Why Doctors Hate Their Computers | New Yorker

If you’ve read my writing over the years, you can probably tell that I’m largely a fan of Atul Gawande and his writing. He weaves these great narratives while tackling complex health policy issues, making such arcane issues accessible for anybody. When I heard about this article in particular, I was very excited to read it. About six months ago, I began a clinical informatics fellowship at Boston Children’s’ Hospital. Almost everything he talks about in this article—from determining how to train users on EMRs to optimizing physician workflows within the EMR—is now part of my formal training in clinical informatics. I devoured the article, trying to see if Gawande’s experience met with my own perceptions and meshed with what I am now learning in my informatics training.

A quick note before I dive into my thoughts on this article; Gawande’s sentiments and analysis are by no means new. As just one example, Christina Farr wrote about this in Fast Company back in in 2016 with a similar title (“How This Technology Is Making Doctors Hate Their Jobs”).

To the article!

He starts off by talking about 16 hours of in-person training to learn Epic. In the three years that Gawande went through this training, many organizations are going away from such extensive and long in-person training. In fact, I am credentialed at Brigham and Women’s Hospital (where Gawande performs surgery) and did not have to do any in-person training on Epic. Their training now consists of a series of on-line modules that you can do at your leisure. Not a perfect system either, but probably better overall.

When it came to viewing test results, though, things got complicated. There was a column of thirteen tabs on the left side of my screen, crowded with nearly identical terms: “chart review,” “results review,” “review flowsheet.”

Achilles’ heel of most EMR designs I’ve seen. There are too many things (in general) and the things have too similar (or completely unrepresentative) names. In the current EMR that I use, there are at least three different ways to view a patient’s vitals. Multiple views of the same information and the resulting naming confusion are due to feature creep [1]. I believe this is ultimately a consequence of the EMR vendors failing to allow people to customize their own personal views. Sure, Epic allows you change the colors, but there is little in the way of allowing an end-user to dictate exactly what they see when they open up an electronic chart.

Many of the angriest complaints, however, were due to problems rooted in what Sumit Rana, a senior vice-president at Epic, called “the Revenge of the Ancillaries.” In building a given function—say, an order form for a brain MRI—the design choices were more political than technical: administrative staff and doctors had different views about what should be included…Questions that doctors had routinely skipped now stopped them short, with “field required” alerts. A simple request might now involve filling out a detailed form that took away precious minutes of time with patients.

I don’t think the problem is necessarily ancillary information that needs to be entered, but more so that entering the information is cumbersome. Order entry on paper is usually just checking boxes or scribbling a prescription on a script pad. For example, you could easily write “amox 500mg PO BID x 10 days” and have a perfectly understandable prescription. Slap a patient’s identifying label on it and you’re done! In a CPOE system, you first have to type “amox” and then choose amongst all of the amoxicillin containing products the specific one you want (which is a nightmare for pediatricians because there are approximately 762 variations of amoxicillin on the market). You would then have to type “500”, followed by choosing “milligrams”, “by mouth”, and “twice per day” each from their own dropdowns, and finally add “10 days” from yet another dropdown [2]. You can see how this adds up and becomes time consuming. I’m not saying we need to go back to paper, but I would like to see a system where I can type (or say) “amox 500mg PO BID x 10 days” into a box and the computer automatically translates it for me.

The problem lists have become a hoarder’s stash.

Certainly people can and do “hoard” diagnoses on the problem list. However, I would argue that the problem list (and the allergy list) suffer more from the tragedy of the commons than hoarding. Who has the time and incentive to accurately update a patient’s problem list or meticulously record an allergy? The advantages of having this information accurately recorded don’t necessarily benefit or accrue to the doctors entering the information. The situation then becomes really untenable when others mindlessly record low quality information in those areas (as Dr Sadoughi describes in the article).

As a program adapts and serves more people and more functions, it naturally requires tighter regulation…There will always be those who want to maintain the system and those who want to push the system’s boundaries. Conservatives and liberals emerge.

I think I’m a liberal in this sense; much to the dismay of some of the people I work with.

