Electronic medical record + analysis = pandemic preparedness

Augmented intelligence supports medical decisions with speed and at large scale

Smiling senior doctor with a digital tablet talking to a female patient at hospital.

Sanford Health was ready to respond when the pandemic hit in 2020 largely because of an investment in data analytics starting around 2015.

That investment included a centralized team of analysts and developers, including math experts who crunch data and guide decisions.

“We decided that this was going to be a data-driven organization,” said Doug Nowak, vice president of data analytics. “Not only centralizing all of our data but also to take it to the next step and get into the world of advanced analytics.

“We consider it ‘augmented intelligence.’ AI is to support the decisions being made, not to make the decisions,” he said.

Nowak likens the AI to a person’s blood pressure or lab results. These algorithms don’t diagnose but rather inform the health care provider by using information that would take a great deal of effort to gather manually.

Algorithms aren’t replacing the provider, said Robert Menzie, lead data scientist on Sanford Health’s advanced analytics group.

“We always say everything we make the computer do the doctor could have done. We just use the computer because we can get a lot more done quicker and sift through a lot more information faster,” said Menzie. “And there might be things in between the details the computer can pull together that the brain can’t.”

Initial uses

Before COVID-19, Menzie and others on Nowak’s team already had started piloting the use of augmented intelligence to help identify some health conditions hidden in patient electronic medical records.

Those algorithms:

“I would consider it population health,” Nowak said. “It’s really to keep someone healthy, not just to fix them once they’re sick but to help determine if they may get sick.”

For example, data in a person’s electronic record may indicate they are prone to falling. That could lead to conversations about looking at their living conditions, medications or other possible contributing factors.

Sanford has “an incredible amount of data” within its electronic medical record, said Erica DeBoer, Sanford Health’s chief nursing officer.

“What augmented intelligence does is organize that information and provide important insights that can inform care,” she said. “AI utilizes all of the incredible data we have within our electronic medical record, pulls it together and can place an alert in front of providers and caregivers at the point of care, so they use historic and current data to make the best possible care decisions.”

Nowak said therein lies the efficiency. While physicians are working with patients about known issues, the health system can be running algorithms against Sanford Health’s entire 1.5 million active patients to flag other concerns.

Pandemic response

When the pandemic hit, Nowak’s team directed its efforts to making sure Sanford Health was better prepared.

Its first model helped determine how many patients may need to be hospitalized, be on ventilators or need intensive care treatment.

It helped determine how much personal protective equipment the system would need and helped with managing staffing levels “to make sure we’ve got the right people on at the right time,” he said.

“It was to help prepare our systems for how many beds we were going to need at any one time throughout the system. It did a good job at that. We had to change it quite regularly because we didn’t know anything about the COVID virus and so we had to keep updating it as we learned more. And as we became smarter, we could help the computer become smarter.”

“We have a pretty strong algorithm. We continue to watch it daily, especially with the impacts of these variances,” Nowak said.

‘We were prepared’

The second model evaluated a patient’s risk of needing hospitalization should they acquire COVID-19. Even before the first vaccine was available, the team built an algorithm to determine who was most at risk, which earned regional and national praise for being proactive.

“We wanted to make sure we get it to the most-risky population first,” Menzie said. “Rather than first come, first serve, we use the at-risk algorithm to determine who should be prioritized to receive the vaccine.”

The algorithm also helped determine how many to invite when vaccines were limited, Nowak said.

“Because not 100% of the people we invite are coming in for the vaccine, we could say we’ve got 1,000 vaccines but with this population we’re inviting, we might have only 50% of them wanting to get the vaccine, so we better invite 2,000. It gave us a tool to do that. And it gave us an idea of who was most likely to get the vaccine,” he said.

That’s no longer an issue because now there’s more vaccine available.

Nowak said it’s impossible to know the entire impact that data analytics played in Sanford Health’s response to the pandemic, but it certainly did help guide the clinical effort because at no point was the health system ill-equipped.

“We were prepared, so I’d like to think that the algorithms did have a hand in that. Which in turn, may have helped save lives and helped patients to recover sooner,” Nowak said.

Learn more

Posted In Coronavirus, Innovations

Leave A Reply