Skip to main content
NCBI home page
Elsevier - PMC COVID-19 Collection logoLink to Elsevier - PMC COVID-19 Collection
. 2023 Feb 13;80:102988. doi: 10.1016/j.healthplace.2023.102988

Modelling the spatiotemporal spread of COVID-19 outbreaks and prioritization of the risk areas in Toronto, Canada

Nushrat Nazia a,∗, Jane Law a,b, Zahid Ahmad Butt a
PMCID: PMC9922578  PMID: 36791508

Abstract

Modelling the spatiotemporal spread of a highly transmissible disease is challenging. We developed a novel spatiotemporal spread model, and the neighbourhood-level data of COVID-19 in Toronto was fitted into the model to visualize the spread of the disease in the study area within two weeks of the onset of first outbreaks from index neighbourhood to its first-order neighbourhoods (called dispersed neighbourhoods). We also model the data to classify hotspots based on the overall incidence rate and persistence of the cases during the study period. The spatiotemporal spread model shows that the disease spread to 1–4 neighbourhoods bordering the index neighbourhood within two weeks. Some dispersed neighbourhoods became index neighbourhoods and further spread the disease to their nearby neighbourhoods. Most of the sources of infection in the dispersed neighbourhood were households and communities (49%), and after excluding the healthcare institutions (40%), it becomes 82%, suggesting the expansion of transmission was from close contacts. The classification of hotspots informs high-priority areas concentrated in the northwestern and northeastern parts of Toronto. The spatiotemporal spread model along with the hotspot classification approach, could be useful for a deeper understanding of spatiotemporal dynamics of infectious diseases and planning for an effective mitigation strategy where local-level spatially enabled data are available.

1. Introduction

The novel Coronavirus-2019 (COVID-19), first reported in Wuhan, Hubei province, China, in 2019, is caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) that causes respiratory illness in humans (Marchand-Senécal et al., 2020). As of July 15, 2022, World Health Organization (WHO) reported over 556.9 million confirmed COVID-19 cases and 6.4 million deaths globally (WHO, 2021). The disease is highly contagious, often without symptoms that seriously challenge the health care system (Smereka and Szarpak, 2020). COVID-19 has also greatly impacted Canada, and the federal and provincial governments have adopted several control strategies to control the spread (Canada, 2022). Due to its fast infection transmission rates, large urban cities such as Toronto are often more affected than rural areas as they are more densely populated, making it hard to control and maintain social distance in the urban areas (Paul et al., 2020). The recommendations of policies and research programs to combat an epidemic are generally based on the perceived standardized definitions of an outbreak. Infections can spread among humans by contacts within the household, community or by random contacts among the general population (Balcan et al., 2010). Timely detection of the spatial spread of infectious disease outbreaks is important for control efforts and can provide clues to the important risk factors (Chowell and Rothenberg, 2018; Steele et al., 2020).

Infectious diseases rarely exhibit simple dynamics (Harapan et al., 2020). COVID-19 is an infectious disease with spatial dimensions that have helped researchers and scientists understand the disease epidemiology of this phenomenon. One of the most common ways to observe outbreak dynamics is the use of visualization techniques that allow us to easily identify spatial patterns and select areas of interest for further studies. Identifying the spatial patterns of the outbreaks and further analysis in those areas can help researchers understand the disease dynamics, which may help in policy-making for controlling the outbreaks (Carroll et al., 2014). These types of maps are often effective in locating an outbreak's origin and depicting outbreak progression over time.

During an epidemic, observing the movement of outbreaks across neighbourhoods in space and time may allow identifying possible sources and directions of the outbreaks. Earlier studies that mapped the spatiotemporal spread of infectious disease outbreaks mostly involved comparing multiple maps to compare the outbreak shifting visually. For instance, a few studies evaluated the spread of COVID-19 over space and time by visually comparing multiple maps to observe the spatial distribution of new or cumulative cases using multiple time points during the early stages of the pandemic (Cuadros et al., 2020; Gao et al., 2021; Yu et al., 2021; Feng et al., 2020). Standard deviation ellipse (SDE), a popular spatial statistical method, is often implemented in outbreak studies to model the spread of disease over time to analyze the spatiotemporal spread of an infectious disease by using the mean centre of the spatial distribution of the geographical elements as the centre and calculates the standard deviation of SDE in the X and Y axis (Wang et al., 2022; Li et al., 2022; Gesler, 1986). Multiple days of SDE maps are often compared to observe the differences in directions between multiple time periods. This method analyzes directional distributions of an outbreak; however, do not provide insights into the magnitude of these transmissions from index to dispersed neighbourhoods.

A study by Gianquintieri et al. in Lombardy, Italy, used choropleth mapping to display the beginning of COVID-19 onset from emergency calls while not including details about the spread mechanism (Gianquintieri et al., 2020). Spread patterns were modelled in other infectious diseases as well, such as the 2009 pandemic influenza outbreaks. A transmission model by Gog et al. showed the spatial spread of the 2009 pandemic influenza but lacked further details about the starting point and the direction of movement of the outbreaks (Gog et al., 2014). Reyes et al. observed the spatiotemporal spread patterns of the 1918 influenza pandemic in British India by using a travel network and a likelihood approach to predict the patterns of pandemic spatial spread. The output map shows the travel network connectivity between districts using nodes across British India, and the nodes with the high excess mortalities are marked in red to explain the spatial spread (Reyes et al., 2018). Another study by daCosta et al. visually examines the spatiotemporal diffusion of influenza A (H1N1), identifies the starting point of an epidemic, and further uses multiple maps of the time of onset and the correlation between effective distance trees across different spatial units (da Costa et al., 2018). However, these existing spatial transmission or spread models often overlook the temporal structures, the direction and the magnitude of the transmission processes within the surrounding neighbourhood networks and primarily compare disease hotspot maps from multiple time points to visualize the dispersion processes. Since the transmission process of the disease may vary geographically and in time, it is important to investigate the transmission dynamics of the diseases using a novel model at a local scale. These types of maps can be effective in locating the origin of an outbreak and depicting outbreak progression over time.

