Critical Dialogue and Capacity-Building Projects Reduced Alcohol and Substance Use in a Randomized Clinical Trial Among Formerly Incarcerated Men

Figures & data

Figure 1. Conceptual model: Community Wise components to reduce alcohol and substance use among formerly incarcerated men. Note: Mediating and moderating effects were not tested in this study.

Table 2. Inclusion and exclusion criteria.

Inclusion	Exclusion
Age >18 years	Gross cognitive impairment
Self-identified as male
Agreed to being recorded during CW^a group sessions	Severe unstable mental illness (e.g. untreated psychotic disorder, suicidality)
Living in Newark, NJ
Released from incarceration <4 years
Ability to read, write, and speak English
Willing to provide informed consent

a. CW, Community Wise.

Table 1. Experimental conditions and cost of delivery per group.

Facilitator	Experimental Condition	Core	Critical Dialogue (CD)	Quality of Life Wheel (QLW)	Capacity Building Project (CBP)	# sessions	Delivery cost*
Licensed facilitator	1: CD alone	Present	Present	Absent	Absent	9	$2,420.00
	2: QLW alone	Present	Absent	Present	Absent	6	$2,005.00
	3: CBP alone	Present	Absent	Absent	Present	6	$2,005.00
	4: CD + QLW	Present	Present	Present	Absent	12	$2,810.00
	5: CD + CBP	Present	Present	Absent	Present	12	$2,810.00
	6: QLW + CBP	Present	Absent	Present	Present	9	$2,420.00
	7: CD + QLW + CBP	Present	Present	Present	Present	15	$3,200.00
	8: Core only	Present	Absent	Absent	Absent	3	$1,640.00
Peer facilitator	9: CD alone	Present	Present	Absent	Absent	9	$1,925.00
	10: QLW alone	Present	Absent	Present	Absent	6	$1,700.00
	11: CBP alone	Present	Absent	Absent	Present	6	$1,700.00
	12: CD + QLW	Present	Present	Present	Absent	12	$2,150.00
	13: CD + CBP	Present	Present	Absent	Present	12	$2,150.00
	14: QLW + CBP	Present	Absent	Present	Present	9	$1,925.00
	15: CD + QLW + CBP	Present	Present	Present	Present	15	$2,375.00
	16: Core only	Present	Absent	Absent	Absent	3	$1,475.00

*Per 10 individuals.

Table 3. Per Protocol analysis: generalized linear mixed model examining effects of components on percentage of alcohol and substance use (%) over 5 months (n = 254).

Effect	Effect	t	P	Cohen’s d
CD alone	−1.85	−3.28	0.001	−0.83
QLW alone	−0.26	−0.44	0.660	−0.11
CBP alone	−0.56	−0.95	0.343	−0.24
CD + QLW	0.65	1.09	0.274	0.27
CD + CBP	0.16	0.27	0.790	0.07
QLW + CBP	0.52	0.91	0.364	0.23
CD + QLW + CBP	0.56	0.96	0.337	0.24
Licensed Facilitator (vs. Peer) alone	0.72	1.24	0.215	0.31
Licensed Facilitator + CD	−0.10	−0.17	0.866	−0.04
Licensed Facilitator + QLW	−0.57	−0.96	0.336	−0.24
Licensed Facilitator + CBP	0.10	0.18	0.859	0.04
Licensed Facilitator + CD + QLW	−0.08	−0.14	0.886	−0.04
Licensed Facilitator + CD + CBP	0.75	1.29	0.197	0.33
Licensed Facilitator + QLW + CBP	0.29	0.5	0.616	0.13
Licensed Facilitator + CD + QLW + CBP	0.57	0.98	0.325	0.25

Regression effects assume a log-normal distribution of the outcome and model the effect of the component relative to individuals who did not receive the component. Given the use of effect coding, the regression effects are calculated as 2 X parameter estimate from the regression model. Effects that meet the criteria for selection (p < 0.1 or absolute value of Cohen’s d ≥ 0.25) are bolded. Cohen’s d was considered a generalizable effect size estimate and calculated as the ratio of the effect to the pooled standard deviation. Components abbreviations: CD = Critical Dialogue, QLW = Quality of Life Wheel, CBP = Capacity Building Project.

