Missing the Point in the Design of Electronic Medical Records
We were lucky to have Dr. David Eibling and Dr. Augie Turano from the Pittsburgh Veterans Administration (VA) hospital come visit MAYA to give us their thoughts on the ongoing Electronic Medical Record (EMR) debate. Dr. Eibling is a tech-savvy ear-nose-throat surgeon, and Dr. Turano is a solution architect and procurement officer with a PhD in Biophysics. For a number of years, they’ve been working hard in their free time (which, given their professions, is not terribly plentiful) to improve the state of medical information systems.
There’s been a good deal of hype about Electronic Medical Records (EMRs) in the last few months, and for good reason. Studies estimate that 98,000 people die every year due to medical errors (though Dr. Eibling thinks that number is closer to 250,000!); as many as 1 in 4 prescriptions result in adverse reactions; and as many as 90% of medical diagnoses are made without sufficient evidence.
EMRs have been portrayed in many circles as a panacea for the problems listed above and spiraling healthcare costs. President Obama has even signed a bill mandating that all hospitals achieve “meaningful use” by 2015 or face non-trivial fines. The naysayers, on the other hand (many of whom are doctors), remain skeptical that the current systems are ready for primetime. Several studies question the effectiveness of EMRs and attribute their limitations to weak system design. Not surprisingly, there is growing interest in EMRs at MAYA, as a critical problem in need of thoughtful, contextual human-centered design.
It’s hard to tell whose opinion to trust on the subject, especially since system price tags can run into nine figures, and everyone seems to want a piece of the pie. EMR providers discuss their close collaboration with doctors, boast about their efficiency gains and tout the future financial benefits of implementation. However, individual doctors and hospital systems give less glowing reviews, citing high customization and upkeep costs, increased errors and process challenges. This is true even at the VA, whose EMR implementation is widely recognized to be well ahead of the game.
Dr. Eibling spoke about Clinical Decision Support (CDS) systems, why they’re falling short of their potential, and why it’s a big problem. He opened with a case study of 370 hypertensive patients who had a life-threatening, adverse reaction to a specific medication. 23 of those patients had previously been hospitalized for a reaction to the drug before, but since the doctor was unaware of it, they were prescribed the drug a second time.
Dr. Eibling presented another equally troubling case study about a man in the VA system who had undergone surgery to remove a cancerous tumor based on lab results in his EMR. It was discovered later, however, that the doctor who checked the biopsy and ordered the surgery was looking at a diagnosis from a previous year, and the patient underwent major surgery to remove a benign tumor! (As it turns out, the biopsy was incorrect; the patient actually did have a cancerous tumor, making the mistake a “near miss.”) While most people blamed the doctor, Dr. Eibling dug a little deeper and showed us how the visual layout of the EMR tool was largely responsible for the mistake. Since that event, the VA has changed the layout of the EMR to prevent similar errors.
Dr. Eibling had many more unsettling stories and statistics (surgery to amputate the wrong leg, surgery completed on the wrong patient…really unpleasant stuff) that could be traced back to a simple misreading of a medical record.
But, how are these errors possible? Doctors are some of the most passionate and intelligent people on the planet; they review patient charts carefully, anticipate (and work to prevent) adverse reactions, and double-check their prescriptions. And what’s more, all of the information needed to make the right decision at the right time was somewhere in the system. Why wasn’t the right decision made?
The answer, of course, is that the data being “somewhere in the system” is no more useful than your keys being “somewhere in your apartment” when you’re already late for work. As it stands, doctors need to dig through a haystack of irrelevant and poorly organized data to find the needle of critical information that could prevent a serious complication. Unfortunately, the systems as they exist today do not have the ability to know what is important, and tend to treat all information as though it were of equal priority. What that means is that the doctor has to shoulder the cognitive load of that “information treasure hunt” when they’d much rather be using their valuable brain space to think about the patient and the diagnosis.
Certainly, the cases of explicit harm are the most worrisome. However, poor EMR design causes problems in more insidious ways as well, producing inefficiencies that adversely affect patient care. We spoke with a doctor who used a new EMR system for patient documentation in the trauma unit of a major hospital. The goal was to speed up the documentation and make the collected data more valuable for diagnosis. When the trauma unit started using the system, however, the average time spent writing up a report for a patient exam went up by nearly a factor of ten. His interns were spending an average of 26 minutes per patient (compared to three minutes per patient under the stone age paper system) writing reports. We live in a world where physicians can only spend about 15 minutes in the exam room with some patients. Spending twice that amount of time on documentation is unacceptable, and drastically reduces the number of patients they can see in a given day (or…perhaps worse, forces them to cut the amount of time they spend with each patient).
What’s even more frustrating is that the collected data was no more valuable to the doctors than what was captured in the paper records. More valuable in terms of optimizing billing and collecting insurance information? Probably. But definitely not in terms of improving care. To keep their schedules on track, doctors resorted to tricking the system into taking meaningless data (entering just ellipses instead of typing out a patient’s medical history, pressing the space bar three times to clear a dialog box, etc.). Not only did documentation slow down with the new EMR, but the content of the reports themselves actually decreased.
But again, this isn’t because doctors are lazy or don’t care about their patients. Quite the contrary, the doctors shortcut the documentation process and find rough workarounds so that they can spend more time in the exam room with sick people and less time at the keyboard doing data entry. The tools ignore the fact that doctors should be spending their time with patients not jumping through hoops to comply with the system demands.