Displaying the outbreaks as they are happening and possible spread to the nearby neighbourhoods can provide deeper insights, allow a more effective response to outbreaks, and provide the ability to understand the intricate details and plan on how to respond to them. In this study, we developed a novel approach for visualizing the spatial spread of the COVID-19 outbreaks at a neighbourhood level that possibly suggests local transmission patterns and the direction of movements of a highly infectious disease in an urban context. The spatial spread of the disease in the neighbourhoods was also classified into priority-based hotspots for facilitating the intervention program for controlling the spread of the disease under resource-constraint situations.

2. Materials and methods

2.1. Study design, study area and population

This study was a retrospective observational study using existing data collected at the neighbourhood level in Toronto (Fig. 1 ). The city of Toronto is on the northern and western shorelines of Lake Ontario in Canada, with a land area size of 630 square kilometres. Toronto is the largest city in Canada and the fourth largest city in North America, with a high population density (Fig. 1) of 4692 persons per square kilometre (City of Toronto, 2021a). Toronto Public Health has reported the highest total number of COVID-19 cases (342,301 cases as of July 15, 2022) and the second-highest incidence rate among all the health regions in the nation (Government of Canada, 2020; City of Toronto, 2020a). Over 4322 deaths have been attributed to COVID-19 in Toronto since March 24, 2020 (City of Toronto, 2020a).

Fig. 1.

Fig. 1

Open in a new tab

The Study Area in Toronto, Canada.

Note: The numbers in the map are the neighbourhood IDs assigned for each of the neighbourhoods.

Our study area consists of 140 geographically distinct social planning neighbourhoods with detailed demographic information for each neighbourhood, as prepared by the city's Social Policy Analysis & Research unit from the data of the 2016 Statistics Canada Census (updated every five years) (City of Toronto, 2020b). The neighbourhoods were defined based on Statistics Canada census tracts for the purposes of statistical reporting (Statistics Canada, 2021). The average population size in a neighbourhood is 19,511 persons (minimum = 6,577, maximum = 65,913) as per the 2016 census.

2.2. Data

The total number of COVID-19 cases per neighbourhood from January 2020 to June 2021 was retrieved from the city of Toronto's open data portal site (City of Toronto, 2020a). The dataset was collected by Toronto Public Health that contains demographic, geographic, and severity information for all confirmed and probable cases. We used the date of episodes as the time of case occurrences of the disease. In the surveillance database, the date of the episode was chosen as the first day that COVID-19 symptoms, the date of specimen collection, or the date of reporting, whichever is the earliest (City of Toronto, 2020a).

The dataset includes the source of infection for each case (City of Toronto, 2022a), which was determined by the Toronto public health authority by assessing the most likely source of infection out of various risk factors such as travel or contact. The common sources of infection included household or close contact such as family members or roommates with confirmed or probable COVID-19 cases, outbreaks in congregate settings such as shelters, correctional facilities, group homes or hostels or healthcare institutions such as long-term care homes, retirement homes, hospitals and chronic care hospitals. Other sources of infection include travel (travelling outside of Ontario in the 14-day prior to their symptom onset or test date), outbreaks in other settings, including workplaces, schools or daycares, and the community (individuals who did not travel outside of Ontario and did not identify being close contact with a COVID-19 case and were not part of a known confirmed COVID-19 outbreak)(City of Toronto, 2022a). Approximately 26% of the total COVID-19 cases had no information on the source of infection.

The population data per neighbourhood were collected from the city of Toronto neighbourhood profiles based on the 2016 census. According to the 2016 census used in our study, the total population of the study area is 2,731,571. The full dataset can be found at the city of Toronto's open data portal site (City of Toronto, 2020b). The neighbourhood-level administrative boundary shapefile for Toronto was extracted from the same site (City of Toronto, 2020b). We aggregated daily COVID-19 cases by the epidemiological week, resulting in a total of 76 epidemiological weeks for each of the 140 neighbourhoods in Toronto. Since this study observes the overall burden of the disease, we included both sporadic and outbreak cases in this analysis.

2.3. Research methods

First, we created the spatiotemporal spread model that identified the index and the dispersed neighbourhoods of COVID-19 outbreaks. We have also analyzed the sources of infections in the dispersed neighbourhoods. Second, the priority-based hotspots were defined in the hotspot classification model based on the two epidemiological parameters: incidence rate and disease persistence in each neighbourhood. Sensitivity analyses were also performed for the spatiotemporal spread model and the hotspot classification model.

2.3.1. The spatiotemporal spread model

We developed a spatiotemporal spread model to visualize the spatial dynamics of COVID-19 across neighbourhoods. At first, the model checked the number of cases per week for each neighbourhood. If the observed number of cases in a week in a neighbourhood was more than the weekly average number of cases of all time points and all neighbourhoods, it was considered an outbreak event in the neighbourhood, a commonly used approach in infectious disease surveillances (Wagner et al., 2001; Stroup et al., 1993; Brady et al., 2015; Badurdeen et al., 2013; Cullen et al., 1984; Nekorchuk et al., 2021; Reintjes and Zanuzdana, 2009). The model then checked the time of the first outbreak in each neighbourhood. Only the first outbreak event in a neighbourhood and the first-order neighbourhood networks sharing a common border of non-zero length (queen contiguity) were considered in evaluating the transmission dynamics in space and time. The index neighbourhoods were considered as the starting point of the dispersion event, and the dispersed neighbourhoods are the neighbourhoods where the first outbreaks happened within two weeks of the outbreak in the index neighbourhood. We used the two-week temporal period, as the World Health Organization and national and international public health sectors have widely accepted the 14-day incubation period for COVID-19 infection spread (Quilty et al., 2021). Also, the two-week time was chosen based on the assumption that further spread would likely be decreased two weeks after the onset of the event. If we get a dispersed neighbourhood, we connect the neighbourhood with the index neighbourhood to show the potential spatial spread of the disease. If there was an outbreak event in a neighbourhood but not dispersed to any neighbourhood, then the neighbourhood was not considered an index neighbourhood. The index and dispersed neighbourhoods were, therefore, determined by

d=wij(oj−oi) (1)

where o i is the start week of the 1st outbreak event in neighbourhood i,

oj is the start week of the 1st outbreak event in neighbourhood j,

wij is a measure of adjacency between neighbourhood i and j and is defined as (1 if i and j are adjacent; 0 otherwise).

