A Multiple Baseline Study of a Brief Alcohol Reduction and Family Engagement Intervention for Fathers in Kenya

Abstract

Objective:

To evaluate a lay provider-delivered, brief intervention to reduce problem drinking and related family consequences among men in Kenya. The 5-session intervention combines behavioral activation (BA) and motivational interviewing (MI). It integrates family-related material explicitly and addresses central cultural factors through gender transformative strategies.

Method:

A nonconcurrent multiple-baseline design was used. We initiated treatment with 9 men ages 30 to48 who were fathers and screened positive for problem drinking; the median Alcohol Use Identification Test score was 17 (harmful range). Participants were randomized to staggered start dates. We measured the primary outcome of weekly alcohol consumption 4 weeks before treatment, during treatment, and 4 weeks posttreatment using the Timeline Followback measure. Secondary outcomes were assessed using a pre–post assessment (1-month) of men’s depression symptoms, drinking- and family-related problem behavior, involvement with child, time with family, family functioning, relationship quality (child and partner), and harsh treatment of child and partner. Men, partners, and children (ages 8–17) reported on family outcomes.

Results:

Eight men completed treatment. Mixed-effects hurdle model analysis showed that alcohol use, both number of days drinking and amount consumed, significantly decreased during and after treatment. Odds of not drinking were 5.1 times higher posttreatment (95% CI [3.3, 7.9]). When men did drink posttreatment, they drank 50% less (95% CI [0.39, 0.65]). Wilcoxon signed-ranks test demonstrated pre–post improvements in depression symptoms and family related outcomes.

Conclusion:

Results provide preliminary evidence that a BA-MI intervention developed for lay providers may reduce alcohol use and improve family outcomes among men in Kenya.

Problem drinking is a critical global mental health problem that accounts for 9.6% of disability adjusted life years worldwide (Whiteford et al., 2013). Men are especially impacted with 7.6% of deaths per year attributable to alcohol use (Grittner, Kuntsche, Graham, & Bloomfield, 2012). Further, consumption above moderate levels and risky drinking patterns are associated with negative consequences across physical, psychological, and psychosocial domains (Rehm et al., 2009, 2010; Steel et al., 2014). Physically, problem drinking is linked to liver cirrhosis, heart failure, and certain cancers (Laonigro, Correale, Di Biase, & Altomare, 2009). Psychologically, it elevates risk for depression, anxiety, and externalizing problems, with men often experiencing more than one of these (Kessler et al., 2011).

Men’s problem drinking also often has a cascade of consequences on family systems and family members (Leonard & Eiden, 2007). For children, male caregivers’ drinking has been directly associated with youths’ drinking, both early onset and excessive patterns, as well as psychosocial, behavioral, and academic problems (Solis, Shadur, Burns, & Hussong, 2012). Within the parent–child dyad, paternal alcohol use increases the likelihood of decreased warmth and harsh parenting (Neger & Prinz, 2015). For couples, men’s drinking has been associated with intimate partner violence, marital conflict, poor communication, and poor coparenting in both high- and low- and middle-income countries (LMICs; Garcia-Moreno & Watts, 2011; Miller et al., 2014). When looking at the broader family system, studies have documented longitudinal pathways of negative consequences stemming from paternal drinking. For instance, a 2-year longitudinal study in the United States found fathers’ problem drinking predicted marital conflict a year later, which in turn predicted reduced parental warmth and increased psychological control, leading to internalizing and externalizing child problems (Keller, Cummings, Davies, & Mitchell, 2008).

Culture and Context

The culture and context in which drinking-related consequences occur are critical to consider given it can shape the nature and severity of outcomes. Aspects of culture, such as gender norms, can influence alcohol use and its impacts. Patriarchal norms, for instance, have been associated with men’s increased alcohol use, violent behavior, poorer mental health, reduced treatment seeking, and mortality (Barker, Ricardo, Nascimento, & World Health Organization, 2007; Garfield, Isacco, & Rogers, 2008). Associations between men’s drinking and intimate partner violence are also often stronger in patriarchal climates, worsening as inequality increases (Jewkes, Flood, & Lang, 2015). Contextually, poverty also exacerbates the impact of alcohol use. In a multicountry study, men with problem drinking in countries with lower gross national income experienced more consequences of drinking, such as social problems, compared with men in high-income countries (Grittner et al., 2012). Another study examining global data demonstrated that the link between units of alcohol consumed and burden of disease was strongest for those who were poorest, putting them at highest risk for alcohol-related consequences (Rehm et al., 2009).

Existing Interventions in LMICs

Overall, there is enormous unmet need for substance use treatment in LMICs (Patel et al., 2018). In LMICs, including Kenya, limited evidence-based treatments, scarcity of human resources, and geographic centralization of available services greatly limit treatment access for the majority of the population (Patel et al., 2018). In a previous qualitative study, men using alcohol did not describe any formal treatment options when asked about sources of assistance to reduce drinking; when asked about experiences with treatment, they exclusively described receiving informal help from family and community members (Patel et al., 2020). To address the growing treatment gap, there is a need to develop and evaluate interventions that can improve access through less centralized, community-based approaches. Steps have already been made in these efforts as researchers, in partnership with nongovernmental organizations and health systems, have evaluated a small number of interventions that include lay provider delivery. Examined lay provider treatments include those based in Motivational Interviewing (MI) that have effectively reduced alcohol use, and pharmacological interventions that have reduced withdrawal symptoms and relapse risk for patients with dependence (Patel et al., 2007; Nadkarni et al., 2017).

A next step is to identify interventions that address both alcohol use and family-related consequences. In high-income countries, this has been done. These include combined interventions, such as Alcoholic Behavioral Couples Therapy, that targets the intersection of drinking and couple-level conflict. This has shown efficacy for reducing alcohol use and improving relationship functioning compared with individual treatment (Powers, Vedel, & Emmelkamp, 2008). Further, some interventions have integrated behavioral parenting skills and restorative parenting to mitigate harmful practices associated with alcohol abuse with promising outcomes on substance use, couples functioning, intimate partner violence, and parenting (Lam, Fals-Stewart, & Kelley, 2009; Stover, 2015). However, in a systematic review of studies in LMICs, no interventions were identified that addressed both men’s drinking and its family-related consequences (Giusto & Puffer, 2018).

