COVID-19 Incidence Following Fan Attendance: A Case Study of the National Football League 2020-2021 Season

Abstract

Background: The 2020/2021 National Football League (NFL) season was unique. Not just because it took place under incredibly difficult circumstances, but because of the contrasting conditions that led to some games having fans attend and others that did not. The conditions are akin to a natural experiment. The treatment, the football games that included fans, and the control, those games played behind closed doors. Here, we exploit this variation to better understand the effect of mass gatherings during sporting events on local incidence of COVID-19.Methods: We conducted a retrospective quasi-experimental study that included daily COVID-19 case reports from 11 March 2020 to 1 March 2021 for 219 U.S. counties (or parishes); attendance numbers from a total of 269 total NFL games, 117 of which had fans and were assigned to the treatment group, and the remaining unattended 152 games, assigned as controls. Separate daily time-series of COVID-19 cases and case rates (derived from 2019 Census county-level residential population estimates) were generated using three separate smoothing parameters (7-, 14-, and 21-day simple moving averages (SMAs)) for every NFL team and were plotted against the actuals (daily cases/rates) to detect potential spikes (outliers) in incidence levels following games. All approaches were applied to the county in which games took place (N = 30), contiguous counties (N = 189) where fans also likely travelled from, and acombination of both. A series of 7-, 14-, and 21-day windows were set up the day following each of the respective 269 games to identify if any potential spike (outlier) in COVID-19 incidence occurred within the incubation period.Outliers flagged in the period following games were recorded. Poisson exact tests were then used to evaluate for differences in case and rate spike (outlier) incidence between treatment and control groups. Additional testing that made use of the same methodology was also carried out to evaluate the potential influence of five different rates of attendance (dosage effect) on generating spikes in case- and rate-levels of COVID-19.Findings: A total of 72 different tests were conducted to determine if fan attended games led to a significantly greater number of cases and rates of COVID-19 at various levels of spatial aggregation than for games in which no fans were present. The evidence overwhelmingly supports that fan attendance at NFL games led to episodic spikes in the count of COVID-19 cases/rates in the 14-day window in-county (rate ratio 1·36 [95% CI 1·00–1·87], pInterpretation: The results of this primary analysis of COVID-19 incidence levels in the context of mass gatherings at sporting events provides compelling evidence that the presence of fans at NFL home games during the 2020/2021 season led to increased levels of COVID-19 cases and rates both in the counties in which the venues are nested within, and the surrounding counties in which fans likely travel from to attend. The increased spikes in COVID-19 cases and rates among those games attended by fans, particularly for large crowds of over 20,000 suggest that return to sporting and other mass gathering events should be handled with extreme caution and may indeed be premature. Within the context of large-scale sporting events, a more moderate, phased-in approach may be required to initially limit crowds until a sufficient level of herd immunity is reached. Still, even then, professional sports leagues and related large indoor/outdoor events need to continue to monitor public health data especially as variants circulate and data from the long-term effectiveness of vaccines emerges.Funding: The authors received no specific funding for this work.Conflict of Interest: None to declare.