This paper is in the following e-collection/theme issue:

Innovative Methods in Public Health and Surveillance (336) Sensors at Home and Domotics (59) Internet of Things (80) Outbreak and Pandemic Preparedness and Management (1743) Infectious Diseases (non-STD/STI) (1592) Mobile Health (mhealth) (2982)

Published on in Vol 10 (2024)

Preprints (earlier versions) of this paper are available at https://preprints.jmir.org/preprint/46903, first published .
Revealing the Mysteries of Population Mobility Amid the COVID-19 Pandemic in Canada: Comparative Analysis With Internet of Things–Based Thermostat Data and Google Mobility Insights

Revealing the Mysteries of Population Mobility Amid the COVID-19 Pandemic in Canada: Comparative Analysis With Internet of Things–Based Thermostat Data and Google Mobility Insights

Revealing the Mysteries of Population Mobility Amid the COVID-19 Pandemic in Canada: Comparative Analysis With Internet of Things–Based Thermostat Data and Google Mobility Insights

Authors of this article:

Kirti Sundar Sahu1 Author Orcid Image ;   Joel A Dubin2 Author Orcid Image ;   Shannon E Majowicz1 Author Orcid Image ;   Sam Liu3 Author Orcid Image ;   Plinio P Morita1, 4, 5, 6, 7 Author Orcid Image
Article Authors Cited by Tweetations (5) Metrics

    Original Paper

    1School of Public Health Sciences, University of Waterloo, Waterloo, ON, Canada

    2Department of Statistics and Actuarial Science, University of Waterloo, Waterloo, ON, Canada

    3School of Exercise Science, Physical and Health Education, University of Victoria, Victoria, BC, Canada

    4Institute of Health Policy, Management, and Evaluation, University of Toronto, Toronto, ON, Canada

    5Research Institute of Aging, University of Waterloo, Waterloo, ON, Canada

    6Department of Systems Design Engineering, University of Waterloo, Waterloo, ON, Canada

    7eHealth Innovation, University Health Network, Toronto, ON, Canada

    Corresponding Author:

    Plinio P Morita, PEng, PhD

    School of Public Health Sciences

    University of Waterloo

    200 University Avenue West

    Waterloo, ON, N2L 3G1

    Canada

    Phone: 1 519 888 4567 ext 31372

    Fax:1 519 746 6776

    Email: plinio.morita@uwaterloo.ca


    Background: The COVID-19 pandemic necessitated public health policies to limit human mobility and curb infection spread. Human mobility, which is often underestimated, plays a pivotal role in health outcomes, impacting both infectious and chronic diseases. Collecting precise mobility data is vital for understanding human behavior and informing public health strategies. Google’s GPS-based location tracking, which is compiled in Google Mobility Reports, became the gold standard for monitoring outdoor mobility during the pandemic. However, indoor mobility remains underexplored.

    Objective: This study investigates in-home mobility data from ecobee’s smart thermostats in Canada (February 2020 to February 2021) and compares it directly with Google’s residential mobility data. By assessing the suitability of smart thermostat data, we aim to shed light on indoor mobility patterns, contributing valuable insights to public health research and strategies.

    Methods: Motion sensor data were acquired from the ecobee “Donate Your Data” initiative via Google’s BigQuery cloud platform. Concurrently, residential mobility data were sourced from the Google Mobility Report. This study centered on 4 Canadian provinces—Ontario, Quebec, Alberta, and British Columbia—during the period from February 15, 2020, to February 14, 2021. Data processing, analysis, and visualization were conducted on the Microsoft Azure platform using Python (Python Software Foundation) and R programming languages (R Foundation for Statistical Computing). Our investigation involved assessing changes in mobility relative to the baseline in both data sets, with the strength of this relationship assessed using Pearson and Spearman correlation coefficients. We scrutinized daily, weekly, and monthly variations in mobility patterns across the data sets and performed anomaly detection for further insights.

    Results: The results revealed noteworthy week-to-week and month-to-month shifts in population mobility within the chosen provinces, aligning with pandemic-driven policy adjustments. Notably, the ecobee data exhibited a robust correlation with Google’s data set. Examination of Google’s daily patterns detected more pronounced mobility fluctuations during weekdays, a trend not mirrored in the ecobee data. Anomaly detection successfully identified substantial mobility deviations coinciding with policy modifications and cultural events.

    Conclusions: This study’s findings illustrate the substantial influence of the Canadian stay-at-home and work-from-home policies on population mobility. This impact was discernible through both Google’s out-of-house residential mobility data and ecobee’s in-house smart thermostat data. As such, we deduce that smart thermostats represent a valid tool for facilitating intelligent monitoring of population mobility in response to policy-driven shifts.

    JMIR Public Health Surveill 2024;10:e46903

    doi:10.2196/46903

    Keywords

    population-level health indicators; internet of things; public health surveillance; mobility; risk factors; chronic diseases; chronic; risk; surveillance; mobility; movement; sensor; population 


    The dynamics of human mobility are currently undergoing a remarkable transition. Rapid global urbanization, sedentary lifestyles, infectious diseases, air pollution, and climate change are some of the factors driving shifts in human mobility. Whether examined at the individual [1,2] or population level [3], human mobility patterns are associated with multiple public health and social issues. Dwindling mobility patterns are linked to chronic diseases including dementia and age-related physical decline. For infectious diseases, individual mobility is linked to the spread of infections such as COVID-19. Human mobility data can therefore be an effective tool for comprehending the complexities of human health and behavior [4].