When d is ≤ 2 (i.e., the difference between the start week of an outbreak between the two adjacent neighbourhoods is within two weeks), then i is the index neighbourhood, and j is the dispersed neighbourhood of the outbreak.

Since disease outbreaks definitions are implicitly variable (Brady et al., 2015), we performed a sensitivity test using observed cases in a week that is greater than the weekly median number of cases but keeping the temporal period to two weeks to observe how the outcomes of our spatiotemporal spread model fluctuate when using the other definition of an outbreak. We also performed another sensitivity test by changing the temporal period from two weeks to one week but keeping the outbreak definition of the main analysis (observed cases in a week is greater than the weekly mean number of cases) to observe the differences in the results of the spatiotemporal model when using a shorter incubation period.

We have further divided the total study period into three temporal periods based on the first 3 COVID-19 waves: Wave 1 (January 23 – July 17, 2020), Wave 2 (July 18, 2020–March 4, 2021) and Wave 3 (March 5 – July 31, 2022) and ran the spatiotemporal spread model for each of this period to understand the variations of the spatial spread of outbreaks across the three waves.

2.3.2. Classification of the hotspots

Priority-based hotspots have previously been used in various infectious disease mapping such (Debes et al., 2021; Ngwa et al., 2021). In this study, we used two epidemiological parameters to define the priority-based hotspots: i) overall incidence rate and ii) disease persistence, i.e., number of weeks with cases. The incidence rate for each neighbourhood was calculated using the total number of cases during the entire study period divided by the 2016 population (the latest available dataset) multiplied by 1000. The disease persistence was the total number of weeks with at least a case in the week (not necessarily the consecutive week). Based on these two epidemiological parameters used elsewhere (Moore et al., 2018; Dom et al., 2010; Mwaba et al., 2020), the neighbourhoods were classified into three categories of priority levels, as shown in Table 1 . The cut-offs for the thresholds were determined based on the distribution of the data. The high priority hotspots consist of a higher incidence rate and longer persistence; the medium priority levels hotspots experience a lower incidence rate, but longer persistence or higher incidence rate but shorter persistence, and the low priority hotspots consist of a lower incidence rate and shorter persistence.

Table 1.

The cut-offs of the epidemiological parameters applied in the priority-based mapping.

Priority type Interpretation Incidence Rate (70th percentile value) Disease Persistence (70th percentile value)
P1 High Priority >76.29 cases/1000 persons >65 weeks
P2 Medium Priority ≤76.29 cases/1000 persons >65 weeks
P3 Medium Priority >76.29 cases/1000 persons ≤65 weeks
P4 Low Priority ≤76.29 cases/1000 persons ≤65 weeks
Open in a new tab

Since prioritizing the hotspots for intervention depends on the availability of the resources, one may choose the cut-off of the epidemiological parameters based on the need and logistical support. Keeping that in mind, we performed sensitivity analyses using different cut-offs of the epidemiological parameters to demonstrate how one can go with the decision-making for prioritizing the hotspots. These cut-offs were: i) 50th percentile values for both incidence rate and persistence (50 cases/1000 persons, 50 weeks), ii) 70th percentile value (76.29 cases/1000 persons) for incidence rate and 50th percentile value (61 weeks) for the persistence, and iii) 50th percentile value for incidence rate (50.37 cases/1000 persons) and 70th percentile (65 weeks) value for the persistence.

2.3.3. Software

The analyses and mapping of this study were conducted using R Studio Version 1.14.1103 and ArcGIS Desktop Version 10.8.1, respectively. We used the following R packages: lubridate, sqldf, data.table, sqldf, rgdal, spdep for writing the codes for the spatial spread model, i.e., identifying index and dispersed neighbourhoods, and the resulting maps were produced using ArcGIS Desktop. The paths of the outbreaks were created in ArcGIS Desktop by connecting the centers of the index and dispersed neighbourhoods. For the priority-based hotspot maps, we used the following R packages: lubridate, sqldf, dplyr, magrittr and grid. The model outcomes were imported into ArcGIS Desktop to create the maps. We used ggplot2 and ggrepel packages in R to create the figures.

3. Results

3.1. Descriptive statistics

During the surveillance period, a total of 169,985 COVID-19 cases were reported. We omitted 2736 (1.6%) cases due to missing neighbourhood names, resulting in 167,249 (146,177 sporadic and 21,072 outbreak-associated) cases for analysis. An average of 19 cases were observed per week in a neighbourhood. The weekly number of cases shows that COVID-19 was reported throughout the study period (Fig. 2 ). There was a sharp increase of cases in week 16 of 2020, which declined in the following weeks. In week 45 of November 2020, the number of cases increased markedly until the first week of 2021. Another sharp increase was observed in week 9 (March) and continued until April 14–21, 2021. The highest number of cases (>8000 cases) were observed in the week of March 14–21, 2021 (Fig. 2).

Fig. 2.

Fig. 2

Open in a new tab

The weekly number of COVID-19 cases during the surveillance period, January 2020–June 2021, Toronto.

