Don’t ‘Break the Bank’ When Innovating for COVID-19 and Beyond. What One Health System Did.

“You don’t have to be a Mayo Clinic, Cleveland Clinic, … or Johns Hopkins [Medicine] with a dramatically complex data warehousing and data engineering environment,” says Millburg. “Those institutions are doing amazing things. But you can do some amazing things as a small- to mid-sized health system. We are living proof of that.”

In the Beginning

“As we came into the COVID-19 pandemic, it became very clear that there was a need for a  lot of information in a short amount of time,” says Millburg. The nonprofit health system, which has five hospitals including the 500-bed Memorial Medical Center, began holding daily command center meetings, and he noticed many people expending great effort to pull together relevant information on a timely basis. Among the information they wanted to know:

• What level of care are patients receiving?
• How many beds are available?
• What is the testing situation?
• What COVID-19 trends are relevant nationally and locally?
• What are the PPE burn rates?

The decision sciences department—which has six data engineers, two data scientists, one statistician, and another manager—along with Millburg, “just jumped in and started tackling problems, building out data pipelines from all of our various systems and putting that information together in a way that cut down on all [demands] for various leaders.”

The Tools Behind the Innovations

One of the novel aspects of this initiative was that the team used ordinary data tools, says Millburg. These included:

• Microsoft SQL Server was the primary tool used for exchanging information, which Millburg describes as “a normal database software that most institutions would have.”
• Python, an open-source data science programming language.
• Tableau, a data visualization tool. Millburg also suggested that Microsoft Power BI might work.

How They Built the Connections

Memorial Health System has five hospitals using seven EHRs because some of the emergency departments use different systems than the hospital they are associated with. There were further complexities because one hospital had just joined the health system and the connection to the Epic EHR that hospital used had not yet been completed. As a workaround, the team pulled spreadsheet information to obtain necessary data.

In addition, vital information was extracted from the ERP system, as well as bed management, lab, pharmacy, and other IT systems. Previous work to build what Millburg calls “data marts” with lab and pharmacy systems helped accelerate the team’s work. “We already had the building blocks of some of these things, which helped a little bit,” he says, “but when you think about all the new labs that were being entered for COVID-19, we had to very quickly be able to flex what we already knew about the various EHRs and systems, and then react to them.”

To help keep the organization’s leaders on top of national, state, and local trends, the team built interfaces though APIs to appropriate external resources, such as the Johns Hopkins Coronavirus Resource Center map.

As a result, “Our leaders don’t have to make 15 different clicks to get the information they need, and our communications folks don’t have to do screenshots from [a variety of] websites,” he says. “All the information is right there in one spot.”

Keep the Focus Narrow

One reason Millburg’s team was able to move quickly was that they kept the nature of their queries to IT systems narrow. “We’ve gotten really good at being able to take a use case and being able to say, ‘Okay, this is the specific thing that we’re after here,’ ” he explains. “Instead of trying to boil the ocean, let’s just pick this one thing. It really allows you to be very agile.” It is always possible to go back later and expand the query, he says.

Homing in on specific data also enabled the decision support department to work with current resources. “We didn’t grow our workforce at all during this time frame,” he says.

Monitoring Lab Turnaround Times

As the crisis began to abate and hospitals began scheduling elective surgeries, a stipulation to have a COVID-19 test within 72 hours of surgery created delays at some lab facilities. To address this issue, the team built a dashboard to monitor turnaround times for the labs the health system interacts with. “It helped us monitor the performance of the different laboratories and redirect some of these tests to other laboratory facilities in our area,” says Millburg. As a result, they were able to expedite testing.

Creating a Contact Tracing System

Beyond the dashboards, another essential tool the team developed was a contact tracing system to determine which employees within the Memorial Health System came in contact with a COVID-19 positive patient. This required pulling information from multiple IT systems.

“We built tools to very easily harvest information from our systems to help better identify situations where there might’ve been exposure,” Millburg explains. “If a patient came into the system before they were identified as having COVID-19, contact tracing was incredibly important for infection prevention to be able to identify how many people came into contact with that patient, quarantine those individuals, or reach out to them.”

The endeavor started by having someone “going into the chart and looking everywhere for this information,” he says. “We built out a tool where people would be able to enter a patient’s ID and [identify] everyone who documented on that patient within the electronic health record. It expedited how quickly we could do contact tracing,” he says.

Developing a COVID-19 Hot Spot Mapping Tool

Another solution the team created was a mapping dashboard using a geospatial tool that identifies COVID-19 hot spots in the community with concentrations of positive patients, based on their street address. To get a more comprehensive picture of the outbreak, the health system works closely with community health departments, which aggregate information from other nearby hospitals, combine it with other data, and share it back to the contributing organizations. “It was a pretty effective way of sharing information,” Millburg says.

Long-Term Implications

The pandemic helped “breed innovation,” says Millburg. While the tools are useful for COVID-19, there are long-term implications.

• “We’ve essentially laid the groundwork for a respiratory surveillance system across our health system,” he says. The tool can be easily scaled to monitor infections such as the flu, pneumonia, and strep throat. “We’ll have the technology to surveil our community on those types of diseases when they come about in the fall. So, I think that is a great development for our institution.”
• In addition, the health system now has enhanced surveillance tools to monitor PPE and other supplies.
• The contact tracing system also can be applied for other use cases, he says, for example tuberculosis or another infectious disease.

“The surveillance tools that we have are great innovations that will serve our health system for years to come,” Millburg says.

Lessons for Other Health Systems

“Small- and medium-sized health systems can do some amazing data and analytics things; they’ve got the ability to take on some pretty cool innovations in this space and do so without breaking the bank,” Millburg says.

“You need to have some type of database system and data visualization system, and then you need to have some talent,” he says. “From my perspective, the tool set is not nearly as important as the talent pool that you have.” He suggests starting with a small use case and build your capabilities from there.  “That’s the best way to get going,” he says.