    The COVID-19 pandemic has served as an ideal case study to develop tools and knowledge in the field of human mobility research [5]. The World Health Organization declared COVID-19 an international public health emergency in late January 2020 and subsequently declared it a pandemic on March 11, 2020. To curb the spread of COVID-19 and avoid overwhelming health care institutions [6], many countries implemented restrictions on human mobility including social distancing, self-isolation, closure of nonessential services, work-from-home policies, and travel restrictions [7,8]. In the early stages of the pandemic, digital data sources found that the decrease in human mobility was closely paralleled by a reduction in the incidence of COVID-19 [1,7,9-15]. Thus, the COVID-19 pandemic provides an ideal opportunity to measure and extract meaning from human mobility and how it is affected by restrictive policies.

    The study of human mobility has historically been retrospective with limited study participants. Interestingly, it was the technological advancements in smartphones and wearable devices that have had the largest impact on the field [16]. These innovations now provide direct access to human location, trajectories, opinions, and interactions [17]. Human mobility can now be tracked passively in real time from various sources: GPS-enabled smartphones, texts or photos via photo-sharing platforms like Twitter (X Corp) and Flickr (Altaba, Ludicorp), geolocation-enabled internet posts, public transport cards, satellite, flight traffic, and even credit card transactions. The data collected have a previously unprecedented level of detail, immediacy, and precision. These large data sets, equivalent to “big data” in volumes, are now being analyzed to describe human movement patterns, characteristics (such as sleep, stress, and activity), and interactions.

    Google Maps is the most popular navigation app in the United States and Canada. The app surpassed 23 million downloads in 2020 with 154.4 million monthly users. Google passively generates and collects over 20 million pieces of mobility data per day. The information has now been made available to researchers and policy makers through Google’s open-source “COVID-19 Community Mobility Reports” [18]. The reported geographic movement, called macromobility, is grouped by category: retail and recreation, grocery and drug stores, parks, transit stations, workplaces, and residences [18]. While macromobility data are readily available through sources such as Google’s “COVID-19 Community Mobility Reports,” our understanding of micromobility at the population level remains limited. This study aims to fill this gap by analyzing smart thermostat data, thereby providing a more comprehensive picture of human mobility patterns. This study leverages the unprecedented access to real-time human mobility data provided by smart thermostats, a source that has not been extensively used in previous research. The high level of detail, immediacy, and precision of these data allows for a more granular understanding of human mobility patterns, particularly in-house mobility (micromobility), which has been less explored compared to macromobility. Unlike Google’s macromobility data, smart thermostat data can provide insights into in-house mobility patterns. This is particularly relevant in the context of the COVID-19 pandemic, where stay-at-home orders and work-from-home policies have significantly altered in-house mobility patterns.

    Modular smart home thermostats offer a novel source of data on in-house human behaviors [19,20]. Not only can they report on indoor temperature, humidity, and air quality, but embedded motion sensors can capture mobility within the home. More than 90% of Canadian households had thermostats in 2018, with most opting for programmable thermostats. Smart thermostats, often known as internet of things (IoT) devices, are a type of programmable thermostat that may be connected to the internet. ecobee has the second-highest market share for smart thermostats in Canada. ecobee has a program called Donate Your Data (DYD), in which subscribers can opt to make their anonymized data available for research purposes. DYD currently has 1 million users including over 172,000 households in Canada. Here, we sought to compare mobility data obtained from ecobee smart thermostats to the “gold standard” mobility data from Google during the COVID-19 pandemic in Canada. We explored day-by-day, week-by-week, and month-by-month seasonality patterns and applied anomaly detection to both data sets.


    Data Sources

    Google Mobility data were collected for each province in Canada from February 15, 2020, to February 14, 2021 [18]. Google Maps uses aggregated and anonymized data to determine the daily total number of visits to specific destinations (Textbox 1) visited by individuals who have enabled their location history [18]. Daily values are compiled across individuals and are compared to the baseline value for that day of the week to determine changes in mobility. The baseline is the median of corresponding days over the 5 weeks from January 3 to February 6, 2020 [18]. Out of the 6 categories, we focused on residential data. We curated mobility data from ecobee’s DYD program from February 15, 2020, to February 14, 2021, to align with the Google Mobility Report publication dates. ecobee’s smart thermostats record in-house mobility via motion sensors. The DYD data set supplied by ecobee is hosted in Google BigQuery. This study analyzed data from 4 provinces in Canada: Ontario, Alberta, Quebec, and British Columbia. This included 12,252 ecobee households. These 4 provinces constitute approximately 87% of the Canadian population [21]. The reliability and validity of the ecobee data set were rigorously assessed in our prior studies, as detailed in our previously published research [20,22,23]. Information regarding Canadian policy implementation dates across provinces was obtained from a dedicated platform managed by the Canadian Government to share COVID-19–related information regularly [24].

    Textbox 1. Google mobility data categories and their description as described on the website [18].

    Grocery and pharmacy

    • “Mobility trends for places like grocery markets, food warehouses, farmers markets, specialty food shops, drug stores, and pharmacies.”

    Parks

    • “Mobility trends for places like local parks, national parks, public beaches, marinas, dog parks, plazas, and public gardens.”

    Transit stations

    • “Mobility trends for places like public transport hubs such as subway, bus, and train stations.”

    Retail and recreation

    • “Mobility trends for places like restaurants, cafes, shopping centers, theme parks, museums, libraries, and movie theatres.”

    Residential

    • “Mobility trends for places of residence.”

