The province of Ontario, Canada is divided into 36 public health units (PHUs) that administer public health services, and of those, Peel (PEL), Toronto (TOR), York (YRK), Halton (HAL), and Durham (DUR) comprise the Greater Toronto Area (GTA), the most densely populated contiguous region in Canada with 6.4 million total inhabitants (548,430 in Halton, 645 862 in Durham, 1,381,744 in Peel, 1,109,909 in York, 2,731,571 in Toronto).11
The effective reproduction number (Rt) is defined as the mean number of secondary cases generated by a typical primary case at a given time t in a population, making it well suited as an indicator of transmission before and after public health interventions.12 After November 7, 2020, Ontario imposed a colour-coded tiered approach to the escalation and de-escalation of regional public health restrictions (Table 1), based on weekly incidence, percent positivity, effective reproduction number (Rt), and outbreak trends. This allowed direct comparison of their relative effectiveness in reducing Rt.
Table 1
Lockdown restrictions according to the colour-coded five-tier COVID-19 framework implemented in Ontario on November 7, 2020. The provincial restriction tiers used prior to November 7 (stage 1, stage 2, and stage 3) were categorized as grey, red, and green, respectively due to their similarities.
Category | Key restrictions |
Green (prevent)/Stage 3 | Maximum 10 people indoors, 25 people outdoors for social gatherings Maximum 50 people indoors, 100 people outdoors for organized events |
Yellow (protect) | Liquor served only between 9 am and 11 pm Limit of 6 persons seated together |
Orange (restrict) | 50-person indoor seated capacity limit Limit of 4 persons seated together Liquor served only between 9 am and 9 pm |
Red (control)/Stage 2 | Maximum 5 people indoors, 25 people outdoors for social gatherings No more than 10 people inside gyms or fitness classes Non-essential retailers operate at 50% capacity Personal care services may operate |
Grey (lockdown)/Stage 1 | Non-essential retailers operate at 25% capacity Indoor and outdoor dining services prohibited Personal care services closed |
Enhanced lockdown | Stay-at-home order Non-essential retailers closed Closure of schools |
Official COVID-19 data (daily PCR-confirmed cases) from March 1, 2020 to March 19, 2021 were obtained at the level of the five PHUs from the official websites of each PHU. By March 13, 589 270 Ontarians (4.0%) had at least one dose of vaccine and 285 667 Ontarians (1.9%) had been fully vaccinated13, representing approximately 1.85% of the population. Increasingly transmissible variants of concern (VOC) in the GTA ranged from 31.4% in Halton to 49.7% in Durham by the week of March 3–9, 2020.14 Weekly estimates of the effective reproduction number (Rt) at the PHU level were calculated using the EpiEstim R package calculator, found at https://github.com/alechay/covid19-rt. Rt was calculated assuming a Poisson distribution, and using a Bayesian framework to estimate credible serial intervals for infections15 with the parametric si option in EpiEstim, where the mean and SD of the serial interval were based on previous studies.16–18
Google Daily Mobility Reports19 are comprised of anonymized and aggregated regional data and use a GPS-linked index of visits and length of stay compared to the pre-pandemic baseline January 3 to March 1, 2020. These reports were collected for each PHU for workplaces, residential, parks, grocery and pharmacy, retail and recreation, and transit stations. A global mobility index (GMI) similar to that used in a previous Australian study10 was calculated to represent global mobility change, as the mean of each type of mobility i in a day t:
GMI(t)=∑6i = 1Mobilityi/6.
Segmented regressions were used to identify breakpoints in the behavior of Rt for each of the PHUs over time (and shifts in COVID-19 transmission trends) using the ‘segmented’ function in R, which employs an algorithm that iteratively fits standard linear regressions to the data and finds points where the properties of the regression (slope, intercept) are significantly changed.20 Only regression segments with at least five data points per segment were retained. Intercepts and slopes were calculated for the best model, using separate intercepts at each different segment, allowing for separate identification of increases or decreases in Rt, and sudden jumps or plunges in daily values.
The median incubation period for COVID-19 is 5.8 days, and 97.5% of patients develop symptoms within 11.7 days of infection.21 Therefore, we selected three scenarios to account for reporting delays from illness onset, testing, and incubation—immediately following policy change, 7 days following policy change, and 14 days following policy change—to relate policy change and mobility change to Rt.
To evaluate the impact of mobility on Rt, generalized linear models were estimated using the ‘glm’ function in R for each mobility variable separately. Models with a 0-day, 7-day and 14-day lag of each mobility variable were estimated. We then extracted the mobility regression coefficients for each model.
We then calculated Pearson correlation coefficients between global mobility and Rt using the ‘cor.test’ function in R. This allowed us to calculate the association between mobility and COVID-19 spread. All calculations were made using R (version 4.02), with code available on GitHub.