Working for the machine in medicine

How electronic medical records have contributed to burnout in doctors

Quick summary:

• Electronic medical records (EMR) are an upgrade from paper files - with a catch

• EMR is an example of partial automation where humans have to fill in the gaps

• Partial automation can be either “machine-assisted” or “machine-managed”

• Machine assisted jobs are what we typically think of with automation efforts

• Machine managed jobs get less attention but surround us and can lead to burnout

• Even highly skilled positions such as physicians can be “working for the machine”

Our not-so-friendly visit to the family doctor

I’ve been fortunate enough to live in the same city now for almost 25 years. Being able to raise a family while putting down roots in a community has many advantages, and one of those has been to rely on the same medical professionals for most of our time here. My optometrist is one of the first people I met when moving to San Antonio back in 1998; Dr. Treviño has seen me get married (twice) and raise my 3 children over the past 2 1/2 decades. We have also relied on the same dentist for virtually the whole time and the same pediatrician, Dr. F. Our doctors feel like extended family members at this point; it is nice to catch up each office visit. We feel comfortable knowing that they know our family and have our best interests at heart.

Dr. F is a well-known pediatrician locally. He is a super high energy person and always makes a point of spending significant time with his patients, which we appreciate. Dr. F is also outspoken and always lets you know what is on his mind. During our last office visit for my youngest daughter, Dr. F was the most agitated I’ve even seen him in 20 years. His office has just “upgraded” to a new electronic medical records (EMR) system and the conversion from the old system was done rather poorly, so much of my daughter’s information was missing or incorrect. It fell upon Dr. F to thoroughly review each patient record when they visited and makes any necessary corrections to ensure their accuracy. In Dr. F’s words, this was the most inefficient process possible - the office had the most highly paid worker in the practice to do the most mundane task: ensure the patient data was accurate in the system. While attempting to introduce better technology through automation and improve efficiency, the group had instead caused more stress and created longer workdays for its doctors.

Is AI assisting you - or are you assisting it?

Dr. F is not alone with his frustrations. A recently published survey shows that physician burnout following the pandemic has reached record levels. According to recent results published in the Mayo Clinic Proceedings, 63% of doctors reported at least one symptom of burnout in early 2022 compared with just 44% five years earlier in 2017. Only 30% reported satisfaction with their work-life balance in 2022 compared with 43% in 2017. Certainly the stress caused by overtaxed healthcare systems during the recent pandemic contributed to these results, but there was a pre-existing source of stress even before the pandemic that has only gotten worse: the move to EMR.

The technological reasons for physician stress and burnout might be best articulated by noted author Dr. Atul Gawande in a 2018 New Yorker article titled “Why Doctors Hate Their Computers”. Dr. Gawande notes that more than 90% of hospitals have computerized in the past decade. This shift was supposed to make life for medical professionals easier and the experience better for patients. Instead, Dr. Gawande states that screens have gotten in between patients and medical professionals, to the detriment of both. Dr. Gawande cites a 2016 study that shows physicians spent 2 hours on record keeping for each hour they spent with a patient, and that even in the examination room, over half of the time spent interfacing with a screen to do electronic tasks.

Something has gone terribly wrong. Doctors are among the most technology-avid people in society; computerization has simplified tasks in many industries. Yet somehow we’ve reached a point where people in the medical profession actively, viscerally, volubly hate their computers. - Atul Gawande

It’s not only the mind-numbing data entry tasks that are frustrating physicians: it is that these AI-based computer systems have advanced to act as “electronic bosses”. Dr. Gawande notes that over 50% of American health records are stored in a major system called Epic. Dr. Emily Silverman wrote about her experiences using Epic for the New York Times in 2019. She shares that her hospitals old computer system was “too disjointed to have a singular personality”, but when they upgraded to Epic it suddenly was admonishing her when records were judged as incomplete. Dr. Silverman shares that Epic’s modern interface has “overwhelming complexity” and she and her colleagues are faced with “relentless reminders of what we haven’t completed, supplications to correct our documentation for billers, and daily, jaundiced reminders” stating that her electronic record keeping is “deficient”.

Automation journeys - are they worth it?

In his recent book Futureproof: 9 Rules for Humans in the Age of Automation, author Kevin Roose discusses at length how automation is rapidly changing our life at work and at home, often in ways that we aren’t aware of. While headlines may scream about how we will all one day be replaced by machines and artificial intelligence (AI), Roose argues that the debates about superintelligence and the singularity mask more mundane yet pressing concerns about automation.

One of the areas Roose explores is partial automation, scenarios where part of a process is automated by computers and part remains for humans to complete. More specifically, Roose distinguishes between two types of partial automation:

1) machine-assisted jobs, where humans direct and oversee the vast majority of work and use computers to assist, and

2) machine-managed jobs, where most of the work is directed and overseen by machines and humans act as the “gap fillers”

Roose contends that “machine-managed jobs are less about collaborating with AI systems and more about serving them”. He goes on to assert that machine-managed jobs are more at risk of becoming obsolete, because there will be a greater desire to reach full automation. Take rideshare drivers for example. These individuals are assisted by technology, namely the software platforms that are used by User and Lyft to cite two examples. Yet the drivers are being directed by the algorithms: the computer decides who to pick up, what route to drive, and what the fare should be. It’s not secret that the biggest liability for these firms is the driver, and if they could instead completely rely on self-driving vehicles, they would.

Roose uses the example of electronic medical records to illustrate the inherent tradeoffs that come with partial automation. One the one hand, EMR is a huge improvement over the old ways of working. When implemented well, EMR improves patient care, reduces cost, and limits errors. On the other hand, too often EMR is taking away from the intimacy of thee doctor-patient human relationship and making doctors feel that they exist primarily to move data from one screen to the next. Dr. Gawande makes a key distinction between medicine and computers. He states that medicine is a complex adaptive system comprised of interconnected, multilayered parts that evolve over time. Computer systems, by and large, are not: they are complex but are not adaptive on their own. The smallest changes require meetings, funding, and approvals - not to mention testing to ensure any new changes do not disrupt other parts that were previously working just fine. So it is us humans that often are the gap fillers that must adapt to the computer systems, and not vice versa. Our machines do not usually make the most empathic bosses.

What is the role of automation in your organization? Does it enable more machine-assisted jobs or more machine-managed jobs? Do you feel like the systems you use are supporting you, or are you supporting them? Do you get automated reminders to complete tasks or achieve metrics or quality standards? How much discretion do you have when interactive with your systems? Are you required to input information or follow processes that make little sense? How much control do people within your organization have to modify the systems you use? Who drives the conversations around automation with your employer - business? IT? HR?