    Workplaces

    • “Mobility trends for places of work.”

    Ethical Considerations

    Per ethical research standards, this study exclusively used secondary data, and therefore, no additional consent from human subjects was necessary for this research. The original informed consent procedures or institutional review board approvals for primary data collection explicitly permitted the secondary analysis conducted in this study. Furthermore, robust privacy and confidentiality safeguards were implemented to ensure the anonymity and deidentification of study data. Ethics approval for this study was duly obtained from the University of Waterloo Office of Research Ethics (#31377), attesting to our commitment to upholding the highest ethical standards in research.

    Data Preparation

    ecobee’s data were transferred from Google’s BigQuery to the Microsoft Azure cloud services platform [25]. Data were prepared using Azure Databricks data analytics platform and Jupyter Notebook (Jupyter Team) using Python; detailed explanation of the whole process has been mentioned in our previous studies [20,23]. Data cleaning, analysis, and visualization were done in R studio (version 1.4.1106; Posit Software, PBC) with R software (version 4.0.5; The R Foundation) and data analysis packages tidyverse [26] and timetk [27].

    Using ecobee data, we established a baseline mobility value akin to the Google data preparation approach detailed earlier for seamless comparison. To ascertain daily in-house mobility, we computed the total number of activated sensors within each 24-hour timeframe by summing all sensor statuses. A table containing date records alongside the average count of activated sensors per date was then saved as a CSV file.

    As the timestamp of the DYD data set was in Coordinated Universal Time format, the time zones in the data set were converted by locating time zone information from the geolocation of the households in the metadata. In Quebec and Alberta, all the cities were in the same time zone. In the province of Ontario, 6 cities have different time zones than EST: namely, Drayton, Kenora, Kenora-Unorganized, Mitchell, Red Lake, and Sioux Lookout cities (1% of the DYD data set for Ontario). For British Columbia, 28 households were from different time zones than Pacific Standard Time and were excluded from the analysis. Once cleaned for the time zone, time-series data analysis was performed on the adjusted data on the households included in this study. These numbers are presented in Table 1.

    Table 1. The number of households selected for the analysis by province.
    LocationHouseholds before time zone cleaning, nHouseholds excluded for time zone difference, nHouseholds after time zone cleaning, nThe proportion of household data by province, n/N (%)
    Canada21,690—a21,690—
    Ontario71451171347134/21,690 (32.9)
    Alberta3989039893989/21,690 (18.4)
    British Columbia44928421421/21,690 (1.9)
    Quebec7080708708/21,690 (3.2)

    aNot available.

    Data Analysis

    To assess mobility variations, we aggregated daily province-level movement into weekly, monthly, and day-of-the-week periods and compared them to the respective baselines in both data sets. Time series plots were generated for each province using Google’s residential and ecobee mobility data. The statistical significance of the relationship between the 2 data sources was determined through Pearson and Spearman correlation coefficients [28].

    We conducted seasonal diagnostic tests using the timetk [27] package in R software, tailored for time-series data analysis. Distinct approaches were employed to investigate daily, weekly, and monthly seasonality. To assess statistical significance [29] 1-way ANOVA was used.

    For anomaly detection, we initially conducted seasonal and trend decomposition using the Loess method [27,30,31]. After removing the trend and seasonality components, we performed anomaly detection on the residual data. Anomalies were identified based on the IQR, specifically the difference between the 75th and 25th percentiles, establishing the distribution of the remaining data. Default boundaries were set at 3× above or below the IQR, designating values beyond these limits as anomalies.

    To examine data granularity, we analyzed mobility data by days of the week for both Google and ecobee data sets. Additionally, we aggregated mobility data into 1-week intervals for all 4 provinces to explore seasonality in mobility changes following the onset of the COVID-19 pandemic.


    Positive Association Between Google and ecobee Mobility Data

    The comparison of Google residential mobility data and ecobee mobility data across each province over a year revealed a positive association (Figure 1). Notably, the Google mobility data exhibited a recurrent weekly pattern, which was closely mirrored by the ecobee data (Figure 1). Both data sources exhibited a significant uptick in residential mobility, both indoors and outdoors, commencing around March 11, 2020, aligning with the declaration of the COVID-19 pandemic in Canada. It is worth noting that given Canada’s provincial health regulation framework [32], the official implementation dates of pandemic-related policies varied among the 4 provinces under investigation (March 17, 2020, for Ontario; March 16, 2020, for Quebec; March 17, 2020, for Alberta; and March 19, 2020, for British Columbia).

    ‎
    Figure 1. Association between Google residential mobility and ecobee mobility for (A) Ontario, (B) Alberta, (C) Quebec, and (D) British Columbia, Canada. A high resolution version is available in Multimedia Appendix 1.

    All provinces exhibited a notable increase in the percentage of mobility change coinciding with the commencement of the pandemic, as indicated by both Google and ecobee mobility data. Subsequently, from March 2020 to September 2020, a gradual decline in this trend was observed across all provinces in both data sets. Notably, a subsequent surge in residential mobility was linked to the onset of the second pandemic wave and the reinstatement of pandemic-related stay-at-home measures on specific dates (December 26, 2020, for Ontario; December 17, 2020, for Quebec; December 13, 2020, for Alberta; and October 19, 2020, for British Columbia).

    For each province, we calculated the correlation between Google’s residential mobility data and ecobee’s mobility data. The trend line displayed a linear relationship between the 2 data sets (see Figure 2). Pearson and Spearman correlation coefficients revealed a statistically significant association, ranging from 0.67 to 0.73 (refer to Table 2). Consequently, the data obtained from ecobee’s smart home thermostat are demonstrably equivalent to Google’s mobility data.

