GE Healthcare

AI-Powered Precision Medicine: Forging a Path to Personalized Health

Today, computers can capture and analyze a staggering amount of clinical, genomic, consumer and imaging data to help clinicians and caregivers make informed care decisions at the individual patient level and, increasingly, prevent disease from occurring at all.

The pace of growth for this new form of care delivery — called precision health — is astonishing. Health care organizations are capitalizing on rapid advancements in artificial intelligence, increasingly cost-effective cloud services and automation. They are also experimenting with holistic approaches to patient care that address social and environmental determinants of health. Together, this is transforming the way care is delivered and the care experience for patients.

Driven by new technologies and a broadening of perspective, a culture shift is happening in how health systems care for patients. Adoption of precision health is increasingly viewed as essential to improving patient outcomes and keeping a competitive edge.

At the American Hospital Association's Leadership Summit in San Diego, four hospital executives from around the country gathered in an Executive Roundtable sponsored by GE Healthcare for a conversation on their approach to precision health and their vision for the future of care.

thumb

Defining precision health

Precision health is the future of patient care, the executives participating in the discussion agreed.

“Using the tools of precision medicine, such as big data analytics and genomics, the goal of precision health is to predict, prevent and cure disease — precisely. Critically, in that order,” said Lloyd Minor, M.D., the Carl and Elizabeth Naumann Dean of the Stanford University School of Medicine in Palo Alto, Calif. “The difference may appear subtle, but in practice, it’s a major shift from the traditional sick care’ model that has driven our [field]. We believe that the full realization of precision health is upstream — before disease has a chance to take hold, and when care interventions can have the greatest impact on a person’s long-term health.”

“Precision health is about helping individual patients along their ‘health care journey,’ while taking into consideration their personal socio-economic and medical circumstances,” said Claire Zangerle, chief nurse executive at Allegheny Health Network in Pittsburgh.

“How are you helping each individual patient along [his or her] health care journey and combating social determinants?” she asked, adding that precision health can be used to more accurately match the clinical workforce with patient needs.

Precision health for inpatient quality improvement

Carilion Clinic, a seven-hospital, nonprofit health system headquartered in Roanoke, Va., is using precision health to react more quickly to hospitalized patients who are at risk of worsening health status. Starting in 2016, Carilion Clinic engaged with clinical surveillance tools linked to the electronic health record system to conduct predictive analytics for inpatient care. These tools are helping clinicians determine when a patient is at high risk of rapid decline. This system helps to determine which patients need to be in the ICU and which patients are ready to go home. More recently, Carilion has begun using a machine learning platform to identify the most impactful interventions to reduce the likelihood of readmissions and inpatient falls for each individual patient.

“We think that the most immediate opportunities for the use of AI and machine learning, are to reduce adverse complex health outcomes that are preventable,” said Jonathan Gleason, M.D., vice president of clinical advancement and patient safety at Carilion Clinic.

Stanford Health Care is using algorithms for patients hospitalized with congestive heart failure to determine whether the patient is ready to go home. The algorithm, created by one of Stanford Medicine’s interns, monitors a patient’s status in real time and delivers to the care team a score on the probability that the patient is ready to be discharged and the risk of being readmitted within 30 days, Minor said.

Similarly, Alleghany Health Network is conducting a pilot program that uses computer algorithms to risk stratify patients for health coaching. Depending on the acuity of a patient's condition, he or she is assigned a health coach or a nurse for regular check-ins, Zangerle said. More complex health needs are escalated to the primary care physician, she added.

To address repeat emergency department visits, CoxHealth, a six-hospital, nonprofit community health system in Springfield, Mo., used analytics to roll out a successful, advanced practice paramedic (APP) program in which APPs go into the homes. “Over a year's time, they looked at visits to the ED, and if a patient had been there five or more times within that year, those patients were the first ones to be enrolled into this program. It’s not meant to be a forever program. You graduate after 90 days and it's been very successful,” said Genny Maroc, vice president of clinical services.

Incorporating consumer-facing devices in individual care decisions

Data from consumer-facing devices are being increasingly woven into care decisions. This is a fast-growing aspect of precision health that providers are testing. In November 2017, Stanford Medicine announced a collaboration with Apple to conduct the Apple Heart Study, during which participants used the Apple Watch's heart rate sensor to collect data on their heart rhythms. The goal is to determine if the Apple Heart Study app can use data from Apple Watch to identify irregular heart rhythms, including those from potentially serious heart conditions such as atrial fibrillation, which is the leading cause of stroke and often goes undiagnosed.

“The Apple Watch is now the most commonly used heart rate monitor in the world,” said Minor. “Hundreds of thousands of people have enrolled, and based on the data from the watch as well as some of your historic data, the watch can give you an indication of [whether] you might be at risk of having AFib.”

Those flagged as at-risk for AFib by the watch can connect with a physician via a partnership with the telehealth company American Well, he added.

“The problem with Fitbits and other [wearables] is they tend not to be sticky,” Minor said. “They wind up on the shelf after a few months. It's been hard for people to integrate them into their lifestyles but, hopefully, that will change over time. ”

Gleason of Carilion Clinic said the wearables market is in an interesting time. “It's not at all clear to me what kind of penetration we are going to get with these technologies,” he said. “Is it going to have a felt impact?”

