Opinion | AI in Primary Care? Sure, Bring It On — But Check With End Users First

Might artificial intelligence (AI) hold some helpful answers for improving outpatient primary care?

Over the past couple of months, interest in AI and its applications in medicine have been surging. We’ve been seeing more and more examples of ways that people have been implementing this in their practice, from the grassroots level all the way up to the startup company level.

Clearly, this technology has the potential to be a really powerful tool to help make the lives of doctors, patients, and all healthcare team members much better. Applications range from assisting radiologists by pre-screening images for abnormalities to speeding up the process of protein folding for basic science researchers seeking candidates for cutting-edge new medications.

But where does primary care fit in? In talking to colleagues, and seeing chatter on social media, lots of folks have been trying to figure this out, and while there have been occasional cool use cases, I think we still have a long way to go before we start seeing really fruitful uses of this amazing and sometimes terrifying new technology.

Right now, a lot of the examples that I’ve seen have to do with simple tasks — automating responses to the stuff we get all day long through the patient portal or in messages from our staff in the electronic health record.

I think this is a lot like the idea of building macros and templated visits for the kind of repetitive tasks we do over and over again, like a standardized normal lab letter that we can send to a patient with the click of a button or insertion of a smart phrase, or crafting intelligible patient education materials for a new medical condition. “Create for me a set of user-friendly instructions for a patient, written at the 6th-grade level, explaining how to do the Epley maneuver at home for benign positional vertigo.”

But I think if we put our minds to it — if we let ourselves dream big — we can probably find more and better ways to have this stuff really make everybody’s lives better. I think there are opportunities for individual patients, for chronic disease management for populations, and for preventive medicine interventions that we can build to help address gaps in care and overcome healthcare inequities.

Think about it. How many times a day do you get a message like this in your electronic health record? “Patient says they are scheduled for their mammogram this week and you need to enter an order.”

True, this is a fairly simple process in the electronic health record. I open the encounter that that message is in, click in the Orders section, type in “mammogram,” select the appropriate test, link it to a diagnosis of “breast cancer screening,” select the future date it is scheduled for, click “Sign,” and the order is routed down to the radiologist’s office who is responsible for getting prior authorization from the patient’s insurance company and aligning that order with the patient’s scheduled appointment.

Whew! Lots of clicks, a few minutes of our precious time, not much clinical brainpower required, and not why we went into the field of medicine.

For simple healthcare maintenance items that are regularly scheduled, rather than even this fairly simple process having to go to a provider, wouldn’t it be great if we could teach the machines to think for us and recognize these gaps in care? They could respond to basic well-circumscribed requests, scrub the charts and find what’s missing or what may in fact have actually been done but just not entered correctly in the chart, and help us get this work done behind the scenes, so we almost never need to see much of anything about it.

I think of this as having a digital medical assistant working by my side, working much faster and more efficiently than any human could, seeing who’s missing their colonoscopy, who never came in for their follow-up labs, who needs a video visit for follow-up for their smoking cessation or their diabetes, and who said they were going to go get their shingles vaccine done at their local pharmacy but apparently never did. This “assistant” could set all these up and line them up for me to sign and approve.

In the future, as we get more sophisticated, perhaps these things will be humming along beside us — listening in on our conversations with our patients; collecting data from our physical exam, laboratory findings, imaging results, and notes; helping us synthesize; acting as a peripheral brain; making suggestions; tweaking our differential diagnosis list; and prodding us and reminding us in ways we won’t find annoying.

Hopefully, smarter people than me are out there working on these things. I would just request that they do this with guidance from the people who are out here doing this work day to day. Talk to us and find out what bugs the heck out of us. Ask whether that system you’re trying to build would really make our lives better, or would just add a lot more busywork to our already hectic, overwhelmed, and burned-out days.

We’ve seen it with things like the electronic health record, the newest telephone answering system tweaks, Open Notes, or immediate and automatic release of labs for patients to view: far too often, when a group of people thinks something’s a really great idea, they forget to check with the people it will impact most — the providers on the front line as well as the patients — to really make sure they’re doing the best thing for everybody involved.

If we move into this brave new world with the right level of guidance and feedback, not just forging ahead blindly because someone thinks it’s a good idea, we’re more likely to end up with a new suite of powerful tools that actually make everybody happy.

And help us take better care of everyone.

Please enable JavaScript to view the

comments powered by Disqus.