    ‎
    Figure 2. Correlation between residential mobility data from ecobee and Google for the province of (A) Ontario, (B) Alberta, (C) Quebec, and (D) British Columbia, Canada. A high resolution version is available in Multimedia Appendix 1.
    Table 2. Correlation between Google and ecobee mobility data for Ontario, Alberta, Quebec, and British Columbia.
    ProvinceHouseholds, nPearson correlation coefficient (95% CI)Spearman rank correlation
    Ontario71340.73 (0.67-0.77)0.75
    Alberta39890.73 (0.69-0.78)0.76
    Quebec7080.67 (0.61-0.73)0.70
    British Columbia4210.69 (0.64-0.74)0.63

    Variations in Mobility by Day of the Week

    There was a significant difference in daily mobility patterns across all 4 provinces when examining the Google residential mobility data set (Figure 3). Greater mobility changes were observed on the weekdays compared to weekends. Further, a 1-way ANOVA test (Table 3) showed that the day of the week had a statistically significant impact on residential mobility for all 4 provinces (Google; all P<.001). On the contrary, these differences in daily mobility were only observed in Quebec when we used ecobee’s mobility data.

    ‎
    Figure 3. Analysis of the (A) Google residential and (B) ecobee mobility data in days of a week in Ontario, Quebec, Alberta, and British Columbia. A high resolution version is available in Multimedia Appendix 1.
    Table 3. The ANOVA test compares the day of the week’s impact on Google and ecobee mobility for the 4 provinces of Canada.
    ProvinceGoogleecobee

    		Sum of squaresF test (df)P valueSum of squaresF testP value
    Ontario
    		17.86 (6, 358)<.001
    		0.458 (6, 358).84

    		Weekday3467

    		10


    		Residual11,579

    		1306

    Quebec
    		9.357 (6, 358)<.001
    		2.364 (6, 358).03

    		Weekday2426

    		130


    		Residual15,471

    		3278

    Alberta
    		8.223 (6, 358)<.001
    		1.123 (6, 358).35

    		Weekday1456

    		22.9


    		Residual10,566

    		1216

    British Columbia
    		6.955 (6, 358)<.001
    		0.371 (6, 358).90

    		Weekday897

    		14.9


    		Residual7673

    		2404

    Monthly Variations in Mobility

    To deepen our understanding of the pandemic-related changes in population behavior, we aggregated the Google and ecobee data to analyze month-by-month mobility changes. Interestingly, in contrast to the differences observed between Google and ecobee mobility data for days of the week, month-by-month patterns were similar between the 2 data sets (Figure 4). Across all provinces, the change in mobility above baseline spiked in April 2020 consistent with the implementation of COVID-19 policies to curb social mobility. Further, the variability within the data, specifically for ecobee data, was reduced drastically from April 2020 onwards. The increase in residential and in-home mobility slowly declined from April 2020 to September 2020. A subsequent rise in mobility from October 2020 to December 2020 corresponded to the second pandemic wave and the reimplementation of pandemic-related stay-at-home policies. Furthermore, 1-way ANOVA test showed a statistically significant change in both residential (Google) and in-house (ecobee) month-by-month mobility across all 4 provinces (all P<.001; Table 4).

    ‎
    Figure 4. Analysis of the (A) Google residential and (B) ecobee mobility data in month-by-month for Ontario, Alberta, Quebec, and British Columbia. A high resolution version is available in Multimedia Appendix 1.
    Table 4. The ANOVA test compares month-by-month impact on Google and ecobee mobility for the 4 provinces of Canada.
    ProvinceGoogleecobee

    		Sum of squaresF test (df)P valueSum of squaresF test (df)P value
    Ontario
    		23.65 (11, 353)<.001
    		26.18 (11, 353)<.001

    		Month6383

    		591.4


    		Residual8663

    		724.8

    Quebec
    		36.64 (11, 353)<.001
    		21.61 (11, 353)<.001

    		Month9541

    		1372


    		Residual8357

    		2037

    Alberta
    		31.16 (11, 353)<.001
    		30.76 (11, 353)<.001

    		Month5923

    		606.4


    		Residual6099

    		632.7

    British Columbia
    		28.83 (11, 353)<.001
    		24.49 (11, 353)<.001

    		Month4062

    		1047


    		Residual4508

    		1372

    Week-by-Week Mobility Changes

    Beginning in February 2020, the initial 5 to 6 weeks had a similar level of mobility across all 4 provinces (Figure 5). Google residential mobility declined from week 7 to week 10. Although there was a high level of data variability at week 7, this was lost in subsequent weeks. ecobee in-house mobility followed a similar pattern but with a lag period of 1 week and a large degree of variability. Beginning at week 12 (corresponding to the week of March 16, 2020), Google residential data witnessed a sharp spike in residential mobility across all 4 provinces. This timing correlates to the date the pandemic was declared in Canada. A similar trend was observed for the ecobee data, however, mobility appeared to increase starting in week 11. Similar to the trends observed in month-by-month data (Figure 4), the elevated mobility steadily declined toward week 25 and stabilized above baseline until approximately week 40 (Figure 5). There was a subsequent steady increase in mobility from week 40 until the end of the year corresponding to the timing of the second wave of the pandemic in Canada. The ecobee mobility data showed a similar trend with the exception that the decline in mobility took place over a longer period and with a shorter period of stabilization before rising again. An ANOVA analysis for the week-by-week mobility data showed a statistically significant difference between weeks for both data sets and for all 4 provinces (all P<.001; Table 5).

