Model citizens: Faculty and staff guide organization through pandemic with mathematical predictions
Providing care for patients through a pandemic — and with a new disease — is a challenge in and of itself. But planning ahead and trying to predict the path the pandemic will take may be even tougher.
Fortunately, Michigan Medicine has been guided through that process by a multidisciplinary team that is using predictive modeling to ensure the organization has enough patient capacity and personal protective equipment to keep everyone as safe as possible.
“When the COVID-19 pandemic was in the early stages and not yet impacting Michigan, a predictive model was being developed to help us plan for the needs of incoming patients,” said Vikas Parekh, M.D., associate chief clinical officer, UH/CVC; medical director, capacity management, UMHS and professor of internal medicine at Michigan Medicine.
In fact, mathematical models have been used for years as a way to help the health care industry understand the dynamics of what is happening in the outside world — and how they would impact patients and providers.
“Predictive mathematical models help us understand what scenarios we are likely to experience in the future,” said Jenny Pardo, an IT project manager and member of Parekh’s team. “Models take some of the guess work out of how systems will work in certain circumstances, allowing us to be better prepared.”
Prior to COVID-19, modeling developed by this team was being used across the health system to plan for things like staffing needs, surgical scheduling and bed availability. Now, models are still being used for those purposes, but have quickly adapted to tracking these needs for COVID and non-COVID patients alike.
Decisions driven by math
The current model used by Parekh’s team was built from a standard epidemiological model that was then modified using data from the current pandemic. Such data included the rate of spread of the illness in other regions or countries, the likelihood of needing hospitalization and the duration of illness.
“Every infection is unique and reacts differently to environmental factors, meaning you have to tailor the model specifically to match the trajectory of the current virus,” said Pardo. “We were able to include a lot of information from trends from COVID-19 spreading through other countries and states before it reached Michigan. Watching how the virus spread in other places gave us a lot of useful information that improved the accuracy of our model.”
Utilizing a well-published, time-tested epidemiological model reduced some of the leg-work necessary to get a model up and running quickly.
“Once COVID-19 hit Michigan, we needed to make a lot of decisions very quickly to ensure we could safely care for our patients,” said Parekh. “Our model helped us plan for the worst-case scenario so we had an idea of how many patients to expect and could plan for the number of beds needed, how much equipment we would need, the amount of PPE and other supplies to order or request. This model really helped drive a lot of those decisions.”
The model also enabled leadership to forecast how quickly the hospitals would run out of bed space, leading to a potential need to open an off-site field hospital.
“Before social distancing measures and ‘stay at home’ orders were in place, our model was showing a large influx of patients very quickly, which led to the development of a task force to plan for opening a field hospital to help care for the expected number of patients,” said Parekh. “Once our model accounted for the decrease in cases due to the interventions, we were able to transition from looking to open a field hospital to simply planning for what one would look like in the event that one would be needed in the future.”
Sharing the knowledge
As the number of COVID-19 cases in Michigan increased, Parekh and his team were able to utilize information shared by health systems across the state to further improve the model.
“Our model has proven to be the ‘gold standard’ in the state and is being used by other health systems as well as state officials including the Governor and Legislative leadership to drive decisions,” said Max Garifullin, a capacity management specialist at Michigan Medicine, and member of Parekh’s team. “Sharing information is vital because gathering inputs from other health systems greatly improves the accuracy of our model and makes it even more useful in making key decisions about treating this new patient population.”
Planning for a second wave
As countries continue working to flatten the curve and reduce the spread of COVID-19, there is a lot of research being done to try and predict if and when a second wave or a second peak will occur.
That research is being incorporated into the current model to help Michigan Medicine plan for future needs in order to keep providing safe, quality care for patients.
“Our goal in tracking a second wave is to predict it early and ensure we are prepared to care for those patients,” said Garifullin. “If we can track one or two very specific measures, like symptoms, we can prepare for an increase of COVID-19 cases if we see an increase in those other metrics.
“This will also help us shift from an epidemic control phase to a containment phase in which we are really focused on trying to control the number of times people come into contact with the virus. That will keep all of us safe — which is the entire goal of our modeling system.”