Previously published literature for passively sensed behavioral measures [9 (link)] has suggested that the passive sensing data within a particular time frame need to be present in sufficient quantities to rule out cases where sensors were not collecting data in a continuous fashion for some reason (eg, criteria of 3 days out of a week and 18 unique hours per day [9 (link)]). Because human-smartphone interaction sensing will only produce data when participants are actively interacting with the phone, obtaining active data for 24 hours is unlikely; therefore, we relaxed the selection criteria for passive sensing data to at least 3 unique hours per day and at least 3 unique days within the 14-day window per DSM L1 measurement.
As a result, the sample that was entered into the multilevel modeling analysis had 142 participants and 984 measurements of depressive symptoms, where each assessment of depressive symptoms had at least 3 different days and 3 different hours of phone interaction data per day prior to the DSM L1 measurement (totaling 12,098 person-days and 196,673 person-hours).
As a result, the sample that was entered into the multilevel modeling analysis had 142 participants and 984 measurements of depressive symptoms, where each assessment of depressive symptoms had at least 3 different days and 3 different hours of phone interaction data per day prior to the DSM L1 measurement (totaling 12,098 person-days and 196,673 person-hours).
Free full text: Click here
