Predictive analytics in healthcare has been a useful tool during the COVID-19 pandemic. It has helped healthcare organizations foresee changes in bed capacity or disruptions in the supply chain for personal protective equipment.
Yet implementing predictive analytics requires clean, consistent data, something that in the early stages of the pandemic was a challenge when diagnosing and treating COVID-19 was constantly changing.
At the recent virtual Annual Thought Leadership on Access Symposium (ATLAS), two healthcare CIOs shared their experiences using predictive analytics models and overcoming data challenges before the work paid off. The event was hosted by Kyruus, a patient experience vendor.
“We’ve probably gone through five or six different predictive models — trying them, seeing if they’re actually providing accurate guidance and then finding out that they’re not really anywhere close,” said Cheryl Hertel, CIO at CoxHealth in Springfield, Missouri. “That was definitely a new area of learning for us.”
Predictive analytics programs
At the beginning of the pandemic, Hertel had to quickly shift her strategic IT priorities to address what was happening with COVID-19. One initiative that became critical was finding a way to predict PPE and other health system supply needs.
“We really needed to develop very quickly how to rely on predictive models for how much supply we were going to need,” she said.
Not only did the testing and refining of the predictive analytics models prove to be a challenge, but so was getting buy-in from leaders within the organization, she said.
To fast-track decisions and get the buy-in she needed, Hertel relied on a new governance structure at CoxHealth, which includes tools to track IT projects and provide data back to the organization on long-term IT projects and day-to-day operations.
“It was really a learning curve for us to get the organization to buy into looking at predictive models, trying them, failing fast and then moving onto the next and landing on something we felt we could rely on to be able to give us some guidance,” she said.
Finding a predictive analytics model that worked for the organization helped CoxHealth keep track of and stock up on PPE. It also helped set supply expectations for the long-term through potential dips and surges of COVID-19 cases, Hertel said.
Predictive analytics also proved to be a challenge for Broomfield, Colo.-based SCL Health, said Craig Richardville, the healthcare organization’s CIO. Part of the problem, particularly at the beginning of the pandemic, was the lack of consistent data about COVID-19, Richardville said.
“We didn’t really have enough science around how to deal with this virus, so as we were going through and predicting what it was going to look like tomorrow, next week, next month, that was probably the piece that as every month came by and better data was actually exposed, we started to understand what are the factors that do have an impact and which factors maybe don’t,” he said.
Richardville said interoperability was the biggest barrier to successfully implementing a predictive analytics program. Moving forward, eliminating data sharing challenges will make it easier to use tools like predictive analytics and better prepare healthcare systems for the next public health crisis.
“As an organization and certainly as a community, the more that we can share and break down some of the interoperability challenges we have and provide that kind of data, the more we’ll be able to help each other out in other parts of the country for other pandemics or situations in the future,” he said.