Geolocation data suggest Lake of the Ozarks’ BikeFest may be a regional coronavirus ‘superspreader.’ These 23 counties should be on alert.
The 14th annual ‘Lake of the Ozarks BikeFest’ concluded this past Sunday in Missouri, a festival which in 2019 was estimated to have attracted 125,000 motorcycle enthusiasts. The festival follows last month’s Sturgis motorcycle rally in South Dakota, which local public health departments have already linked to hundreds of new COVID-19 cases in 8 states and a widely reported study claimed could see as many as 266,000 new cases result from a ‘worse case scenario.’
The COVID Alliance has analyzed geolocation data, finding that BikeFest may contribute substantially to spreading COVID-19 through several Midwestern states, though not with the same magnitude or geographic reach as Sturgis.
Unlike Sturgis, where the South Dakota Department of Transportation identified 462,182 attendees, no official count for BikeFest has been made available. The COVID Alliance counted devices attributed to attendees from both rallies, finding 28.2% as many devices marked to have attended BikeFest. If attendees of both events were equally likely to be included in our geolocation data sample, that would suggest between 100,000 to 130,000 attendees to BikeFest. That may still allow for substantial transmission spread from the event, but not at the same scale as Sturgis.
To support the ongoing COVID-19 public health response, the Alliance has been leveraging our vast geolocation data to develop tools to support clear understanding of and rapid response to risky mass gathering events, such as Sturgis and BikeFest. We are now launching our Superspreader Rapid Response Program to give hard-working public health officials quick access to data they can leverage to understand and act upon risks facing their communities.
What makes BikeFest a potential coronavirus superspreader event?
Missouri already ranks near the top of all states for new COVID-19 cases per capita in September, and both Camden and Miller counties, which cover Lake of the Ozarks, are in The White House Coronavirus Task Force’s “red” zone for dangerously high virus transmission.
Though the risk of transmission when individuals are actually riding motorcycles may be quite low, the surrounding festival activity presents ideal circumstances for COVID-19 transmission. Crowded indoor bars are particularly notable, as they often lack proper ventilation and make proper social distancing difficult.
In defiance of direct White House guidance to close bars and dining, an official BikeFest “passport” program encouraged the estimated 100,000 attendees to visit more than 300 bars, restaurants, live music and entertainment shows, many of which were indoors and at full capacity throughout the weekend. Unlike nearby urban areas such as Kansas City and St. Louis, few areas around the lake have required face coverings — and none had put in place limits on mass gatherings. On-the-ground reports suggest only a handful of visitors were seen wearing masks.
What geographic areas are at the highest risk from BikeFest?
An analysis of individual-level geolocation data undertaken by the COVID Alliance has found that BikeFest attendees hail from at least 34 states and 13.5% of all U.S. counties. This is in contrast to Sturgis attendees, who were found to have arrived from far and wide, representing all 50 states and 55% of all U.S. counties.
The top origins for Sturgis attendees included Great Plains states such as South Dakota, Wyoming, Nebraska, North Dakota, Minnesota, Wisconsin, Iowa, and Colorado. BikeFest attendees tended to come from Midwestern states such as Missouri, Kansas, Illinois, and Iowa. Iowa was one of the few states that contributed substantial numbers of attendees to both rallies.
While it’s too early to know how widely spread coronavirus might have been at this particular event, the reported high-risk behavior of attendees suggests that public health officials should still take appropriate precautions for returning residents. For that reason, the Alliance has listed 23 “Counties of Concern” in four states — Missouri, Kansas, Illinois, and Iowa — whose residents were highly represented at Lake of the Ozarks’ BikeFest.
- Camden County, MO — #1 total attendance & per-capita attendance
- St. Louis County, MO — #2 total attendance & per-capita attendance
- Miller County, MO — #3 total attendance
- Jackson County, MO — #4 total attendance
- St. Charles County, MO — #5 total attendance
- Johnson County, KS — #6 total attendance
- Cole County, MO — #7 total attendance & #6 per-capita attendance
- Jefferson County, MO — #8 total attendance
- Boone County, MO — #9 total attendance
- Greene County, MO — #10 total attendance
- Morgan County, MO — #11 total attendance & #3 per-capita attendance
- Madison County, IL — #12 total attendance
- Pulaski County, MO — #13 total attendance & #8 per-capita attendance
- Laclede County, MO — #14 total attendance & #5 per-capita attendance
- Polk County, IA — #15 total attendance
- Callaway County, MO — #16 total attendance & #9 per-capita attendance
- St. Louis City, MO — #17 total attendance
- Clay County, MO — #18 total attendance
- Franklin County, MO — #19 total attendance
- Adams County, IL — #20 total attendance
- Brown County, IL — #4 per capita attendance
- Moniteau County, MO —#7 per-capita attendance
- Scott County, IL — #10 per-capita attendance
How was the analysis of BikeFest undertaken?
