Click here to preview the new Fast Company

Want to try out the new

If you’d like to return to the previous design, click the yellow button on the lower left corner.


Badly Designed Electronic Medical Records Can Kill You

When doctors are overloaded with data, they can make errors that endanger your health, writes Continuum's Rick McMullen.

Last May, the University of Pittsburgh Medical Center shut down its living donor kidney transplant program because of a serious medical error. A man in need of a transplant was given a kidney from a donor who did not know she had tested positive for hepatitis C. The positive test was, however, included in the donor’s medical record, and an automated alert saying as much was missed by all six members of the team reviewing the procedure. The incident resulted in the immediate shutdown of transplants, the demotion of the department head, the suspension of a nurse, and the initiation of two federal investigations.

There are two reasonable explanations for this incident. First, the six medical professionals reviewing the transplant records were incompetent and inattentive. Second, the system that was supposed to warn practitioners of the infection did not do its job, even though it did show a highlighted alert. Neither is easy to swallow. But there’s a growing body of evidence showing that the electronic medical records software designed to prevent precisely these types of errors may be contributing to them.

Good Medical Decisions?

Health care in the 21st century is reliant on the availability of accurate information. Electronic medical records (EMRs) have the potential to improve the safety, quality, and efficiency of health care dramatically by providing information when and where it is needed. One critical aspect of EMR systems is the promise that they will provide caregivers with information that wasn’t available before, such as drug allergies and patient histories, and that this information will help them make good medical decisions. This capability, termed Clinical Decision Support (CDS), has been heralded by many as the single greatest benefit of EMR systems.

"Information is not the scarce resource. Attention is."

As health care becomes more complex and EMRs get more and more sophisticated, doctors are now provided with more and more CDS information aimed at helping them improve care and reduce medical errors. Intuitively, the more a doctor can access about their patient, the better he or she can care for them. But there are some trade-offs that we will need to keep in mind in order to maximize the effectiveness of these systems. There is, after all, such a thing as too much information. Conrad Wai said it well in a recent article in The Economist, "Information is not the scarce resource. Attention is."

A Tension of Attention

Information overload is all too common today, and physicians aren’t immune to it, either. According to multiple recent studies, doctors ignore between 49–96% of all CDS alerts that EMRs give them. There is even a well-documented phenomenon referred to as "alert fatigue," when physicians ignore EMR alerts without reading them as a knee-jerk reaction to the system constantly crying wolf or generating information that is useful for someone but not the current user.

Constantly clicking through spurious alerts and sifting through a haystack of extraneous stuff for a needle of useful information is irritating. But the consequences of information overload are much more severe in a clinical setting: According to recent research, information overload is to blame for a rising number of medical errors. A study by the Center for Studying Healthcare System Change showed that unintentional information overload from EMRs actually served to complicate the providers’ efforts to discern key clinical information.

The Blame Game

Does this mean that clinical decision support systems, which are supposed to improve patient care, are actually harming patients?

Well, not really. We want medical practitioners to make medical decisions, not computers. So in that sense, computers are supposed to show information and let the professionals (i.e., doctors) do their jobs (i.e., make medical decisions). You can’t blame CDS systems, per se.

What you can blame, however, is the tendency for these systems to show all of the information, all of the time, as opposed to only what is relevant. It’s probably well intentioned, but it’s essentially a rejection of responsibility by software designers.

Said another way, designers are covering their asses. Nobody wants to be the one to say, "We should limit the amount of information the doctor sees," even if they know that too much information is as dangerous as too little, because that one time the system doesn’t show something that turns out to be necessary, it’s easy to point a finger at the software and say, "The information I needed wasn’t there!" In the case where the info the doctor needs is there, but they miss it, the software isn’t the obvious culprit; the doctor is. The physician who makes a medical error is legally liable, even if a large part of the reason he or she makes that error is poor information design. This is exactly what happened in Pittsburgh, and similar things are happening across the country with alarming frequency. It would be pretty lame to get sued for malpractice because a software company’s design department passed the buck onto you.

Organizational Underload

In defense of the software companies, their hands are tied in a couple ways. First of all, they have to sell these systems. If you don’t present every imaginable piece of information to the physician, you’ll lose in the marketplace. The competition will be able to say, "We give you more information," which leads the customer to think your product is less sophisticated. The result is an apparent incentive mismatch. That is, if you present your user with a glut of information, it increases their odds of making a medical error; if you don’t, they may very well buy a competitive product that has more features.