3.2. The spatiotemporal spread model

The resulting map from our spatiotemporal model is shown in Fig. 3 , where the black dots represent the index neighbourhoods (62 (44%) neighbourhoods), which dispersed the outbreaks to their bordering neighbourhood(s). The red dots represent dispersed neighbourhoods (61 (43%) neighbourhoods), implying that the outbreak was possibly dispersed from the index neighbourhoods, as indicated by the path. In some cases, the dispersed neighbourhoods became the index neighbourhoods (27 (19%) neighbourhoods), potentially spreading the outbreak to nearby neighbourhoods. The earlier onset of outbreaks was observed in the western and southern parts of Toronto (shown in lighter blue gradient colours), whereas the later onsets were seen in the central parts of Toronto (shown in darker blue gradient colours). In the northwestern, northeastern, and southern parts of Toronto, several dispersed neighbourhoods became index neighbourhoods and further spread the disease to its nearby neighbourhoods. A maximum of 4 dispersed neighbourhoods was observed bordering the index neighbourhood. During the study period, the last dispersion of the COVID-19 outbreak occurred between 18 and 24 April 2021. No dispersion events were observed in 90 (64%) neighbourhoods within two weeks of the onset of the outbreak in these neighbourhoods. Four neighbourhoods in the center of Toronto never experienced an outbreak event, as per our definition of the outbreak.

Fig. 3.

Fig. 3

Open in a new tab

Spatiotemporal spread of the COVID-19 outbreaks in Toronto, January 2020–June 2021.

Note: The lighter the colour the earlier the onset of the first outbreak in the neighbourhood.

Some dispersed neighbourhoods, particularly in northwestern, southern and central Toronto, had multiple index neighbourhoods, as shown in Fig. 3. Since these neighbourhoods were shown to have dispersed infections from multiple index neighbourhoods, we also created a distance-based analysis where only one neighbourhood was determined as the index neighbourhood for a dispersed neighbourhood (Fig. S3). If a dispersed neighbourhood had multiple index neighbourhoods, only the neighbourhood that was the closest (based on the linear distance from the centers of the neighbourhoods) was considered the index neighbourhood.

3.2.1. Sensitivity analyses of the spatiotemporal spread model

The results of the sensitivity analyses using the outbreak definition as the observed number of cases greater than the median number of cases show a similar spatial spread of the disease with a few exceptions (Fig. S4). However, unlike the main analysis that used the mean number of cases in the outbreak definition, all neighbourhoods experienced an outbreak event using the median as the definition of an outbreak in the sensitivity analysis since the median is only 9 cases, whereby the mean is 19 cases, much higher than the median number of cases. While reducing the temporal period from two weeks (61 dispersed neighbourhoods) to one week, we observed a lower number (31 neighbourhoods) of dispersed neighbourhoods (Fig. S5).

3.2.2. Sources of infection

To identify the sources of infection in the dispersed neighbourhoods, we included cases only for the 1st outbreaks in the dispersed neighbourhoods. A total of 26% of the cases did not have the source of infection in those neighbourhoods in the week of the first outbreaks, thus excluded from the analysis. About 40% of the source of infections in the dispersed neighbourhoods was attributed to the healthcare institution (Fig. 4 ). The majority of the sources of infection were household, close contacts and communities (49%) and after excluding healthcare institution-related outbreaks) it became 82%. Travel (1.4%), congregate settings (i.e., shelters, correctional facilities) (3.9%), and other settings (i.e., workplace, daycare) (5.4%) were not the major sources of infection in the study area.

Fig. 4.

Fig. 4

Open in a new tab

Sources of infection of COVID-19 among the dispersed neighbourhoods in Toronto, January 2020–June 2021.

3.2.3. Spatiotemporal spreads during the first three COVID-19 waves

Fig. S2 displays the spatiotemporal spread during the first three COVID-19 waves of the pandemic. In the first wave period, 58 (41%) index neighbourhoods and 55 (39%) dispersed neighbourhoods were observed. 23 (42%) of these dispersed neighbourhoods also became an index neighbourhood and dispersed to surrounding neighbourhoods. Similarly, in the second wave period, 54 (39%) index neighbourhoods and 52 (37%) dispersed neighbourhoods were observed. 22 (42%) of these dispersed neighbourhoods also became an index neighbourhood and dispersed to the surrounding neighbourhoods. Out of these three waves, wave 3 experienced the maximum number of dispersion events throughout Toronto. During this third wave, 81 (58%) index neighbourhoods and 67 (48%) dispersed neighbourhoods were observed. 45 (67%) of these dispersed neighbourhoods also became an index neighbourhood and dispersed to surrounding neighbourhoods.

3.3. Classification of the hotspots

The incidence rate and disease persistence of COVID-19 varied widely in the study area during our study period (Fig. 5 ). The northwestern part of Toronto had a higher incidence rate with a long persistence. The southern part of Toronto had a relatively lower incidence rate, particularly in the neighbourhood no. 77 (Waterfront Communities – the island) but experienced a longer persistence of COVID-19. The spatial patterns of the incidence rate show that majority of the neighbourhoods had an overall incidence rate between 20 and 90 cases/1000 persons. Only a few neighbourhoods had an incidence rate of over 100 cases/1000 persons (Fig. S1).

Fig. 5.

Fig. 5

Open in a new tab

Spatial distribution of incidence rate and persistence of COVID-19 in Toronto, January 2020–June 2021.

The hotspot classification chart displays how many neighbourhoods fall in each priority category based on epidemiological parameters used to classify the hotspots (Fig. 6 b). A total of 21 (15%) neighbourhoods were classified as high priority (P1) hotspots. These neighbourhoods had a higher incidence rate and a higher persistence and were observed in the northwestern and northeastern corners of Toronto (Fig. 6a). According to the 2016 census, a total of 530,119 population lived in the P1 hotspots. In total, 16 (11.4%) were classified as medium-priority (P2/P3) hotspots. These neighbourhoods had either higher incidence rates with a shorter persistence or lower incidence rates with a longer persistence. P2/P3 hotspots were observed in different parts of Toronto. A large number of neighbourhoods [82 (58%)] were classified as low priority (P4) hotspots because the neighbourhoods experienced a lower incidence rate with a shorter persistence. Most of these hotspots were in the central and southwestern parts of Toronto.