Adapting Evidence-Based Approaches for LMICs

To address the lack of community-based approaches for drinking and family consequences, we developed an intervention to reduce alcohol use and improve family directed behaviors for fathers with problem drinking. The program is designed specifically for lay providers in low-resource contexts. We used a multistepped development process beginning with qualitative research on contextually specific drivers of fathers’ alcohol use and family disengagement (Giusto, Ayuku, & Puffer, 2020). This was followed by: (a) a systematic review of alcohol use and family-related interventions (Giusto & Puffer, 2018) that informed the content and manual (Babor, Del Boca, & Bray, 2017; Daughters, Magidson, Lejuez, & Chen, 2016; Kato-Wallace, Aguayo, & Mendoza, 2013); (b) initial drafting; and (c) collaborative and iterative revisions with a Kenyan team before the study and during the first two cases. During this process, MI and behavioral activation (BA) were identified as optimal intervention components alongside discussion-based strategies for expanding conceptions of masculinity called gender transformative strategies (Dworkin, Treves-Kagan, & Lippman, 2013), which we refer to here and in our intervention as gender norm-transformative strategies for clarity (GTS).

MI was chosen based on the following: demonstrated efficacy reducing problem drinking with lay providers in multiple settings (World Health Organization [WHO], 2017b); effectiveness engaging hard-to-reach populations and increasing treatment engagement (Lawrence, Fulbrook, Somerset, & Schulz, 2017); a client-centered ethos that allows culturally rooted values and beliefs to dictate treatment content; and amenability to integration with other treatments (Balán, Lejuez, Hoffer, & Blanco, 2016; Stormshak, DeGarmo, Garbacz, McIntyre, & Caruthers, 2020). MI works to increase participants’ intrinsic motivation for change and increase ambivalence about problematic behaviors like drinking (Rollnick & Miller, 1995). Yet, MI alone may not be sufficient for changing behaviors beyond its direct target—in this case, drinking.

As such, BA was integrated to target both drinking and family-directed behaviors. BA was chosen given its efficacy reducing substance use and other problems, such as depression (Daughters et al., 2008); effectiveness across contexts (Collado, Castillo, Maero, Lejuez, & Macpherson, 2014); and emerging use for family directed behaviors (Benson-Flórez, Santiago-Rivera, & Nagy, 2017). Further, BA uses a straightforward structure, targets specific patterns of reinforcement, is anchored in participant values, and is conducive to addressing negative emotional triggers for drinking. BA, as applied to alcohol use, works to increase men’s engagement in healthy, value-driven activities to increase positive reinforcement for behaviors inconsistent with drinking (Kanter et al., 2010). MI-informed strategies were blended throughout BA content to maintain treatment engagement and build self-efficacy and motivation for completing activities outside of sessions (Balén et al., 2016).

Lastly, GTSs were selected for two reasons. First was the finding from the multistep development process that prescriptive gender roles, such as the man holding the responsibility to be the provider, influenced men’s drinking. Specifically, participants described that the inability to provide led to purposelessness/frustration that led to drinking. Second, GTS strategies have been shown to hold promise for reducing family violence and increasing family engagement in LMICs specifically (Dworkin et al., 2013; Van den Berg et al., 2013). Throughout the intervention, men are guided to consider family-related values, goals, and activities alongside self-related values, goals, and activities.

Present Study

The purpose of this study was to conduct the first evaluation of the intervention to generate preliminary data on its potential for reducing men’s alcohol use and improving family outcomes in a resource-scarce context. To evaluate our primary outcome—men’s alcohol use—we used a nonconcurrent multiple baseline design, which is designed to analyze intervention effects in small samples through numerous repeated measurements before, during, and after the intervention. This design allows us to examine within-subject change across conditions, replicated across different individuals using a number of essentially “mini” ABA designs through repeated measures. We hypothesized that fathers would show a decline in alcohol use, both amount consumed and days drinking, once starting the intervention. Second, we measured 1 month pre–post changes to explore preliminary findings on secondary outcomes of men’s mental health and family outcomes. For these, we gathered data from multiple reporters—men, their partners/coparents, and one of their children. We hypothesized small improvements across these outcomes given their dependence on men’s behavior changes over time and the short follow-up period.

Method

Setting

We conducted this study in Eldoret, Kenya—a lower middle-income country— in collaboration with Moi Teaching and Referral Hospital (MTRH) and AMPATH (the Academic Model Providing Access to Health Care), which consists of MTRH and a consortium of North American schools in partnership with the Kenyan Ministry of Health. Eldoret is the fifth largest town in the country located in the Rift Valley Province. In Kenya, the behavioral health care system is weak with care focusing on severe mental illness and limited outpatient services for common mental disorders (WHO, 2017a). Services for substance use disorders, especially those that do not meet levels of severe dependence, are extremely scarce. For instance, Eldoret, despite being the location of a large teaching and referral hospital, does not have public outpatient services related to substance use as part of the health care system—only one inpatient addiction rehabilitation center for severe dependence (National Authority for the Campaign Against Alcohol and Drug Abuse [NACADA], 2017). The cost and wait for rehabilitation services, as well as a lack of awareness in the community about services, further restricts access to care (Patel et al., 2020). The Duke Institutional Review Board and the Institutional Research and Ethics Committee at MTRH approved the study.

Recruitment Procedures

Two community leaders recruited the lay counselors. A larger group of leaders, as well as lay counselors themselves, then participated in recruiting participants. First, to find potential lay counselors, we asked leaders to identify adult men who were role models and fathers with no prior training in mental health. The goal was to identify a final group of three counselors for the pilot study. The leaders nominated 12 men, and 10 were interviewed; the other two were under age 18. Of these, six were then selected to be trained on recruitment procedures and the intervention; of the four not selected, three were not available due to other time commitments, and one exhibited a harsh communication style in role plays during the interview.

Participants were recruited by these six counselor trainees and five leaders recommended by the Kenyan research team and our community partners; leaders were individuals with designated community positions, longstanding community knowledge, and representatives of diverse sectors of the community (2 Christian pastors, 1 Muslim imam, 2 community policymakers/elders). These recruiters invited prospective clients to meet with study staff to learn more about the intervention and study. They were given a recruitment script and were trained on: (a) how to identify eligible participants based on eligibility and (b) nonjudgmental techniques for approaching them. Recruiters were instructed to “think of men who were fathers and whose drinking caused small and big problems at work or home, but not men who needed alcohol to wake up and fall asleep or who are drunk and sleep outside almost all the time, or who are home brewing.” This type of community-based approach was chosen and employed for two reasons. First, no formal behavioral health systems exist in the area that interact regularly with this population. Second, a previous qualitative study we conducted suggested men with problem drinking primarily sought help at community and family levels, making local leaders the best option for reaching those in need (Patel et al., 2020). For men who expressed interest in hearing more, a Kenyan research team member completed a formal eligibility assessment and informed consent procedures. If the man consented to participate, a Kenyan research team member then explained the study to his partner/coparent and child to obtain their informed consent/assent to participate in assessments. If the man had more than one child aged 8 to 17 years, we invited the oldest child to participate.

