Table 5.
Notification typea | Estimated notification effect | Estimated effect moderation | ||
|
Relative risk (95% CI) | Multiplicative difference in effect (95% CI) | ||
|
No | Yes | Ratio (yes or no) | |
Habituation—“Did the user receive a notification the day before?” | ||||
|
Pooled | 3.645 (2.665-4.987) | 3.449 (2.761-4.308) | 0.946 (0.655-1.368) |
|
Standard | 3.935 (2.837-5.458) | 3.497 (2.739-4.465) | 0.889 (0.601-1.314) |
|
New message | 3.357 (2.381-4.732) | 3.401 (2.692-4.298) | 1.013 (0.677-1.517) |
Already engaged—“Did the user open the app between 8 PM and 9 PM the day before?” | ||||
|
Pooled | 3.620 (2.908-4.508) | 3.168 (2.314-4.336) | 0.875 (0.616-1.242) |
|
Standard | 3.720 (2.938-4.710) | 3.580 (2.542-5.043) | 0.962 (0.652-1.420) |
|
New message | 3.521 (2.794-4.439) | 2.820 (2.015-3.946) | 0.801 (0.548-1.168) |
aAll models are adjusted for the continuous variables of age; Alcohol Use Disorders Identification Test score; days since download; the categorical variables of sex and employment type; the time-varying variables “Did the users open the app before 8 PM that day?” and “Did the users open the app after 9 PM the day before?”; and the effect moderation variable of habituation.