Fig. 6.

Fig. 6

Open in a new tab

Hotspots and classification chart of COVID-19 in Toronto, January 2020–June 2021.

Note: The numbers in the chart and inside the map are the neighbourhood IDs.

3.3.1. Sensitivity analyses of the hotspots

The results of the sensitivity analyses using different cut-offs of the epidemiological parameters show mostly similar results in identifying and classifying the hotspots, particularly in northwestern and northeastern Toronto (Fig. S6). However, in contrast to the main analysis, a relatively higher number of P1 hotspots were observed mostly in the western and eastern parts, and a lower number of P4 hotspots were observed in the central part of Toronto.

4. Discussion

The model for the spatiotemporal spread of the disease developed in this study helped us visualize the spatial and temporal dynamics of COVID-19 in a densely populated urban city and provided a deeper insight into the spread of the COVID-19 outbreaks in the neighbourhoods of the city. This geographic model can be applied to any infectious disease to observe community-level spread patterns by updating the local definition of an outbreak and the incubation period of the disease for a geographic setting. This model can retrospectively observe the movement of outbreaks across neighbourhoods in space and time to identify possible sources and directions of the outbreaks. Understanding the spatial dynamics of the disease may help assist in planning timely interventions. Note that control, mitigation and eventually elimination of a highly transmissible disease such as COVID-19 require careful and comprehensive planning during the initial stages of the epidemic so that timely and efficient intervention plans can be executed. Our study suggests that the spread of the disease was not spatially random. Therefore, it is important to identify the areas where the targeted inventions are essential. This kind of model helps us identify the places where control efforts can also be initiated under resource constraint situations.

We observed that spatial dynamics of the COVID-19 outbreaks varied across the neighbourhoods in Toronto. An index neighbourhood possibly dispersed the outbreaks to a maximum of four neighbourhoods within two weeks. Some dispersed neighbourhoods became index neighbourhoods and further dispersed the outbreaks to their nearby neighbourhoods. For example, in the south of Toronto in the Waterfront communities-the island (#77), the outbreak dispersed to five surrounding neighbourhoods within two weeks during the early stages of the pandemic and further continued to spread. Waterfront communities-island has a high population density (8943 persons per sq km), with a majority (69%) of the population at working age (24–52 years). This neighbourhood became an index neighbourhood and experienced dispersion to surrounding neighbourhoods during all three COVID-19 waves. The same neighbourhood also demonstrated a high disease persistence. In the northwestern part of Toronto, in the West Humber-Clairville neighbourhood bordering the Peel region, the outbreak spread to multiple neighbourhoods during the first wave of the pandemic. This neighbourhood with a relatively larger spatial size has a lower population density (1117 persons per sq km) but a high population count of 33,112 with 82% of the population being a visible minority. Some similar spatiotemporal spread patterns were observed across the three COVID-19 waves, primarily in the southern, eastern and northern areas in Toronto where multiple dispersion events are observed. While the first two waves showed similar number of index and dispersed neighbourhoods, the third wave starting from March 5, 2021, was driven by the more transmissible Alpha (B. 1.1.7) and showed a much higher dispersion events throughout Toronto.

The results suggest that ecological phenomena for creating a risk for the population are not random in space. Thus, understanding local-level variation is important for an effective control program. In our study, the first COVID-19 outbreak dispersion occurred between March 22 and March 28, 2020, in the Morningside neighbourhood (neighbourhood#135) and then spread to the Rouge neighbourhood (neighbourhood#131), where it became an index neighbourhood of another dispersion event. The Morningside neighbourhood, located in the eastern part of Toronto, has a relatively low population density (3041 persons per sq km) and 73% of the population has a lower level of education. The northwestern and southern parts of Toronto experienced multiple outbreak dispersions during the early stages of the pandemic. Glenfield-Jane Heights neighbourhood (neighbourhood#25) in northwestern Toronto experienced the spread of the outbreak to four neighbourhoods between April 5 and April 11, 2020. This neighbourhood has a high population density with over 5864 persons per sq km, with 76.6% of the population being a visible minority and 89% having a lower level of education. These findings may imply that early intervention at the index and its nearby neighbourhoods may have controlled the magnitude of the spread of the disease in the study area.

In our study, we attempted to provide an effective model for visualizing the spatiotemporal dynamics of the outbreaks, which is different from earlier studies that visualized infectious disease outbreaks. We believe that our model improves the traditional way of visualizing disease outbreaks using choropleth mapping or other methods such as SDE by including temporal structure, identifying index and dispersed neighbourhoods, dispersion directions and magnitude of the outbreaks in a single map. This kind of map may also be useful for evaluating the impact of an intervention, showing where intervention efforts are successful and where they are not. This could help to identify the barriers to a successful intervention program.

By analyzing the sources of infections of the COVID-19 outbreaks in the dispersed neighbourhoods, we identified a few factors that played a key role at a spatial scale in the spread of COVID-19 in Toronto. Based on the assessment of the sources of infection information from the COVID-19 dataset, we observed that apart from outbreaks from healthcare institutions, most of the sources of infection in the dispersed neighbourhoods were from close contact either through individuals or from communities. Travel or workplace outbreaks had very little to contribute to the spread of the disease in the dispersed neighbourhoods in our study area. This suggests that close attention to the index and its nearby neighbourhoods could have controlled the spread of the disease in our study area.

