Close

Presentation

Household and population level behavioural changes due to Covid-19 pandemic: A smart thermostat based comparative data analysis
Event Type
Oral Presentations
TimeTuesday, April 132:20pm - 2:40pm EDT
LocationDigital Health
DescriptionThe World Health Organization (WHO) declared the coronavirus outbreak as a pandemic on 11th March 2020. Not a single country was ready to face this scale of a pandemic. The ability to carry out daily life activities has significantly impacted. Each country has witnessed a phase-wise lockdown, total shutdown and several new additions to hygiene standards, including a requirement of more frequent hand wash, use of a face mask, reduced social activities and gatherings. To inhibit the spread of COVID-19, governments around the globe have implemented physical distancing and lockdown measures. Considering the implications of the government guidelines, mandating physical distancing and implementing a work-from-home protocol for most jobs and promoting online schooling, it is natural to assume that such changes will pose some consequences on adults, children, and the youth population. Physical activity participation is limited indoors, and achieving the recommended physical activity targets may be difficult.
Physical activity is the foundation of a healthy lifestyle. Despite several attempts to encourage physical activity, it is found that people of all ages are choosing sedentary behaviour over an active lifestyle. Maintaining a healthy and physically active lifestyle with the right balance of exercise and rest is crucial to achieve and maintain an overall state of health, well-being and improving the population's quality of life irrespective of age, sex, and other sociodemographic indicators.
Scientific studies show that sleep, sedentary behaviour and physical activity are associated with a broad spectrum of chronic disease, not just limited to diabetes, hypertension, cancer or mental health problems. Recently, very mild physical activity, including daily living activities in the household, has also been reported to have positively affected health. An increase in the amount of time spent in any one of these movement-related behaviours that comprise a 24-hour day will change the amount of time spent sedentary. Time spent at home and engagement with different household activities has emphasized the importance of movement behaviours spread over 24 hours of the day. In line with these reports, recently, Canada has developed a 24-Hour Movement Guideline for all ages laying guidance on the ideal amount of physical activity, sedentary behaviour, and sleep for an individual in a day. To monitor and measure the range of associated indicators, Canada's Public Health Agency has prepared PASS (Physical Activity, Sedentary Behaviour and Sleep) indicators. Repeated publication of this indicator provides the changes in population-level behaviour and other associated factors that directly or indirectly influence them.
The purpose of this study was to investigate changes on the household and population-level in lifestyle behaviours, including physical activity, sleep, sedentary behaviour and time spent indoors at the household-level, following the implementation of physical distancing protocols and stay-at-home guidelines. Our data analysis shows how Covid-19 impacted household behaviours and patterns of change.

Objectives

This study's primary goal is to compare household and population behaviours before and during the covid-19 pandemic using household-level data collected in the home via smart thermostats. The approach is to (a) identify room occupancy patterns and trends using the motion and thermostat sensor data from a day-over-day, weekend vs weekday, week-over-week, month-over-month pattern, (b) determine how policy-level changes during the pandemic such as lockdown implementation impacted household behaviours, (c) identify the impact of lockdown policy on a sleep schedule, work-from-home schedule and physical activity pattern. As a result, it will be possible to understand the household level and population-level behaviours and, consequently, understand the lifestyle habits or population. A prototype system will be built composed of machine learning algorithms that will mine the data to find patterns and deliver the findings to stakeholders. The data will be presented to public health officials in a user-interface designed using user-centred design methods, presenting big data in a dashboard format.

Methods

Zero-effort-technologies are critical in the digital world and for the future of public health surveillance. Datasets like such could be integrated with other technologies to tackle unsolved remote monitoring issues that challenged the traditional data collection method barriers. For example, when the remote sensor (RS) is placed in the household, they can provide insights on occupancy and indirectly different household activities. This addresses the challenges of participant's declining engagement, low response rates in surveys and focus groups, and technical barriers to wearable technology. This also eliminates recall bias, common when asking participants to quantify the amount of sleep and questions about behaviours such as the amount of physical activity they engage in within a day. Using motion data, we can quantify the amount of sleep by using the absence of movement and, similarly, increased sensor activation will show a longer duration of household occupancy. Through a research partnership with ecobee, a Canadian smart wi-fi thermostat company, we aim to leverage data from over 100,000 households in North-America collected through the Donate Your Data (DYD) program. The length of the dataset is for five years (2016-2020). For this study, we will use a dataset limited to 2019 and 2020 to compare before and during the pandemic lockdown. The DYD dataset was analyzed using Python and R. This method will enable personalized insights to monitor individual and population-level health behaviours.

Results

Our previous studies presented in the past two HFES healthcare symposiums have shown that there was a positive association between the wearable data (Fitbit) and smart thermostat (ecobee) data was found (Spearman's Correlation coefficient = 0.8, p > 0.001). Since presenting at HFES 2018 and 2019, we have improved our data analysis using data visualization techniques to capture 24-hour household behaviour and trends at each household level. The 24-hour household activity includes sleep, sedentary behaviour, as well as physical activity. Our preliminary findings show significant changes at the household and population level for these indicators due to pandemic related policy changes. People stay at home for more extended periods, and time away from home has been significantly reduced. A month by month and intermittent time series analysis of the motion sensor data shows the pattern, intensity and duration of the changes. It also explores the pandemic fatigue-related reversal pattern of the households. Further analysis of the Donate your data for the year 2020 is under process, and the results will be available before the final presentation.

Impacts (Long and short-term)

This project is innovative, (i) population-level comparison of household behaviours for 24 hours, (ii) using novel data source(s) for population-level surveillance, (iii) measuring new indicators, (iv) using novel tools/solutions to capture the data. Data collection on a granular level in real-time will now be a reality. The methods proposed here will enable access to a much larger sample size and increase behavioural changes' generalizability.