident reporting FDA database related to HIT, including 6 deaths and 43 injuries. However, a closer looks reveals that only a small fraction of these reports are actually related to EHRs per se. Most reports involve PACS, medication dispensing systems, blood banks and other FDA regulated equipment. Out of the 6 reported deaths (2 of which occurred in 2006), one was related to a PACS system latency, another to human error in labeling an x-ray cassette and another to a hospital pharmacy system. 2 deaths were attributed to system wide failures of CPOE and one to lack of intuitiveness in display of notes. As to injuries, out of the 43 reported, I could only count 17 directly related to EHR software and most have to do with CPOE.

Is this the tip of the iceberg, as some contend? Are there many more unreported deaths caused by EHR software? There may be, but frankly, the evidence of massive numbers of adverse events is not there. It does, however, stand to reason that voluntary reporting would be incomplete and the fact that only a couple of EHR vendors are represented in the FDA database is suspicious to say the least. On the Health Care Renewal blog they are engaging in what I think they know are rather creative mathematics, to project hundreds of thousands of injuries per year if, and when, EHR adoption really takes off.
If EHRs become as pervasive in everyday medicine as ONC is proposing, every patient will eventually be touched by an EHR.  It is very likely, that some errors will be prevented by the sheer existence of an EHR, but new and unfamiliar errors will also be introduced as side effects. Of course, the potential benefits must be shown to significantly outweigh the hazards, and we already have accepted mechanisms for such assessments.

While ONC is exploring collaboration with the FDA, and the FDA seems willing to engage, the customary counterargument is that FDA processes will stifle innovation and make EHRs unaffordable. There is validity to such arguments, but as long as money seems to be no object for HITECH, maybe we can spend some of it on devising reasonable and affordable methods of testing patient safety, both pre and post market. Innovation will take care of itself and the alternative is unconscionable.


  1. 'The EMR as a medical instrument', that's an interesting connection. With anything, if there's a middle layer (EMR) between the patient and provider there is some level of interpretation. But what happens when an MD writes an illegible Rx on a paper notepad? Is the notepad an instrument?
    HITECH was designed as a catalyst, allowing states and the free market to shape the EMR. It's up to MDs on advisory panels of software companies to take more responsibility and give better feedback as though they themselves were the 'medical instrument'.

  2. I think, Faisal, that there is a distinction here. The illegible note is fully within the doctor's control. It is up to him/her how he/she writes the note and up to the reader to decide if it is readable. The information in and within itself is all there.
    On the other hand, if the computer is placed in between users, the computer may "decide" what is worthy to show and how it is shown. This makes the computer a provider of care, IMHO.

  3. Once again, a great assessment of a complex issue. I was about to bring up Health Care Renewal until you mentioned it yourself; certainly, while I agree in some measure with those authors that harm can occur and is probably not be self-reported, I'm also wary of anyone offering "projections" of numbers without a solid base to calculate from.

    The line here in what makes the computer a partner is blurry, to say the least. For example, does my use of blogging software constitute a partnership in presenting information? What about public works like power stations and water treatment plants, or transportation centers like airports or train stations? When a mistake is made, how far does the culpability of the computer reach in all those instances (not just medical)?

    I can see the reasoning behind your response to Faisal's comment, but I think a distinction should be made: a computer doesn't "decide" anything, it merely follow the course that it has been programmed to take. Yes, it seems to determine what is shown and what is not shown, but that is really a choice a person made at some point in development, or a user at some point in installation.

  4. Great article Margalit - I touched a little on this during a presentation at HIMSS Virtual and one of the things that struck me in researching various errors is the idea that the system has now become the sole funnel and filter for patient information.

    The fact that HealthIT related adverse events are only voluntarily reported troubles me. The lines between device-system-interface are blurring and taking silo-based approach can only lead to confusion.

  5. I think Michelle that the culpability of the "computer" depends on whether the computer is functioning as designed and if, a big if, the user is aware of the design principle.

    If the software design is faulty and if the user is for example not aware that he has to click on buttons to see addenda while viewing a progress note, than it may be a training issue. If the addendum button requires scrolling to get to, or it appears in different places on different screens, it's a design/usability issue. If the button fails to appear if there's only one addendum because the developer started counting the array from 1 instead of 0 (common stuff), it's a bug.
    In all cases the responsibility lies with the computer, a.k.a. vendor (developers, UI designers, QA engineers, acceptance testers, trainers, technical writers and implementers).
    If the computer has the ability to obscure information, which a paper chart cannot do, then the computer is responsible for what the doctor sees and as Jasmin said, that is mighty powerful stuff.
    If the computer assumes to perform calculations, than it is the computer's responsibility to be correct. If the clinician has to check all calculations by hand just to be sure, then why have a computer?
    If the computer loses data entered by a clinician, it should suffer the same consequences as the medical records clerk who misplaced records - she usually got fired.

    Of course, there could be user errors too. Garbage in garbage out. We need to differentiate though between what is under direct control of the user and what is not.

  6. Margalit:

    Do cars kill people? Do airplanes?

    We all know there are tens of thousands of traffic fatalities each year in the US, but we still drive cars, because we know the risk of death (per mile driven) is tiny and the benefits of the technology outweigh the risks, by far.

    Plus, there are things we can do, like wearing seatbelts and not texting while driving, that mitigate the risks of a traffic fatality.

    It's the same thing with EHRs. They are a technology with risks and benefits. You should decide to use them if you believe the benefits outweigh the risks, and you should do whatever is possible to mitigate the risks.

    Glenn Laffel, MD, PhD
    Sr. VP Clinical Affairs
    Practice Fusion EMR

  7. I absolutely agree, Glenn.

    We should use EHRs and we should strive towards HIT adoption and we should try to realize all the benefits of interoperability.
    But just like cars and airplanes, there should be some sort of regulatory safety requirements in place to ensure that not too many people are harmed. It will never be zero and that's understandable.