At HIMSS21, Analytics Experts Look at the Shifting Landscape Around AI and ML

It’s complicated, it’s messy, but it’s happening: that was the core takeaway from the opening panel of the AI and Machine Learning Forum on Monday, August 9, at the start of HIMSS21, which is taking place in Las Vegas and is sponsored by the Chicago-based Healthcare Information & Management Systems Society (HIMSS).

The daylong AI and Machine Learning Forum, one of a number of such programs that took place on Monday, focused on the complexities, nuances, and yes, progress, taking place around artificial intelligence and machine learning.

And the first panel, entitled “Leadership Panel: State of the Industry,” brought together key leaders in AI from across U.S. healthcare. James Gaston, senior director in the Healthcare Advisory Services Group at HIMSS Analytics, a division of HIMSS, moderated; he was joined by Kerri Webster, chief analytics officer at Children’s Hospital Colorado (based on the Denver suburb of Aurora); Vikas Chowdhry, chief analytics and information officer at the Parkland Center for Clinical Innovation, a spinoff of Parkland Hospital (Dallas); and Aaron Martin, executive vice president and chief digital officer at the Seattle-based Providence Health.

Early on in the discussion, Gaston asked the panel participants to share with their audience some highlights of their journeys around AI and analytics.

“We have a rich history of using analytics, but until three-and-a-half years ago, it was siloed across seven different areas,” Webster said. “Since October 2021, we’ve pulled together all those people into an analytics division, separate from IT, Quality, etc. We started as a division of one; today, we’re 60-plus people. The hunger and need for data across the organization have been huge from day one—to support research, etc.,” she said. “I joke that we never found an Excel spreadsheet we didn’t love. But this organization has enabled data-driven decision-making and a culture to support research, analytics, and decision-making. We achieved HIMSS Analytics Stage 7.”

Meanwhile, Webster said, “COVID hit and accelerated beyond my wildest dreams how we leveraged analytics. And in a children’s hospital, COVID didn’t really impact kids so much, but our partner hospitals were full. We were the first hospital system in our area to open up a COVID testing system. And they used me and my team to help predict how many beds we could put up to support adults who came to our system, how many test kits to bring together, etc. Analytics really became a cornerstone for decision-making. We had a dashboard up within three days of the command center opening. And that dashboard became instrumental for us.”

The Parkland Center’s Chowdhry said that “The journey of data science at Parkland started back in 2010, when a group of leaders came together and developed a model around the SDOH. And the program has really grown by leaps and bounds. And it was set up as a separate not-for-profit organization called the Parkland Center for Clinical Innovation, where I work. Since then, there are mainly three areas where predictive analytics, machine learning and AI have been leveraged, with Parkland Health as our main partner. Parkland has the number-one trauma center in the area. And we’ve been leveraging AI to develop predictive models.”

Referring to a specific area of activity, Chowdhry reported that, “In terms of sepsis, we developed the model from scratch. Secondly, we have a Medicaid HMO plan with 200,000 members across seven counties in north Texas. They don’t have the resources to do machine learning on their own. So we develop models for care of pediatric asthma and maternal health. And with COVID, we decided not to use AI models to diagnose COVID. We went down the path of operationally using data analytics with COVID. We created a partnership with Parkland and the Dallas County Health Department, and, using information to help Parkland to operationally see which patients might be coming in from COVID hotspots, and we would have information right there. As community spread was taking place in the US, data on traveling abroad was not really helpful. So we developed models to see whether patients had been in hotspot areas. And we used that to direct some patients to telehealth. And our clinicians and managers are understanding that there are many advantages, both clinical and operational, in develop predictive models with public health.”

Meanwhile, Providence Health’s Martin, noting that his health system encompasses 54 hospitals and 1,000 clinics across eight states, said that “Our CIO, BJ Moore, owns the infrastructure. And Ari Robicsek, M.D.,” senior vice president, research, and chief medical analytics officer, he said, has a strong handle on clinical analytics. Meanwhile, “My team, the digital team, is responsible for anything consumer-facing.” One area in which he and his team innovated, he reported, was around analyzing PPE (personal protective equipment) availability and resources during the first, very intense, months of the COVID-19 pandemic in the spring of 2020.

A shifting landscape

“How has your competitive environment changed?” Gaston asked.

“I always say we live in a pretty tough neighborhood,” Martin said, referring to the intensity of the presence of organizations with analytics capabilities. “I can wave from my house to the guy who runs Amazon Care. And kitty corner to that is 98.6—so we’ve got two disruptors in our neighborhood. And COVID was a boon to disruptive organizations in healthcare. Because normally, you need to fund experiments to get consumers to try something. But the clinics got shut down and payments were equalized, and that created a massive marketing campaign for digital-first; patients loved it. While we were caring for the very sick patients, the digital-first disruptors got a leg up on trying out their wares. So if you’re in a health system as I am, you really have to think about your level of funding of digital, because it’s on,” he said, referring to the acceleration of analytics development.

“Digital care is the future,” Gaston agreed.

“At Parkland, it’s an interesting challenge,” Chowdhry said, “because of our unique patient care mix… At the most basic level, you can have innovation that really improves your cost profile”

“I always say, never waste a crisis,” Webster said. “And we’ve learned a lot in the past year-and-a-half. And we’ve had a thousand-fold increase in telehealth-based care delivery, particularly in mental health. So the development will all increase.” Meanwhile, she said, changes in inpatient admissions patterns because of the COVID-19 pandemic have meant that “We’ve been challenged to really predict what our census and capacity will be. Our challenge has been to predict the future. So we’re predicting about 18 months out, and so far, our models have been pretty accurate, in terms of census. But our beds our full right now in August, as they would be in the winter. We’re working to make sure we have enough nurses, enough hospitalists, etc. And I’m a nurse by background; it’s a great feeling that you’re helping our nursing and medical staffs.”

Meanwhile, Chowdhry underscored, “From a strategic perspective, health equity and access have become absolutely a focus. And the answer will not be in building more facilities. We have to come up with a solution. So we’re going to use AI and machine learning to help improve access to patients, with the right focus. One of the projects I’m most excited about,” he said, “is that, last year, our leaders wanted to find out what the personas of people were, who were coming into the ED? So put our heads together and, it’s a long process, but we’re trying to use some machine learning metrics to really understand those cohorts, such as patients coming in through the Parkland ED. And we developed four personas in terms of different types of people with different demographics and needs, including those who could have been helped through primary care or urgent care,” in order to analyze different patient cohorts in order to understand patients better both demographically and clinically.

“And the personal engagement involved is so very important,” Gaston noted. “And how do we begin to help the industry to help reform itself? We need to think about health and health maintenance, too.”

“I think the expectations of healthcare have changed drastically,” Chowdhry said. “Consumers are demanding more.”

“At the forefront of my mind,” Gaston said, “is that most of us in this room are probably very analytical and think about things in an analytical way; but in reality, we have to deal with people, with stakeholders. What’s that like?”

“Without engagement from your bedside clinicians and from your operational stakeholders, you have nothing,” Webster emphasized. “So how do you connect them with the power of analytics? That’s the trick, and that’s challenging. We’ve spun off a machine learning partnership with our clinical analytics group. And this has to be driven by the people who are going to consume the models and the data.”