The Healthy Behaviour Data Challenge explores behavioural patterns to enable remote healthy behaviour monitoring using sensor data from smart-home technology.
The data collected will help to understand individual and population-level behaviours that provide real-time and granular information regarding the indicators of sleep, sedentary behaviour, and activity in the house. This information will help the PHAC (Public Health Agency of Canada) to generate real-time population-level insights on Canadian health behaviours.
Our platform leverages off-the-shelf technology such as smart Wi-Fi thermostats, associated remote thermostat sensors (ecobee), as well as fitness trackers (Fitbit), to enable real-time, and more efficient health behaviour data collection.
The thermostat sensors were originally designed to monitor motion in the house to maintain a comfortable temperature in rooms that were in use, aiming to save energy. We have decided to use this in-home technology and their data to monitor Canadians’ health behaviour.
The remote sensors are capable of detecting motion in the house and provide a real-time and continuous assessment of the patterns of movement between rooms, which can be used to understand health behaviours such as physical activity in the home, sleep quality, sleep duration, and sedentary behaviour.
This project aims to enable the UbiLab Public Health Surveillance Platform (UPHSP) for population-level surveillance to monitor health behaviour. To achieve this objective, we have taken a two-step approach:
- Deployed a pilot study to demonstrate direct the relationship between ecobee’s remote sensor activation and movement/physical activity in the house (measured by the number of steps on a Fitbit), serving as evidence to the capability of the ecobee’s dataset for population-level surveillance.
- Used the algorithms developed in the proof of concept on a larger dataset secured with ecobee, providing evidence of the scalability of our platform.
We ran the Pilot Study amongst our team (five participants) for the initial development of our algorithms. We illustrate the power of using a small number of households and the data that can be collected by using sensors and fitness trackers together to create an algorithm to monitor healthy behaviour at a population level.
The homes were occupied by a single resident wearing a fitness tracker throughout the data collection period, and data were collected both from the ecobee thermostat and the Fitbit.
Each house was equipped with an ecobee thermostat and had between 5 and 29 remote sensors installed (depending on the size of the house). Participants wore a Fitbit for about one week to collect data in parallel with ecobee’s sensor data.
Our assumptions were validated. The correlation tests indicate that there is a strong correlation between the ecobee’s sensor activation and the number of steps taken by the participant, providing evidence of the potential of using this data as a PHAC indicator of physical activity.
These results indicate that our sensor-based platform can be utilized as a substitute for the individual fitness tracker based dataset. This approach is cost-effective, easy to deploy, and a very good example of zero-effort technology.
With our protocols and algorithms developed in the pilot study, the Donate Your Data (DYD) dataset provided by ecobee was used to illustrate the deployment of this technology across thousands of households to display population-level surveillance at a national level. This voluntary, opt-in program contains over 10,251 homes in the United States and Canada and is run by ecobee, which includes 1,302 Households from Canada and 7,888 from the USA that have consented to the DYD data-sharing program.
According to ecobee, these countries have over 1 million thermostats installed, providing an immense potential dataset with the development of PHAC curated data-sharing programs. Existing federal and provincial programs are installing free smart thermostats in over 100,000 residences in Ontario, demonstrating the future potential of our platform.
With an ageing population and the rising rates of chronic diseases, we must look to prevention and positive messaging to ensure that Canadians have relevant health information to make informed decisions about their lifestyle choices.
This technology presents an opportunity for the PHAC to generate personalized medicine algorithms and provide personal evidence-based health recommendations at a population-level.
Using knowledge translation frameworks and building on the data collected, non-expert audiences will have access to high-quality information generated by themselves.
The data can be presented in raw format and shared with users that indent to know more about their health behaviour, recommendations for physical activity, sleep and sedentary behaviour improvements, that can be shared via web/mobile dashboard applications, the thermostats themselves, and even AI-enabled personal assistants (Amazon Alexa, Google Home). Raw data will do little to encourage action and will require AI-based algorithms developed by health experts to interpret and present information positively and easy to understand terms.