Statistical procedures for evaluating trends in coronavirus disease-19 cases in the United States

OBJECTIVES: In late 2019, a novel respiratory disease was identified as it began to spread rapidly within China’s Hubei Province soon thereafter, being designated coronavirus disease 2019 (COVID-19). Unfortunately, trends in cases and rates of infection have been consistently misunderstood, particularly within the media, due to little, if any, statistical analysis of trends. Critical analysis of data is necessary to determine how to best manage local restrictions, particularly if there are resurgences of infection. As such, researchers have been calling for data-driven, statistical analysis of trends of disease to provide more context and validity for significant policy decisions. METHODS: This quantitative study sought to explore different statistical methods that can be used to evaluate trend data to improve decision-making and public information on the spread of COVID-19. Analyses were conducted using Spearman’s rho, Mann-Whitney U tests, Mann-Kendal tests, and Augmented Dickey-Fuller tests with follow up Kwiatkowski–Phillips–Schmidt–Shin tests. RESULTS: The results indicated a mix of both surprising and expected findings. Variations among COVID case reporting for each day of the week were identified but not deemed significant. Spearman correlation data appeared to perform well in identifying monotonic trend while Mann-Kendal tests appeared to provide the most intelligible results. CONCLUSIONS: This study provides examples of statistical tools and procedures to more thoroughly examine trends in COVID-19 case rate data. It is advocated that such metrics be made available to health and policy stakeholders for potential use for public health decisions.