In most doctors’ offices these days, you’ll find a pattern: Everybody’s Googling, all the time. Physicians search for clues to a diagnosis, or for reminders on the best treatment plans. Patients scour WebMD, tapping in their symptoms and doomscrolling a long list of possible problems.ĚýĚý

But those constant searches leave something to be desired. Doctors don’t have the time to sift through pages of results, and patients don’t have the knowledge to digest medical research. Everybody has trouble finding the most reliable information.ĚýĚý

Optimists believe artificial intelligence could help solve those problems, but the bots might not be ready for prime time. In a , , a 91±¬ÁĎ research professor of environmental & occupational health sciences and of neurology in the 91±¬ÁĎ School of Medicine, described a troubling experience with Google’s Gemini chatbot. When Franklin asked Gemini for information on the outcomes of a specific procedure – a decompressive brachial plexus surgery – the bot gave a detailed answer that cited two medical studies, neither of which existed. Ěý

Franklin wrote that it’s “buyer beware when it comes to using AI Chatbots for the purposes of extracting accurate scientific information or evidence-based guidance.” He recommended that AI experts develop specialized chatbots that pull information only from verified sources.ĚýĚý

One expert working toward a solution is , a 91±¬ÁĎ assistant professor in the Information School who focuses on making AI better at understanding and relaying scientific information. Wang has developed tools to , , and .Ěý

91±¬ÁĎ News sat down with Franklin and Wang to discuss how AI could enhance health care, what’s standing in the way, and whether there’s a downside to democratizing medical research.ĚýĚý

Each of you has studied the possibilities and perils of AI in health care, including the experiences of patients who ask chatbots for medical information. In a best-case scenario, how do you envision AI being used in health and medicine?Ěý

Gary Franklin: Doctors use Google a lot, but they also rely on services like which provide really great summaries of medical information and research. Most doctors have zero time and just want to be able to read something very quickly that is well documented. So from a physician’s perspective trying to find truthful answers, trying to make my practice more efficient, trying to coordinate things better — if this technology could meaningfully contribute to any of those things, then it would be unbelievably great.Ěý

I’m not sure how much doctors will use AI, but for many years, patients have been coming in with questions about what they found on the internet, . AI is just the next step of patients doing this, getting some guidance about what to do with the advice they’re getting. As an example, if a patient sees a surgeon who’s overly aggressive and says they need a big procedure, the patient could ask an AI tool what the broader literature might recommend. And I have concerns about that.Ěý

Lucy Lu Wang: I’ll take this question from the clinician’s perspective, and then from the patient’s perspective.ĚýĚý

From the clinician’s perspective, I agree with what Gary said. Clinicians want to look up information very quickly because they’re so taxed and there’s limited time to treat patients. And you can imagine if the tools that we have, these chatbots, were actually very good at searching for information and very good at citing accurately, that they could become a better replacement for a type of tool like UpToDate, right? Because UpToDate is good, it’s human-curated, but it doesn’t always contain the most fine-grained information you might be looking for.Ěý

These tools could also potentially help clinicians with patient communication, because there’s not always enough time to follow up or explain things in a way that patients can understand. It’s an add-on part of the job for clinicians, and that’s where I think language models and these tools, in an ideal world, could be really beneficial.Ěý

Lastly, on the patient’s side, it would be really amazing to develop these tools that help with patient education and help increase the overall health literacy of the population, beyond what WebMD or Google does. These tools could engage patients with their own health and health care more than before.ĚýĚý

Zooming out from the individual to the systemic, do you see any ways AI could make health systems as a whole function more smoothly?Ěý

GF: One thing I’m curious about is whether these tools can be used to help with coordination across the health care system and between physicians. It’s horrible. There was a book called “” that argued the main problem in American medicine is poor coordination across specialties, or between primary care and anybody else. It’s still horrible, because there’s no function in the medical field that actually does that. So that’s another question: Is there a role here for this kind of technology in coordinating health care?Ěý