Table 4. Generalized linear mixed model examining effects of components and facilitator type on percentage of alcohol and substance use (%) over 5 months following multiple imputation of missing data (n = 602).

Effect	Effect	t	P	Cohen’s d
CD alone	−1.15	−4.76	0.00	−0.03
QLW alone	0.13	0.51	0.61	0.16
CBP alone	−0.52	−2.08	0.04	−0.39
CD + QLW	0.34	1.39	0.16	0.07
CD + CBP	−0.60	−2.46	0.01	−0.14
QLW + CBP	0.57	2.38	0.02	0.39
CD + QLW + CBP	−0.10	−0.42	0.67	−0.07
Licensed Facilitator (vs. Peer) alone	0.47	1.95	0.05	0.03
Licensed Facilitator + CD	−0.05	−0.20	0.84	−0.22
Licensed Facilitator + QLW	−0.23	−0.96	0.34	−0.03
Licensed Facilitator + CBP	−0.04	−0.16	0.87	−0.03
Licensed Facilitator + CD + QLW	−0.33	−1.35	0.18	−0.22
Licensed Facilitator + CD + CBP	0.21	0.88	0.38	0.14
Licensed Facilitator + QLW + CBP	−0.05	−0.21	0.83	−0.03
Licensed Facilitator + CD + QLW + CBP	0.37	1.53	0.13	0.25

Multiple imputation with pattern-mixture model was applied over 5 imputations. The imputed data were analyzed as for the primary analysis using generalized linear mixed models for each imputation. Results were then integrated to examine model effects. Regression effects assume a log-normal distribution of the outcome and model the effect of the component relative to individuals who did not receive the component. Given the use of effect coding, the regression effects are calculated as 2 X parameter estimate from the regression model. Effects that meet the criteria for selection (p < 0.1 or absolute value of Cohen’s d ≥ 0.25) are bolded. Cohen’s d was considered a generalizable effect size estimate and calculated as the ratio of the effect to the pooled standard deviation. Components abbreviations: CD = Critical Dialogue, QLW = Quality of Life Wheel, CBP = Capacity Building Project.

Figure 2. Consort Diagram for Groups.Note. Bolded values = Pearson chi-square tests; p < 0.05. *Number of sessions varied from 3 to 15.

Table 5. Participants’ characteristics at baseline (N = 602).

Characteristic	N(%) / Mean (±SD)
Heterosexual	542 (90.0%)
Age	45.11 (11.30)
Race
Black/African American	485 (80.5%)
White	34 (5.6%)
Ethnicity
Hispanic	46 (7.6%)
Not Hispanic	556 (92.4%)
Household yearly income	$5,865.02 ($1,8273.9)
Religion
Christian	319 (53.0%)
Muslim/Islam	152 (25.4%)
Jewish	4 (0.7%)
None	104 (17.3%)
Never married	402 (66.8%)
Unemployed	468 (78.9%)
Under supervision (parole, drug court, halfway house)	498 (82.7%)
Number of months since release from last incarceration	14.34 (15.3)
Substance use (mean # of days in past 30 days)
Alcohol use	9.98 (11.6)
Cannabis use Cocaine use	5.63 (10.6) 4.45 (9.0)
Heroin use	9.39 (12.8)
Opiate use	1.13 (5.0)
Prevalence of current substance use
Alcohol	402 (66.8%)
Cannabis	211 (35.0%)
Cocaine	202 (33.6%)
Heroin	256 (42.5%)
Opiate	49 (8.1%)
Received other SUD treatment	263 (43.7%)
Received welfare support	189 (31.4%)
Homeless living in the street	81 (13.5%)
Lived in structured living situation	8 (1.3%)
Lived in homeless shelter	224 (37.2%)
Mental health
Traumatic Stress Scale (TSS)	4.93 (3.99)
Somatic Symptom Index (SSI)	1.51 (1.5)
Depressive Symptom Index (DSS)	3.27 (3.18)
Length of incarceration (in month)	9.6 (24.5)
Criminal Justice System Index (CJSI)	7.08 (17.15)
Readiness for abstinence (out of 100)	71.44 (30.84)

Table 6. Description of raw average percentage of alcohol and substance use (ASU) over time (N = 602).

