Passive digital phenotyping: Objective quantification of human behaviour through smartphones

Passive digital phenotyping is defined as the passive quantification of human behaviour through devices such as the smartphone. Due to important advantages compared to traditional questionnaires this approach is rapidly gaining traction in recent years as a research tool. This trend particularly prevalent in research disciplines that are concerned with studying psychiatric disorders and their underlying biological pathways. The two most important advantages of this approach are 1) that behaviour is quantified without any active input of the participant (i.e. passive) and in a longitudinal manner and 2) that the collected data is less subjective. However, a major challenge in using this approach is how to derive valid and clinically relevant phenotypes from complex smartphone-based sensor data.
in this thesis, we first introduce and validate several statistical methods that can be used to derive behavioural phenotypes from smartphone data. Subsequently, we assess if the derived phenotypes are clinically relevant in the context of psychiatric disorders. All chapters together suggest these so-called digital phenotypes can reliably be derived and that they are mainly associated with several aspects of daily social functioning. In addition, we also show that these phenotypes can be used in a classification approach to differentiate between healthy controls and individuals diagnosed with either schizophrenia or Alzheimer's.