When an outbreak hits a large area, it may not be practical to initiate an intervention in all the areas simultaneously. For instance, several efforts have been adopted by the Toronto local authorities to control the pandemic since the cases were reported at the Sunnybrook Health Sciences Centre in Toronto on January 23, 2020 (Silverstein et al., 2020). According to BBC News, on May 24, 2021, Toronto is considered to have the longest continuous COVID-19 lockdown among any major city around the world (Toronto lockdown, 2021). On March 23, 2020, the Toronto Mayor declared a citywide emergency, restricting businesses from operating. Because of the spread of the disease, more regionally targeted restrictions and policies were applied. These measures included lockdowns, mask mandates, closure of schools, indoor gathering limits, limited non-essential business, limiting shopping malls to curbside pickup, and preventing any subsequent large gatherings (Soucy et al., 2021; Long et al., 2021). The city of Toronto prioritizes the postal code areas with high neighbourhood-level case rates for mobile and pop-up clinic vaccinations (City of Toronto, 2022b; City of Toronto, 2021b). Given all those efforts, there were no signs of declining the disease in Toronto. Past studies have found that spatially targeted interventions can be highly effective for controlling infectious disease outbreaks (Franch-Pardo et al., 2020; Finger et al., 2018; Khundi et al., 2021; Cudahy et al., 2019). The results of our hotspot analysis provided important data by which one could prioritize areas for local interventions. Along with the incidence rate, our approach includes the persistence of the cases during the study period in defining the hotspots. In our study, the high-priority areas were observed near the western part of Toronto, bordering the Peel region, which had reported a high COVID-19 incidence rate in Canada. It is possible that Toronto was affected by the high number of cases in the Peel region. The central regions of Toronto near downtown observed a few dispersions but were mostly at low risk, whereas a higher risk was observed near the outer boundaries.

In an urban area, various risk factors such as socioeconomic inequalities, demographic, environmental factors or pre-existing health conditions in a population are often used to explain the variations in risk and outbreak dynamics (Vaz, 2021; News · and C. B. C, 2020; Ingen et al., 2021). Population density is an important factor in the spread of COVID-19 incidences (Bhadra et al., 2021; Kadi and Khelfaoui, 2020; Sy et al., 2021; Pequeno et al., 2020). We observed that many of the spread of outbreaks occurred in the highly densely populated areas (Central and Southern Toronto) with a population density ranging from 6500 to 44,321 persons per sq km. However, in the northeastern and northwestern corners of Toronto, where the population density is lower (1000–3000 persons per sq km) than in the other part of Toronto, we observed the high-risk priority neighbourhoods and multiple dispersions, suggesting that population density alone may not explain the increased risk of COVID-19. Targeting the high-priority hotspots first, as defined in this study, and subsequently to medium-priority hotspots and then the low-priority hotspots would be the most efficient strategy in controlling the disease.

This study is not free from limitations. We used the commonly used definition for the outbreak. The recommendations of international policy guidelines and research agendas are based on the perceived standardized definition of an outbreak characterized by a prolonged, high-caseload, extra-seasonal surge. Any changes in the definition may change the dynamic patterns to some degree. However, the results of our sensitivity analysis showed only minor variations in the spatiotemporal dynamics of the outbreak. Second, asymptomatic cases are not included in the analysis, as these data were unavailable for our study area. Additionally, access to testing and testing resource allocation across neighbourhoods may have varied over time (CBC News, 2020; Ontario, 2022). These missing numbers of cases could potentially result in an underestimation of disease transmission. Finally, we did not have the individual or neighbourhood-level mobility data to explain some disease spread processes. However, we have seen travel contributed only a little in explaining the spread of the outbreak.

In a wider outbreak control effort, research agendas and subsequent policy guidelines have heavily focused on methods to identify and predict outbreaks and how to respond appropriately by outbreak response protocols to better plan for future outbreak occurrences using effective control measures. With appropriate and timely control, the spread of the disease outbreak can be minimized. Our study offers a new approach to a deeper understanding of the spatial dynamics of infectious disease outbreaks, which would facilitate the intervention program for controlling the spread of the outbreaks. The approach could be useful for understanding spatial dynamics of other infectious diseases where local scale spatially enabled data are available.

Authors' contributions

NN made substantial contributions to the conception and design of the project. NN has performed the data acquisition, data processing, formal analysis, visualization, and writing the first complete version of the manuscript. JL has supervised the research, reviewed and edited the manuscript. ZAB has reviewed and edited the manuscript and provided important intellectual contributions. All authors were involved in revising the manuscript for important intellectual contact. All authors read and approved the final manuscript.

Ethics approval

Not applicable.

Consent to participate

Not applicable.

Availability of data and materials

All data generated or analyzed during this study are included in this published article (and its supplementary information files).

Author's statement

The study is in accordance with relevant guidelines and regulations.

Declaration of competing interest

The authors declare that they have no competing interests.

Acknowledgements

The authors would wish to acknowledge the city of Toronto, Toronto Public Health, the Ontario Government and Statistics Canada for the free open access to the data.

Footnotes

Appendix A

Supplementary data to this article can be found online at https://doi.org/10.1016/j.healthplace.2023.102988.

Appendix A. Supplementary data

The following is the Supplementary data to this article:

Multimedia component 1
mmc1.docx (3MB, docx)

