BENEFITS TO CHILDREN WHO PARTICIPATE IN FAMILY THERAPY WITH THEIR SUBSTANCE-USING MOTHER

Abstract

It is rare that family members other than the identified patient are followed over time in studies of therapy effectiveness. Family therapy is believed to be effective because it targets processes within the system that maintain symptoms. If these processes are changed, then all family members can benefit. Using a sample of 183 mother–child dyads from a study comparing family therapy for adult substance use versus an attention control, change in child’s substance use (tobacco, alcohol, and marijuana) was estimated. Children who participated in family therapy with their mothers showed greater decreases in alcohol and tobacco use and were less likely to begin using compared to children whose mothers participated in the attention control condition.

It is rare that family members other than the identified patient are followed over time in clinical trials of psychotherapy or family therapy more specifically. Systems-based family therapy is believed to be effective because it targets processes within the system that maintain symptoms. If these processes are changed in the course of treatment, then other family members may benefit as well. Horigian et al. (2014) demonstrated that Brief Strategic Family Therapy targeted at an adolescent’s substance use, also decreased parents’ alcohol use. However, no study to date, has investigated the influence of family therapy targeted at a mother’s substance use on children’s substance use.

Children whose parents use substances are at higher risk for problematic substance use and earlier onset of substance use compared to their peers (Biederman, Faraone, Wozniak, & Monuteaux, 2000; Bu, Watten, Foxcroft, Ingebrigtsen, & Relling, 2002). Half of these children will develop a substance use disorder by young adulthood (Chassin, Pitts, DeLucia, & Todd, 1999) and they are at greater risk for experiencing other emotional, social, and behavioral adjustment problems (Hussong et al., 2008; Osborne & Berger, 2009). Additionally, these children are more likely to have poor relationships with their parents compared to other children (Ryan, Jorm, & Lubman, 2010). This is of particular importance because poor quality family relationships can further increase the risk that a child begins to use substances (Van Ryzin, Fosco, & Dishion, 2012). The purpose of this project was to explore changes in a child’s substance use when their substance-abusing mothers were randomly assigned to family therapy in which the child participated, or an attention control condition, in which the child did not participate.

Substance Use in Childhood and Adolescence

Childhood substance use continues to be a problem in the United States (Elliott, Huizinga, & Menard, 1989). A majority of youth report experimentation with alcohol and/or illicit drugs. In particular, 61.2% of youth report alcohol use, 44.5% report marijuana use, 28.3% report tobacco use, and 18% report prescription drug use prior to high school graduation (Monitoring the Future, 2016). Substance use, especially early onset, is associated with a variety of negative consequences such as early engagement in sexual activity, high-risk sexual behavior, disengagement from school, and mental health disorders (NIDA, 2014). Therefore, intervention that seeks to prevent childhood substance use is of paramount importance, especially for children who are at high risk for developing problematic substance use. Since children with substance-abusing parents are at high risk for developing problematic substance use, including children in their parent’s substance use treatment offers a unique opportunity to engage children in prevention and intervention services. Specifically, family systems therapy may address risk and protective factors among children who may not already be using substances, and who might not have other opportunities for intervention. The current project tested the influence of family therapy on children’s substance use whose mothers were randomly assigned to a family therapy condition versus an attention control condition.

The Family Context of Young Substance Users

There is ample evidence that childhood substance use is strongly related to family context variables including family structure, disruptions in family structure, as well as quality of family relationships (Engels & Bot, 2006; Kuendig & Kuntsche, 2006; Nash, McQueen, & Bray, 2005; Shaw, 2006). Researchers consistently note parenting deficits among adult substance abusers, including a higher incidence of child abuse and neglect (Kelleher, Chaffin, Hollenberg, & Fischer, 1994; Raitasalo, Holmila, Autti-Rämö, Notkola, & Tapanainen, 2015), and overly punitive or lax monitoring (Lang, Pelham, Atkeson, & Murphy, 1999). Within substance-using families, the parent–child relationship is often characterized by inconsistent care and unstable attachment (Lussier, Laventure, & Bertrand, 2010). Low quality parent–child relationships such as these increase risk for problem substance use among children (Hoffmann & Cerbone, 2002; Kuntsche & Silbereisen, 2004). However, studies have shown that improving family relationships such as through establishing a healthy balance of closeness and independence between parent and child can serve as a protective factor for a child’s substance use (Bartle & Sabatelli, 1989; Hoffmann & Cerbone, 2002; McElhaney, Allen, Stephenson, & Hare, 2009). This evidence may suggest that processes within the system that support substance use, may also be associated with poor parenting and poor parent–child relationships. Thus, family therapy could disrupt those processes and all those in the family would benefit from this change in family processes. To date, however, very few clinical trials for family therapy effectiveness have assessed change in members of the family other than the identified patient. This study sought to provide initial evidence that family therapy with a substance-abusing mother and her children could also benefit those children. Even for children who have only just begun to use substances, research indicates that intervening early in a child’s substance use behaviors can lead to improved short and long-term outcomes (Dennis & Scott, 2007). Thus, including children in their parent’s substance abuse treatment, may provide this early intervention.

