Adam Edelman
May 10, 2020

How a Rise of COVID-19 Infections at Large Meat Processors Affected Local Communities

The Washington Post recently reported that three of the nation’s largest meat processors failed to provide protective gear to all workers, and some employees were told to continue working in crowded plants even while sick. This turned the facilities into infection hot spots with hundreds of COVID-19 cases at many of the plants.

In this blog post we analyze how the spike of infections at meat processing plants affected the surrounding communities which tend to be smaller, rural, and where a large percentage of the population is employed at these large meat processing facilities.


To start, we’ll take a look at some of the plants mentioned in the Post article. At a JBS facility in Greeley, Colorado county health officials identified at least 277 employees or dependents with cases of suspected or confirmed COVID-19.  The Denver Post reported on a culture of “working while sick” which led to the company backtracking from the promise of testing every employee when the first day of testing revealed that a significant percentage of plant managers and supervisors were positive for coronavirus.


To analyze activity around these plants we’ll use OmniSci in conjunction with massive mobile device location data and point of interest data, courtesy of our partners at X-Mode and SafeGraph.   


X-Mode’s dataset used in this research is aggregated and generalized. X-Mode does not collect or share any personally identifiable information such as name, email, or phone number. All devices have given consent to location collection.



First, we identified the plants mentioned in the Washington Post article to create a cohort of the mobile devices that were active in and around the plant.  To determine the community impact we need to follow that set of devices into the surrounding communities.  We can do this with the OmniSci Cohort Builder which allows us to select that set of devices and determine where they went over five weeks in February and March, 2020.



As you can see in the map above, there’s an extensive range of travel for this cohort.  Next, we’ll use a simple bar chart to group where most of the devices went by county. In addition, we’ll bring in county level infection data from the New York Times to determine if the counties visited by the devices in the cohort also showed a high level of COVID-19 infection.



From the chart on the right we find that the devices overwhelmingly visited locations in Weld County, Colorado.  That’s not much of a surprise given that Weld County is also where the JBS plant is located. 


Let’s match that up with the infection data in the chart on the left.  Note that Weld County has both the highest rate of infections and deaths in the surrounding counties (per 1 million residents). These rates are higher than even Denver and Boulder counties which contain higher population densities. To visualize this on a map we’ll color the counties in the area by infection rate.




Let’s take a look at some other meat processing plants to see if there’s a similar pattern. On April 21st local health officials in Grand Island, Nebraska confirmed roughly 237 cases tied to a JBS plant. This was up from only 10 JBS workers who tested positive on April 3rd. According to the Washington Post, JBS confirmed that it did not receive masks for its employees until April 2 and did not mandate their use until April 13. 


Going back to the map we see that the counties surrounding Grand Island have some of the highest infection rates in the area.



Like our previous example we can build a cohort around the factory in Grand Island and determine where most of those devices went in the surrounding counties. Again, it’s not a surprise that most of the devices stayed in Hall County (home to Grand Island). We’ll compare it to the county infection rate where we can see the high prevalence of infections in Hall County. It’s worth noting that Hall County has the highest infection rate in the state of Nebraska.  



Let’s look at one more facility, this one a Smithfield Foods factory in Sioux Falls, South Dakota. The factory was closed on April 12th after state health officials confirmed over 200 COVID-19 cases are related to the plant, most of which involved Smithfield employees. On May 4th the New York Times reported that the plant is now the “country’s biggest coronavirus hot spot” responsible for more than 640 cases.  


Those infections led to Minnehaha County becoming second highest in the state for COVID-19 cases. On the map it’s easy to visualize stark contrast between Minnehaha and the surrounding counties but how can we measure the impact back at the local level?  With OmniSci it’s easy to switch between data layers so we can go from county level infection data for South Dakota to the building level visit data for the city of Sioux Falls.




Here we’re using data from Microsoft on US building footprints and coloring the buildings based on the number of times a device that visited the Smithfield Foods plant also visited that location (red meaning more visits). When we zoom into Sioux Falls we can see the factory in the upper right in red but we also notice a cluster of red and orange as we move the map over the downtown area. To identify the businesses most visited by our cohort of devices we’ll bring in places of interest data from SafeGraph. Looking at the graphic below we can see that besides the local park the most popular destination was a local tavern located near the factory. With a lot of devices visiting the tavern this could be the next hot spot. This level of contact tracing will be critical as public health officials struggle to get ahead of local outbreaks.



In summary, we leveraged the parallel processing power of GPUs to cut through over 16 billion location records from 6 million mobile devices to investigate an outbreak of COVID-19. We showed how these communities experienced disproportionately higher COVID-19 infection rates compared to nearby counties. 


Finally, we used this data to hone-in on locations in the community where risk of disease transmission might be higher.


For those healthcare professionals with the arduous task of tracking and notifying individuals potentially exposed to a virus, big data analytics can be a major leap forward in fighting disease spread through contact tracing and notifying potentially infected individuals or groups.


To learn more about how big data analytics can help fight the spread of COVID-19, visit omnisci.com today!


About the Author

Adam Edelman is a senior technical architect who has over 15 years of hands-on experience in the Federal technology sector. Adam currently focuses on delivering leading edge big data analytics into the US public sector.