The Relationship between Clinical, Momentary, and Sensor-based Assessment of Depression

The clinical assessment of severity of depressive symptoms is commonly performed with standardized self-report questionnaires, most notably the patient health questionnaire (PHQ-9), which are usually administered in a clinic. These questionnaires evaluate symptoms that are stable over time. Ecological momentary assessment (EMA) methods, on the other hand, acquire patient ratings of symptoms in the context of their lives. Today’s smartphones allow us to also obtain objective contextual information, such as the GPS location, that may also be related to depression. Considering clinical PHQ-9 scores as ground truth, an interesting question is to what extent the EMA ratings and contextual sensor data can be used as potential predictors of depression. To answer this question, we obtained PHQ-9 scores from 18 participants with a variety of depressive symptoms in our lab, and then collected their EMA and GPS sensor data using their smartphones over a period of two weeks. We analyzed the relationship between GPS sensor features, EMA ratings, and the PHQ-9 scores. While we found a strong correlation between a number of sensor features extracted from the two-week period and the PHQ-9 scores, the other relationships remained non-significant. Our results suggest that depression is better evaluated using long-term sensor-based measurements than the momentary ratings of mental state or short-term sensor information.