Table 3.

Estimation results of two-part model on staff total training hours^* by intervention status

	RS		CEP		Part-1 logit			Part-2 among users
Variable	Estimated hours	95%CI	Estimated hours	95%CI	t	df	P	t	df	p
Primary outcome
Total hours in training	.19	.03-.36	2.35	1.09-3.62	3.90	94	<.001	4.67	94	<.001
Healthcare programsa	.35	.03-.66	1.88	.31-3.46	1.39	54	ns	3.08	54	.004
Social-community programsb	.1	−.03-.22	2.91	1.53-4.29	4.51	39	<.001	3.29	39	.003
Secondary outcomes-collaborative care component
Total hours in CBTc training	.12	.01-.22	.92	.24-1.59	2.39	94	.019	3.90	94	<.001
Total hours in CMd training	.04	0-.07	.79	.39-1.19	4.88	94	<.001	4.02	94	<.001
Total hours in mede training	.01	−.01-.03	.08	.03-.12	2.47	94	.015	.32	94	ns
Total hours in other typesf of training	.05	.01-.09	.58	.22-.94	2.81	94	.006	7.06	94	<.001

^*Two-part model, the first part is the probability of any positive hours of training using logistic regression, the second part estimates the total training hours, if positive, is the log of hours given any, using ordinary least squares controlled for community and accounted for the design effect of the cluster randomization; full sample analyses were also controlled for sector.

^ahealthcare programs: primary care, public health, mental health, substance abuse treatment programs.

^bsocial-community programs: community-based (social services agencies, faith-based, senior centers, parks and recreation, barber shops) and homeless serving programs. .

^cCBT= cognitive behavioral therapy.

^dCM=care management.

^emed training=antidepressant medication management and education.

^fOther types of training= administrator training in team building, grant writing for programs, active listening, “resiliency class”