LLW: There’s been a lot of work on tools that can summarize a patient’s medical history in their clinical notes, and that could be one way to perform this kind of communication between specialties. There’s another component, too: If patients can directly interact with the system, we can construct a better timeline of the patient’s experiences and how that relates to their clinical medical care.Ěý

We’ve done qualitative research with health care seekers that suggests there are lots of types of questions that people are less willing to ask their clinical provider, but much more willing to put into one of these models. So the models themselves are potentially addressing unmet needs that patients aren’t willing to directly share with their doctors.Ěý

What’s standing in the way of these best-case scenarios?ĚýĚý

LLW: I think there are both technical challenges and socio-technical challenges. In terms of technical challenges, a lot of these models’ training doesn’t currently make them effective for tasks like scientific search and summarization. Ěý

First, these current chatbots are mostly trained to be general-purpose tools, so they’re meant to be OK at everything, but not great at anything. And I think there will be more targeted development towards these more specific tasks, things like scientific search with citations that Gary mentioned before. The current training methods tend to produce models that are instruction-following, and have a very large positive response bias in their outputs. That can lead to things like generating answers with citations that support the answer, even if those citations don’t exist in the real world. These models are also trained to be overconfident in their responses. If the way the model communicates is positive and overconfident, then it’s going to lead to lots of problems in a domain like health care.ĚýĚý

And then, of course, there’s socio-technical problems, like, maybe these models should be developed with the specific goal of supporting scientific search. People are, in fact, working toward these things and have demonstrated good preliminary results.Ěý

GF: So are the folks in your field pretty confident that that can be overcome in a fairly short time?Ěý

LLW: I think the citation problem has already been overcome in research demonstration cases. If we, for example, hook up an LLM to PubMed search and allow it only to cite conclusions based on articles that are indexed in PubMed, then actually the models are very faithful to citations that are retrieved from that search engine. But if you use Gemini and ChatGPT, those are not always hooked up to those research databases.ĚýĚý

GF: The problem is that a person trying to search using those tools doesn’t know that.Ěý

LLW: Right, that’s a problem. People tend to trust these things because, as an example, we now have AI-generated answers at the top of Google search, and people have historically trusted Google search to only index documents that people have written, maybe putting the ones that are more trustworthy at the top. But that AI-generated response can be full of misinformation. What’s happening is that some people are losing trust in traditional search as a consequence. It’s going to be hard to build back that trust, even if we improve the technology.Ěý

We’re really at the beginning of this technology. It took a long time for us to develop meaningful resources on the internet — things like Wikipedia or PubMed. Right now, these chatbots are general-purpose tools, but there are already starting to be mixtures of models underneath. And in the future, they’re going to get better at routing people’s queries to the correct expert models, whether that’s to the model hooked up to PubMed or to trusted documents published by various associates related to health care. And I think that’s likely where we’re headed in the next couple of years.ĚýĚý

Trust and reliability issues aside, are there any potential downsides to deploying these tools widely? I can see a potential problem with people using chatbots to self-diagnose when it might be preferable to see a provider.Ěý

LLW: You think of a resource like WebMD: Was that a net positive or net negative? Before its existence, patients really did have a hard time finding any information at all. And of course, there’s limited face time with clinicians where people actually get to ask those questions. So for every patient who wrongly self-diagnoses on WebMD, there are probably also hundreds of patients who found a quick answer to a question. I think that with these models, it’s going to be similar. They’re going to help address some of the gaps in clinical care where we don’t currently have enough resources.ĚýĚý

For more information or to reach the researchers, email Alden Woods at acwoods@uw.edu.Ěý

Older adults throughout Washington are increasingly isolated during the COVID-19 pandemic, challenged by the even greater reliance on technology and often neglecting chronic health conditions, according to new research from the 91±¬ÁĎ.

And though senior-services agencies are adapting and innovating to meet the needs, researchers say a lack of consistent funding, coupled with the general uncertainty of the pandemic, could destabilize the care system for the state’s older adults.

“The web of services in the aging network is so critical and during the pandemic a lot of nonprofits are struggling. But the people they serve are at particular risk, because we already had a growing crisis of isolation among older adults,” said , assistant professor of social work at the 91±¬ÁĎ and the study’s lead author.

