Demographic characteristics of participants.

Background
Geriatric depression often goes unnoticed due to recall bias and overlapping symptoms with normal aging by using questionnaire screening tool. Consequently, this study aims to explore the associations between passive sensing parameters, such as physical activity and sleep characteristics collected from smart devices, and depression screening scores, aiming to validate its efficacy as a detection tool for the Thai elderly.
Methods
The prospective cohort study was conducted from July to September 2023. One hundred and seventy-seven elderly individuals were purposefully selected from the main districts of each province across five regions in Thailand. Inclusion criteria required participants to be aged 60 years or older, socially active, and free from diagnoses of cognitive impairment and mental health disorders. Participants were required to wear an Actigraph wGT3X-BT collecting data on physical activity and sleep characteristics. The Patient Health Questionnaire (PHQ-9) was used to screen for depression every two weeks. Univariate and multivariate regression analyses were performed to identify association between passive sensing parameters and PHQ-9.
Results
There is an association between physical activity parameters and depression score. A One- unit increase in Vector Magnitude in Counts per Minute (VM CPM) and step count, PHQ-9 score would be statistically significant reduced by 0.001 and 0.00007 to 0.00008 score over both two-week periods (p-value <0.1). Only Wakefulness After Sleep Onset (WASO) of all sleep variables showed an association with PHQ-9 in univariate analysis but it did not show further relationship after adjusting with baseline PHQ-9 score and other confounders. In addition, participants who lived outside Bangkok and being younger were more likely to have lower PHQ-9 score.
Conclusion
There is a potential to apply passive sensing data in mental health issues in Thailand. However, more evidence is needed to ensure the accuracy of sensor detection and appropriate algorithm to predict depression. Moreover, the implication of passive sensing data from smart devices supporting public health policy should be explored.