Background: Despite the presence of robust global public health surveillance mechanisms, the COVID-19 pandemic devastated the world and exposed the weakness of the public healthcare systems. Public health surveillance has improved in recent years as technology evolved to enable the mining of diverse data sources, for example, electronic medical records, social media, to monitor diseases and risk factors. However, the current state of the public health surveillance system depends on traditional (e.g., Canadian Community Health Survey (CCHS), Canadian Health Measures Survey (CHMS)) and modern data sources (e.g., Health insurance registry, Physician billing claims database). While improvement was observed over the past few years, there is still a room for improving the current systems with NextGen data sources (e.g., social media data, Internet of Things data), improved analytical mechanism, reporting, and dissemination of the results to drive improved policymaking at the national and provincial level. With that context, data generated from modern technologies like the Internet of Things (IoT) have demonstrated the potential to bridge the gap and be relevant for public health surveillance. This study explores IoT technologies as potential data sources for public health surveillance and assesses their feasibility with a proof of concept. The objectives of this thesis are to use data from IoT technologies, in this case, a smart thermostat with remote sensors that collect real-time data without additional burden on the users, to measure some of the critical population-level health indicators for Canada and its provinces.
Methods: This exploratory research thesis utilizes an innovative data source (ecobee) and cloud-based analytical infrastructure (Microsoft Azure). The research started with a pilot study to assess the feasibility and validity of ecobee smart thermostat-generated movement sensor data to calculate population-level indicators for physical activity, sedentary behaviour, and sleep parameters for Canada. In the pilot study, eight participants gathered step counts using a commercially available Fitbit wearable as well as sensor activation data from ecobee smart thermostats.
In the second part of the study, a perspective article analyzes the feasibility and utility of IoT data for public health surveillance. In the third part of this study, data from ecobee smart thermostats from the “Donate your Data” program was used to compare the behavioural changes during the COVID-19 pandemic in four provinces of Canada. In the fourth part of the study, data from the “Donate your Data” program was used in conjunction with Google residential mobility data to assess the impact of the work-from-home policy on micro and macro mobility across four provinces of Canada. The study's final part discusses how IoT data can be utilized to improve policy-level decisions and their impact on daily living, with a focus on situations similar to the COVID-19 pandemic.
Results: The Spearman correlation coefficient of the step counts from Fitbit and the number of sensors activated was 0.8 (range 0.78-0.90; n=3292) with statistically significant at P < .001 level. The pilot study shows that ecobee sensors data have the potential to generate the population-level health indicators. The indicators generated from IoT data for Canada, Physical Activity, Sleep, and Sedentary Behaviours (PASS) were consistent with values from the PASS indicators developed by the Public Health Agency of Canada.
Following the pilot study, the perspective paper analyzed the possible use of the IoT data from nine critical dimensions: simplicity, flexibility, data quality, acceptability, sensitivity, positive predictive value, representativeness, timeliness, and stability. This paper also described the potential advantages, disadvantages and use cases of IoT data for individual and population-level health indicators. The results from the pilot study and the viewpoint paper show that IoT can become a future data source to complement traditional public health surveillance systems.
The third part of the study shows a significant change in behaviour in Canada after the COVID-19 pandemic and work-from-home, stay at home and other policy changes. The sleep habits (average bedtime, wake-up time, sleep duration), average in-house and out-of-the-house duration has been calculated for the four major provinces of Canada (Ontario, Quebec, Alberta, and British Columbia). Compared to pre-pandemic time, the average sleep duration and time spent inside the house has been increased significantly whereas bedtime, and wake-up-time got delayed, and average time spent out-of-the-house decreased significantly during COVID-19 pandemic.
The result of the fourth study shows that the in-house mobility (micro-mobility) has been increased after the pandemic related policy changes (e.g., stay-at-home orders, work-from-home policy, emergency declaration). The results were consistent with findings from the Google residential mobility data published by Google. The Pearson correlation coefficient between these datasets was 0.7 (range 0.68-0.8) with statistically significant at P <.001 level. The time-series data analysis for ecobee and google residential mobility data highlights the substantial similarities. The potential strength of IoT data has been demonstrated in the chapter in terms of anomaly detection.
Discussion and Conclusion: This research's findings demonstrate that IoT data, in this case, smart thermostats with remote motion sensors, is a viable option to measure population-level health indicators. The impact of the population-level behavioural changes due to the COVID-19 pandemic might be sustained even after policy relaxation and significantly affects physical and mental health. These innovative datasets can strengthen the existing public health surveillance mechanism by providing timely and diverse data to public health officials. These additional data sources can offer surveillance systems with near-real-time health indicators and potentially measure the short- and long-term impact of policy changes.