Participants

To be eligible, men had to: (a) be responsible for the care of at least one child aged 8 to 17 years; (b) engage in problematic drinking indicated by a score of 8 to 20 on the Alcohol Use Disorder Identification Test (AUDIT; Babor, Higgins-Biddle, Saunders, Monteiro, & World Health Organization, 2001) that assesses frequency and patterns of drinking over the past year; the AUDIT has been validated in Kenya (Saunders, Aasland, Babor, de la Fuente, & Grant, 1993); (c) report consuming alcohol within the past 2 months; and (d) agree to have one child and a partner/coparent participate in assessments, if relevant. We excluded men who: (a) lived in a home that brews alcohol; and (b) exhibited indicators of possible dependence, including visible symptoms of withdrawal assessed with an observational checklist (WHO, 2017b), history of hospitalization when trying to reduce drinking, and/or scoring greater than 20 on the AUDIT. For possible dependence, staff with mental health experience provided treatment referrals and brief psychoeducation about alcohol’s health consequences as informed by the Screening, Brief Intervention and Referral to Treatment guidelines previously used in LMICs (Babor et al., 2017).

Intervention

As described above, the intervention integrates behavioral activation (BA), motivational interviewing (MI), and gender norm-transformative strategies (GTS) to treat problem drinking and drinking-related family problems. The intervention is guided by a structured manual and includes five sessions delivered weekly lasting 60–90 min each. Figure A1 in the Appendix depicts the treatment’s theory of change. Table A1 in the Appendix provides the detailed intervention activities schedule. Further, a detailed description of intervention development and the treatment is described elsewhere (Giusto et al., 2020).

The first steps are MI-informed and include building readiness to change, engaging the client in treatment, and establishing commitment. Men then identify values for themselves and their families that could motivate change. Using these, counselors discuss the connections between events in one’s life, feelings, urges, and drinking behavior using a path metaphor, which sets the stage for BA elements. These foundational components are introduced in Session 1.

In the remaining four sessions, counselors employ BA strategies, including activity scheduling, to increase value-driven, rewarding behaviors to replace and reduce drinking behavior and increase family-directed behaviors. Homework is assigned to engage in these behaviors and to track both the behaviors and drinking. Throughout counselors use MI-integrated strategies to review homework, assess readiness for further change, and discuss instances when the client drank versus instances they resisted with goals of bolstering efficacy and reinforcing new behaviors. Each session also has some unique content. In Session 2, counselors introduce GTS content, including discussing men’s past relationship models, male role models, and what it means to “be a man.” In Sessions 3 and 4, they review refusal skills and, in Session 5, they discuss lessons learned and identify a person in their family or community to support them to maintain changes (i.e., relapse prevention). (The two first cases received only 4 sessions, not including refusal skills, as these were added based on their experiences).

Lay Counselor Training and Supervision

The six selected lay counselors completed a 10-day (80 hr) training in order to identify a final group of three counselors for the study. Training focused on general clinical skills (e.g., validation, active listening); MI, BA, and GTS theory and strategies; specific session content; and emergency and safety planning procedures. Activities included interactive didactics, modeling, role play, and participatory games to test skills. This training length and procedure was developed after reviewing training procedures for other task-shifted alcohol use interventions in LMICs, which were of similar or higher length and intensity (Barnett, Gonzalez, Miranda, Chavira, & Lau, 2018; Nadkarni et al., 2017; Papas et al., 2011). Training was co-led by a master’s-level doctoral student from the United States and a trainer from Kenya with a bachelor’s-level psychology degree; doctoral-level psychologists from the United States and Kenya provided consultation. All content was provided in Kiswahili directly or through live translation. Three final counselors and one alternate were then selected to deliver the intervention based on: (a) post-training clinical skills assessed during role plays with the ENhancing Common Therapeutic Factors Scale developed for use by lay counselors (ENACT; Kohrt et al., 2015), a midrange score of 2.5 or above were required; (b) knowledge evaluated with a written test (70% correct required); and (c) ability to use the manual, receptivity to feedback, and willingness to learn observed during training; this was tracked and discussed using trainers’ daily structured notes. The three counselors were then assigned one initial case followed by two to three additional. This allowed for continued learning during the first case. The low caseload is also part of the implementation model with community-based lay providers; this mirrors local helping practices (community members helping one another outside of the health system) and allows providers to manage counseling responsibilities with other life and work demands.

We used a multistage process for supervision since this was the first pilot of the intervention. Supervision followed a tiered approach including mental health professionals and local supervisors similar to procedures often used in evaluation of other task-shared interventions (Murray et al., 2011; Puffer, Friis-Healy, Giusto, Stafford, & Ayuku, 2019). Three local supervisors were selected from a pool of nine candidates who all held bachelor’s degrees in psychology. All nine received a 3-day training in MI, BA, GTS, and the intervention manual. The final three were then selected based on highest scores on a knowledge test about the intervention content (>75%) and best demonstrated general and MI-specific counseling competencies during structured role plays observed by the lead author. Key observed factors included appropriate use of reflection, affirmation, open-ended questions, validation, summary, and elicitation of change talk.

These local supervisors provided in-person supervision for lay counselors. For each counselor’s first case, the supervisor and counselor role played each session prior, and the supervisor provided live supervision (via video feed) during sessions with immediate feedback. For subsequent cases, supervision was provided in three steps before and after each session. First, after a session, supervisors debriefed with counselors (10–15 min). Second, local supervisors consulted with the master’s-level doctoral student (who was supervised by the clinical psychologists in the United States and Kenya) to confirm recommendations for next steps. Local supervisors then completed in-depth supervision sessions with counselors (~ 45 min) to plan the next session and practice skills as needed.

Research Design

We used a nonconcurrent multiple baseline design to examine the impact of the intervention on alcohol use: drinking frequency and amount consumed. In a multiple baseline design, participants begin the study as controls, complete baseline observations, and then cross over into the intervention. In the nonconcurrent variant of this design, participants are enrolled on a rolling basis and are randomized to cross over to the intervention at different times. This creates several independent ABA tests over time. The inferential logic is that: (a) the cause (intervention) precedes the emergence of the effect and (b) historical threats to internal validity are reduced by the staggered introduction of the treatment with randomized ordering to start dates. In other words, the design allows a cause-and-effect association to be established between treatment and outcomes within a small sample of participants serving as their own control (Barlow & Nock, 2009). In this study, men were randomized to staggered intervention start positions by enrollment order except for the first two men who were not randomized. We replaced two randomized participants with newly enrolled clients who assumed their starting position because these men no longer met inclusion criteria; one stopped drinking during the waiting period (to total 2 months not drinking) and the other was found to live in a home where alcohol was brewed.

