There is a lot of hype around using artificial intelligence (AI) technologies in care provider settings. Last week, the University of Pittsburgh Medical Center’s (UPMC) partnered with Microsoft, an expert in AI tools, to launch the first project under the new Healthcare NExt initiative aimed at improving clinicians’ workflow, signaling continued interest and investment in the healthcare AI space.
The nonprofit, which operates 25 hospitals and 600 doctors’ offices and outpatient facilities, is at the forefront of the healthcare AI wave. It invests in the companies that create AI capabilities and launches its own startups, in addition to implementing co-created technologies for its clinicians to use with their patients.
Dr. Rasu Shrestha, UPMC chief innovation officer and executive vice president of UPMC Enterprises, the institution’s commercialization arm, told Healthcare Dive that its AI technologies has resulted in “quite the transformation” as they have allowed for an increased focus on what is in a patient’s best interest.“This person-centered approach to care is really critical because we believe that’s the best way to drive costs down and improve the quality overall of care,” Shrestha said.
"I do believe that there is a huge role for technology to humanize healthcare processes."

Dr. Rasu Shrestha
UPMC Chief Innovation Officer and UPMC Enterprises Executive Vice President
From voice recognition software to clinical decision support systems, AI has been helping to streamline workflow processes in healthcare. It can also be used to improve care delivery and the patient experience.
Reducing clinician burnout
Market research company, Frost & Sullivan, recently projected the AI market to reach $6 billion by 2021 – up from $600 million in 2014. In addition, AI is included in the ECRI Institute's 2017 Top 10 Hospital C-suite Watch List of emerging healthcare technologies. But many AI products still require weekly updates.
Currently, healthcare organizations' increased adoption of technologies, particularly of electronic medical record systems (EMRs), has been having an adverse effect on clinician burnout.
"Patients are more detached than ever before from the physicians because physicians are spending up to 40% of their time entering notes into the EMR."

Dr. Rasu Shrestha
UPMC Chief Innovation Officer and UPMC Enterprises Executive Vice President
However, AI can "positively support physicians' workloads and improve their overall job satisfaction by alleviating administrative burdens," according to a new report from computer software company, Nuance Communications, and Becker's Hospital Review.
With MACRA's recent implementation of the Quality Payment Program, Medicare and Medicaid providers will feel increasingly pressured to have good quality scores because their payments are tied to the value of their services, rather than quantity, in advanced alternative payment models (APMs) and in the Merit-based Incentive Payment System (MIPS). Yet the cause of a poor quality score is sometimes a result of inadequate EMR data entry. "Documentation methodologies that more accurately capture severity can reduce variability in a physician's observed mortality metrics and patient outcomes," the report states.
A 2016 study conducted by Nuance and California-based Quantros found hospitals that use Nuance's clinical documentation improvement software have outperformed other hospitals every year since 2011 in terms of quality.
Evolving the roles of clinicians
The use of AI is already changing the roles of some healthcare employees, perhaps for the better. Over the past decade or so, medical transcriptionists have experienced a gradual replacement as software like voice recognition continued to improve, according to Shrestha.
"In radiology departments across the country and internationally, a lot of transcriptionists evolved into becoming PACS (Picture Archiving and Communications System) administrators and risk administrators," said Shrestha, who is also a radiologist by background. "These healthcare workers come with a lot of knowledge as well about the specifics of terminology and workflow so there's always going to be an evolution in the way that we're doing work. The only thing constant is change and change onto itself."
Crafting personalized care plans
UPMC believes that having data is good, but not enough, according to Shrestha. "What we really need is not data,” he said. “What we need is insights and not just insights, but actionable insights at the point of decision-making.” These actionable insights are necessary when crafting a well-informed, highly personalized care plan for a patient, leveraging the specifics of precision medicine.
Shrestha argues care providers should at this point be asking themselves: What is the genotypic makeup of a patient that makes him or her uniquely special and what needs to be done to ensure that a care plan is tailored to what motivates the patient? "An understanding of patients’ social determinants of health is required because what may work for someone may not work well for others," Shrestha said.
Unfortunately, these kinds of data points are currently not included in EMRs.
UPMC, the University of Pittsburgh, and Carnegie Mellon University formed an alliance in 2015 with the goal of revolutionizing healthcare and wellness by using machine learning, as well as researching and inventing new technologies for patients and doctors to use. “We're betting on creating these technologies and solutions that leverages the data that we have but takes that data and generates insights leveraging machine learning and pattern recognition and deep learning to solutions that we hope we'll be using in the next couple of years at UPMC and throughout the marketplace,” Shrestha said.
Improving remote patient monitoring
Texas-based provider and insurer Vivify Health partnered with UPMC last year to begin monitoring the health of the nonprofit's patient population remotely. The technologies from Vivify, which has signed contracts with more than 500 hospitals, allow UPMC clinicians to perform virtual care visits and receive patients' health data thanks to the health kit or the bring-your-own-device (BYOD) solutions that are provided upon discharge. "We're looking at patients with chronic conditions that come to UPMC and we're leveraging the set of solutions that we're co-creating to risk stratify these patient populations," Shrestha said.
At UPMC, patients who are in the highest risk tier of ending back in the hospital are discharged with certain types of technologies, such as apps, wearables, and sensors. Patients who want to use their own devices, including smartphones and tablets, bring them in so that UPMC staff can add the appropriate monitoring solutions.
"We're using these algorithms to inform us of where these patients are at any given time in terms of their health and wellbeing."

Dr. Rasu Shrestha
UPMC Chief Innovation Officer and UPMC Enterprises Executive Vice President
"Even before patients feel ill and end back up in the emergency department, we know that they're going to fall of the rails," Shrestha said. "We're then able to actively, not passively, but actively intervene and take care of them and guide them back toward health and wellbeing."
A recent case study at Intermountain Healthcare revealed that after a 57-year-old patient began using Vivify’s remote monitoring technologies, his/her emergency room and hospital use decreased, while adherence to medications and diet improved.
Humanizing healthcare
Humanization is an area that Shrestha is particularly passionate about. He believes AI can help care providers humanize care settings tremendously. "If we can leverage the latest groundbreaking technologies to create solutions, that would really make physicians not just an electronic documentor or a scribe," he said.
As clinicians' roles continue to evolve, experts expect them to become part of a more collaborative model and their incentives to rely more heavily on population health management. Value-based care models, which are becoming more and more prevalent, are encouraging increased collaboration among care providers. Shrestha argues "you can't get at quality and outcomes with just one physician taking entire charge of what may happen in a complex patient or a complex workflow."