    ‎
    Figure 5. Analysis of the (A) Google residential and (B) ecobee mobility data in week-by-week for Ontario, Alberta, Quebec, and British Columbia. A high resolution version is available in Multimedia Appendix 1.
    Table 5. The ANOVA test comparing the week-by-week impact on Google and ecobee mobility for the 4 provinces of Canada.
    ProvinceGoogleecobee

    		Sum of squaresF test (df)P valueSum of squaresF test (df)P value
    Ontario
    		12.4 (52, 312)<.001
    		18.43 (52, 312)<.001

    		Week10,140

    		992.9


    		Residual4906

    		323.2

    Quebec
    		18.88 (52, 312)<.001
    		13.79 (52, 312)<.001

    		Week13,582

    		2375


    		Residual4316

    		1034

    Alberta
    		17.57 (52, 312)<.001
    		17.84 (52, 312)<.001

    		Week8961

    		927.3


    		Residual3061

    		311.9

    British Columbia
    		19.49 (52, 311)<.001
    		13.53 (52, 312)<.001

    		Week6558

    		1675.8


    		Residual2012

    		742.9

    Social Mobility Policy Restrictions and Festive Periods Create Mobility Anomalies

    To determine whether Google and ecobee data analysis could pick out behavioral anomalies in mobility associated with policy changes, we performed an anomaly detection analysis (Figure 6). Neither Google residential data nor ecobee in-house mobility data showed any anomaly within 2020 for Ontario. For Quebec, Alberta, and British Columbia, anomalies were found at the beginning period in both the Google and ecobee data. Notably, these corresponded with the dates of COVID-19–related policy changes. In Alberta and British Columbia, ecobee’s in-house mobility analysis captured anomalies in May and June 2020, which correspond to phase-wise reopening plans and the lifting of social restrictions. Interestingly, Google residential data were able to capture anomalies corresponding to festive periods such as Christmas and New Year’s Eve in Quebec, Alberta, and British Columbia. These anomalies were not seen in the ecobee data. Overall, these results demonstrate the ability of Google’s residential data and ecobee’s thermostat data to capture notable shifts in population behavior as a result of policy changes and cultural festivities.

    ‎
    Figure 6. Analysis of the (A) Google residential and (B) ecobee mobility data in days of a week for Ontario, Alberta, Quebec, and British Columbia. A high resolution version is available in Multimedia Appendix 1.

    This study aimed to investigate the potential application of in-house mobility data obtained from ecobee sensors in comparison to residential Google mobility data. The most significant differences between data sets were observed, particularly when mobility was analyzed by the day of the week. While Google’s residential data exhibited a significant correlation with the day of the week across all 4 provinces, ecobee’s data showed a significant impact only in Quebec. However, when scrutinized on a monthly and weekly basis, consistent findings were observed across all provinces and data sets. A notable surge in in-house and residential mobility occurred between March and April 2020, coinciding with the initiation of pandemic-related policy changes in Canada. These mobility shifts gradually diminished until September 2020, and the onset of the second wave of the pandemic, coupled with the reinstatement of social policies, corresponded with increased mobility from October to December 2020. In summary, a statistically significant association between the 2 data sets was identified. Anomaly detection analysis provided evidence supporting the capability of both data sets to detect deviations in population mobility, capturing events such as emergency declarations, reopening phases, and festive days such as December 25 (Christmas) and January 1 (New Year’s Eve). These findings underscore the benefits of employing public health surveillance mechanisms, especially during health emergencies of pandemic proportions.

    Historically, monitoring individual mobility has been hampered by insufficient sample sizes, difficulty in collecting data, recall bias, and privacy concerns [33-35]. The advent of smartphones, wearable devices, and the IoT has paved the way for researchers to capture individual geolocation and movement within the home (micromobility) [36-40]. For both micro- and macromobility, it is necessary to integrate data from multiple sources [41]. Thermostat-based IoT data, such as the ecobee DYD data reported here, can provide new opportunities for calculating population-level mobility indicators. IoT is a modern, passive sensing tool that can quantify in-house movement. Understanding in-house movement is particularly relevant given that people now spend more than 80% of their time indoors [4,42]. The use of these large data sets, combined with geographical data and timestamps [16,17], has the potential to unravel different dimensions of human behavior and lifestyle. In-house mobility data can measure changes in sleep [5,43], physical activity [44], sedentary behaviors [45-47], and movement patterns [48] with greater detail and granularity. Motion sensors are a next-generation tool with a wealth of opportunities for application in the field of public health. Integrated into smart cities, motion sensors have the power to detect overall human mobility from various sources. Data collection will enable effective planning and implementation of responsive and preventative public health strategies [37].

    Monitoring real-time population mobility is important in public health as it plays a significant role in both chronic and infectious diseases. A dwindling mobility pattern is a predisposition to various chronic diseases including dementia and age-related physical decline. For infectious diseases, individual mobility is directly proportional to the transmission of infections such as COVID-19 [49]. Although the “stay-at-home” and “work-from-home” strategies were promoted globally to curb the spread of COVID-19, it is still unclear to what extent individuals complied with policy restrictions. Mobility data analysis has the potential to provide real-time information about the impact of such policies on individual and population behavior [12,50,51].

