COVID-19 vaccine program eliminates law enforcement workforce infections: a Bayesian structural time series analysis

ABSTRACT

COVID-19 has created tremendous operational difficulties for law enforcement agencies, with substantial portions of their staff quarantined for either exposure or infection. With the rollout of a vaccine beginning in early 2021, there is hoped for relief on the horizon. However, to date, no study has reported the vaccine’s effect on infection rates within the law enforcement workforce. We address that gap with a report on a single large agency’s experience, using data on officer positivity rates gathered over 341 days. During the immunization period, employees accepted vaccination at over 70% uptake. Results show the vaccine eliminated new cases of COVID-19 among the agency’s nearly 700 employees within weeks. As other agencies consider their vaccination programs, they should consider communicating early and often about the impact of the pandemic on operations and the efficacy of vaccination, including the results reported here.

KEYWORDS:

• COVID-19
• vaccination
• police
• Bayesian structural time series

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. To calculate the police department’s infection rate, we used the authorized number of employees (691) for the denominator across the entire period. Using the largest denominator possible provides the most conservative estimate for the effect of the vaccination program.

2. The Salt Lake County Health Department provided several vaccination sessions after the initial vaccinations for those who did not receive the vaccine originally. At the completion of these make-up sessions, approximately 75% of the Salt Lake City Police Department’s workforce was vaccinated. If the number of employees who acquired a level of natural immunity from infection, but did not receive the vaccine, are included, the percentage of the workforce that has some level of immunity is nearly 80%.

3. Bayesian Structural Time Series modeling was chosen over the more common ARIMA method for a number of reasons. First, BSTS models are flexible and modular. This allows for researchers to determine the structure of the model by considering whether and how to include regressors, whether short- or long-term predictions are more important, and whether seasonal model components are necessary (Scott, 2017). Further, by working in a Bayesian framework, investigators can better acknowledge and incorporate uncertainty into statistical models and discuss outcomes in terms of probabilities, which tend to be more intuitive.

4. The BSTS model was estimated with a local linear trend state component and a regression component for the control variables with spike-and-slab priors over coefficients. Ten thousand Markov Chain Monte Carlo (MCMC) samples were drawn, with 89% posterior distribution credible intervals generated.

Additional information

Notes on contributors

Scott M. Mourtgos

Scott M. Mourtgos is a PhD student in the Department of Political Science at the University of Utah. His research focuses on policing and criminal justice policy.

Ian T. Adams

Ian T. Adams is a PhD candidate in the Department of Political Science at the University of Utah. His research interests include body-worn cameras, policing policy, and public workplace surveillance.