The , funded by a 91±¬ÁĎ Population Health Initiative grant, is the result of interviews with leaders of nearly four dozen organizations around the state during July and August — just as COVID-19 cases spiked, many lockdown restrictions remained in place, and the social and physical effects of the first few months of quarantines became apparent.

According to the Centers for Disease Control and Prevention, eight in 10 deaths from COVID-19 occur in people ages 65 and older. Berridge and her team, all 91±¬ÁĎ faculty and doctoral students who study aging from health and policy-related fields, wanted to examine how Washington’s older adults were faring through the lens of the agencies that serve them, so as to inform future policy and budget decisions around social services.

Washington is home to an estimated 1.7 million people over age 60, and a variety of public and private entities that serve them, from regional Area Agencies on Aging to smaller community-level organizations and senior centers that offer meals, case management and social activities. 91±¬ÁĎ researchers sought a geographically and demographically diverse sampling of organizations in order to get a broad scope of the needs and how they’re being addressed.

Among the findings:

• The COVID-19 pandemic has increased social isolation among older adults, primarily due to their higher susceptibility to the disease and the restrictions that encourage them to stay put, with few, if any visitors.
• Not going out also may mean not going to the doctor, affecting physical health. Agency representatives report new or worsening health conditions among older adults.
• Many communities — low-income older adults; Black, Indigenous and people of color; those with limited English proficiency; and those experiencing homelessness — are at risk of being overlooked and underserved.
• With so many of life’s routines moved online during the pandemic — including medical appointments — the “digital divide,” or lack of Internet access or devices such as phones and laptops, has widened among older adults. And many may not have the resources, or ability or interest to learn, especially among those with dementia or sensory loss.

“The digital divide has always been there. It just matters so much more now, because everything depends on being online,” Berridge said.

The agencies, in turn, are struggling mainly due to higher costs, falling revenues and an uncertain future. Not every service can be seamlessly transitioned to remote operations; several organizations reported laying off staff, while the volunteer corps has shrunk due to quarantine restrictions and infection risks. Many organizations realized they needed to acquire more resources quickly, whether additional PPE for employees or refrigerators to store meals for delivery. Agencies and their clients in rural areas have been particularly vulnerable to resource shortages and technology gaps.

At the same time, the nature and purpose of these organizations — knowing and serving their clients — has been their greatest strength, Berridge said. The sudden shift to remote operations propelled many agencies to collaborate with local businesses or other organizations and come up with creative ways to try to meet the need. Popular innovations included meal delivery in partnership with restaurants and with transit agencies and offering activities such as games and mindfulness exercises over Zoom, Berridge said.

Telemedicine appointments, for example, are helpful to reach isolated seniors but are limited in the scope of conditions that can be managed.

“Telemedicine is a great option for some, but there are certain conditions that cannot be evaluated adequately over a video visit,” said study co-author , an assistant professor of neurology in the 91±¬ÁĎ School of Medicine.

But sustaining such innovations indefinitely, while meeting the health needs of the growing population of older adults, will require budgetary and policy commitments to the state’s aging network, Berridge said. With technology alone, communities need expanded WiFi access and distribution of devices and the means to instruct older adults in how to use laptops, tablets, phones and apps.

By 2030, in a majority of Washington counties. And while the state is recognized as a leader in elder care, inconsistent funding and the potential impact on health could jeopardize its aging social and health care services sectors’ ability to meet the growing need.

“Funding stability is huge. Many of these organizations operate on a shoestring,” Berridge said. “Having the funds, knowing they’re going to be secure, would allow them to maintain their services and to act in more permanent, innovative ways.”

Other co-authors of the study were of the Department of Health Services, of the School of Social Work, and and of the Evans School of Public Policy and Governance.

For more information, contact Berridge at clarawb@uw.edu.

Determining whether your snoring is merely annoying, or crosses the threshold into a life-threatening problem, isn’t convenient or cheap.

