Predictive Analytics Streamline Hospital Operations at Seattle Children’s

What is a digital twin and how does it improve efficiency?

“Discrete event simulation allows me to turn the hospital into a video game,” explains Eugene Day, DSc, data science manager at Seattle Children’s. “I play the video game to determine the best way to run the hospital and then translate those results back into the real world.”

For more data-minded folks, Day elaborates, “discrete event simulation creates a digital sandbox that is a computerized copy of the system that is useful for understanding, modeling, and generating hypotheses about the hospital.”

“You can use the digital twin to implement potential improvements, address policy changes, or answer questions about the real world. Fundamentally, it is a way of creating the ability to perform scientific experiments on a complex system,” Day continues.

As a trailblazer of the discrete event simulation movement in healthcare, Day uses digital twins to test how something like an admission boarding area in the emergency department would impact operational efficiencies.

“In the real world, it would take far too much time, effort, and money to experiment whether or not we should include an admission boarding area, for example,” Day states. “If we were to run the experiment in the real world, we would have to build one for millions of dollars and let it run for months or years in order to collect enough data to determine whether it is working as we thought it would.”

A discrete event simulation can create a digital twin to run that experiment for the hospital in a risk-free environment leveraging the existing data.

Like almost every other hospital, Seattle Children’s houses vast amounts of data from its various health IT systems, including the EHR which is home to a plethora of patient and utilization information. Using an analytics platform from BigBear.ai, Day and the team of data scientists, engineers, and analysts at Seattle Children’s tap into this wealth of information to create a digital twin with, say, an admission boarding area in the emergency department.

“We can run [the experiment on the digital twin] off multiple years’ worth of data to answer whether the admissions boarding area will improve flow or not,” Day says. “Obviously, [the simulation] isn’t perfect but it is very close and there are systems for validating it. So, it allows me to make informed, evidence-based decisions about operational and policy changes impacting my system.”

Tightening the purse strings

Hospital operational efficiency is key in the wake of the COVID-19 pandemic. Hospitals and health systems are still reporting financial troubles years after the initial surge of the virus, with the average hospital facing negative operating margins for the majority of 2022.

Expenses across the board remain high after the pandemic, while revenues in certain areas have stayed low. This combination could make 2022 one of the worst financial years for hospitals in a long time since hospitals do not have the same level of financial support from the government to offset losses, experts predict.

Hospitals and health systems are having to tighten the purse strings in order to cushion their bottom lines to maintain smooth operations for their patients. They cannot afford to make a wrong move, whether that be altering a workflow in hopes that it streamlines patient care or taking one of the ultimate financial risks in healthcare and opening another clinic or hospital.

Discrete event simulation has helped hospitals and health systems overcome operational challenges over the last couple of years in a less financially risky manner. Seattle Children’s, for example, tapped its predictive analytics platform at the height of the COVID-19 pandemic.

“I used the BigBear.ai platform to answer all kinds of questions at the beginning of the pandemic,” Day states. “How rapidly is our PPE going to diminish if we’re running three operating rooms instead of five? How rapidly can we fill the hospital again once we are able to restart elective surgeries? I was able to answer these questions, or at least give really good estimates. You can’t guess those answers to make operational decisions.”

And when the children’s hospital found that it had more bed capacity than it expected, discrete event simulations helped Seattle Children’s fill gaps for neighboring general acute care hospitals, which were seeing an influx of sicker adults.

Moving beyond the pandemic, discrete event simulation has helped hospitals to target one of their biggest expenses and operational challenges: labor and workforce.

Seattle Children’s creates a capacity action score using the predictive analytics platform to better inform leadership and specific departments about their bed capacity and labor resources for the day.

“Several times a day we predict what the capacity pressure is going to be in 12 hours and 24 hours,” Day explains. “We generate a color code for every unit in the hospital—blue, green, yellow, orange, red—based on capacity and, say, the number of nurses who called in sick that day. We use this highly specific quantitative information and various other sources and combine it into a qualitative measure. Then, we distribute that to leadership and nursing.”

Now, many hospitals have a similar approach to understanding capacity. However, Day stresses that the predictive analytics piece at Seattle Children’s allows him to code units not for their capacity pressure now, but what it will be.

“We may only be yellow at the moment, but we know we are going to be orange in 12 hours,” Day says. “Therefore, what are the levers we pull when we expect to be orange in order to manage it? That may mean calling on-call nurses or opening up a unit that has been closed. Each color code as a specific set of levers that we pull automatically.”

Not planning appropriately will cost a hospital. For example, postponed and cancelled elective procedures are a major cost to hospitals despite being potentially avoidable. By predicting capacity, Seattle Children’s has minimized the number of canceled or postponed elective surgeries because of insufficient bed space. Day also reports that predicting capacity has streamlined workflows for nurses, leading to increased inefficiencies and savings.

“We used to have about 1,500 person hours per year spent on daily huddles just to essentially inform staff about the number of surgeries happening that day, what the hospital’s operations were looking like, and information like that,” Day explains. “All of those are gone now because we have a system in place.”

The hospital hasn’t officially calculated just how much eliminating daily huddles has saved, but a back-of-the-envelope calculation done by Day using publicly available salary data for staff typically in daily huddles shows that the hospital is saving hundreds of thousands of dollars a year.

ROI also stems from right-sizing the workforce for a particular day. Labor expenses have risen 37 percent since 2019, landing at nearly $5,500 per adjusted discharge, according to a recent Kaufman Hall report. Having an evidence-based capacity prediction can ensure hospitals have the right number of staff working to meet patient needs that day.

Predictive analytics are key to hospital operations

Predictive analytics platforms can be a major investment for hospitals. However, the return can make implementation well worth it as Seattle Children’s has found out. Day only sees the value of predictive analytics increasing as hospitals seek more operational efficiencies.

“Adopting predictive analytics of whatever kind in hospital operations is a critical way forward,” Day explains. “Healthcare is only getting more expensive and more capacity constrained. We need ways to understand what’s coming and react to it before it happens. Tools like ours are inevitably going to become more and more part of the landscape.”

Seattle Children’s intends to continue tapping into this “wide open field,” according to Day. The hospital is looking into ways predictive analytics can be used for bedside analytics, for example.

“We get data from patient monitors in order to look at kidney function and things like that to predict organ failure,” Day says. “That’s one area where predictive analytics is going.”

There is also an opportunity for discrete event simulation to optimize emergency department workflows. The emergency department is a fast-paced environment with a lot of flow hurdles. For example, a patient with a broken bone and diabetes might bounce between orthopedics and endocrinology, ultimately landing in general medicine after some time so a provider can address both aspects.

“But there is a lot of competition for general medicine beds,” Day says. “If we can identify which of our patients are the likeliest to have a challenging time before being admitted to the hospital, we can find the right bed faster. These kinds of predictive analytics are going to make a big difference.”

Ultimately, Day believes predictive analytics will bring health systems closer to public health.

“My long game is public health,” Day states. “I work at the hospital operations level, but what I really want is a functioning healthcare system that takes care of the kids in my community and optimizes healthcare for them. Having a smoothly operating hospital that provides the best care and the quickest care for most kids generates capacity so that more kids in the community can be seen when and where they need to. That leads to a healthier community.”