News

Sandia’s COVID-19 Medical Resource Modeling

Team: Laura Swiler, Teresa Portone, Walter Beyeler

Contributing Writer: Whitney Lacy

When the Ebola outbreak of 2014 ravaged West Africa, blood samples from ailing people would often be sent to a laboratory for testing, but the closest lab was hundreds of miles away. In more urban areas, blood samples were sent to the closest labs, but those labs were often already overwhelmed by the sheer volume of samples to test. Staff had no way of knowing that another lab, a little further away, had plenty of capacity. Sandia recognized this problem and quickly stepped in to help. Using the power of HPC, Sandia created transportation models to quickly determine possible locations for mobile diagnostic laboratories that could better support regions most affected by the outbreak.

In late 2019, a new coronavirus outbreak was quickly becoming a global concern, resulting in COVID-19, a deadly disease caused by the coronavirus with no vaccine in sight. In early 2020, federal and state officials worried that what was happening in Italy—overwhelmed hospitals with too few medical resources—would soon be happening in the United States. How could officials know, in over 3,000 counties across the country, which hospitals had enough of the right medical resources on hand to protect frontline workers and treat infected patients? Which states had excess capacity and might be able to lend resources to other states? In order to provide decision makers with the best answers to these critical questions, Sandia volunteered the power of their HPCs.

Leveraging their experience in Africa, Sandia researchers began their effort by developing discrete event mathematical models to track patient progress through a hospital treatment system. As a patient enters the medical system, many factors affect possible treatment paths, such as a patient’s underlying health conditions or a hospital’s medical resources on hand. Researchers had to incorporate this treatment path uncertainty and the ranges of resource use per patient to provide risk indicators. For patients arriving at hospitals with COVID-19 symptoms, how many will need advanced care?  Will the hospitals have enough consumable resources on hand (masks, gowns, gloves, face shields, sedatives) to protect the frontline staff? Will hospitals have enough fixed resources (regular or ICU beds, ventilators) for patients? And importantly, will hospitals have enough personnel resources (physicians, ICU nurses, respiratory therapists) to handle an influx of critically ill patients?

The power of HPC allowed quick delivery of results that can inform decision makers across the country.

Using data from an epidemiological (“epi”) model, which determines the spread of a virus, Sandia researchers created a “resource demand” model to predict medical resource needs at any geographic scale of information available. While a resource model can take any epi model data as input, for these studies Sandia used county-level patient streams generated from Los Alamos National Laboratory’s existing EpiGrid model.

The resource model predicts the medical resources (consumable, fixed, and personnel) needed based on the epi model’s patient arrival stream predictions.

Using two of Sandia’s institutional program HPC clusters—Ghost and Uno—the generated patient streams were run through the resource model for each of 3,145 counties in the United States, where each county-level run involved 100 samples per scenario to perform uncertainty analysis. Three different social distancing scenarios were investigated. This resulted in approximately 900,000 individual runs of the medical resource model, requiring 15 node hours (540 processor hours), on the HPCs. The results included mean estimates per resource per county, as well as uncertainty in those estimates (e.g., variance, 5th and 95th quantile, and exceedance probabilities).

Within a few weeks of starting this study, Sandia was able to determine the maximum number of resource needs accounting for parameter uncertainties, such as the probability that a patient goes into the ICU, needs a ventilator, the length of stay, etc. Resource needs over time (i.e., the number of ICU beds needed over time) were calculated with uncertainty bounds. Lastly, Sandia calculated state and county risk indicators, such as the percentage of ICU beds available depending on capacity needed.

Sandia’s resource modeling proved valuable in this very critical time in U.S. history. The power of HPC allowed quick delivery of results that can inform decision makers across the country.  As updated patient stream projections become available from the latest epidemiology models, the analysis can be re-run quickly to provide resource projections in rapidly changing environments.

Unless otherwise indicated, this information has been authored by an employee or employees of National Technology and Engineering Solutions of Sandia, operator of Sandia National Laboratories with the U.S. Department of Energy. The U.S. Government has rights to use, reproduce, and distribute this information. The public may copy and use this information without charge, provided that this Notice and any statement of authorship are reproduced on all copies.

While every effort has been made to produce valid data, by using this data, User acknowledges that neither the Government nor operating contractors of the above national laboratories makes any warranty, express or implied, of either the accuracy or completeness of this information or assumes any liability or responsibility for the use of this information. Additionally, this information is provided solely for research purposes and is not provided for purposes of offering medical advice. Accordingly, the U.S. Government and operating contractors of the above national laboratories are not to be liable to any user for any loss or damage, whether in contract, tort (including negligence), breach of statutory duty, or otherwise, even if foreseeable, arising under or in connection with use of or reliance on the content displayed in this report.