References

  1. Badurdeen S., et al. Sharing experiences: towards an evidence based model of dengue surveillance and outbreak response in Latin America and Asia. BMC Publ. Health. 2013;13:607. doi: 10.1186/1471-2458-13-607. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Balcan D., et al. Modeling the spatial spread of infectious diseases: the GLobal Epidemic and Mobility computational model. J. Comput. Sci. 2010;1:132–145. doi: 10.1016/j.jocs.2010.07.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Bhadra A., Mukherjee A., Sarkar K. Impact of population density on Covid-19 infected and mortality rate in India. Model. Earth Syst. Environ. 2021;7:623–629. doi: 10.1007/s40808-020-00984-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Brady O.J., Smith D.L., Scott T.W., Hay S.I. Dengue disease outbreak definitions are implicitly variable. Epidemics. 2015;11:92–102. doi: 10.1016/j.epidem.2015.03.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Canada . 2022. P. H. A. of. Summary of evidence supporting COVID-19 public health measures.https://www.canada.ca/en/public-health/services/diseases/2019-novel-coronavirus-infection/guidance-documents/summary-evidence-supporting-covid-19-public-health-measures.html [Google Scholar]
  6. Carroll L.N., et al. Visualization and analytics tools for infectious disease epidemiology: a systematic review. J. Biomed. Inf. 2014;51:287–298. doi: 10.1016/j.jbi.2014.04.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. CBC News . CBC; 2020. https://www.cbc.ca/news/health/covid-testing-shortages-1.5503926 (Why It's So Difficult to Get Tested for COVID-19 in Canada). [Google Scholar]
  8. Chowell G., Rothenberg R. Spatial infectious disease epidemiology: on the cusp. BMC Med. 2018 doi: 10.1186/s12916-018-1184-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. City of Toronto . 2020. Coronavirus disease 2019 (COVID-19): epidemiology update.https://www.toronto.ca/home/covid-19/covid-19-latest-city-of-toronto-news/covid-19-status-of-cases-in-toronto/ [Google Scholar]
  10. City of Toronto . 2020. Toronto neighbourhood profiles.https://www.toronto.ca/city-government/data-research-maps/neighbourhoods-communities/neighbourhood-profiles/ [Google Scholar]
  11. City of Toronto. Toronto at a Glance; 2021. [Google Scholar]
  12. City of Toronto . 2021. COVID-19: where to get vaccinated. City of Toronto.https://www.toronto.ca/home/covid-19/covid-19-vaccines/covid-19-how-to-get-vaccinated/ [Google Scholar]
  13. City of Toronto . 2022. COVID-19 cases in Toronto.https://open.toronto.ca/dataset/covid-19-cases-in-toronto/ Open Data Catelogue. [Google Scholar]
  14. City of Toronto . 2022. City of Toronto and its community partners continue to support at-risk communities in fight against latest wave of COVID-19. City of Toronto.https://www.toronto.ca/news/city-of-toronto-and-its-community-partners-continue-to-support-at-risk-communities-in-fight-against-latest-wave-of-covid-19/ [Google Scholar]
  15. Cuadros D.F., et al. Spatiotemporal transmission dynamics of the COVID-19 pandemic and its impact on critical healthcare capacity. Health Place. 2020 doi: 10.1016/j.healthplace.2020.102404. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Cudahy P.G.T., et al. Spatially targeted screening to reduce tuberculosis transmission in high-incidence settings. Lancet Infect. Dis. 2019;19:e89–e95. doi: 10.1016/S1473-3099(18)30443-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Cullen J.R., Chitprarop U., Doberstyn E.B., Sombatwattanangkul K. An epidemiological early warning system for malaria control in northern Thailand. Bull. World Health Organ. 1984;62:107–114. [PMC free article] [PubMed] [Google Scholar]
  18. da Costa A.C.C., Codeço C.T., Krainski E.T., da Costa Gomes M.F., Nobre A.A. Spatiotemporal diffusion of influenza A (H1N1): starting point and risk factors. PLoS One. 2018 doi: 10.1371/journal.pone.0202832. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Debes A.K., et al. Cholera hot-spots and contextual factors in Burundi, planning for elimination. Tropical Med.Infect. Dis. 2021;6:76. doi: 10.3390/tropicalmed6020076. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Dom N.C., Ahmad A.H., Adawiyah R., Ismail R. 2010 International Conference on Science and Social Research. CSSR 2010; 2010. Spatial mapping of temporal risk characteristic of dengue cases in Subang Jaya; pp. 361–366. [DOI] [Google Scholar]
  21. Feng Y., et al. Spatiotemporal spread pattern of the COVID-19 cases in China. PLoS One. 2020;15 doi: 10.1371/journal.pone.0244351. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Finger F., et al. The potential impact of case-area targeted interventions in response to cholera outbreaks: a modeling study. PLoS Med. 2018;15 doi: 10.1371/journal.pmed.1002509. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Franch-Pardo I., Napoletano B.M., Rosete-Verges F., Billa L. Spatial analysis and GIS in the study of COVID-19. A review. Sci. Total Environ. 2020;739 doi: 10.1016/j.scitotenv.2020.140033. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Gao X., Li G., Wang J., Xu T. Spatiotemporal evolution, pattern of diffusion, and influencing factors of the COVID-19 epidemic in Hainan Province, China. J. Med. Virol. 2021 doi: 10.1002/jmv.27502. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Gesler W. The uses of spatial analysis in medical geography: a review. Soc. Sci. Med. 1986;23:963–973. doi: 10.1016/0277-9536(86)90253-4. [DOI] [PubMed] [Google Scholar]
  26. Gianquintieri L., et al. Mapping spatiotemporal diffusion of COVID-19 in Lombardy (Italy) on the base of emergency medical services activities. ISPRS Int. J. Geo-Inf. 2020 doi: 10.3390/ijgi9110639. [DOI] [Google Scholar]
  27. Gog J.R., et al. Spatial transmission of 2009 pandemic influenza in the US. PLoS Comput. Biol. 2014;10 doi: 10.1371/journal.pcbi.1003635. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Government of Canada . 2020. Coronavirus disease 2019 (COVID-19): epidemiology update.https://health-infobase.canada.ca/covid-19/epidemiological-summary-covid-19-cases.html [Google Scholar]
  29. Harapan H., et al. Coronavirus disease 2019 (COVID-19): a literature review. J. Infect. Public Health. 2020;13:667–673. doi: 10.1016/j.jiph.2020.03.019. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Ingen T. van, et al. 2021. Neighbourhood-level risk factors of COVID-19 incidence and mortality.https://www.medrxiv.org/content/10.1101/2021.01.27.21250618v1 2021.01.27.21250618. 10.1101/2021.01.27.21250618. [Google Scholar]