Only a very small number of studies have examined the impact of parental substance abuse treatment on their children’s psycho-social functioning (Copello, Copello, Velleman, & Templeton, 2005; Cuijpers, 2005). A small number of studies have tested interventions for substance-abusing women with children in their care (Suchman, Decoste, Mcmahon, Rounsaville, & Mayes, 2011). These studies ranged in focus from behavioral skills training to advocacy and education, and have reported little change in parent–child interactions or in child adjustment. Suchman and colleagues conclude that given that women with substance use disorders often have poor attachment histories and exposure to childhood and recent trauma, interventions fostering behavioral management skills will do little to strengthen the parent–child relationship without improving their capacity to recognize and respond to the emotional cues of their children.

Family Therapy and Substance Use

Interventions, such as family systems therapy, that have the potential to improve family relationships and reduce problem substance use, have been found among the most effective interventions in reducing substance use for adolescent substance users (Becker & Curry, 2008; Dakof et al., 2010; Hogue & Liddle, 2009; Rowe, 2010). Also, family systems therapy that engages other family members such as children has been shown to be one of the most successful forms of intervention for adult substance users (Fals-Stewart, Lam, & Kelley, 2009; Gruber & Fleetwood, 2004; O’Farrell & Fals-Stewart, 2006). It is believed that family systems therapy targets substance use indirectly by helping families develop new ways of interacting that improve family functioning and support reduced substance use. That is, it is the change in the family processes that provides the mechanism of change for decreased substance use. Very few studies have assessed for changes in family processes but when they have, the evidence suggests that aspects of family processes change as a result of family therapy interventions (i.e., cohesion; Santisteban et al., 2003; family functioning; Horigian et al., 2014). However, no study to date has examined the impact on children’s substance use when they are engaged in family therapy as part of their mothers’ substance use treatment.

Family Processes and Family Therapy

Family systems therapy offers a distinct advantage to other treatment modalities in that it targets family processes such as family distance regulation that impact the maintenance of substance use. Family distance regulation refers to the balance of separateness and connectedness in patterns of interaction between members of a family. Bowen (1976) referred to this as differentiation, while Barton and Alexander (1981) in Functional Family Therapy speak of distance regulation specifically. For effective family functioning, it is believed that distance regulation needs to provide both intimacy and individuality experiences in the lives of family members (Allison & Sabatelli, 1988). There is evidence that distance regulation patterns within the family are related to substance use and fewer negative consequences due to substance use among older children (Bartle & Sabatelli, 1989). In this study, it was expected that parent–child dyads with an effective balance of separateness and connectedness would delay children’s initiation of alcohol and/or drug use, and those in families with more effective distance regulation would be more likely to stop using if they had begun.

Current Study

Using a longitudinal design, this study examined the impact of family systems therapy on children involved in their mother’s substance use treatment. It was expected that the inclusion of children in their mother’s treatment would interrupt childhood substance use. Specifically, children engaged in family therapy treatment who had already begun to use were expected to decrease their use of tobacco, alcohol, or marijuana at follow-up compared to children not involved in family systems therapy. Further, healthy separateness and connectedness between mother and child were expected to interrupt child substance use.

METHODS

Participants

This study utilized data from a larger randomized clinical trial testing family systems therapy with women seeking treatment for a substance use disorder (N = 183). Women were recruited from a substance use treatment facility in a large Midwestern city. In order to be eligible for the study, mothers had to meet criteria for a substance use disorder (SUD) as defined by the DSM IV, be seeking outpatient treatment for their SUD, and have a biological child between the ages of 8–16 years in their custody. In this study, the Computerized Diagnostic Interview Schedule (CDIS; Shaffer, 1992) was used to determine if participants met criteria for a SUD. CDIS is a computerized structured interview containing 263 items pertinent to a comprehensive psychiatric diagnostic interview based upon DSM IV criteria, and includes modules to diagnose alcohol, marijuana, tobacco, and other substance abuse and dependence.

