Skip to content Skip to footer

Loading Results

COVID-19: Q&A on building a COVID-19 local planning model for health systems

Start adding items to your reading lists:
Save this item to:
This item has been saved to your reading list.

March 30, 2020

PwC’s Health Research Institute spoke with PwC Director Sierra Hawthorne about Behavior Predictor, a PwC tool that uncovers localized insights about populations.

PwC Health Research Institute (HRI)

You’ve been working on developing Behavior Predictor for four years. Tell us about how it works and what makes it a good tool for capacity planning during the COVID-19 pandemic.

Sierra Hawthorne, Director, PwC Health Industries Advisory

Behavior Predictor gives us a look at the entirety of the population that resides within a certain area, like a ZIP code. We combine multiple data sources to create a synthetic population with the intent of giving insight into each person’s demographics, socioeconomic context, social connections, health conditions and personal health behaviors, such as smoking or diet and exercise.

These are the factors you want to understand about the people that could be using a facility at specific times during a pandemic. And it’s not limited to past information; it also gives us insight into what may happen in the future, which is especially useful for health systems, payers, policymakers and community organizations as they plan for capacity needs.

Most health systems only have the ability to look at existing data they have on hand, and that’s limited to the patients that already go to their hospitals and clinics. We’ve seen some emerging models on COVID-19 case severity based on data that can be collected at patient intake, for example pneumonia and/or acute respiratory distress syndrome (ARDS) among patients that are already presenting in an emergency room. They may even have some information on patients’ personal history, such as smoking status or diet and exercise.

But in a pandemic scenario, many patients getting routed to facilities may not have been seen there before, and many wouldn’t be coming in with a history of pneumonia or ARDS (as that often happens once people are already infected). That means that these clinical severity predictions are useful in triaging in the moment, but not as useful in proactive planning for organizations looking to act—at scale—over a period of several weeks.

That’s why we used simulation on synthetic data at the person level to look at a combination of likely person-level case severity (before people are sick in real life), alongside different transmission scenarios for COVID-19 to help health systems and governments plan. It’s like a virtual laboratory for us to study the future spread and severity of the pandemic at a localized level.

When planning at scale and making trade-offs between different facilities within a health system’s footprint or for different policies that could impact different geographies, it’s better to understand potential patients before they walk through your facility’s door. And that’s what our new local capacity planning model allows us to do.

HRI: What is the SIR (susceptible, infectious, recovered) framework and why did you model capacity against ICU beds?

Sierra Hawthorne: By understanding the potential mix of case severity among those infected, we can estimate capacity needs at a local level. We followed the SIR—or susceptible, infectious, recovered—framework and are continuing to evolve it as we get more information on real-time transmission in communities.

This approach helped us solve for the number of people falling into any one of these three categories to understand the rate of change at which people are moving between each category. There are a lot of different permutations of this framework that are very useful. Any of those would allow us to modify different model assumptions, like the rate of contact between people based on changing social distancing policies, or how long people remain sick based on new therapies that become available. We’ll expect these transmission projections to continue to change as we get additional data on the true number of cases at a local level, day by day.

We think it’s critical to combine this type of transmission framework with an estimate of severity for each case, based on population mix, to get more precise around planning for acute healthcare utilization. As we’re comparing severity and transmission scenarios, we’re running that up against ICU bed capacity at the local level to help project future capacity gaps.

Moving forward, we’ll also be looking at other types of equipment and staffing support, all of which is changing day by day as new bed capacity and equipment is coming online. This view of localized demand/supply insight then provides a view into how different mitigation and suppression strategies can help make these gaps more manageable.

We are still learning about the spillover effects the COVID-19 pandemic is having on people’s everyday behaviors that will also affect ER load and ICU admissions. For example, with less traffic on the roads due to people sheltering in place, we likely can expect fewer traffic accidents and potentially fewer traumas as a result. Less outdoor pollution from less traffic might also mean fewer heart attacks and asthma attacks for some groups.

On the flip side, we may see more people going without prescription refills to manage chronic diseases, and we could see an increase in alcohol consumption and depression as people feel more isolated, which could influence patterns of self-harm. We just don’t know how a lot of this will play out yet, but what we do know is there will be a lot of feedback loops that impact available supply.

When planning at scale and making trade-offs between different facilities within a health system’s footprint or for different policies that could impact different geographies, it’s better to understand potential patients before they walk through your facility’s door. And that’s what our new local capacity planning model allows us to do.

HRI: Were you surprised by any of the initial outputs from the model?

Sierra Hawthorne: I was struck by the concentration of severity risk in specific ZIP codes. Within Dallas, for example, we see populations only three ZIP codes apart with 300% differences in severity ratios, driven largely by chronic disease burden rather than age. For urban areas, we really see a clear distinction based on underlying socioeconomic factors that have already been driving disparities in health outcomes for a long time.

We also saw rural areas as being potentially hard hit by a higher proportion of severe cases, where age and chronic disease patterns drive up the severity of cases. For example, there’s a higher burden of chronic obstructive pulmonary disease (COPD), diabetes, obesity and cardiovascular disease—and a higher proportion of older Americans—than you would see in some healthier coastal states that are getting hit hard now by the pandemic.

There’s a counterbalancing effect, though; people in rural areas are less likely to come in contact with an infected person than people living in urban areas. So they have a lower likelihood of getting the condition and will be likely to get infected later than more closely connected urban areas, but the severity of the cases may be higher. This means a higher ratio will need ICU care.

The challenge for many of these areas is that they overlap, particularly in rural areas, with communities that have seen a number of hospital closures and face substantial healthcare access issues, even outside of a pandemic scenario.

HRI: Any final thoughts?

Sierra Hawthorne: This is a challenging time for many reasons, but it’s also an extraordinary moment in our communities where we’re seeing the best of collaborative and open-source knowledge-sharing between academia, healthcare organizations and community organizations. We’re all on the same team here, trying to understand what the future holds and shape that future to the best of our collective ability. We’re seeing a huge surge in innovation and progress as a result, and that is extremely encouraging.

Read our research

Contact us

Trine K. Tsouderos

HRI Regulatory Center Leader, PwC US

Tel: +1 (312) 241 3824

Crystal Yednak

Senior Manager, Health Research Institute, PwC US

Erin McCallister

Senior Manager, Health Research Institute, PwC US

Follow us