Process Outcomes

Process evaluation measures included men’s attendance, intervention dosage (hours exposed), and indicators of treatment adherence and clinical competency. Attendance and dosage were tracked throughout and averaged across treatment completers. For adherence, an assessment tool was developed for the intervention and completed by a Kenyan research assistant based on session recordings and transcripts. The tool assessed completion (yes/no) of the steps in the manual, and, for completed steps, the quality with which each step was completed ranging from poor (1) to excellent (4). This yields: (a) Percentage of steps correctly completed and (b) Mean quality score across counselors. Adherence and quality of delivery were tracked and rated weekly to inform supervision. For clinical competency, counselors’ use of general clinical skills was measured using eight items from the ENACT scale developed for lay counselors in LMICs (Kohrt et al., 2015). Example items include verbal communication skills and rapport building. Items were rated from poor/not meeting patient needs (1) to excellent/optimally meeting patient needs (4); scores were summed separately for fidelity and competency. The Kenyan research assistant, who had been previously trained in ENACT coding, completed ratings using transcripts and audio recordings. Adherence and competency measures were rated for all counselors and all sessions.

Primary Outcome

Daily alcohol consumption: Men’s report.

The primary outcome was men’s self-reported alcohol consumption. We used an adapted Timeline Followback (TLFB; Sobell, Sobell, Leo, & Cancilla, 1988) instrument to estimate the volume of alcohol each participant consumed on a daily basis throughout the study period (nine recall periods covering 1 month prior to the intervention, during the intervention, and 1 month after the intervention). This is a self-reported, calendar-based measure validated by Papas and colleagues (2010) for this setting. Using the TLFB, we first asked clients to recall the number of standard commercial drinks they consumed during the recall period (typically 5 to 8 days). We used this count to calculate the volume of ethanol consumed using the standard conversion of 17.7 mL ethanol per drink (Sobell et al., 1988). We asked clients to recall their spending on locally made alcohol—changaa and busaa—using their own money or borrowed funds. We converted expenditures to mL ethanol using conversions of 0.57 mL per 1 KSh busaa and 0.67 mL per 1 KSh changaa (Papas et al., 2010) and calculated the daily consumption of local brew in mL ethanol. Summing these two values enabled us to estimate the volume of commercial and local drinks to get a total daily volume of alcohol consumed (mL ethanol). Additionally, a question was added to assess amount spent on commercial drinks. This total was summed with amount spent on locally made alcohol resulting in a total amount spent on alcohol per day. Data also yielded a dichotomous variable (yes/no) of whether a participant engaged in any drinking on a given day (“drinking days”).

Men’s alcohol consumption: Partner report.

As a check on men’s self-reported drinking, we developed a calendar-based instrument to assess partners’ reports of men’s drinking. The structure and prompts mirrored the TLFB used with men, but partners reported on whether the man drank each day: “Yes”, “No.” If yes, she indicated whether the man was drunk on that day (yes/no). This yielded a proportion of days the man drank and was drunk (yes/no).

Secondary Outcomes

Secondary outcomes were: depression, drinking-related problem behavior, missed family time, parental involvement, family functioning, household decision making, couple relationship quality, couple harsh treatment, father–child relationship quality, and child maltreatment. See Table 1 for measures descriptions.

Table 1.

Measures of Family Functioning Domains

Construct	Name	# of item	Scale	Example item	Reporter	Cronbach alpha
Depression	Patient Health Questionnaire–9**,a	9	0–3 [Not at all–Often]	Little interest or pleasure in doing things	M	0.91
Drinking-related problem behavior	Locally-developed items	17	0–4 [Never–4 or more times per week]	How often have you quarreled with children when drinking?	M	0.95
					P	0.94
					C	0.95
Missed family time	Locally-developed items	2	0–4 [Never–4 or more times per week]	How often have you missed dinner with your family due to drinking?	M	0.92
					P	0.28
					C	0.95
Paternal involvement	Alabama Parenting Questionnaire (subscale)*b		0–4 [Never—All the time]	How often does your partner help your child with homework?	M	0.78
					P	0.85
					C	0.92
Family functioning	Kenyan Family Functioning Scale†	25	1–10 [A little–A lot]	How often does your family have quarrels?	M	0.95
					P	0.96
					C	0.95
Household decision making	Decision makingc	3	0-You/Father		M	0.66
			1- Spouse/Mother		P	0.74
			2-Together		C	0.85
Couple RQ	Dyadic Adjustment Scale*,d; 11 LD items†	26	0–5 [All the time–None of the time]	How often do you and your spouse talk freely?	M	0.86
					P	0.94
Couple harsh interactions	Conflict Tactics Scale*,e	8	0–5 [Never–More than 8 times]	How many times did your husband threaten to hit you?	M	0.44
					P	0.72
Father–child RQ	Parental Rejection & Acceptance Questionnaire*,f 10 LD items+	35	1–4 [Almost never true–Almost always true]	You enjoy having your child around.	M	0.93
					P	0.98
					C	0.97
Child maltreatment	Discipline interview*,g	15	0–4 [Never–Almost always]	How often does your father beat you on your body?	M	0.84
					C	0.85

Note. M = men; P = partners; C = child; LD = locally developed; RQ = relationship quality.

^aKroenke & Spitzer, 2002.

^bShelton, Frick, & Wootton, 1996.

^cLeon & Foreit, 2009.

^dSpanier, 1976.

^eStraus, Hamby, & Warren, 2003.

^fRohner & Khaleque, 2005.

^gLansford et al., 2005.

^†locally-developed based on qualitative work and used in a prior family intervention study in Kenya and tested.

^*Prior to use, items were adapted and assessed for acceptability, relevance, comprehensibility, and completeness using an iterative process that began with lexical back translation and expert review and a final stage of focus grouping with local community members.

^**Validated in Kenya; Monahan et al., 2009.

Data Collection

Trained Kenyan research assistants administered surveys verbally in private locations in/or around families’ homes, recording data using pencil and paper. Intervention sessions were audio recorded, and a staff member translated and transcribed them verbatim into English.

Analysis

Alcohol consumption was operationalized as: (a) the proportion of days drinking (yes/no) and (b) mL ethanol consumed on days of use. Daily substance use count data typically include a large number of observations, stacked zeros, and skew (Atkins, Baldwin, Zheng, Gallop, & Neighbors, 2013), and our study was no exception. Our data includes a large number of observations per participant (M = 74.3, min = 57, max = 90; observation = single day), a clear stack of zeroes (65% of observed days), and skewed distribution of consumption (γ₁ 2.8, range: 6.7 mL–352 mL). To account for these characteristics, we fit a mixed-effects negative binomial hurdle model to examine the impact of treatment on alcohol use during the intervention and follow-up period. In this approach, an outcome is modeled in two parts: (a) a logistic regression that models the probability of having a nondrinking day (zero vs. nonzero consumption) and (b) a negative binomial regression for nonzero counts to model the amount of alcohol consumed on drinking days. Random intercept terms were included in both to account for variance of observations nested within a person. The model was fit in R v2.13.0 (R Core Team, 2013) using the glmmTMB package (Magnusson et al., 2017).