    However, a notable limitation of our study was the absence of sociodemographic information in the ecobee data set. Consequently, our analysis was confined to spatiotemporal dimensions, and we were unable to examine the impact of sociodemographic features, including age, gender, and occupation. Varied perceptions surrounding infectious diseases, vaccinations, social mobility, and government policies across cultural and socioeconomic groups highlight the need for caution when generalizing the results [52]. Although the data were collected across 4 Canadian provinces, encompassing 87% of the population, the demographic represented by those with smart home thermostats likely skews toward a specific group—namely, young, tech-savvy individuals with higher socioeconomic status, who may be more inclined to work from home. Therefore, caution should be exercised in generalizing the reported results. Additionally, challenges arise in separating mobility patterns in multiperson households and eliminating sensor activation due to factors such as animals, rapid airflow, or other noises.

    In conclusion, the real-time monitoring of population-level mobility using smartphones and IoT sensors has emerged as a recent development in public health, primarily in response to the COVID-19 pandemic. This study investigates the utility of IoT-based mobility data, specifically from smart thermostats, for assessing individual mobility within the context of social isolation policies. The findings demonstrate a close alignment between thermostat mobility data and the data presented in the Google Mobility Report, affirming its value as a mobility monitoring tool. The acquisition of real-time mobility data from smart thermostats has the potential to enhance our understanding of the intricate social determinants of health, providing valuable insights for the formulation of public health policies.

    Acknowledgments

    The authors would like to thank ecobee for sharing the data with us. The authors would also like to thank the teammates at the Ubiquitous Health Technology Lab (UbiLab) for their help in data management, quality checking of the code, and visualizations.

    Data Availability

    The Google mobility data can be readily accessed by the public [18]. In contrast, access to ecobee data is restricted to researchers and scientists through a data-sharing agreement. If you are interested in accessing ecobee data, please email the research division of ecobee.

    Authors' Contributions

    KSS and PPM conceived and designed this study, KSS performed data extraction, cleaning, and analysis. KSS wrote the paper under the supervision of PPM, JAD, SEM, and SL. All the authors have reviewed and approved the final manuscript and provided edits and feedback to strengthen it.

    Conflicts of Interest

    KSS is an Editorial Board member of JMIR Public Health and Surveillance at the time of this publication. The other authors have no conflicts of interest to declare.

    Multimedia Appendix 1

    High resolution images of Figures 1-6.