Burnout seemed to vary by specialty. Surgical professions such as neurosurgery had especially poor ratings of work-life balance and yet lower than average levels of burnout. Emergency physicians, on the other hand, had a better than average work-life balance but the highest burnout scores. The inconsistencies began to make sense when a team at the Mayo Clinic discovered that one of the strongest predictors of burnout was how much time an individual spent tied up doing computer documentation.

I’m not familiar with this research from Mayo, but I think a physician’s own sense of personal agency in relation to caring for their patients is a strong confounding factor when talking about burnout and EMR’s contribution to burnout. As medicine has increased in complexity it has necessarily evolved (and rightly so) to be a team sport. Thus, the centrality of the physician and consequently his or her ability to direct a patient’s care has decreased. Surgical specialties have been less affected by this de-centralization of the physician because they generally only take care of patients outside of the OR in the immediate post-op period when they are the absolute domain experts. So, neurosurgeons are largely the ones directing almost all of a patient’s care, while ER doctors function in a patient’s larger care team which may include their primary care doctor, multiple subspecialists, physical therapists, social workers, etc and consequently they feel like they have less control over how to make the patient better. On top of this, there are insurance companies and the ever-present social determinants of health for which physicians are unable to impact whatsoever. So, while EMRs have undeniably led to poor workflows, laying all blame for the epidemic of physician burnout solely on EMRs ignores some much larger forces at work.

As I observed more of my colleagues, I began to see the insidious ways that the software changed how people work together. They’d become more disconnected; less likely to see and help one another, and often less able to. Jessica Jacobs, a longtime office assistant in my practice—mid-forties, dedicated, with a smoker’s raspy voice—said that each new software system reduced her role and shifted more of her responsibilities onto the doctors. Previously, she sorted the patient records before clinic, drafted letters to patients, prepped routine prescriptions—all tasks that lightened the doctors’ load. None of this was possible anymore. The doctors had to do it all themselves.

If it’s not obvious already, I will just point out that EMRs are designed to obviate some of this more tedious work by automating it. So, while Gawande seems to indicate here that we should be changing the systems to allow more office assistants to do this work, I (and I think the EMR vendors would agree) would argue that we need to improve the systems to eradicate such work from any human workflow.

Adaptation requires two things: mutation and selection. Mutation produces variety and deviation; selection kills off the least functional mutations. Our old, craft-based, pre-computer system of professional practice—in medicine and in other fields—was all mutation and no selection. There was plenty of room for individuals to do things differently from the norm; everyone could be an innovator. But there was no real mechanism for weeding out bad ideas or practices.

Computerization, by contrast, is all selection and no mutation. Leaders install a monolith, and the smallest changes require a committee decision, plus weeks of testing and debugging to make sure that fixing the daylight-saving-time problem, say, doesn’t wreck some other, distant part of the system.

Sorry to quote at length, but I think these two paragraphs are great.

It is then reviewed by a second physician for quality and accuracy, and by an insurance-coding expert, who confirms that it complies with regulations—and who, not incidentally, provides guidance on taking full advantage of billing opportunities.

Ha! Everything truly does revolve around billing in the US health care system. I wonder how much of the virtual scribe company’s pitch is maximizing billing.

What’s more, she now has the time and the energy to explore the benefits of a software system that might otherwise seem to be simply a burden. Kong manages a large number of addiction patients, and has learned how to use a list to track how they are doing as a group, something she could never have done on her own.

This is what I and I think many other doctors want from our EMRs. Simplify our workflows, make creating orders and notes easier, display patient-level data intuitively and meaningfully, and help us treat our patients’ as a large group overall better.

As usual, Gawande weaves a masterful story with just the right ending. Overall, I felt this article was a great representation of what the informatics field is trying to shepherd–dealing with the increasing complexity between humans, computers, and the practice of medicine. Gawande should apply for a clinical informatics fellowship…

  1. Dilbert has a great series on feature creep  ↩

  2. This is a bit of an exaggeration because this tediousness can be mitigated by creating “order sentences” which is essentially creating the full prescription for the most common indications into a preformed order where you can just check a box to fill in all the relevant information. This, however, has its own set of problems.  ↩