To examine case patterns, we describe proportion of days abstinent and total spending on alcohol during the baseline, intervention, and follow-up periods. For mL ethanol, we calculated the percent decrease in amount consumed from baseline to intervention and follow-up periods. We also identified how many men fell into a low-risk for alcohol use disorder category postintervention as defined by the National Institute of Alcohol Abuse and Alcoholism; low-risk equates to approximately four standard drinks per day (~ 68.8 mL; National Institute on Alcohol Abuse and Alcoholism [NIAAA], 2017).

To explore secondary outcomes, we aggregated data by period and conducted pre–post comparisons. We conducted Wilcoxon signed-ranks test and calculated effect sizes (r) by dividing the z-score by the square root of the observations (Field, Miles, & Field, 2012).

Results

Participants

Men’s mean age was 38 years (range: 30–48), and all except one were married. Children’s ages ranged from 8 to 15 years (M = 11). Self-reported median monthly income was KSh 1,500 (approximately $15 USD). The median AUDIT score at enrollment was 17 (harmful range). Table 2 shows baseline characteristics. Figure 1 presents the participant flow.

Table 2.

Participant Demographics

ID	Age	Education level	Monthly income*	Religion	Marital status	HH size	Child gender	Child age	AUDIT score	Employment	Completed treatment
1a	30	Vocational school	2,000	AIC	Married	4	Female	9	17	Casual worker	Yes
2a	39	Secondary	1,000	Muslim	Married	5	Female	8	18	Permanent job	Yes
3	38	Primary	1,000	Catholic	Married	6	Female	8	19	Casual worker	Yes
4	36	Primary	3,000	None	Married	6	Female	10	17	Casual worker	Yes
5	36	Did not answer	1,800	Catholic	Living with someone	4	Female	12	18	Casual worker	Yes
6	48	Secondary	1,500	Muslim	Married	6	Male	13	16	Casual worker: Farmer	Yes
7b	45	Primary	1,000	Catholic	Married	5	Male	15	9	Casual worker	Partial
8	41	Secondary	6,000	Catholic	Married	3	Male	15	11	Permanent job	Yes
9	34	Primary	500	Catholic	Married	5	Male	13	19	Casual worker	Yes
Grp mean	38.6	—	1,500 (median)	—	—	4.9		11.4	16.0		—

Note. Grp mean = Group mean; AIC = Africa Inland Church (Christian denomination); HH = household; AUDIT = Alcohol Use Disorder Identification Test.

^afirst two cases.

^bchild did not participate.

^*Reported in Kenyan Shillings.

Figure 1.

Participant flow. The men who ultimately participated in treatment were recruited by six different recruiters; two were counselors and four were community leaders. AUDIT = Alcohol Use Disorder Identification Test. * Participants were listed and identified by lay counselor trainees and community leaders.

Process Outcomes

The eight treatment completers came to all weekly sessions with a high rate of attendance (97.6%), with one person missing one session that was rescheduled. One man withdrew from treatment after three sessions (ID 7: Tables 2 and 4) due to reported scheduling difficulties, though endline interviews suggested other factors may have contributed: his partner’s ambivalence about treatment, living close to an alcohol brewing home, and his son’s circumcision (a celebratory period). Mean session length was 91 min; average total intervention exposure was 7.3 hr.

Table 4.

Father Report of Alcohol Consumption, Days Drinking, and Amount Spent on Alcohol Before, During, and After the Intervention

ID	Mean milliliters ethanol consumed per day					Proportion of days NOT drinking			Amount spent on alcohol: Total and by day (USD)
	Pre	During treatment	% change	Post	% change	Pre	During treatment	Post	Pre: Day	Post: Day	Total difference
1	56.70	58.30	3.00	11.16	−80.32	0.24	0.69	0.71	1.47	0.60	−0.87
2	113.96	55.18	−52.58	57.51	−49.53	0.00	0.28	0.20	1.69	1.11	−0.58
3	11.16	3.19	−71.41	2.21	−80.20	0.71	0.91	0.91	0.30	0.08	−0.22
4	75.92	32.00	−57.85	11.84	−84.40	0.29	0.67	0.59	2.50	1.07	−1.43
5	40.38	0.00	−100.00	0.00	−100.00	0.63	1.00	1.00	0.61	0.00	−0.61
6	21.88	2.46	−86.24	2.72	−88.76	0.27	0.93	0.90	0.28	0.18	−0.10
7a	25.80	33.92	31.47	18.76	−27.29	0.62	0.38	0.60	0.10	0.28	0.18
8	11.42	0.19	−98.33	0.00	−100.00	0.21	0.97	1.00	0.15	0.00	−0.15
9	61.61	13.46	−78.50	0.00	−100.00	0.10	0.64	1.00	1.92	0.00	−1.92
M	46.54	22.08	−56.72	11.58	−78.94	0.34	0.72	0.77	1.00	0.37	−0.63
SD	34.10	23.53	45.45	18.47	24.98	0.25	0.26	0.27	0.90	0.45	0.68

Note. mL = milliliters; M = Mean; SD = standard deviation.

^aParticipant 7 did not complete treatment.

Treatment adherence, averaged across sessions and counselors, was 93.8% (SD = 1.62; range = 82–98%) with overall high adherence to treatment content by counselor and session; only one session by one counselor fell below 90% adherence. Quality of step completion, averaged across sessions and counselors, was 3.67, reflecting most scores were in the good to excellent range (SD = 0.18). Clinical competency scores fell primarily in the “moderately” to “optimally” meeting needs range (M = 3.2, SD = 0.27; Scale = 0–4).

Primary Outcome: Impact on Daily Alcohol Consumption

Mixed-effects hurdle model.

Table 3 presents results of the fitted mixed-effects hurdle model. In the logit model, there was a treatment effect on the odds of not drinking on any given day. The odds of not drinking were 4.2 times higher during the intervention compared to the baseline period (95% CI [2.8, 6.4]). After the intervention, they were 5.1 times higher (95% CI [3.3, 7.9]). When examining covariates in the count model, there was a significant effect of the intervention on the amount of alcohol consumed on drinking days during the intervention and follow-up. During the intervention, on drinking days, men drank 81% less than they did prior to treatment (95% CI [0.66, 0.99]). After the intervention, men drank 50% less on drinking days compared with preintervention amounts (95% CI [0.39, 0.65]). Partners also reported a decrease in men’s drinking from 54% of days during baseline to 34% of days during follow-up (r = 0.63).

