Table 2.

Summary of effect sizes

Outcomes (number of studies)	Levels	Number of studies	Number of effect sizes	Hedges’ g	Standard Error	95% CI	P	Q	Total I2 (%)	Level 2 I2 (%)	Level 3 I2 (%)
Attendance
All outcomes (1)	3	30	65	0.49*	0.08	0.33, 0.64	<.001	299.88, p <.001	83.14	18.64	64.50
Primary outcomes (2)	2	24	24	0.53*	0.09	0.34, 0.72	<.001	114.06, p <.001	80.83
Primary, no outlier (3)a	2	23	23	0.49*	0.09	0.31, 0.67	<.001	88.10, p < .001	77.18
Only randomized studies (4)	3	16	34	0.52*	0.11	0.29, 0.74	<.001	213.88, p < .001	86.36	29.37	56.99
Only high-quality studies (5)	3	21	48	0.51*	0.09	0.33, 0.69	<.001	257.03, p < .001	84.16	24.03	60.13
Medication adherence
All outcomes (6)	3	6	14	0.95*	0.22	0.47, 1.44	<.001	65.01, p < .001	87.73	16.87	70.56
Primary outcomes (7)	2	6	6	0.89*	0.29	0.33, 1.45	.002	43.56, p < .001	89.48
Only high-quality studies (8)	3	5	12	0.83*	0.22	0.35, 1.31	.003	52.64, p < .001	86.57	22.69	63.89
Only randomized studies	All studies were randomized
Treatment Goal Completion
All outcomes (9)	3	5	7	0.61*	0.16	0.22, 0.99	.008	14.97, p = .021	60.55	60.55	0.00
Primary outcomes	Not analyzed – only 3 studies
Only randomized studies	Not analyzed – only 3 studies
Only high-quality studies	Not analyzed – only 3 studies
Symptom and functional outcomes
Attendance, survey (10)	3	6	16	0.10	0.05	0.00, 0.20	.060	6.93, p = .960	0.00
Attendance, toxicology (11)	3	6	10	0.23	0.10	−0.01, 0.46	.054	8.49, p = .486	36.15	0.00	36.15
Medication adherence, survey (12)	3	4	10	0.18	0.20	−0.27, 0.63	.393	23.26, p = .006	78.43	0.00	78.43
Goal completion (13)	3	4	14	−0.02	0.08	−0.19, 0.15	.791	16.07, p = .245	23.52	23.52	0.00

Note. CI = confidence intervals. Q = Cochran’s Q test for heterogeneity. Total I² = for 3-level models, the proportion of total variance attributed to within- and between-study variance. For 2-level models, the proportion of total variance attributed to between-study variance. Level 2 I² = proportion of variance attributed to within-study variance (effect sizes within studies) for 3-level models; Level 3 I² = proportion of variance attributed to between-study variance (effect sizes across studies) for 3-level models. All outcomes = models all non-overlapping outcomes provided in the studies; Primary outcomes = models used the one primary outcome measure common across all studies; Only randomized studies = models included only between-subjects designs in which participants were randomized to conditions. Survey = studies that assessed symptom or functional outcomes with self-reported or clinician-rated measures. Toxicology = studies that assessed substance use with urine toxicology screens. Toxicology screen analyses were not run for medication adherence or goal completion outcomes as there were too few studies (< 4) in these categories.

^aKidorf et al., 2009 emerged as an outlier in model 2

^*p < .01