The gold standard for diagnosing sleep apnea — a disease which affects roughly 1 in 13 Americans — requires an overnight hospital stay and costs thousands of dollars. The patient sleeps in a strange bed, gets hooked up to a tangle of wires and undergoes an intensive polysomnography test to count how many times a night he or she struggles to breathe.

By contrast, uses a smartphone to wirelessly test for sleep apnea events in a person’s own bedroom. Unlike other home sleep apnea tests in use today, ApneaApp uses inaudible sound waves emanating from the phone’s speakers to track breathing patterns without needing special equipment or sensors attached to the body.

In a at the in May, ApneaApp captured sleep apnea events as accurately as a hospital polysomnography test 98 percent of the time. An abstract will also appear at the in Seattle this June. Researchers believe the app could be available to consumers in the next year or two.

“Right now we don’t have enough sleep clinics, sleep laboratories and sleep specialists in the country to address all the sleep apnea that is out there,” said co-author , professor of neurology and co-director of the .

“These initial results are impressive and suggest that ApneaApp has the potential to be a simple, noninvasive way for the average person to identify sleep apnea events at home and hopefully seek treatment.”

Obstructive sleep apnea affects more than 25 million Americans. It’s characterized by an obstructed airway that causes people to snort or gasp, along with breathing pauses and shallow breathing. The lack of oxygen often causes people to wake up, at least briefly.

In addition to making it harder to stay awake or focus at school or work, untreated sleep apnea can put people at higher risk for high blood pressure, stroke, heart problems, depression and diabetes.

To determine if a person is experiencing sleep apnea events, ApneaApp transforms an Android smartphone phone into an active sonar system that tracks tiny changes in a person’s breathing movements. The phone’s speaker sends out inaudible sound waves, which bounce off a sleeping person’s body and are picked back up by the phone’s microphone.

“It’s similar to the way bats navigate,” said , lead author and a doctoral candidate in the 91±¬ÁĎ’s . “They send out sound signals that hit a target, and when those signals bounce back they know something is there.”

Because the sound waves are at a frequency adults can’t hear, the app easily screens out audible background noise from people talking, cars honking or a bedroom fan. The 91±¬ÁĎ team developed new algorithms and signal-processing innovations to push the boundaries of what people previously believed smartphones could do.

“Right now phones have sensing capabilities that we don’t fully appreciate,” said co-author assistant professor of computer science and engineering and director of the . “If you can recalibrate the sensors that most phones already have, you can use them to achieve really amazing things.”

The clinical study tested ApneaApp on 37 patients undergoing sleep studies at Harborview Medical Center because of apnea concerns. Researchers put a Samsung Galaxy S4 smartphone on a corner of the bed during the overnight sleep study. During nearly 300 hours of testing, the app tracked various respiratory events including central apnea, obstructive apnea and hypopnea with between 95 and 99 percent accuracy, compared to intensive polysomnography.

Patients are also scored using an “apnea-hypopnea index” based on the number of breathing disruptions in the night. Doctors use that calculation to diagnose patients as having no, mild, moderate or severe sleep apnea.

ApneaApp correctly classified 32 out of 37 patients in the clinical study. Four of the five patients that the app put in a different category were on the boundary between no apnea and mild apnea. In those cases, doctors typically have to make a judgment call about treatment.

Tests in a home bedroom setting showed ApneaApp works efficiently at distances of up to 3 feet, in any sleeping position and even when the person is under a blanket. The app is much simpler to use than other home sleep apnea tests. Those typically require a patient to pick up special equipment from a hospital or clinic and wire themselves up at home to multiple instruments and sensors, which can become detached and throw off results.

Because insurance companies typically only pay for a single sleep test, polysomnography results only offer a one-night snapshot. Using ApneaApp at home over the course of several nights or weeks could produce a more complete picture of real-life sleeping patterns.

The researchers are exploring the process of getting federal Food and Drug Administration approval, validating results with more laboratory and in-home testing and determining whether the sonar technology can track other minute body movements during sleep.

Funding for the project came from the National Science Foundation and the 91±¬ÁĎ.

For more information, contact Nandakumar, Gollakota and Watson at apnea@cs.washington.edu.