Love & Wisdom
Dr. AI? Not So Fast
Health systems across the state are grappling with how to use AI
By Nat Rubiolicht July 31, 2023
For more than a year, Kaiser Permanente has been using artificial intelligence chatbots to help patients navigate care options. Leaders at UW Medicine are working to understand how AI applies to both clinical care and medical research. Overlake Medical Center is integrating AI into many of its business practices.
ChatGPT and Midjourney have garnered headlines for their capability to write work emails or generate realistic-looking photos of, for example, the Pope wearing a Puffer jacket, but AI reaches far beyond comedy and convenience. Medical professionals across Washington state say AI will eventually reshape how providers treat and diagnose patients.
AI is making its way into the health care space in a similar manner to that of other workplaces: Finding ways to automate tedious day-to-day tasks, like paperwork and record keeping, to improve productivity. Long gone are the days of folders piled high behind a reception desk, filled with patient notes and insurance forms. The vast majority of all health care notation is now done electronically.
“One of the biggest potentials for AI is to assist in sifting through that massive amount of data that a physician has to manage for each patient,” says Dr. Chris Cable, senior medical director of digital health and access management for Kaiser Permanente Washington, a nonprofit that serves more than 700,000 members across the state. “AI tools are primed to take advantage of that and present just the right data to a physician just when it’s needed.”
Kaiser last year began pilot testing an AI-based program that essentially works as an automated advice nurse. Patients send a text about a medical problem and the service asks follow-up questions “very similar to what would be asked in a physician visit,” Cable says. Based on the user’s responses, the tool suggests a potential list of conditions and recommended care options, ranging from self-care to an urgent care visit.
It was first launched to a small group of users last year, but in March became available to all Kaiser members across the state. “Importantly, this tool doesn’t replace the interaction between a patient and physician,” Cable adds. “It’s built to guide the patient to connect with us to solve their needs in the way that’s easiest for them.”
Bellevue-based Overlake is following the lead of its health tech vendor partners and is working to weave AI into its business and enterprise practices, but is taking a more measured approach with chatbots and electronic medical records because “the risks are higher in patient care versus business operations,” says Chief Information and Technology Officer Scott Waters. “I think it’s up to the health care systems to educate the patients and give them a choice on whether they want to have that technology utilized in the process of delivering care to them.”
Overlake, which employs more than 3,000, including more than 1,000 medical professionals, is considering how to best integrate AI into clinical settings, especially on ways to reduce the burden on documentation for physicians. One solution, Waters adds, revolves around ambient listening devices.
“It’s this concept of a virtual scribe,” he says. “A listening device sits in the exam room with the provider and the patient, and uses natural language processing to pick up on a conversation and create clinical documentation.”
This is beneficial in more ways than one. AI that can accurately transcribe all appointment notes allows a doctor to have a more personal connection with a patient, removing the “barrier of technology in between them.” This is also a major time saver, reducing clerical work and allowing providers to focus more on patients.
These adoptions represent major steps for AI’s place in health care. But they’re still only a fraction of the potential that this tech holds in the medical field. Looking ahead, it’s inevitable that AI will become more ingrained in clinical processes and patient care as it develops.
AI has the ability to break down a wealth of information quickly and effectively, so that providers can have access to exactly what they need at any given time, whether that be best practices, a patient’s health history, or specialized medical information.
At the very highest level, AI can help providers make better decisions at the right time.
For example, AI could monitor patient vitals and flag a nursing team when someone is starting to get sicker. Machine learning could be used to assist doctors in more quickly diagnosing medical conditions and finding the best treatment options for patients. Neural networks could help a radiologist interpret medical scans, or help a geneticist better understand the human genome.
“There are thousands of diagnosable conditions, far more conditions than any single providers could know how to diagnose on their own,” says Sean Mooney, chief research information officer at UW Medicine. “At the very highest level, AI can help providers make better decisions at the right time.”
UW Medicine is already working on bringing AI into the clinical space for what Mooney calls “clinical decision support.” While this tech doesn’t make decisions on behalf of providers, it can give an early warning when data indicate that something is wrong with a patient’s health.
The UW School of Medicine — the only five-state medical school in the United States, serving Washington, Wyoming, Alaska, Montana, and Idaho — is also exploring ways to implement AI topics into its student curriculum. It recently launched the Institute for Medical Data Science, a collaboration between the medical school, the college of engineering, and the school of public health.
UW Medicine is partnering with researchers from schools across the country to understand how AI can be applied in both clinical health care and medical research, as well as studying both its positive and negative impacts.
AI, with its ability to automate often tedious but necessary tasks, also has the potential to reduce physician burnout amid the ever-worsening shortage of health care providers. As a result of the pandemic, providers abandoned the health care field in staggering numbers. Between the first quarter of 2020 and the fourth quarter of 2021, around 117,000 physicians left the profession, according to analytics firm Definitive Healthcare. The Association of American Medical Colleges estimates that the country could have a shortage of 124,000 physicians in the next decade.
Definitive Healthcare also estimates that more than 50,000 nurses left the field in the first two years of the pandemic, and the U.S. Bureau of Labor Statistics predicts an impending shortage of more than 275,000 nurses by 2030.
“We view the ‘A’ in AI as ‘assisted,’” Kaiser Permanente’s Cable says. “We don’t plan to use AI to replace that critical and human interaction between a doctor and a patient. We see the potential to remove some of that clerical work, which occupies a lot of the physician’s day. There’s an opportunity in AI to actually give people more time to interact as humans.”
Many patients are excited about the prospects of AI or see it as a “novelty,” Waters notes, adding that he hears lots of questions about the protection of sensitive patient information. Early research shows patients are wary: A survey conducted in December by Pew Research Center showed that 60% of respondents were uncomfortable with the idea of their physician relying on AI for their medical care.
Another major consideration involves the ethics behind using AI models. AI is ultimately as fallible as the people who build it and the data it is based around. If left unchecked, these AI tools can easily exhibit the biases of those developers or data, Waters adds. The medical industry should create an industry-wide ethical framework for “creating equitable care and removing bias from those algorithms.”
“These are learning programs, so they do evolve over time,” Waters says. “There have to be spot checks built in to go back and see if new biases developed. That should be job one, and it should happen now, before we deploy it into something as high stakes as health care.”
At the end of the day, health care is a human-focused, person-to-person business. While AI can lend a helping hand, it can’t take over a doctor’s entire job. Delivering care is about far more than the diagnosis itself.
“I don’t think AI is ever going to, or should ever, replace the personal touch of health care,” Waters says. “We are literally in the people business, and if you take the people out of that business, then it’s a very different environment that we’ll all be living in.”