Data were collected from the children engaged in the larger study. Children ranged in age from 8 to 16 years (M = 11.54; SD = 2.55), were primarily White (53.6%) or African American (42.6%). However, one participant was American Indian and 25 participants (13.6%) identified as “other”. Children were evenly divided across sex with 95 males (51.9%) and 88 females (48.1%). Twenty-nine (15.84%) children reported substance use prior to baseline.

Procedures

Mothers were screened for eligibility and interest in the research through the community treatment center. The initial screening was followed by signing of the informed consent, obtaining parental permission for child’s participation, and completion of baseline assessments. In cases where more than one child was eligible to participate, the child that reported more severe substance use as reported on the Form 90 (Miller, 1996) or reported higher rates of problem behavior as per the Youth Self Report (Achenbach & Edelbrock, 1982) was selected to participate in the study. All children, mothers’ romantic partners, and other caregivers were invited to participate in the family systems therapy sessions. In particular, 36% of all families included more than one child in the therapy sessions, 9% included mother’s romantic partner, and 10% included another family member.

Participating mother–child dyads completed baseline assessments including several individual and family measures. Dyads were then randomly assigned to one of three intervention conditions: (a) home-based family systems therapy, (b) office-based family systems therapy, or (c) Women’s Health Education (WHE, mothers only). Follow-up assessments were administered at 3, 6, 12, and 18 months post-baseline. Participating mothers received a $75 gift card to Walmart and their children received a $40 gift card to Walmart upon completion of their baseline assessment and each of their follow-up assessments.

Measures

A demographic questionnaire that assessed core variables necessary to characterize and compare samples was administered at baseline, 3, 6, 12, and 18 months post-baseline. Child’s age, race/ethnicity, and sex were collected. Additionally, information regarding family structure and changes in family structure were collected over the course of the study (see Table 1).

Table 1.

Descriptive Statistics of Age at First Use Variables

Variable	% of children that used during the study	Reported age at first use	Age of users	Age of nonusers
Tobacco	9.0	11.79 (2.46)	11.56 (2.39)	11.55 (2.59)
Alcohol	10.7	12.74 (2.24)	11.57 (2.36)	11.54 (2.6)
Marijuana	11.8	13.52 (1.37)	12.04 (2.46)	11.48 (2.58)

Children’s substance use.

Children’s substance use was measured using the Form 90 (Miller, 1996). The Form 90 is a structured interview that measures daily drug and alcohol use for the past 90 days using a timeline follow-back approach (Sobell & Sobell, 1992). Thus, a comprehensive report of children’s alcohol and drug use was generated. The Form 90 is internally reliable and sensitive to adolescents’ substance use (Waldron, Brody, & Slesnick, 2001) and has demonstrated high test-retest reliability for alcohol (r = .78) and illicit drugs (r = .92) (Carlson et al., 2001). We created categories for the number of days within the last 90 days since the data were highly skewed with a plethora of zeros. The categories were 0 = 0 days; 1 = 1 day; 2 = 2 days; 3 = 3–5 days; 4 = 6–8 days; 5 = 8–10 days; 6 = 10 or more days for alcohol use. The same set of categories was used for marijuana use and tobacco use with the addition of 6 = 10–15 days; 7 = 16–20 days; and 8 = 21 or more days.

Family distance regulation.

Family distance regulation was measured with two scales that were revised for this project. The first was the Healthy Separation subscale from the Separation Individuation Test for Adolescence (Levine, Green, & Millon, 1986). This subscale was designed to measure an adolescent’s sense of self within relationships with parents. The five items were reworded to refer to specific people, in this case from the mother’s perspective of her sense of self with her child. Sample items were “Even though I’m very close to my child, I feel I can be myself,” and “My child and I have some common interests and some differences.” These same items were included from the child’s perspective as well, such as “Even though I’m very close to my mother, I feel I can be myself.” Items were scored on a five-point Likert scale with “5” representing “strongly agree” and “1” representing “strongly disagree”. The internal consistency reliability for this measure was 0.65 for mothers and 0.67 for the youth. Higher scores represented higher healthy separateness.

The second scale that was used to assess Family Distance Regulation was a revised version of The Social Connectedness Scale-revised (Lee, Draper, & Lee, 2001). This is a measure of a sense of connection to people in general, while the version used in this study was reworded to reflect a sense of connection to the youth from the mothers’ perspective and a sense of connection to the mother from the child’s perspective. Sample items included “I feel distant from my mother/child,” and “I feel understood by my mother/child.” The nine items were scored on a five-point Likert scale with “5” representing “strongly agree” and “1” representing “strongly disagree”. The internal consistency reliability for this measure was 0.83 for mothers and 0.79 for youth. Higher scores represented higher connectedness.

