Action recommendations review in community-based therapy and depression and anxiety outcomes: a machine learning approach

Table 2 Results from Generalized Estimating Equations modeling for the depression and anxiety datasets

Variable	PHQ9		GAD7
Variable	Coefficient (p-value)	Coefficient CI [0.025 0.975]	Coefficient (p-value)	Coefficient CI [0.025 0.975]
Intercept	-1.32 (0.561)	[-5.79, 3.14]	-1.77 (0.319)	[-5.25, 1.71]
Log (Review Percentage)	0.03 (0.659)	[-0.11, 0.17]	-0.01 (0.846)	[-0.14, 0.12]
Number of Dialogues/Session	0.03 (0.029*)	[0.00, 0.06]	0.01 (0.384)	[-0.01, 0.04]
Number of Action Recommendations/Previous Session	0.01 (0.819)	[-0.06, 0.07]	0.03 (0.408)	[-0.04, 0.09]
Client's Initial Score	0.06 (0.023*)	[0.01, 0.12]	0.13 (0.000*)	[0.08, 0.19]
Days Examined	0.02 (569)	[-0.06, 0.10]	0.02 (0.518)	[-0.04, 0.08]
Therapist Fixed Effects	–**	-	–**	-
Organization Fixed Effects	–**	-	–**	-

^*denotes p < 0.05. **There were 67 & 72 significant therapist fixed effects and 7 & 9 significant organization fixed effects found in the GEE models of PHQ-9 and GAD-7, respectively. Variables transformed: review percentage logged, score change cube-rooted (the dependent variable). All significant associations were verified via bootstrap resampling (1000 samples)

ISSN: 1471-244X