Setting priorities for precision health with limited resources

With the surge in availability of analytics tools to drive precision health adoption, it can be difficult to know where to place resources for maximum benefit.

CoxHealth has an Innovation Accelerator, where about 50 employees present improvement ideas and those ideas are whittled down to several winners. “The system makes a commitment that whatever comes out of it is approved and funded,” said Maroc.

Often, priorities depend on the funding mechanism and what is possible within the operating budget or with external sources such as grants, the executives said.

“I think centralizing it works so that you have a repository and you have multiple disciplines working together to get the job done,” said Zangerle. “It's better if it's more than one person and more than one entity because it is a joint effort.”

It comes down to what is the biggest problem a hospital or health system is facing, or where it is lagging on quality control, she said.

“The first thing you have to do is look at your biggest pain points,” Zangerle said. “Say your biggest pain point is very high readmissions. That's where you go back to the readmission algorithm. Or you might have a particular disease process that is the issue.”

Once one area of precision health is implemented, the same process or workflow often can be transferred to another problem area for easier implementation, she added.

Challenges: reimbursement, interoperability, data integrity

Often, the biggest roadblock to getting a precision health idea off the ground is the lack of traditional reimbursement from payers, the executives said.

Although all participating executives work at mission-driven organizations, they still face the realities of a fee-for-service reimbursement environment.

“That's one of the challenges. The application of machine learning to improve wellness and health overall is not being reimbursed,” said Gleason. “We are looking at creative ways to overcome it.”

Interoperability across the care continuum and among various providers is also a problem, some executives said.

“We don't have the ability to easily aggregate meta-data in a way that would make analysis much more meaningful,” Gleason said. “The real power is not from looking at data from any individual system, but from looking at data across multiple systems, and from outside of health systems, and being able to pick out trends and patterns that otherwise would not emerge.”

Zangerle described it as ‘friction’ in the health system. “We’ve got to figure out how to take the friction out,” she said. “And I think there is value in asking patients what they want.”

Rob Nelson, global marketing director for AI and analytics solutions at GE Healthcare, agreed that data interoperability is important to move the field forward and improve the overall patient experience.

“From a device manufacturer and technology provider standpoint, we are trying to build AI algorithms to do diagnostic care,” Nelson said. “One of our challenges is that we need global data to train algorithms. We are looking to advance more collaborative approaches to data sharing so we can build algorithms faster and better.”

The road ahead for precision health

Precision health won't replace physicians and nurses, but it will help them do their jobs more efficiently, allow more time at the bedside and solve the hardest cases, the executives agreed.

“I don't think radiologists are going to be put out of business, for example,” said Minor. “But what AI is going to do is prioritize things.”

Precision health could help reduce physician burnout and inefficiencies, he added. “Now we get no support as physicians from our EHR in thinking about diagnoses,” Minor said.

Gleason agreed. “Technology that provides immediate, usable, aesthetic access to care is likely to accelerate,” he said.

Health systems that adopt precision health must have a strong commitment to patient safety first, and the people in place to check the results delivered by machines, said Gleason. “We're very selective and thoughtful about where we are using AI and machine learning,” he said. “And we are not totally trusting our machines right now. When you are exploring machine learning and AI in health care, it is necessary to have a strong safety culture, with active operational surveillance focused on emerging technology.”

One of the lessons learned from the passage and implementation of the Affordable Care Act was the importance of balancing the number of caregivers with patients, as many areas experienced an influx of insured patients, but not enough providers and nurses, said Zangerle.

“There's not enough focus on workforce development,” she said. “Having a nurse vacancy rate of 8 percent across my health system, you have to think about these things. I think we have the opportunity, but they go hand in hand.”

Nelson added, “Hopefully, AI and technology can help us reduce some of the quality issues, and then the reality is probably that we can't solve the human shortage alone. We'll need the technology to supplement and complement what the caregiver is doing because we have a big shortage.”

Precision means putting patients at the center

Ultimately, the adoption of precision health is about doing the right thing for patients, the executives agreed.

It's widely known that hospitals wrestle with how best to care for patients at the end of life and still meet their wishes. More than 80 percent of Medicare beneficiaries prefer to die at home, but one-third die in the hospital, according to government studies.

Precision health can alert clinicians to patients who can benefit the most from palliative or hospice care and who aren't receiving it.

Stanford Medicine developed an algorithm to identify patients with severe end-stage disease, such as cancer, who are likely to benefit from palliative care based on their health histories. The algorithm then triggers a message to a palliative care physician.

“Now, we've been very careful about the design of this approach because ultimately our goal is to encourage dialogue, not to discourage treatment,” said Minor. “One of the palliative care doctors calls the attending physician and says, ‘You know, it seems as though your patient’s not doing well. Is there anything we can do to help?’ And often the answer is, ‘Sure.’”

Research indicates that a patient’s quality of life improves after a palliative care consultation, said Zangerle. “Using that [computer] intelligence to be able to manage that would be very helpful,” she said.

It goes back to supporting the patient to make informed care decisions wherever they are on their health journey, said Zangerle. “Gathering the information is very helpful,” said Zangerle. “Patients want a plan and help to achieve that plan.”