Background: Major depressive disorder (MDD) is a serious and prevalent disease with an unpredictable course. Smartphone technology is ubiquitous and can assist doctors by monitoring patients’ symptoms and behavioral patterns, and may eventually be useful in the prediction of depressive episode time courses. However, the extent to which the course of depression can be predicted with cell phone data remains unknown. Withdrawal, reduction in activity, and anhedonia are among common symptoms of MDD. These behavioral and motivational symptoms can influence mobility and result in lower location variation. Thus, the quantitative measurement of movement patterns extracted from location providers including WiFi, global positioning system (GPS), and cellular network may be useful for the prognostication of the course of MDD. This study assessed the extent to which objective location change measures could detect differences among individuals with MDD. We hypothesized that variability in location, specifically during weekends, would be negatively associated with depression scores. Methods: Between April 2016 and January 2017, patients with MDD (n=10) and healthy volunteers (n=3) completed an 8week protocol that involved tracking depressive symptoms and mobile phone usage. All patients were assessed at 2week intervals for depression symptoms as measured with the Hamilton Depression Rating Scale (HDRS). Movisens (an Android application) was used to passively capture GPS location changes. We calculated the total standard deviation of location data – average of latitude and longitude standard deviation – in the week prior to clinical assessment. To remove the effect of the time spent at home around nighttime and while sleeping, we constrained the hours to between 9 AM and 6 PM to estimate the location changes only throughout the day (weekly_location_std_9to6). We also calculated a similar feature, only considering the weekends (weekend_location_std_9to6). The latter feature would better represent the user behavior when not obliged to showup at work. We used linear mixedeffects models with random intercepts and slopes to assess the relationship between the HDRS total score and the two mentioned features. This model is an extension of the linear regression model for data that are collected and summarized in groups. Measurements from each user are considered as a group. Results: There was a statistically significant relationship between weekly_location_std_9to6 and HDRS total scores (coef. = 4.633, p = 0.040). A more significant relationship between weekend_location_std_9to6 and HDRS total scores was also observed (coef. = 7.525, p = 0.022). These results indicate that location variability is negatively associated with HDRS scores. Conclusion: As hypothesized, our analyses revealed that individuals with more variable movement patterns in the week, and specifically on the weekend, prior to their study visit had significantly less subjectively reported symptoms of MDD. These findings unveil the potential of passive smartphone sensor monitoring for extracting meaningful behavioral markers that are related to self-reported depression ratings. Analysis of commodity smartphone sensor data for at risk populations can assist clinicians to remotely monitor the status of the patient, and provide grounds for just-in-time interventions while reducing the burden for both patients and clinicians.