Artificial intelligence in primary care

Physicians could have new assistants in the future: artificial intelligence (AI). AI could be capable of screening patients, offering diagnoses, and suggesting optimal treatments as well as triaging patient inquiries and processing patient claims.

AI is a technology that mimics human thought processes by finding patterns, then using what it finds to make decisions. AI is also defined by its ability to learn as it processes more data over time.

AI, in fact, promises to impact nearly every aspect of primary care, according to health IT experts.

Those knowledgeable about the technology say AI-enabled computer applications will collect and synthesize disparate patient data and present it to physicians alongside insights into the patient’s medical needs.

It will help physicians better identify patients who need more attention and provide personalized recommendations on what protocols would work best for each individual. And AI will even help diagnose patients by analyzing quantities of information that would be too difficult for any human to assess.

“There are very high hopes for AI to improve the lives of physicians as well as the lives of their patients,” says Anil Jain, MD, FACP, an internist who is vice president and chief health information officer for Watson Health at IBM in Armonk, N.Y. “As a primary care doctor, I’m encouraged by this because it could make my life easier.”

AI is gearing up

Several factors have come together recently to support the quickening pace of AI developments for healthcare and, more specifically, for primary care.

Leading the way is the amount of healthcare data collected in recent years. The high level of computing power now available to process large data sets, the ubiquity of computer systems and EHRs, and overall advances in computer technologies also have fueled AI’s advancements in medicine, according to experts.

Still, AI in healthcare is in its early stages, says Teresa Zayas Cabán, Ph.D., chief scientist at the Office of the National Coordinator for Health Information, which funded the 2017 report “Artificial Intelligence for Health and Health Care.”
Zayas Cabán says technology vendors, medical institutions, and other health organizations are working to identify the tasks that would be suitable for AI to ­handle and bring together the right data needed to ­develop, or “train,” AI to handle those tasks.

Anticipated uses

Even though AI is in its early stages, experts say it’s poised to impact both clinical care and administrative tasks in primary care practices and thereby allow more physicians to practice to the top of their license, says Erik Louie, MD, chief medical officer at HealthBox, a healthcare innovation consulting and fund management services firm in Chicago, Ill.

Louie and others envision a variety of different uses. AI could screen patient data from physician notes, tests results, and other data sources and combine that with information from protocols, clinical studies, and recommendations to identify a patient’s condition, which follow-up tests are needed, and which medicines work best. In other words, AI will be significantly more sophisticated versions of today’s clinical decision tools.

Meanwhile, AI could collect and analyze patient data as it is generated from multiple sources, including fitness trackers and at-home monitors, thus aiding physicians as they monitor patients’ health in ways that time and resources without AI don’t allow. “AI lets us look at data and help think about the patient’s total picture. In the future, it will help [physicians] to get insights, rather than just data,” Jain says.

Moreover, AI is expected to enable primary care physicians to engage in precision medicine, where care protocols culled from large-scale studies are tailored to the individual patient based on analysis of his or her own conditions, genetic makeup, and even social circumstances. It’s a level of analysis that no human could do, Jain says.

Experts predict physicians will use AI to help them more accurately identify patients who need specialty care. They also expect physicians will use it to better manage patients by directing those who need higher-level care to specialists while being able to more accurately monitor at-risk patients in their own practice-a critical benefit as the U.S. healthcare system moves more patients to value-based care models.

As an example, Louie points to an AI application to screen for diabetic retinopathy as a tool for primary care physicians. The federal Food and Drug Administration in April approved IDx, a diagnostic system that uses AI to detect diabetic retinopathy. IDx does not require a specialist to interpret the images or results, making it the first such system cleared for use by the FDA.

Physicians and their care teams, even those not normally involved in eye care, can use the technology to screen their patients for the condition during routine office visits.

In addition, physicians soon could have AI take their notes, analyze their discussions with patients, and enter required information directly into their EHR systems. Physicians might even use AI to analyze indicators that they can’t, such as patients’ voices to detect anxiety or depression.

IT forecasters predict that this emerging technology will also improve many front- and back-office functions. For example, AI will guide patients through intake forms that it tailors based on responses, and AI will analyze patient records to determine how to most accurately submit claims for timely reimbursements with less chance of denials.

Numerous benefits expected

AI has the potential to positively impact healthcare in several ways, all delivering real returns, health IT leaders say. Consider, for instance, that research shows that clinicians have an average of 15 minutes with each patient but spend nine of those minutes entering data in their EHRs, says R “Ray” Wang, MPH, principal analyst and founder of Constellation Research Inc. of Monta Vista, Calif. Imagine, he says, how much more a clinician can discuss with the patient if AI takes over those mundane data-entry tasks.

Wang also notes that AI can access and learn new research at a volume and speed impossible for any human to match. That means a physician using AI can expand his or her breadth of knowledge and ensure his or her decisions draw on the latest research as well.

