David Y. Ting, chief medical information officer of the Massachusetts General Physicians Organization in Boston and co-director of the MGH Center for Innovation in Digital HealthCare, discusses technologies aimed at easing administrative burden and provider burnout.
Across the US, physician burnout has reached legendary proportions — between 30% and 65% report symptoms of burnout, a problem affecting more early-career physicians than mid-career or late-career colleagues. Surveys show marked declines in work-life satisfaction among doctors (see here and here). As the tide of administrative burden crests, doctors and their patients express stunning levels of dissatisfaction. Nearly every practicing physician has a dismal tale to share about wrestling with electronic health care records (EHR) as patients explain simple or complex health issues. On the plus side, the EHR promises to enable the sharing of vital information, reduce medical errors, and improve diagnostics and outcomes. Too often, though, it requires physicians to focus on a monitor and keyboard rather than on their patients. Nascent solutions now being explored include human medical scribes or hybrid medical scribes and ways to weave voice recognition and artificial intelligence (AI) systems into virtual assistants.
David Y. Ting, MD, FACP, FAAP, is chief medical information officer of the Massachusetts General Physicians Organization, co-director of the MGH Center for Innovation in Digital HealthCare, and a practicing primary care internist and pediatrician. He answers our questions about developing technologies to ease administrative burden and provider burnout.
Interview edited and condensed for clarity
When we ask doctors, “What is burning you out?” it’s the administrative burden of taking care of patients, not the hands-on patient care. It’s all the trappings of documentation, charge capture, electronic prescribing, managing the medication list — all things that doctors know are right to do, but in aggregate feel like a burden. So, how do we return doctors to being one-on-one with their patients instead of one-on-one with a 27-inch monitor?
Technical solutions can help. One major burden for doctors is execution. The amount of information any practitioner needs to deal with for even a straightforward medical issue is astounding. A patient may come in with a simple sore foot, but you have to know all the medications they’ve been prescribed to make sure there are no interactions, their history, their allergies. You have to check for imaging. If the patient is outside of your health organization, you have to connect with their home EHR. Imagine trying to do all that while having a conversation and connecting as a doctor and patient. It’s crazy. The patients can’t stand it. They go into the exam room, and the doctor is just glued to the screen, hands on the keyboard.
Now imagine if the doctor had a virtual assistant taking visual and verbal cues from the doctor. “Can you pull up Ms. Smith’s allergies from the last time she was seen at the other hospital system?” “Does she have any imaging?” “Does she have a medication list? Just queue that up for me and tell me what I need to know.” And the assistant kind of whispers in my ear, “She has an allergy to amoxicillin.” Ah. “Can you look up a good alternative to amoxicillin?”
The verbal cues are like using Alexa: “Alexa, order ciprofloxacin 500 mg twice a day. Please send it to the Walgreens on Main Street and update her medication list.” You can say that so much more naturally than typing it on the computer. And as you’re saying it, the patient is hearing that instruction. It becomes part of the patient-doctor interaction. Now imagine we bring in computer visualization, and all I need to do is gesture with my hands. It reads my body language and knows I’m going to order something, so it pays attention to the next thing I say. Those technologies are already being piloted and tested in a number of health care organizations. And the rudimentary form of a medical or HIPAA-compliant version of Amazon Echo, I would say will be available within the next six months.
Another area of solution is the use of automated virtual medical scribes that harness AI to reduce or eliminate the documentation burden. A virtual scribe adds a new dimension of real time, synchronous documentation. We’ve got to get away from the “courtroom transcript” that traditional speech recognition technology delivers. AI can read the transcript of an entire visit and then move details around, edit paragraphs, and delete what’s irrelevant to come up with a crisp, concise note. I’ve seen promising early demonstrations by a number of vendors ranging from small startups to large industry players. I would predict they’re another five to seven years away from full automation.
Cloud computing has completely changed the ability to deal with the nuances of accents. It iterates and improves so much faster than the traditional voice recognition system. All the storage and processing is on the vendor cloud — a secure, private, encrypted cloud. Not only is this much faster, but the vendor is also taking voice files from every user of that system.
Just as an example, Google Voice has tremendously improved its voice recognition accuracy compared even to Siri. That’s because Apple made a conscious decision to lock down their files to Apple users. Google has many millions more users in their cloud, so now the race for improved voice recognition in the consumer space is a matter of who has more customers.
Again, what is the problem we’re trying to solve? Decreased doctor-patient interaction — that’s what is burning out our clinicians and our patients. Therefore, the metrics need to reflect how much or little opportunity the clinician has to directly interact with the patient as opposed to the monitor, keyboard, and mouse. You want to measure how much time the doctor is interacting with the EHR versus the patient. Also, you want to look at specific modules in the EHR to understand the time spent writing orders, clinical messages, documentation, reviewing, or gathering information. You want to adjust the ratio spent on different activities so that it makes sense. For example, if a doctor is spending an inordinate amount of time writing clinical messages and answering messages from nurses compared with time actually talking to their patients, there’s something broken about the workflow.
Another big burnout issue is pajama time. There’s such an overwhelming amount of work doctors need to do aside from examining patients, they’re taking it home. We have those measures, too: if people are logging in at 11 pm, chances are they’re not sitting in their offices. So, being able to measure and reduce pajama time is a big, big deal.
—Francesca Coltrera