The second barrier to good information design here is regulatory (surprise!). Simply put, nobody but a doctor is allowed to make a decision about what medical information is important and what information isn’t. As such, no one who creates software can really make a call about when to notify or not notify a doctor of a potential issue that the system recognizes. Don’t get me wrong, it’s mostly a good policy; I definitely don’t want my life in the hands of some software engineer’s diagnostic algorithm.

But there may be a way to have all of the valuable information but reduce the degree to which it overwhelms providers and hurts decision-making. The key to mitigating information overload is not to reduce the amount of information but rather to increase the extent to which that information is organized. Famed information designer Edward Tufte has been known to coyly refer to information overload as "organizational underload."

You can have all the vital information without hurting decision-making.

Good systems know something about the user, what they are doing, and what they care about to curate only the information that is most valuable in the moment. Amazon, Google, and others have made a killing by displaying relevant advertisements and products in real-time. They do this by tracking their users’ behaviors over time and also by making use of behavioral economics research on how people use information and make decisions. You could imagine that with some changes in policy, EMR/CDS systems could employ similar techniques to curate information for reasons other than selling products.

In fact, there are a number of efforts under way to do just that. Drs. David Eibling and Augie Turano from the Pittsburgh VA are working on an intelligent system that allows physicians to rapidly search and filter through well-formatted patient data for improved and unencumbered clinical decision-making. Dr. Jeff Belden from the University of Missouri is doing funded research on how to structure EMR information such that the amount of detail shown is appropriate for the particular person who is looking at it. But they’re the minority, and importantly, they don’t work for software companies. We can only hope that the great work being done by folks like them will make it to the industry.

There is a very precarious balance between providing enough appropriate information to be valuable but not so much so as to overload the decision maker. It’s another one of those problems where design is a critical piece of a larger puzzle that includes bold regulatory change and business innovation. But whoever gets it right is going to win the market and make a sizeable impact on the lives of patients (and their doctors) across the country. Doing it wrong is setting doctors up to fail — and then letting them take the heat for it.

[Photo courtesy the Otis Historical Archives National Museum of Health and Medicine]

Add New Comment


  • DeLuca

    I've worked in healthcare for 30plus years, in a clinical role (technology), an operational role and a marketing role. It is my opinion that system designers do not employ those who think like end-users. Designers are not like the rest of us and tech companies might do well to hire a few sharp liberal arts types to pull it all together.

  • Nick Romero

    @Chi – This topic is far too advanced and/or scientific to
    be pin pointing any type of business strategy. In the grand scheme of things,
    the context demonstrates a dilemma in how humans use informatics to provide safe
    and efficient healthcare. Tossing in business aspects only compounds the disarray
    existing over the underlying phenomenon.

    The two systems that are under the microscope in this
    article are robust enough to remain separate entities. By dividing/conquering services--or applications if you will--might help us find the core
    challenge.  In my mind, I believe it to
    be the human element, causality being human error as depicted by the kidney
    scenario. I sense these
    individuals confided in their systems so much that they neglected the oath and fundamentals
    in which they were taught.

    @Rick – I think the principle of what you mean by “balance”
    needs more clarity but great article. I am not sure that I would advocate the
    notion of regulating information as an approach to a solution.  Limiting any medical related information on
    behalf of a patient is erroneous. However, I would agree on governing “INITIAL”
    information presented to healthcare providers. Healthcare providers should always have
    access to as much information as possible.  Obtaining, processing, and comprehending large amounts of information is part of a physicians specialty. Med school naturally builds that endurance for them.

    For example, let’s say a physician has an interactive list
    to assist in identifying names and rooms of patients to perform rounds on in a
    hospital.  Key word-initially, this list
    displays the vital information (i.e. name and room). Once the physician crosses
    that bridge they can revert back to the device and discover more details on the
    patient’s pertinent data from a selectable region on the interactive list next
    to the said name. Thus, all the information is there but not shown all at once.
    This concept does in fact point to the critical issues of designing healthcare
    systems. Ultimately, the flow of information and how users handle it becomes a
    topic of cognitive science.  I am sure
    you’re already aware of that.

     In the article the word
    ‘design’ seemed to be used very loosely and I think our approach should be far
    more specific. This is an avenue for defining human-centered healthcare systems.
    I can only imagine a result coming from a very granular health-informatics campaign that uses careful
    heuristic research and strict contextual design. I don't believe a solution would come from an outside development group that gets lucky in striking an ideal information threshold and/or superficial design.

  • Chi

    I do believe the EMR system should be display in a fashion that will allow users to customize , what information is displayed and how that information will be displayed. This effort should be organized between Medical University and private company with some government funding. I don't think the medical organization will pay for a EMR system, that will track a users behaviors in the system and custom the display of patient information. At the end of the day private companies aren't going to do invest money in something, they don't think they can sell.