Table 3.

Alcohol Changes During and After Treatment: Mixed-Effects Hurdle Model

Variable	ORa	B	p value	SE
	Odds of NOT drinking alcohol on a given day (yes/no) Logit submodel
Intercept (Baseline)	0.73	−0.31	.02	0.14
Treatment	4.18	1.43	<.001	0.21
Follow-up	5.10	1.63	<.001	0.22
Variable	RRa	B	p value	SE
	Volume of mL ethanol consumed on days drinking Negative binomial submodel
Intercept (Baseline)	73.70	4.31	<.01	0.15
Treatment	0.81	−0.21	.04	0.10
Follow-up	0.50	−0.69	<.01	0.13

Note. OR = odds ratio; RR = rate ratio; B = coefficient on linear-predictor scale (i.e., log of outcome).

^aRRs are unit-specific (or conditional) estimated, as opposed to population average.

Exploring individual change patterns.

Table 4 presents each man’s average amount of mL ethanol consumed per day, proportion of days abstinent, and the average amount spent on alcohol pre- and postintervention. Figure 2 visually depicts consumption by time period. All men reduced the amount of alcohol (mL ethanol) consumed during the intervention except one completer (ID: 2) and the noncompleter. During follow-up, all completers reported reductions in amount consumed. For days drinking, all completers reported increases in the number of nondrinking days during the intervention. During follow-up (M = 23.7 days), all completers reported continued increases in nondrinking days. Three men reported complete abstinence after the intervention, and all but one reported drinking less than 50% of days both during and after the intervention. The exception was Participant 2; while he experienced a pattern of reduction, he reported drinking every day at baseline and drinking around 75% of days during and after treatment.

Figure 2.

Treatment completing participants’ milliliters of ethanol consumed across time periods. mL = milliliters; the x-axis refers to the number of observations collected per person divided by baseline, treatment, and follow periods. (Some participants completed longer baseline repeated assessments than others (i.e., more than 30 observations)).

The average number of drinks consumed per day prior to treatment across men was 2.7, and the posttreatment average decreased to 0.67 drinks per day. Severity of drinking at baseline is a bit difficult to categorize. The AUDIT places all men in the “harmful drinking” range based on overall drinking patterns—frequency, amounts, and consequences. Based on the NIAAA (2017) categories, three men’s average daily drinking fell into a high-risk range (i.e., more than 4 drinks per day) and six men’s into low-risk (1–4 drinks per day), two of which were drinking above recommended levels (2–3 drinks per day). After treatment, all men except one were drinking less than two drinks per day, falling in the recommended range; the one remaining decreased from 6.6 drinks per day to 3.4 at endline, falling in the low-risk range. All completers also reported spending less on alcohol during and after treatment. Among them, the average amount spent per day at baseline was $1.00 USD (KSh 103.91) compared to $0.37 USD (KSh 38.5) during follow-up.

Secondary Outcomes

Appendix tables A2, A3, and A4 show Wilcoxon’s results and effects across secondary outcomes and reporters.

Depression symptoms.

Eight of nine men’s symptoms decreased from mild symptoms to none/minimal. One father fell in the moderate range at baseline and remained in that range after treatment, though with some symptom improvement. At the group level, symptoms improved with an effect size of 0.61.

Drinking- and family-related problem behavior.

Wilcoxon’s tests indicated reductions in problematic behaviors associated with drinking (e.g., men fighting with family members when drunk) across all reporters. Children reported the largest improvement (r = 0.64), followed by the men (r = 0.61) and their partners (r = 0.59). Partners’ and children’s responses reflected that, after the intervention, most problematic behaviors were occurring very infrequently or not at all.

Missed family time.

There were pre–post effects observed for all reporters (men r = 0.51; children r = 1.0; and partners r = 0.21). According to all reporters, after treatment, men rarely missed time with family due to drinking.

Paternal involvement.

On paternal involvement—men’s help and engagement in their children’s lives—there were pre–post improvements observed for all reporters (men r = 0.67; children r = 0.67; and partners r = 0.39).

General family functioning.

Mean scores for overall family functioning suggested improvement across all reporters (children r = 0.62; men r = 0.45; partners r = 0.25). Regarding joint decision making, children and men’s means showed a slight improvement while partner means did not change.

Dyadic couples’ and father–child relationships.

Couples’ relationship quality improved according to men (r = 0.63) and their partners (r = 0.63). Regarding harsh marital treatment, men reported a decrease (r = 0.29), while partners’ mean scores did not show change (r = 0.19). Importantly, baseline reports of harsh treatment were low for both men and partners. For father–child relationship quality, improvements were seen across all reporters (men r = 0.50; partners r = 0.54; children r = 0.76). For harsh discipline, children reported a reduction (r = 0.67) but not men or partners who reported very low use of harsh discipline at baseline; children’s baseline scores were also low though higher than the adults’ scores.

Discussion

This study addresses a gap in the literature by testing an intervention that targets both alcohol use and family related problems in a low-resource community setting with lay counselors in Kenya (Giusto & Puffer, 2018). It also adds to the evidence supporting alcohol reduction interventions in LMICs by examining an intervention designed for men that combines behavioral activation (BA), motivational interviewing (MI), and gender norm-transformative strategies (GTS). This was the first evaluation of this intervention to gather preliminary data. For our primary outcome of alcohol use, we maximized learning through a multiple baseline single case series design that takes a rigorous, repeated measures approach to be able to make causal inferences between treatment and outcome with a small sample. However, for secondary outcomes—mental health and family related outcomes—we were only able to conduct an uncontrolled, exploratory pre-post evaluation; meaning that these results must be interpreted with greater caution.

Results of the multiple baseline study suggest that the intervention is helpful for reducing men’s alcohol use—both in terms of how much and how often they drink. Post intervention, participants were 5.1 times more likely to abstain from alcohol on a given day after treatment compared with baseline. On occasions participants did drink after the intervention, they drank 50% less than the amount consumed during baseline. Partners’ reports of men’s percent days drinking supported these results. These findings add to the low-resource setting literature demonstrating that MI can be effective for reducing alcohol use when delivered by lay providers (Nadkarni et al., 2017; Patel et al., 2017). Our results also begin to expand evidence for the use of BA to treat substance use to LMICs and community settings. While BA is less commonly used, our preliminary findings are consistent with the work of Daughters and colleagues (2008, 2018) in the United States using BA to address illicit drug use; they demonstrated improved depression and reward value of nondrug activities (Daughters et al., 2008) and later demonstrated increased abstinence rates among substance-dependent users in residential treatment (Daughters et al., 2018). Effects on drinking in this study were comparable in size to those documented for other alcohol use interventions in LMICs tested in randomized controlled trials that showed large effects on the likelihood of not drinking, although there were differences in the measurement of drinking. These trials evaluated brief blended MI plus problem solving therapy (Sorsdahl et al., 2015), motivational enhancement-based (MEB) therapy (Nadkarni et al., 2017), and eight-session cognitive–behavioral therapy (Papas et al., 2011). For instance, the average percent days abstinent postintervention (76.8%) in our study is similar to the 69.4% abstinence rate in a trial examining a four-session, lay provider-delivered MEB treatment in India (Nadkarni et al., 2017).