This analysis makes use of the COVID Alliance Research Platform (CARP) and a geolocation dataset (provided via X-MODE) with 25.8 million unique mobile devices in the USA, including 3,434 anonymized individuals from our sample determined to have attended BikeFest in and around Lake Ozarks, MO.
To assess event attendance, Alliance data scientists “geo-fenced” 63 individual businesses and downtown Lake Ozark, including venues listed on the official “passport map.” Devices were tagged if they “dwelled” at or in the immediate vicinity of a BikeFest venue for a considerable period of time. This helps to ensure that those driving through, but not participating in BikeFest were not captured in the analysis.
To assess the transmission risk of observed mobility behavior, device data was used to create individual-level metrics that have been found by some researchers as leading indicators for COVID-19 cases or morbidity: daily distance travelled and the amount of time individuals spent at home.
The COVID Alliance will be publishing a follow-up post with with full methodology later this week.
How can public health officials leverage this data to protect residents?
To mitigate COVID-19 contagion, public health officials in counties with large numbers of attendees may wish to enhance case-detection and contact-tracing efforts, issue targeted stay-at-home directives for returning attendees, and refine public health messaging campaigns.
Contact-tracing is a powerful way to mitigate COVID-19 contagion, however very few public health agencies in the United States have managed to implement such a system at full-scale and therefore at full effectiveness. The data and analysis made available by the COVID Alliance relating to BikeFest attendance may help local public health departments ramp up and target such efforts, and to assess the number of cases that may have gone undetected.
Prior research by the COVID Alliance relating to the Sturgis motorcycle rally suggests very poor compliance by rally returnees in states like Minnesota, which issued targeted rally-specific stay-at-home directives — rendering these stay-at-home directives ineffective if not irrelevant. Therefore, counties with many rally attendees may wish to spend more time and resources designing public health messaging campaigns specially to this audience.
All public health officials in “Counties of Concern” are being offered free access to the COVID Alliance‘s interactive dashboards, which include the ability to view and download aggregate data. The technology and data underpinning this platform are generously provided by the Alliance’s partners, including X-Mode, Immuta, Saturn Cloud, Pachyderm, and Snowflake.
To request access, please contact daniel@covidalliance.org from your official work email account.
Given the data being used, how is individual privacy being protected?
Enabling research is a key focus area of the COVID Alliance, but it must be done appropriately with proper protections and controls. This is why we’ve built the COVID Alliance Research Platform (CARP) by embedding privacy and security into the product design, operations, infrastructure, IT systems, and business practices from its beginning. That means no individual’s data can be traced back to them, even for contact-tracing efforts.
By ensuring privacy and security through every phase of the data lifecycle, CARP enables processing of novel and timely data relevant to solving the pandemic while remaining uncompromising on the protection of individual privacy. This will bridge the gap between researchers with the expertise to tackle quantitative questions relevant to stemming the pandemic, and data-driven analytics based on real-world information about citizen mobility and health.
To learn more about how the COVID Alliance undertakes privacy-preserving research in the age of COVID-19, including a technical walkthrough of our privacy-first approach towards analytics on geolocation data, read our previous blog post on Medium. And to learn more broadly how this geolocation data is being used to advance public health capabilities, read this profile in The Wall Street Journal.
How can academics undertake similar research?
The COVID Alliance has launched a Rapid Request for Proposal (RFP) for qualified academic researchers.
To learn more and apply, visit covidalliance.org/rfp.
About the Authors
Dr. Steven Davenport is a data scientist, policy researcher, and Chief of Staff for the COVID Alliance. He holds a PhD in Policy Analysis from the Pardee RAND Graduate School, and has reviewed for academic journals such as PLoS One and the American Journal of Public Health.
Ryan Naughton is the founder and Co-Executive Director of the COVID Alliance.
Abraham Fraifeld is a data science lead with the COVID Alliance. He holds a MS in Computer Science from Georgetown University.
