Table 1.
This table describes prior studies that have used predictive modeling to predict future depressive symptoms and have reported predictive outcome metrics among general samples or samples high in clinical depression, but not meeting criteria for other primary diagnoses (e.g., bipolar disorder).
| Study | Timescale | Design Summary | Result |
|---|---|---|---|
| [18] | 8 weeks | smartphone mobile sensing and support | 59.1–60.1% accuracy, 62.3–72.5% sensitivity, 47.3–60.8% specificity |
| [24] | 13 weeks | smartphone sensors and wearable sensors | R2 = 0.44, F1 = 0.77 |
| [11] | 3–6 weeks | mobile phone sensors and location data | AUC = 0.88 |
| [30] | 1–14 days later | location data from smartphone sensors | 71–74%-sensitivity, 78–80% specificity |
| [13] | daily | smartphone sensors | median area underthe curve [AUC] > 0.50) for 80.6% of persons |
| Current Study | Hourly | smartphone sensors | r = 0.587 across persons, r = 0.376 within persons |