Reduced spending on alcohol—a secondary outcome—was another finding particularly relevant for low-resource settings. While spending was self-reported and measured only pre–post in this study, these preliminary findings should prompt further study given its importance. Reduced spending has multiple implications for the family such that men may be able to provide more financially, possibly leading to more positive relationship and mental health outcomes. In populations where poverty is a pervasive stressor, one related future direction could also be pairing an intervention like this with livelihoods programming (Lund, 2014). This would be consistent with literature suggesting economic disenfranchisement may be an important intervention target to broaden ideas of masculinity (Gibbs, Jewkes, Sikweyiya, & Willan, 2015).

Beyond alcohol-related outcomes, results also showed pre–post improvements in secondary family-related outcomes, most of which improved according to multiple reporters—the men, their partners, and one of their children. While findings are very preliminary given the small sample and uncontrolled design, seeing family-level improvement across reporters is encouraging. One important question for this study was whether an intervention with only fathers had potential to address any family-related alcohol use consequences. Our findings suggest that it could, showing positive movement on: men’s family-related problem behaviors attributed to drinking, parental involvement, time with family, and overall family functioning. The directions and magnitudes of change across outcomes support moving forward with a larger efficacy trial.

These signals of change at the family level are particularly promising given one motivation for this study was to consider the intervention as a possible adjunct to family-based interventions designed to address pervasive family problems and promote child well-being (Puffer et al., 2019). Using this intervention with men prior to initiating other family based treatments may help reduce substance use-related barriers to men’s participation and build motivation to engage in the process of family intervention. Exploring this could fill a gap in family research given fathers are often absent from parenting and family treatments, in part due to engagement difficulties (Panter-Brick et al., 2014). Their inclusion has been called a potential “game changer” for improving family-level and family member outcomes (Panter-Brick et al., 2014).

Men’s depression symptom scores also suggested improvements pre- to posttreatment. On average, men reported mild symptoms at baseline and minimal to none at endline. Although depression was not explicitly targeted, the intervention did target negative emotions tied to drinking urges, and BA was chosen in part to address men’s feelings of purposelessness. Results, while based only pre–post data, are consistent with the evidence of BA for depression globally (Cuijpers, van Straten, & Warmerdam, 2007; Patel et al., 2017). Further, reductions in even mild depression symptoms alongside changes in drinking suggest BA may be functioning in part by increasing positive reinforcement of healthy, valued behaviors and targeting a shared underlying vulnerability of lower experienced reinforcement (Lewinsohn, 1974; Magidson, Robustelli, Seitz-Brown, & Whisman, 2017). Targeting reinforcement through BA may be a parsimonious and cross-cutting approach for targeting drinking and mental health in resource restricted settings.

Lastly, it is important to consider the treatment’s sustainability and scalability even at this early stage. How to sustain and scale treatments while maintaining adequate adherence are critical challenges facing the field of global mental health and implementation science (Patel et al., 2018). At the outset of this study, an effort was made to mirror natural help-seeking processes by using community-based recruitment and identification of counselors and participants. The promise of such an approach was supported by counselor and participant engagement and qualitative findings that counselors expressed interest in carrying more cases. However, major challenges to scalability in the field are training and supervision. In the broader literature on task-shifted mental health services, initial training can vary widely depending on provider type and treatment (Barnett et al., 2018; Van Ginneken et al., 2013). For instance, training can range from 6 hours for a very brief three-session, 15-min brief treatment delivered by nurses in Thailand for problem drinking to 6 months for generalist training in common mental health disorders among community health workers in Malawi (Barnett et al., 2018; Caulfield, Vatansever, Lambert, & Van Bortel, 2019; Van Ginneken et al., 2013). In trials of treatments most similar to ours—task-shifted treatments for alcohol use in India and Kenya (Nadkarni et al., 2017; Papas et al., 2011)—our training and supervision were comparable, if not less intensive. Of these, one trial testing a four-session MI-based program included a 6-month internship and intensive 2-week training (Nadkarni et al., 2017), while the other, testing CBT for alcohol use and HIV adherence, included 175–300 training hours (Papas et al., 2011). This is compared to the 80-hr training in this study. That said, the fact that all of these require significant inputs reflects the broader need for implementation research to identify how to optimize these processes without sacrificing quality, adherence, and support for lay providers. Hybrid-effectiveness designs will allow us to continue studying efficacy while also examining implementation questions; future studies could test training and supervision thresholds and formats, including creative uses of technology; evaluate delivery within other settings; and measure cost effectiveness (Curran, Bauer, Mittman, Pyne, & Stetler, 2012).

Several important limitations should be considered when interpreting the results of this study. First, while a single-case series design is appropriate for an initial examination of a novel intervention, the small sample clearly limits generalizability and, though it controls for time, the design does not account for other factors that may have driven change (e.g., therapeutic alliance). Related to generalizability, fathers who chose to participate in treatment may have had higher readiness to change than a randomly selected sample. Future research could explore how initial readiness before treatment may affect outcomes, as the intervention may be most effective for those at some level of readiness. It is also important to consider that recruiter selection may have biased the sample, though the recruiters themselves were chosen to represent multiple sectors of the community. We also only had partial randomization, with two participants replaced, which deviated from our initial design. However, for this design, the rationale behind randomization was to prevent readiness bias (i.e., earlier participants having higher readiness to change); this would have been important if we had done a visualization analysis. Ultimately, though, we adopted an analysis approach that was not dependent on the randomization. The 1-month postassessment also limits the ability to detect longer-term outcomes, and the number of observations during each time period varied slightly across participants, possibly influencing between-case comparisons. For secondary outcomes, no causal inferences can be made given the pre–post design with no control. Men and therapists also were not blind to study goals as they intersected with intervention goals. Regarding measures, multiple reporters was a strength, though biases in survey data are still present. In future work, biometric measures could strengthen conclusions.

