Objective: R0, the number of cases resulting from one infectious person in a susceptible population, is a common metric in infectious disease research. Newly-developed methods for real-time approximations (Rt) offer promise for monitoring COVID-19 trends using real-world data. The objective was to compare available methods of calculating Rt and understand drivers in variability, using available regional public surveillance data.

Methods: Longitudinal public health data describing infections, recoveries, and deaths from New Jersey (NJ), British Columbia (BC), and Ontario were visualized. Two methods for calculating Rt, by Cori et al. (growth in cases) and Contreras et al. (cases, recoveries, deaths), were implemented and compared. The sensitivity of estimated Rt parameter to changes in infection rates, duration of infection, and mortality rates was investigated, and the relationship with public health measures explored visually.

Results: Rates of COVID-19 infections per 100,000 residents from May to December ranged from 2,856 (NJ) to 622 (BC), and total cases ranged from 253,696 (NJ) to 31,782 (BC). Using Cori et al. method, Rt estimates ranged from 0.5 to 1.5 across jurisdictions; even with large rises in incidence, estimates remained relatively stable. In contrast, Rt estimates based on Contreras et al. markedly fluctuated. Estimates for BC and Ontario were generally 0.5 to 3, but exceeded 10 for NJ for much of the observation period due to limited reporting of recoveries.

Conclusions: Rt estimates for COVID-19, calculated using two recent methods, vary; and each method considers different parameters in its derivation. The Cori et al. method may be better suited to tracking COVID-19 burden in jurisdictions with incomplete recovery data reported; the choice of measure should depend on accuracy and timing of the reporting of required metrics. Accurate estimates and visualizations of Rt will be informative for understanding changing COVID-19 burden and comparing the impact of interventions across jurisdictions.