    ZIP File (Zip Archive), 7855 KB
    1. Marcus MK, Lucas L, Scott J. The impact of COVID-19 restrictions on individual mobility. BRUEGEL. 2020. URL: https://www.bruegel.org/2020/05/the-impact-of-covid-19-restrictions-on-individual-mobility/ [accessed 2020-05-14]
    2. González MC, Hidalgo CA, Barabási AL. Understanding individual human mobility patterns. Nature. 2008;453(7196):779-782. [CrossRef] [Medline]
    3. Kang Y, Gao S, Liang Y, Li M, Rao J, Kruse J. Multiscale dynamic human mobility flow dataset in the U.S. during the COVID-19 epidemic. Sci Data. 2020;7(1):390. [FREE Full text] [CrossRef] [Medline]
    4. Wang A, Zhang A, Chan EHW, Shi W, Zhou X, Liu Z. A review of human mobility research based on big data and its implication for smart city development. ISPRS International Journal of Geo-Information. Dec 31, 2020;10(1):13. [FREE Full text] [CrossRef]
    5. Ong JL, Lau T, Massar SAA, Chong ZT, Ng BKL, Koek D, et al. COVID-19-related mobility reduction: heterogenous effects on sleep and physical activity rhythms. Sleep. 2021;44(2):zsaa179. [FREE Full text] [CrossRef] [Medline]
    6. WHO director-general's opening remarks at the media briefing on COVID-19—11 March 2020. World Health Organization. 2020. URL: https:/​/www.​who.int/​director-general/​speeches/​detail/​who-director-general-s-opening-remarks-at-the-media-briefing-on-covid-19---11-march-2020 [accessed 2021-05-14]
    7. Nouvellet P, Bhatia S, Cori A, Ainslie KEC, Baguelin M, Bhatt S, et al. Reduction in mobility and COVID-19 transmission. Nat Commun. 2021;12(1):1090. [FREE Full text] [CrossRef] [Medline]
    8. Flaxman S, Mishra S, Gandy A, Unwin HJT, Mellan TA, Coupland H, et al. Estimating the effects of non-pharmaceutical interventions on COVID-19 in Europe. Nature. 2020;584(7820):257-261. [FREE Full text] [CrossRef] [Medline]
    9. Kishore K, Jaswal V, Verma M, Koushal V. Exploring the utility of google mobility data during the COVID-19 pandemic in India: digital epidemiological analysis. JMIR Public Health Surveill. 2021;7(8):e29957. [FREE Full text] [CrossRef] [Medline]
    10. Zeng C, Zhang J, Li Z, Sun X, Olatosi B, Weissman S, et al. Spatial-temporal relationship between population mobility and COVID-19 outbreaks in South Carolina: time series forecasting analysis. J Med Internet Res. 2021;23(4):e27045. [FREE Full text] [CrossRef] [Medline]
    11. Hanvey L. The road to recovery: what mobility data is telling us about COVID-19 and how it can help us plan economic recovery. UrbanLogiq. 2020. URL: https:/​/urbanlogiq.​com/​the-road-to-recovery-what-mobility-data-is-telling-us-about-covid-19-and-how-it-can-help-us-plan-economic-recovery/​ [accessed 2020-05-15]
    12. Yilmazkuday H. Stay-at-home works to fight against COVID-19: international evidence from Google mobility data. J Hum Behav Soc Environ. 2021;31(1-4):210-220. [FREE Full text] [CrossRef]
    13. Badr HS, Du H, Marshall M, Dong E, Squire MM, Gardner LM. Association between mobility patterns and COVID-19 transmission in the USA: a mathematical modelling study. Lancet Infect Dis. 2020;20(11):1247-1254. [FREE Full text] [CrossRef] [Medline]
    14. Kraemer MUG, Yang CH, Gutierrez B, Wu CH, Klein B, Pigott DM, et al. The effect of human mobility and control measures on the COVID-19 epidemic in China. Science. 2020;368(6490):493-497. [FREE Full text] [CrossRef] [Medline]
    15. Buckee CO, Balsari S, Chan J, Crosas M, Dominici F, Gasser U, et al. Aggregated mobility data could help fight COVID-19. Science. 2020;368(6487):145-146. [FREE Full text] [CrossRef] [Medline]
    16. Thums M, Fernández-Gracia J, Sequeira AMM, Eguíluz VM, Duarte CM, Meekan MG. How big data fast tracked human mobility research and the lessons for animal movement ecology. Front Mar Sci. 2018;5:21. [FREE Full text] [CrossRef]
    17. WuFang J, ChenYing J, SouSok I. CoCo: quantifying correlations between mobility traces using sensor data from smartphones. ACM Transactions Int Things. 2021;2(3):1-22. [CrossRef]
    18. Google COVID-19 community mobility reports. Google LLC. URL: https://www.google.com/covid19/mobility/ [accessed 2020-05-13]
    19. Sahu KS, Majowicz SE, Dubin JA, Morita PP. NextGen public health surveillance and the internet of things (IoT). Front Public Health. 2021;9:756675. [FREE Full text] [CrossRef] [Medline]
    20. Sahu KS, Oetomo A, Morita PP. Enabling remote patient monitoring through the use of smart thermostat data in Canada: exploratory study. JMIR mHealth uHealth. 2020;8(11):e21016. [FREE Full text] [CrossRef] [Medline]
    21. Population estimates, quarterly. Statistics Canada. 2021. URL: https://www150.statcan.gc.ca/t1/tbl1/en/tv.action?pid=1710000901 [accessed 2024-02-03]
    22. Jalali N, Sahu KS, Oetomo A, Morita PP. Understanding user behavior through the use of unsupervised anomaly detection: proof of concept using internet of things smart home thermostat data for improving public health surveillance. JMIR mHealth uHealth. 2020;8(11):e21209. [FREE Full text] [CrossRef] [Medline]
    23. Jalali N, Sahu KS, Oetomo A, Morita PP. Usability of smart home thermostat to evaluate the impact of weekdays and seasons on sleep patterns and indoor stay: observational study. JMIR mHealth uHealth. 2022;10(4):e28811. [FREE Full text] [CrossRef] [Medline]
    24. COVID-19 intervention scan. Canadian Institute for Health Information. 2021. URL: https://www.cihi.ca/en/covid-19-intervention-scan [accessed 2021-07-09]
    25. What is Azure Databricks? Microsoft. 2021. URL: https://docs.microsoft.com/en-us/azure/databricks/scenarios/what-is-azure-databricks [accessed 2021-08-09]
    26. Wickham H, Averick M, Bryan J, Chang W, McGowan LD, François R, et al. Welcome to the tidyverse. JOSS. 2019;4(43):1686. [FREE Full text] [CrossRef]
    27. Matt D, Davis V. R package timetk version 2.6.1-a tool kit for working with time series in R. Comprehensive R Archive Network (CRAN). timetk for R. 2021. URL: https://business-science.github.io/timetk/ [accessed 2024-02-03]
    28. Sedgwick P. Pearson's correlation coefficient. BMJ. 2012;345(jul04 1):e4483. [CrossRef]
    29. Larson MG. Analysis of variance. Circulation. 2008;117(1):115-121. [FREE Full text] [CrossRef] [Medline]