“AI will be going through information-the patient’s genome, history, current medications, all the things the physician doesn’t have the time to be on top of and information that might not even be accessible [to him or her],” says John Huffman, chief scientific officer of data science and AI at Philips, a Dutch company whose portfolio of integrated health tech products includes AI-enabled platforms.

AI at work today
Numerous initiatives and technology vendors are introducing more AI products into a variety of healthcare disciplines, including primary care.

One example of AI at work in primary healthcare is the nonprofit Human Diagnosis Project, or Human Dx, which uses machine learning algorithms to analyze patient conditions.

This is how it works: a physician inputs his or her patient’s background and medical findings into the Human Dx system, which then invites other physicians to review the case and enter their recommended tests and diagnosis. The system uses AI to combine and analyze the information from the contributing physicians along with the patient’s medical records; it also factors in what it has learned from other cases analyzed on the Human Dx system.

In the future, Human Dx will also incorporate genomics, epigenomics, proteomics, published medical research, and health outcomes data. Already, though, thousands of physicians around the world are using Human Dx to gain insight on their patient cases, according to a spokesperson with Human Dx.

Robin Healthcare, a Berkeley, Calif.-based technology company, has developed a voice-enabled AI device that drafts clinical notes for clinicians as they speak and enters them directly into their EHRs. A spokesman says the company is currently serving only orthopedists but plans to expand to other outpatient specialists and eventually primary care physicians.

IBM Watson Health also uses AI to help physicians and researchers derive insights from vast data sets to support cancer care in more than 230 hospitals and health organizations worldwide, according to IBM.

Healthcare organizations, particularly some of the larger institutions, also have implemented AI for administrative needs, such as using chatbots to answer and direct patient inquiries and to handle financial tasks, says Kaveh Safavi, MD, JD, senior managing director for the global health practice at Accenture, a global management consulting firm.
Safavi says EHR vendors and technology giants like Apple and Google as well as numerous startups around the globe are exploring ways that they can develop AI-powered products for the healthcare industry.

Patients seem comfortable interacting with a computer instead of a doctor. Global consulting firm PwC, in its 2017 report, “What doctor? Why AI and robotics will ­define new health,” found that 54 percent of the 12,003 respondents in Europe, the ­Middle East, and Africa said they’re willing to engage with AI and robotics for their healthcare needs, while 38 percent said they’re unwilling to do so. (Some 7 percent said they were neither willing nor unwilling.)

Barriers to widespread use

EHRs and other computerized medical systems have created vast quantities of data over the past few decades, but not all that data is ready to be used to teach AI systems, Zayas Cabán says.

For AI to work, the data and the algorithms that process it need to be perfect, with no “biases,” meaning algorithms producing skewed results due to either ­being trained with faulty data or developed in ways that ­incorporate the developers’ own human misconceptions. She says technology vendors, medical institutions, and other health organizations are still working to “train” AI applications to handle certain tasks by “feeding” systems the large data sets they’ve gathered in recent years.

That work takes time, so health IT experts say physicians shouldn’t expect a plethora of additional AI applications to arrive soon.

Moreover, they predict that AI will work its way into EHRs and other applications gradually, with vendors building AI into existing systems or offering AI functions as add-on components one or two functions at a time. “You won’t go to the AI system; there isn’t going to be an AI product. AI will always be embedded in some operational system,” Huffman says.

Even if AI functions are embedded in EHRs and other applications that physicians already use, health IT leaders say that’s no guarantee physicians will quickly adopt them. They say vendors will have to develop AI applications that integrate well into physicians’ workflow and that are easy to use.

More important, AI vendors will have to demonstrate to physicians and even to their patients-who may eventually interact with the AI applications directly-that the systems are based on solid science and produce consistently accurate results.

Regulators and medical leaders will want to test AI functions and ensure they’re safe, secure, and reliable–requirements that could slow AI’s move into the exam room. “The algorithms have to be transparent, and they have to be explainable,” Wang says.

Furthermore, given how most physicians struggled for many years with EHRs and other health IT components without seeing real benefits, physicians will need to know that AI will help them be better, more efficient clinicians and not become another mundane computer function that they feel obligated to use, says Scott Shipman, MD, MPH, director of primary care initiatives and workforce analysis at the Association of American Medical Colleges.

Artificial intelligence vs. augmented intelligence

Already, however, examples exist of AI stepping into the role of primary care physician.

London-based Your.MD offers a free service that employs AI to help users find health information. Ada Health of Berlin, Germany, similarly uses AI alongside physicians for its app that helps consumers manage their health, an application that also includes some diagnostic capabilities.

Still, health IT experts see a future that involves physicians leveraging AI for clinical and administrative tasks but not one where physicians are actually replaced by AI. As such, Jain says AI should be thought of as augmented intelligence.
Safavi has similar thoughts, saying that AI won’t ever get good enough to replace a physician’s capabilities for higher-level decision-making-even as it starts to help the clinician with his or her job.

“We never thought an AI agent would get good enough to replace the judgment of a well-trained physician,” he adds. “What I would say is we expect the technology to take the simpler and more routine tasks away.”