Geoprocessing-enabled COVID-19 map aids resource allocation amid pandemic
A new spatially enabled coronavirus map leverages multiple data models to more aptly pinpoint areas in need of limited medical equipment and resources.
Months after the initial cases of COVID-19, the pandemic continues to spread around the globe. During that time, a seemingly infinite scroll of coronavirus infographics and forecasting models detailing ways to “flatten the curve” have become evening news mainstays. In lieu of a vaccine, these models exist as the only feasible front-line strategy to mitigate the spread of the virus and salvage any semblance of a functional healthcare system in the months ahead.
Recently, we had a chance to speak with Lauren Bennett who is the spatial analysis and data science software development lead at Esri. Bennett’s team of geographers and statisticians are working to transition this armamentarium of models into sophisticated mapping solutions to aid coronavirus response efforts. Did we mention there are lots of models?
“If you want the top 10 most used models, it’s 10 different answers, right? They all have similarities, they’re not like 10 completely different answers, but they’re all different. In some cases, it’s levels of magnitude different, depending,” Bennett said. “They have pros and cons.”
At the moment, there are innumerable models for agencies and health officials to leverage. As is the case with any forecasting tool, the model is only as smart as its inputs and the underlying algorithm. For increased utility and presumed accuracy, more organizations are deploying a suite of tools to guide their approach to the pandemic.
“There’s a lot of models and they all have weaknesses and our users, what we’re seeing from the local county level, at the state level, and at the federal level, no one wants to make decisions using just a single model, which is great. They shouldn’t, right?” Bennett said rhetorically.
SEE: COVID-19: A guide and checklist for restarting your business (TechRepublic Premium)
Understanding these forecasting tools
The epidemiologists often utilize mathematical “SIR” models that account for the susceptible, infectious, and recovered within a given population. The Predictive Healthcare Team at Penn Medicine modified a SIR to create their own model known as the COVID-19 Hospital Impact Model for Epidemics, or simply CHIME. CHIME then uses available data to forecast hospitalizations and, consequently, the overall demand for critical medical resources (ICU beds, ventilators, etc.)
Unlike other traditional epidemiological SIR models that simply churn out projections with minimal or user inputs, CHIME allows users to add more data to the model. This is invaluable for local agencies with a keen understanding of hyper localized data in their area.
“Your county who knows literally the exact things that are going on in your county, you may want to put in data because you know more of what’s going on on the ground than anyone else. That’s a pro of being able to plug in your own data,” Bennett said.
CHIME also enables organizations to incorporate hypothetical data to run a series of what-if scenarios to test the forecasted effects of public policy such as lifting lockdown measures. This scenario-specific component will become critical as more states begin to ease restrictions in the weeks ahead.
Adding the spatial piece to the epidemiological puzzle
Bennett’s team incorporated their geospatial expertise to these open source tools to help organizations visualize this hyper localized data within a geographic information system (GIS) for forecasting purposes. In more simplified terms: Esri parlays info from this enhanced SIR model used in tandem with localized health data to create a map. Simply put, all of the data in the world means nothing if it isn’t packaged in a helpful and meaningful way.
“There are very few industries where there isn’t a need for an understanding of the spatial patterns in data. While it is easy to overlook space, it’s usually a critical piece of how things work and truly understanding what’s going on,” Bennett said.
SEE: Top 100+ tips for telecommuters and managers (free PDF) (TechRepublic)
Visualizing the data
It’s this helpful visual of space and relation to place that can help pilot an effective, collaborative multi-agency emergency response. Esri’s geospatial interface program works in tandem with the CHIME forecasting model and transitions these forecasts into spatially oriented maps to aid strategy. These geospatial maps can then be used for overcapacity planning, especially as states begin to tweak social distancing protocols. This all boils down to the R-naught or the infection rate for a given contagion.
“If you can change that number, you change how many people get sick, you change how many people are in the hospital at any given time. You change how much of a burden you have in your healthcare system,” Bennett said. “So being able to see exactly how that plays out and how small changes to policies will impact things like hospitalizations can help these agencies make more informed decisions.”
For added benefit, this all-in-one component acts as a one-stop-shop for organizations, allowing them to work within a single platform rather than juggling a full suite of models and forecasting tools.
Managing critical resources during hospitalization surges
One of the more daunting challenges during a surge of hospitalizations is the sheer logistics of resource allocation. This involves strategically placing overcapacity medical resources in areas where they can be most efficiently deployed to hospitals in a given area. It’s essentially epidemiological supply-and-demand where distance matters and lives are at stake.
“The demand could be how many beds we think we’re going to be overcapacity potentially in the worst-case scenario that day. And they say OK, well, where would we put extra capacity? So that it’s within a reasonable drive of the places that expect overcapacity so that we equitably serve the underlying population,” Bennett said.
This regional and local data-enhanced spatial map can then be leveraged to determine where to place “surge sites” for medical resources at the county and state levels.
“If you’ve got 10 counties all next to each other that are all going to be high, that’s problematic. If you’ve got one, but the rest of them don’t look like they’re going to be quite as bad you can think about how do we move things around to do things as efficiently as possible,” Bennett said.
The ability to visualize this forecasted demand will be imperative moving forward, especially as businesses resume operations and employees slowly return to the workplace amid a modern plague.
Preserving the healthcare system
The initial wave of coronavirus gave rise to grizzly images worldwide. Makeshift field hospitals peppered the globe, and in some cities, refrigeration trucks were hauled in acting as impromptu morgues for facilities beyond operational thresholds. In some countries, overcapacity hospitals with quickly depleting resources were left in a precarious ethical situation with tough decisions to make.
For now, there’s no telling when a vaccine or cocktail of effective treatments will be readily available. That said, the infographics will continue to air on nonstop blocks for the foreseeable future and for good reason.
“There’s a lot of work being done about antivirals and [a] vaccine. Those are very important things that will change our response, but these models are all about protecting the healthcare system,” Bennett said.
In the interim, these forecasting tools are all that remain between a functioning system and a system in shambles.