  31. Kadi N., Khelfaoui M. Bulletin of the National Research Centre; 2020. Population Density, a Factor in the Spread of COVID-19 in Algeria: Statistic Study. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Khundi M., et al. Effectiveness of spatially targeted interventions for control of HIV, tuberculosis, leprosy and malaria: a systematic review. BMJ Open. 2021;11 doi: 10.1136/bmjopen-2020-044715. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Li W., et al. The analysis of patterns of two COVID-19 outbreak clusters in China. Int. J. Environ. Res. Publ. Health. 2022;19:4876. doi: 10.3390/ijerph19084876. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Long J.A., Malekzadeh M., Klar B., Martin G. Canada. Health & Place; 2021. Do Regionally Targeted Lockdowns Alter Movement to Non-lockdown Regions? Evidence from Ontario. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Marchand-Senécal X., et al. Clinical Infectious Diseases; 2020. Diagnosis and Management of First Case of COVID-19 in Canada: Lessons Applied from SARS-CoV-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Moore S., et al. 2018. Epidemiological Study of Cholera Hotspots and Epidemiological Basins in East and Southern Africa - In-Depth Report on Cholera Epidemiology in Angola. [Google Scholar]
  37. Mwaba J., et al. Identification of cholera hotspots in Zambia: a spatiotemporal analysis of cholera data from 2008 to 2017. PLoS Neglected Trop. Dis. 2020;14 doi: 10.1371/journal.pntd.0008227. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Nekorchuk D.M., et al. Comparing malaria early detection methods in a declining transmission setting in northwestern Ethiopia. BMC Publ. Health. 2021;21:788. doi: 10.1186/s12889-021-10850-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. News ·, C. B. C . CBC; 2020. https://www.cbc.ca/news/canada/toronto/low-income-immigrants-covid-19-infection-1.5566384 (Lower Income People, New Immigrants at Higher COVID-19 Risk in Toronto, Data Suggests | CBC News). [Google Scholar]
  40. Ngwa M.C., et al. The cholera risk assessment in Kano State, Nigeria: a historical review, mapping of hotspots and evaluation of contextual factors. PLoS Neglected Trop. Dis. 2021;15 doi: 10.1371/journal.pntd.0009046. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Ontario . ontario.ca; 2022. http://www.ontario.ca/page/covid-19-testing-and-treatment (COVID-19 Testing and Treatment). [Google Scholar]
  42. Paul R., Arif A.A., Adeyemi O., Ghosh S., Han D. Progression of COVID-19 from urban to rural areas in the United States: a spatiotemporal analysis of prevalence rates. J. Rural Health. 2020 doi: 10.1111/jrh.12486. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Pequeno P., et al. Air transportation, population density and temperature predict the spread of COVID-19 in Brazil. PeerJ. 2020;8 doi: 10.7717/peerj.9322. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Quilty B.J., et al. Quarantine and testing strategies in contact tracing for SARS-CoV-2: a modelling study. Lancet Public Health. 2021;6:e175–e183. doi: 10.1016/S2468-2667(20)30308-X. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Reintjes R., Zanuzdana A. Outbreak investigations. Mod. Infect.Dis.Epidemiol. 2009:159–176. doi: 10.1007/978-0-387-93835-6_9. [DOI] [Google Scholar]
  46. Reyes O., et al. Spatiotemporal patterns and diffusion of the 1918 influenza pandemic in British India. Am. J. Epidemiol. 2018 doi: 10.1093/aje/kwy209. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Silverstein W.K., Stroud L., Cleghorn G.E., Leis J.A. First imported case of 2019 novel coronavirus in Canada, presenting as mild pneumonia. Lancet. 2020 doi: 10.1016/S0140-6736(20)30370-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Smereka J., Szarpak L. COVID 19 a challenge for emergency medicine and every health care professional. Am. J. Emerg. Med. 2020;38:2232–2233. doi: 10.1016/j.ajem.2020.03.038. [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Soucy J.-P.R., et al. 2021. Increased interregional travel to shopping malls and restaurants in response to differential COVID-19 restrictions in the greater Toronto area.https://www.medrxiv.org/content/10.1101/2021.04.23.21255959v1 2021.04.23.21255959. 10.1101/2021.04.23.21255959. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Statistics Canada . 2021. Census Tract (CT) - Census Dictionary. [Google Scholar]
  51. Steele L., et al. Earlier outbreak detection—a generic model and novel methodology to guide earlier detection supported by data from low- and mid-income countries. Front. Public Health. 2020;8 doi: 10.3389/fpubh.2020.00452. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Stroup D.F., Wharton M., Kafadar K., Dean A.G. Evaluation of a method for detecting aberrations in public health surveillance data. Am. J. Epidemiol. 1993;137:373–380. doi: 10.1093/oxfordjournals.aje.a116684. [DOI] [PubMed] [Google Scholar]
  53. Sy K.T.L., White L.F., Nichols B.E. Population density and basic reproductive number of COVID-19 across United States counties. PLoS One. 2021;16 doi: 10.1371/journal.pone.0249271. [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Toronto Lockdown - One of the World's Longest? BBC News; 2021. [Google Scholar]
  55. Vaz E. COVID-19 in Toronto: a spatial exploratory analysis. Sustainability. 2021;13:498. [Google Scholar]
  56. Wagner M.M., et al. The emerging science of very early detection of disease outbreaks. J. Publ. Health Manag. Pract. 2001;7:51–59. doi: 10.1097/00124784-200107060-00006. [DOI] [PubMed] [Google Scholar]
  57. Wang X., Wang L., Zhang X., Fan F. The spatiotemporal evolution of COVID-19 in China and its impact on urban economic resilience. China Econ. Rev. 2022;74 doi: 10.1016/j.chieco.2022.101806. [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. WHO . 2021. World health organization:COVID-19 dashboard.https://covid19.who.int/ [Google Scholar]
  59. Yu H., Li J., Bardin S., Gu H., Fan C. Spatiotemporal dynamic of COVID-19 diffusion in China: a dynamic spatial autoregressive model analysis. ISPRS Int. J. Geo-Inf. 2021;10:510. [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Multimedia component 1
mmc1.docx (3MB, docx)

Data Availability Statement

All data generated or analyzed during this study are included in this published article (and its supplementary information files).

Articles from Health & Place are provided here courtesy of Elsevier

RESOURCES