    30. Vallis O, Hochenbaum J, Kejariwal A. A novel technique for long-term anomaly detection in the cloud. Berkeley, CA, United States. USENIX Association; Presented at: HotCloud'14: Proceedings of the 6th USENIX conference on Hot Topics in Cloud Computing; June 17-18, 2014, 2014; Philadelphia PA. URL: https://dl.acm.org/doi/proceedings/10.5555/2696535 [CrossRef]
    31. Robert C, William C, Irma T. STL: a seasonal-trend decomposition procedure based on loess. J Off Stat. 1990;6(1):3-73. [FREE Full text]
    32. Canada's health care system—Canada.ca. Government of Canada. URL: https:/​/www.​canada.ca/​en/​health-canada/​services/​health-care-system/​reports-publications/​health-care-system/​canada.​html#a4 [accessed 2022-02-17]
    33. Banerjee S, Hemphill T, Longstreet P. Wearable devices and healthcare: data sharing and privacy. Inf Soc. 2018;34(1):49-57. [CrossRef]
    34. Mooney SJ, Pejaver V. Big data in public health: terminology, machine learning, and privacy. Annu Rev Public Health. 2018;39:95-112. [FREE Full text] [CrossRef] [Medline]
    35. Lin C, Song Z, Song H, Zhou Y, Wang Y, Wu G. Differential privacy preserving in big data analytics for connected health. J Med Syst. 2016;40(4):97. [CrossRef] [Medline]
    36. Charu V, Zeger S, Gog J, Bjørnstad ON, Kissler S, Simonsen L, et al. Human mobility and the spatial transmission of influenza in the United States. PLoS Comput Biol. 2017;13(2):e1005382. [FREE Full text] [CrossRef] [Medline]
    37. Rosa L, Silva F, Analide C. Mobile networks and internet of things infrastructures to characterize smart human mobility. Smart Cities. 2021;4(2):894-918. [FREE Full text] [CrossRef]
    38. Nižetić S, Šolić P, de-Ipiña González-de-Artaza DL, Patrono L. Internet of things (IoT): opportunities, issues and challenges towards a smart and sustainable future. J Clean Prod. 2020;274:122877. [FREE Full text] [CrossRef] [Medline]
    39. Trivedi A, Silverstein K, Strubell E, Iyyer M, Shenoy PJ. WiFiMod: Transformer-based Indoor Human Mobility Modeling using Passive Sensing. New York, NY, United States. Association for Computing Machinery; Presented at: COMPASS '21: Proceedings of the 4th ACM SIGCAS Conference on Computing and Sustainable Societies; 28 June 2021-2 July 2021, 2021;126-137; Virtual Event Australia. URL: https://dl.acm.org/doi/proceedings/10.1145/3460112 [CrossRef]
    40. Yechezkel M, Weiss A, Rejwan I, Shahmoon E, Ben-Gal S, Yamin D. Human mobility and poverty as key drivers of COVID-19 transmission and control. BMC Public Health. 2021;21(1):596. [FREE Full text] [CrossRef] [Medline]
    41. Huang X, Li Z, Jiang Y, Ye X, Deng C, Zhang J, et al. The characteristics of multi-source mobility datasets and how they reveal the luxury nature of social distancing in the U.S. during the COVID-19 pandemic. Int J Digit Earth. 2021;14(4):424-442. [FREE Full text] [CrossRef]
    42. Robillard R, Dion K, Pennestri MH, Solomonova E, Lee E, Saad M, et al. Profiles of sleep changes during the COVID-19 pandemic: demographic, behavioural and psychological factors. J Sleep Res. 2021;30(1):e13231. [FREE Full text] [CrossRef] [Medline]
    43. Asgari F, Gauthier V, Becker M. A survey on human mobility and its applications. ArXiv. Preprint posted online on 02 Jul 2013. 2013. [FREE Full text]
    44. Simone T, Marzolini B, Bulgheroni M. Smartphone-based passive sensing for behavioral and physical monitoring in free-life conditions: technical usability study. JMIR Biomed Eng. 2021;6(2):e15417. [FREE Full text] [CrossRef]
    45. Stockwell S, Trott M, Tully M, Shin J, Barnett Y, Butler L, et al. Changes in physical activity and sedentary behaviours from before to during the COVID-19 pandemic lockdown: a systematic review. BMJ Open Sport Exerc Med. 2021;7(1):e000960. [FREE Full text] [CrossRef] [Medline]
    46. Falck RS, Davis JC, Khan KM, Handy TC, Liu-Ambrose T. A wrinkle in measuring time use for cognitive health: how should we measure physical activity, sedentary behaviour and sleep? Am J Lifestyle Med. 2023;17(2):258-275. [FREE Full text] [CrossRef] [Medline]
    47. Sañudo B, Fennell C, Sánchez-Oliver AJ. Objectively-assessed physical activity, sedentary behavior, smartphone use, and sleep patterns pre- and during-COVID-19 quarantine in young adults from Spain. Sustainability. 2020;12(15):5890. [FREE Full text] [CrossRef]
    48. Gordon BA, Bruce L, Benson AC. Physical activity intensity can be accurately monitored by smartphone global positioning system 'app'. Eur J Sport Sci. 2016;16(5):624-631. [FREE Full text] [CrossRef] [Medline]
    49. Gottumukkala R, Katragadda S, Bhupatiraju RT, Kamal AM, Raghavan V, Chu H, et al. Exploring the relationship between mobility and COVID-19 infection rates for the second peak in the United States using phase-wise association. BMC Public Health. 2021;21(1):1669. [FREE Full text] [CrossRef] [Medline]
    50. Kartal MT, Depren Ö, Depren SK. The relationship between mobility and COVID-19 pandemic: daily evidence from an emerging country by causality analysis. Transp Res Interdiscip Perspect. 2021;10:100366. [FREE Full text] [CrossRef] [Medline]
    51. Engle S, Stromme J, Zhou A. Staying at home: mobility effects of COVID-19. SSRN Electronic J. 2020.:1-16. [CrossRef]
    52. Bayeh R, Yampolsky MA, Ryder AG. The social lives of infectious diseases: why culture matters to COVID-19. Front Psychol. 2021;12:648086. [FREE Full text] [CrossRef] [Medline]

    ‎
    DYD: Donate Your Data
    IoT: internet of things

    Edited by A Mavragani; submitted 02.03.23; peer-reviewed by N Maddah, Y Wu; comments to author 08.06.23; revised version received 27.09.23; accepted 03.01.24; published 20.03.24.

    Copyright

    ©Kirti Sundar Sahu, Joel A Dubin, Shannon E Majowicz, Sam Liu, Plinio P Morita. Originally published in JMIR Public Health and Surveillance (https://publichealth.jmir.org), 20.03.2024.

    This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR Public Health and Surveillance, is properly cited. The complete bibliographic information, a link to the original publication on https://publichealth.jmir.org, as well as this copyright and license information must be included.