Table 3.

Results from multivariate analysis of variance looking at usage behaviors based on cluster membership of 861 myCompass users who completed the symptom screener at sign-up and at 28 days.

Usage variablesa (count data)	Moderates (n=479), mean (SE)	Trackers (n=224), mean (SE)	Super users (n=158), mean (SE)	F test (df)	P value
Modules completed	0.18 (0.48)b	0.23 (0.40)b	2.08 (0.62)c	651.68 (2,858)	<.001
Logins	4.92 (0.33)b	19.02 (0.27)c	13.25 (0.43)d	960.89 (2,858)	<.001
Tracking	3.44 (0.34)b	17.96 (0.28)c	10.44 (0.44)d	930.79 (2,858)	<.001
Modules started	0.85 (0.07)b	0.77 (0.06)b	3.16 (0.09)c	443.56 (2,858)	<.001
Notifications received	5.22 (0.31)b	9.05 (0.25)c	6.28 (0.39)d	78.79 (2,858)	<.001

^aInput variables are sorted by overall importance in this analysis, from highest to lowest.

^b,c,dDiffering superscripts indicate a significant difference at P<.05 for Bonferroni corrected multiple comparisons.