	Average percentage of ASU in past 30 days (median [min, max])
	B	F1	F2	F3	F4	F5	B	F1	F2	F3	F4	F5
Total	20 (0,100)	13 (0,70)	17 (0,58)	16 (0,80)	18 (0,73)	17 (0,100)	--	--	--	--	--	--
By intervention component	Received						Did not Receive
Critical Dialogue (CD)	20 (0,100)	13 (0,53)	16 (0,53)	16 (0,51)	15 (0,52)	15 (0,80)	20 (0,100)	14 (0,70)	18 (0,58)	16 (0,80)	20 (0,73)	20 (0,100)
Quality of Life Wheel (QLW)	20 (0,80)	13 (0,70)	17 (0,53)	16 (0,51)	15 (0,54)	20 (0,100)	20 (0,100)	14 (0,57)	17 (0,58)	17 (0,80)	20 (0,73)	17 (0,80)
Capacity Building Project (CBP)	20 (0,100)	13 (0,57)	17 (0,58)	15 (0,57)	16 (0,60)	19 (0,100)	20 (0,100)	14 (0,70)	17 (0,51)	17 (0,80)	20 (0.73)	17 (0,80)
By facilitator type	Peer Facilitator						Licensed Facilitator
	20 (0,100)	13 (0,57)	15 (0,58)	15 (0,47)	17 (0,60)	17 (0,100)	19 (0,83)	14 (0,70)	20 (0,52)	18 (0,80)	20 (0,73)	18 (0,100)

B = baseline; F1-F5 = follow-up (post-baseline) months 1 to 5. No differences were observed by intervention component or facilitator.

Table 7. Modified intent to treat analysis: generalized linear mixed model examining effects of components on percentage of ASU (%) over 5 months (N = 602).

Effect	Effect	t	p	Cohen’s d
CD alone	−1.21	−3.96	<0.0001	−0.65
QLW alone	0.25	0.80	0.424	0.13
CBP alone	−0.68	−2.23	0.026	−0.36
CD + QLW	0.28	0.91	0.364	0.15
CD + CBP	−0.62	−2.02	0.043	−0.33
QLW + CBP	0.56	1.83	0.067	−0.29
CD + QLW + CBP	−0.09	−0.28	0.776	−0.05
Licensed Facilitator (vs. Peer) alone	0.49	1.62	0.105	0.26
Licensed Facilitator + CD	−0.19	−0.62	0.538	−0.10
Licensed Facilitator + QLW	−0.10	−0.34	0.735	−0.06
Licensed Facilitator + CBP	−0.13	−0.43	0.668	−0.07
Licensed Facilitator + CD + QLW	−0.24	−0.81	0.419	−0.13
Licensed Facilitator + CD + CBP	0.01	0.04	0.965	0.01
Licensed Facilitator + QLW + CBP	0.18	0.60	0.552	0.09
Licensed Facilitator + CD + QLW + CBP	0.25	0.81	0.416	0.13

Note. Regression effects assume a log-normal distribution of the outcome and model the effect of the component relative to individuals who did not receive the component. Given the use of effect coding, the regression effects are calculated as 2 X parameter estimate from the regression model. Effects that meet the criteria for selection (p < 0.1 or absolute value of Cohen’s d ≥ 0.25) are bolded. Cohen’s d was considered a generalizable effect size estimate and calculated as the ratio of the effect to the pooled standard deviation. Components abbreviations: CD = critical dialogue, QLW = Quality of Life Wheel, CBP = capacity-building project.

Data availability statement

Due to confidentiality and privacy concerns, the data will not be deposited for public access. Those interested in obtaining the data can contact the corresponding author to request the data and sign a data sharing agreement.