Using Passive Sensing to Identify Depression

Abstract

Depression is a common mental health issue that affects people’s thoughts, behaviours, and feelings. However, depression can often be under-diagnosed or under-treated. Early identification of depression can help to reduce the severity of the condition. Passive sensing, which captures data through mobile applications and wearable devices, has been shown effective in monitoring and identifying mental health problems, including depression. In line with the scope of AISoLA for Digital Humanities to explore the challenges and opportunities of interdisciplinary action to develop better practices in research, this paper explores the efficacy of passive sensing through mobile applications and fitness trackers to identify signs of depression among 52 adults in a three-week study. Sensing data captures calls, text messages, locations, nearby devices, usage of social applications, physical activity, sleep duration and quality through the AWARE and FitBit applications. The paper also investigates differences in the behaviour between people without depression and people with symptoms of depression, and it explores which sensor data can help to accurately identify depression. The results show high accuracy of certain sensing data to identify symptoms of depression. Depression is associated with reduced physical activity, higher sleep efficiency, increased number of incoming calls, increased number of visited places and reduced application use. Differences between behaviours show that people with symptoms of depression are less active, have a higher sleep score and receive more calls compared to people without symptoms. These findings should be interpreted within the methodological issues that are discussed in this paper in relation to wider research in sensing technology that aims to identify and monitor depression, including small sample sizes and lack of information about participants.