This study provides preliminary evidence for a brief, lay counselor-delivered intervention addressing alcohol use and family-related problems with applicability to Kenya as well as other low-resource settings in which mental health care is scarce. Results prompt the pursuit of a larger trial to establish treatment efficacy and acceptability and feasibility in a more diverse sample. This will also afford the opportunity to explore mechanisms of action. If effective, this intervention could provide one approach for improving multiple outcomes simultaneously—men’s alcohol use, mental health symptoms, and interactions with families. Improvements in these outcomes could promote more protective environments for children in which they are more likely to thrive. Research to assess these outcomes and to maximize effectiveness and efficiency will benefit from multiple trial designs, including those that allow a focus on optimizing intervention components (Collins, Murphy, & Strecher, 2007) and those that allow examination of clinical and implementation outcomes simultaneously (Curran et al., 2012).

What is the public health significance of this article?

There are few interventions that aim to reduce men’s alcohol use and the problems it causes in families. This is especially true in low-resource settings, including Kenya, that do not have adequate mental health services. In this study, a brief intervention delivered by nonprofessionals reduced alcohol use among fathers in Kenya and led to promising changes in family relationships.

Appendix

Figure A1.

General treatment theory of change.

Table A1.

Intervention Schedule

Session	Treatment content
1	Pros and cons of drinking: Self & family Importance scaling question
MI	Readiness scaling question Confirm commitment program Vision for self Vision for family Value selection: Self & family (value cards) Program invitation Path introduction & discussion (treatment rationale) Homework: Track drinking. Notice positive nondrinking activities.
2	Drinking and mental health assessment
BA + MI	Homework and path review: Drinking event and resisted drinking event
GTS*	Client’s relationship with father: Positive & negative* Client’s parents’ relationship: Positive & negative* Conceptions of masculinity: Helpful & hurtful* Value selection: Self & family Activity scheduling Homework: Complete activities. Track drinking.
3	Drinking and mental health assessment
BA + MI	Homework and path review: Drinking event and resisted drinking event Refusal skills
	Value selection: Self & family Activity scheduling Homework: Complete activities. Track drinking.
4	Drinking and mental health assessment
BA + MI	Homework and path review: Drinking event and resisted drinking event Refusal skills
	Value selection: Self & family Activity scheduling Homework: Complete activities. Track drinking. Drinking and mental health assessment Homework and path review: Drinking event and resisted drinking event
5	Value selection: Self & camily (client-directed)
BA + MI	Activity scheduling (client-directed) Staying on the path: People and strategies to continue change (relapse prevention) Lessons learned: Treatment review Graduate

Note. From “An intervention to reduce alcohol use and improve family engagement for fathers in low-resource settings: Development and feasibility testing in Kenya,” by Giusto, A., Ayuku, D., & Puffer, E., 2020, doi: 10.31219/osf.io/wbyct. Copyright CC0 1.0 Universal. Adapted with permission. MI = motivational interviewing; BA = behavioral activation; GTS = gender norm-transformative strategies; Underlined activities are standalone components.

^*GTS elements.

Table A2.

Group-Level Outcomes: Male Caregiver Report [Treatment Completers]

Measurement (valence of higher scores)	Score range	Pre	Post	Wilcoxon z	ES (r)
Primary outcomes
Alcohol use: mL ethanol (−)	0−Any #	46.99	8.69	−2.52**	0.63
Alcohol use: % Days drinking (−)	0.00–1.00	0.58	0.21	−2.52**	0.63
Secondary outcomes
Father family-directed behavior
Missed family time (−)	0–4	1.50	0.50	−1.76**	0.51
Drinking-related problem behavior: Family (−)	0–4	1.49	0.54	−2.29**	0.61
Father parental involvement (+)	0–4	2.04	2.70	1.91**	0.67
Family indicators
Family functioning (−)	1–10	3.82	2.92	−1.70*	0.45
Household decision making (+)	0–3	1.29	1.50	1.41	0.38
Couple
Relationship quality (−)	0–5	1.91	1.00	−2.37**	0.63
Harsh treatment (−)	0–5	0.35	0.06	−1.10	0.29
Child
Relationship quality (+)	1–4	3.35	3.70	2.10*	0.50
Harsh discipline (−)	0–4	0.33	0.26	0.00	0.00
Individual
Mental health (PHQ-9; −)	0–27	7.00	1.43	−2.30**	0.61
Child mental health (−)	0–38	0.50	0.09	−2.37**	0.59

Note. ES = effect size; PHQ = Patient Health Questionnaire.

^*p < .10.

^**p < .05.

Table A3.

Group-Level Secondary Outcomes: Female Caregiver Report [Treatment Completers]

Measurement (valence of higher scores)	Score range	Pre	Post	Wilcoxon z	ES (r)
Primary outcome check
Alcohol use: MCG % days drunk (−)	0.00–1.00	0.54	0.34	−2.37**	0.63
Father family-directed behavior
Missed family time (−)	0–4	1.08	0.81	−0.77	0.21
Drinking-related problem behavior: Family (−)	0–4	1.87	0.98	−2.20**	0.59
Father parental involvement (+)	0–4	1.81	2.20	1.16	0.39
Family indicators
Family functioning (−)	1–10	4.01	3.30	−0.98	0.25
Household decision making (+)	0–3	1.66	1.66	0.29	0.08
Couple
Relationship quality (−)	0–5	2.51	1.44	−2.37**	0.63
Harsh treatment (−)	0–5	0.45	0.51	0.71	0.19
Father–child
Relationship quality (+)	1–4	2.77	3.35	2.03**	0.54
Harsh treatment (1 item)	0–4	0.24	0.24	—a	—
Individual
Mental health: Self (PHQ-9; −)	0–27	7.29	1.90	−2.31**	0.62
Child mental health (−)	0–38	4.12	1.43	−1.96**	0.53

Note. ES = effect size; PHQ = Patient Health Questionnaire; MCG = male caregiver.

^aAll pairwise differences = 0, so unable to complete test.

^**p < .05.

Table A4.

Group-Level Secondary Outcomes: Child Report [Treatment Completers]

Measurement (valence of higher scores)	Score range	Pre	Post	Wilcoxon z	ES (r)
Father family-directed behavior
Missed family time (−)	0–4	1.75	0.00	−2.02**	1.01
Drinking-related problem behavior: Family (−)	0–4	1.90	0.50	−2.02**	0.64
Father parental involvement (+)	0–4	1.30	2.40	1.83*	0.65
Family indicators
Family functioning (−)	1–10	4.54	1.97	−1.76*	0.62
Household decision making (+)	0–3	1.40	2.20	1.70*	0.53
Parent–child
Relationship quality (+)	1–4	2.58	3.60	2.02**	0.76
Harsh treatment (−)	0–4	0.68	0.23	−1.91*	0.67
Individual
Child mental health (−)	0–38	3.80	2.6	−0.96	0.34

Note. ES = effect size.

^*p < .10.

^**p < .05.