The use of artificial intelligence is happening in “unlikely places,” according to a recent working paper published by the National Bureau of Economic Research. While much of the country debates the future of AI, America's rural health care providers are pioneering new uses of AI in their care. As the leader of the largest rural health care system in the United States, I predict that AI's impact on rural health care will be significant.
After a decade of decline before the pandemic, rural populations are growing slightly, according to a recent USDA study: Nationwide, rural counties (defined as those with cities with populations of 50,000 or less) grew 0.25% from 2020 to 2022.
However, this modest population growth is unlikely to solve rural health care's biggest demographic challenge: recruiting and retaining enough clinicians to work in medical specialties that are in short supply in rural America.
Three-fifths of federally designated health care worker shortage areas are in rural areas. Rural areas have only 30 specialists per 100,000 people, compared with 263 specialists per 100,000 people in urban areas. Meanwhile, due to an aging workforce approaching retirement, rural areas are expected to see 25% fewer physicians practicing in practice by 2030.
Paradigm Shift
The relentless accelerating pace of medical practice in recent years has led to burnout for nearly all U.S. physicians, but many are encouraged by the potential for AI to make them more efficient, allowing them to refocus on patients instead of chasing electronic medical records.
A paradigm shift is occurring in rural America as rural health care providers embrace the idea that what we do won’t change, but how we do it must change.
Clinicians at Bryan Health in the Nebraska cornfields are now using AI-enabled software that allows them to take notes on their phones with the push of a button, so they can look patients in the eye instead of clicking on a keyboard. The technology uses “ambient listening” to securely record doctor-patient conversations during consultations and transcribe them into the electronic medical record.
Bryan Health CEO Russ Gronewold told me that one doctor in Grand Island called it a “career-extending technology,” and another in Lincoln said he'll look back on the moment as “one of the most important moments” of his medical career.
In the coming weeks, Bryan Health plans to roll out a new generative AI tool in its electronic medical record that is designed to drastically reduce the time it takes physicians to respond to high volumes of patient messages. The new tool will “pre-populate responses,” but physicians will be able to customize and edit each message before sending it.
Recent studies have shown that using large-scale language models may not actually save clinicians time, but time savings may not be the only metric that matters. As Dr. Michael Pfeffer, chief information officer and vice dean at Stanford Health Care and the Stanford School of Medicine, recently announced, his team found that generating draft replies to patient messages “reduces cognitive burden” by giving doctors a starting point.
Marshfield Clinic Health System, which serves the dairy-rich Wisconsin and northern Michigan peninsulas, is taking a slightly different approach to combating physician burnout related to patient messaging: To help with “noise reduction” for physicians, Marshfield will soon be introducing AI technology into its electronic medical records that will sort and route messages to the appropriate members of the care team.
According to Jeri Koster, Marshfield's chief information and digital officer, about 60% of the messages patients send to the system are related to tasks that a nurse or medical assistant could manage, such as prescription refills, schedules and filling out forms. The new tool will “remove the clutter” from doctors' inboxes, allowing them to focus on clinical or emergency messages that require their expertise and immediate attention.
When the U.S. Preventive Services Task Force recommended that screening for colorectal cancer begin at age 45, Sanford Health's clinical team came together to figure out how to manage screening for 100,000 newly eligible people in rural Dakota, where there are limited gastroenterologists. They developed a proprietary AI model that incorporated additional risk factors that may increase risk of colorectal cancer. The model was designed to help doctors understand the risk of the patient in front of them without having to scroll through medical records, saving doctors time they could spend with the patient instead.
The next chapter in AI
AI can do much more than streamline administrative tasks. These technologies will soon serve as another tool in clinicians' black bags, regardless of whether they operate in a metropolitan, suburban, or rural setting. AI-enabled clinical decision support tools can help identify serious health threats, improve diagnoses, and refine medical procedures.
A new White House-led effort is focused on creating a voluntary framework for addressing medical AI, but it doesn’t sidestep the trust barriers that must be addressed for both consumers and healthcare providers, such as ensuring that the data used to train AI models is representative of the population, which can serve as a safeguard against bias.
To date, 38 payers and health systems have joined the collaboration to determine how to leverage AI models safely, securely, and transparently. Bringing diverse voices to the table is a critical component of this work.
The challenges in urban and rural health systems are not the same.
Rural America has some of the highest rates of late-stage breast cancer diagnoses in the country. A recent report from the Centers for Disease Control and Prevention found that rural Americans are more likely to die prematurely from preventable causes like cancer. The possibility of reversing this trend with new AI techniques that predict disease risk would be a game changer for rural clinicians and the patients they treat.
Premature birth is also a rural issue. In rural northern Minnesota, some of the poorest and most geographically isolated areas of the state, obstetrician-gynecologists at Sanford Bemidji Medical Center have launched a pilot program using an FDA-cleared AI non-stress test belt to monitor fetal heart rate and the presence of contractions in patients who may be at high risk for premature birth, allowing them to intervene earlier and ensure the best possible maternal health outcomes.
My kids are obsessed with the movie The Croods, which tells the story of a family living in a prehistoric cave, the furthest thing from high-tech life you can imagine. In the original 2013 film, the stubborn and cautious father (“Fear is what will keep us alive!”) won't let anyone leave the cave except for brief periods at dawn to gather food, and he tells his kids to “never be afraid.” But the kids defy his orders and eventually make it to the other side of the mountain, where they discover a peaceful paradise.
No matter where you live or practice, it’s natural to be cautious about AI in healthcare. But some healthcare providers, including in some of the most remote and rural parts of the U.S., have already turned the tide and are seeing a hopeful new world.
Bill Gassen is president and CEO of Sanford Health, the nation's largest regional health care system, headquartered in Sioux Falls, South Dakota.
