Table 2 Multilevel model results predicting PHQ-8 from sensing data across shifting prediction windows.

From: Differential temporal utility of passively sensed smartphone features for depression and anxiety symptom prediction: a longitudinal cohort study

Predictor	Sensing predicting PHQ-8 with 2-week lag (R² = 0.049)			Sensing predicting PHQ-8 with 1-week lag (R² = 0.048)			Sensing predicting PHQ-8 with 0-week lag (R² = 0.053)
Predictor	Estimate	SE	p-value	Estimate	SE	p-value	Estimate	SE	p-value
Home duration—B	0.089	0.206	0.666	0.112	0.203	0.582	0.113	0.201	0.575
Home duration - W	0.219	0.087	0.012*	0.198	0.087	0.022*	0.183	0.091	0.045*
Circadian movement - B	−0.323	0.266	0.226	−0.189	0.273	0.490	−0.375	0.266	0.159
Circadian movement - W	0.034	0.062	0.577	−0.089	0.060	0.138	−0.131	0.062	0.035*
Location variability - B	−0.030	0.227	0.893	−0.149	0.230	0.516	−0.043	0.226	0.848
Location variability - W	−0.133	0.130	0.306	−0.129	0.129	0.316	−0.110	0.131	0.400
More frequent venues - B	−0.116	0.232	0.619	−0.092	0.232	0.693	−0.073	0.237	0.758
More frequent venues - W	−0.064	0.063	0.308	−0.185	0.062	0.003**	−0.168	0.063	0.007**
Less frequent venues - B	−0.345	0.213	0.106	−0.244	0.215	0.256	−0.295	0.220	0.179
Less frequent venues - W	−0.022	0.060	0.714	−0.069	0.058	0.234	−0.019	0.056	0.729
GPS variability and mobility - B	−0.464	0.258	0.073	−0.503	0.252	0.046*	−0.424	0.252	0.093
GPS variability and mobility - W	−0.052	0.050	0.302	−0.086	0.050	0.083	−0.037	0.050	0.458
Call and text communication - B	−0.079	0.206	0.702	−0.181	0.211	0.391	−0.197	0.211	0.350
Call and text communication - W	−0.077	0.081	0.340	−0.043	0.081	0.592	−0.049	0.079	0.534
App-based messaging - B	0.317	0.235	0.178	0.373	0.238	0.117	0.408	0.238	0.087
App-based messaging - W	0.059	0.067	0.385	0.115	0.066	0.083	0.162	0.066	0.015*
Social media - B	−0.289	0.211	0.171	−0.157	0.210	0.454	−0.241	0.206	0.243
Social media - W	−0.008	0.070	0.904	0.022	0.069	0.753	0.019	0.068	0.785
Screen-on time - B	0.503	0.208	0.016*	0.272	0.211	0.196	0.541	0.214	0.012*
Screen-on time - W	−0.009	0.050	0.857	0.000	0.047	0.995	0.037	0.047	0.424
Browser - B	0.219	0.206	0.287	0.308	0.206	0.135	0.367	0.205	0.075
Browser - W	−0.052	0.065	0.430	−0.032	0.064	0.617	0.024	0.064	0.709
Email - B	−0.107	0.199	0.592	−0.132	0.203	0.517	−0.102	0.200	0.611
Email - W	0.076	0.068	0.269	0.026	0.064	0.689	0.030	0.065	0.647
Game - B	0.006	0.205	0.978	−0.007	0.203	0.974	−0.044	0.203	0.826
Game - W	0.017	0.058	0.775	0.024	0.059	0.681	−0.028	0.059	0.634
Launcher - B	−0.596	0.223	0.008**	−0.525	0.222	0.018*	−0.653	0.224	0.004**
Launcher - W	−0.040	0.069	0.561	−0.111	0.070	0.115	−0.161	0.071	0.023*
Age	−0.603	0.174	0.001**	−0.573	0.174	0.001**	−0.581	0.173	0.001**
Male (vs0. female)	−0.521	0.169	0.002**	−0.542	0.170	0.001**	−0.563	0.169	0.001**
Urban (vs0. rural)	−0.311	0.168	0.065	−0.320	0.168	0.058	−0.307	0.168	0.068
Study week	−0.119	0.026	<0.001***	−0.107	0.026	<0.001***	−0.112	0.025	<0.001***
(Intercept)	9.458	0.163	0.000	9.467	0.163	0.000	9.460	0.163	0.000

Features highlighted in bold have at least one significant relationship to the outcome.
B between, W within.
*p < 0.05, **p < 0.01, ***p < 0.001.

Back to article page

Table 2 Multilevel model results predicting PHQ-8 from sensing data across shifting prediction windows.

Search

Quick links