Systems can’t keep up with the pace (and nuance!) of medicine
It is in the nature of medicine that even the perfect tool will not remain perfect for long. In many offices and hospitals, the exam forms for most procedures must be changed every three to six months to address procedural changes and improvements. In an earlier era, these changes meant adjusting a form on the computer (or in an even earlier era, a typewriter) and distributing the new version. With the electronic systems, the doctors need to put in requests to vendors or their IT staff to make these changes. Each of those changes adds to the customization and upkeep the hospital needs (and pays for) from the vendor.
In some cases, EMR vendors have said it can take them 18 months to make changes (even simple ones, like the labeling of buttons). If the forms need to be updated every three to six months, it’s easy to understand what the doctors and medical staff feel hampered by their existing tools.
Where do we go from here?
Now, I don’t want to scare you (well…maybe a little), but these problems aren’t getting better. In fact they might be getting worse, as many hospitals are shackled into staying with their huge investments in the existing EMRs.
So why is this happening? Why are systems that are supposed to help hospitals provide care actually alienating physicians and causing harm to patients?
The answer is because many aspects of CDS and EMR software are being designed without patient care as the primary focus. Certainly, the system vendors work with doctors and do some contextual research, but there is still a significant gulf between the capabilities that are needed and the systems that are currently on the market.
The gap is not solely from a lack of communication between doctors and vendors. In many cases, the recommendations from medical providers are added into the system in successive versions. However, in many cases, the basic design requirements for medical IT systems stem from the needs of insurance companies, whose priorities have much more to do with collecting data for billing purposes than providing doctors with tools to support their work. Each new requirement is addressed individually, like a Band-Aid, to “fix” the limitations of the system.
The series of Band-Aids, however, doesn’t ensure that the overall system improves. There is no synthesis of the varied requirements, workflows and recommendations; the pieces are never distilled into a clear, logical system that works in the real-life context of the medical community and its patients. The inconsistent, awkward, and complex software that results causes the errors noted earlier, slowing care, harming patients and causing frustration.
The message here isn’t that we should abandon CDS systems or other EMR projects. The message here is that this technology has unthinkable potential, but it will never materialize if we continue to cut corners in the design process. Asking doctors what they’d like changed about their current tools isn’t enough: their needs must be synthesized and prioritized into real requirements; the resulting design concepts must be prototyped and refined to ensure they are usable; and the system must be piloted and studied to ensure it actually improves patient care. In short, they need real human-centered design. Spending the time and money for HCD is not going to be the straw that breaks the multi-billion dollar budget’s back. If anything, it will save money by reducing maintenance costs, increasing efficiency, and (most importantly) preventing errors.
Positive incremental change is great, what we really need is a reinvention of Electronic Medical Records. It isn’t too difficult to imagine that a system could take doctors 20 minutes to learn instead of 20 hours, know what information was important when, prioritize alerts, save time while increasing useful documentation, and allow doctors and nurses to flexibly manipulate data.
However, this fluffy vision of the future isn’t going to happen without a very significant research and design effort. It certainly isn’t going to emerge out of systems that are rapidly becoming band-aid tapestries.
Glimmers of hope
What is frustrating for MAYAns is that we know how realistic and effective this sort of undertaking is. In fact, we’ve done it with Command and Control (C2) for the Army. Instead of just building a better C2 system, we worked with generals to take the whole process apart and discover the key issues. The mission was to close the “last 18 inches” of C2, namely between the computer screen full of data and the commander’s brain. After a few years of dedicated research and design, we, and the many wonderful partners we teamed with both in the government, in academia, and in industry, were able to produce the Command Post of the Future (CPOF).) CPOF is currently deployed in theater, and well on its way to being distributed throughout the entire US Army. Reports from DARPA note that they saw an increase in decision-making and “missions planned” of up to 400%, and an increase in “situational awareness” of 300%. The current system is widely deployed, takes less than a day to learn (the old system took two weeks), and doesn’t even show up in the top ten of bandwidth usage on the network. “A few years” may sound like a long, expensive time, but it’s well worth the investment when you consider both the scale of universal EMRs and the payoffs they could offer, if done correctly. It is important to note that when we started the CPOF program the goal was to have it deployed by 2010 (this was in 1998). It deployed 6 years ahead of time at a fraction of the cost of the old system.
So don’t let the fact that we’ve already sunk a lot of money into EMRs mean we shouldn’t start over and do it right. It’s been done before, and it can be done again. It’s one of the biggest and most important technology projects in recent history, so let’s take it seriously.
The strong push for EMRs and smart CDS systems is encouraging, but the message is loud and clear that it has fallen prey to hasty design. EMRs are already benefiting insurance companies by providing more complete documentation, but we’ve got a long way to go before they start truly benefiting the patient. Which is the point of healthcare in the first place…right?
Further Reading (since I’ve already vastly over-shot my page limit):
If you need convincing that bad human-centered design can have horrific results (or if you ARE convinced and need ammunition to convince others), read Set Phasers to Stun, by Steven Casey.
If you want more insight into some of the important ways that Clinical Decision Support systems can make medical practice safer, and some of the challenges that need to be overcome, read How Doctors Think, by Jerome Groopman, M.D. It’s a really fascinating study of the ways that doctors make decisions, and some of the ways that they can be set up to fail.
Many thanks to Dr. David Eibling, Dr. Augie Turano, and Katie Scott for their contributions.