The two subscales were correlated (r = .38, p < .05) for mothers, and (r = .49, p < .05) for youth. These correlations were in the expected direction and significant. The two scales were meant to measure the balance between separateness and connectedness within the relationship between mother and youth.

Therapy Conditions

Ecologically Based Family Therapy (EBFT; Slesnick & Prestopnik, 2005) is a 12-session family systems therapy based on the social ecological theoretical perspective (Bronfenbrenner, 1979). Dysfunctional interactions that contribute to the development and maintenance of problem behaviors such as substance use are targeted. The intervention is based on the recognition that substance use and related individual and family problems derive from many sources of influence and occur in the context of multiple systems. Therefore, intervention does not focus solely on the individual, but on the social interactions among all participants that impact successful interactions within and across systems. Treatment sessions use techniques such as reframing and interpretations, interrupting problem behaviors through communication and problem-solving skills training, and assisting families in obtaining services such as medical care, job trainings, or self-help programs. Ongoing supervision and independent treatment fidelity coding by advanced graduate students were employed to assure implementation quality and adherence to the intervention protocol. Originally developed for substance-using runaway adolescents and their families (Slesnick, Guo, Brakenhoff, & Bantchevska, 2015; Slesnick & Prestopnik, 2005, 2009), the intervention has been rated as a promising evidence-based practice by the National Institute of Justice and as a supported evidence-based practice by the California Evidence-Based Clearinghouse.

Women’s Health Education (WHE) is a 12-session manualized education-based intervention that was used as an attention control (Miller, Pagan, & Tross, 1998). WHE helps women understand the body, human sexual behavior, pregnancy and childbirth, STD’s, HIV, and AIDS. WHE provided equivalent therapist attention and expectancy of benefits, but did not include family systems therapy techniques, and children were not engaged in the therapy. WHE has been used as an attention control and has been shown to improve maternal outcomes in high-risk populations of women (Hien et al., 2009).

Data Analysis

Given the large number of zero’s in the data for youth’s use of substances and the count nature of the data (number of days used), we performed zero inflated Poisson latent growth curve analyses in Mplus 8.0 (Muthen & Muthen, 2017). In this analysis, the count portion of the data is modeled along with the inflated or zero portion of the data. An intercept and slopes for the count portion are estimated and an intercept and slopes for the inflated part are also estimated. The results of the count portion are the log of the expected count, and the results of the inflated part are the odds of being at 0 over time (Atkins & Gallop, 2007; Liu, 2007). We used this model and regressed these intercepts and slopes onto sex of the youth, age of the youth, change in family structure, treatment condition, number of treatment sessions attended, and the separateness and connectedness subscales of the FDR from both the mother and the youth’s perspectives. Although alcohol, marijuana, and tobacco use are more than likely related, given the nature of the data analysis, we could not estimate a model with all three outcomes. When conducting these analyses, the residual variances of the latent variables are typically set to zero in order to reduce the computation load (Liu, 2007). Since that is the case, we would have not been able calculate the correlations among the residual variance for the three outcomes, thus we elected to estimate three separate models: one for alcohol use, one for marijuana use, and one for tobacco use. This is a limitation to the study, since we needed to use the same predictor variables for all three outcomes as well.

RESULTS

Table 1 provides the percentage of the youth who reported using tobacco, alcohol, or marijuana at any point in the study as well as their reported average age at first use, the average age of use for those who reported using in the study, and the average age of nonusers. Table 2 provides the means, standard deviations, and ranges of the predictor variables in the models. Table 3 provides the frequency of use for all three substances over time.

Table 2.

Descriptive Statistics of Predictor Variables in the Analysis

Variable	Percentage	Mean (SD) Median	Skewness (standard error)	Kurtosis (standard error)	Range
Males	50.6%
Attention control	32.6%
Family therapy	67.4%
Sessions attended		5.28 (4.90)	0.317 (0.182)	−1.58 (0.362)	0–12
		4.00
Change in family structure	30.3%
Child age		11.55 (2.57)	0.229 (0.182)	−1.18 (0.362)	8–16
		11.00
Child separateness		2.17 (0.76)	0.527 (0.187)	0.017 (0.373)	1–4.4
		2.00
Child connectedness		1.94 (0.78)	0.928 (0.188)	0.597 (0.374)	1 –5
		1.77
Mother separateness		2.14 (0.68)	0.281 (0.185)	−0.158 (0.367)	1 –4
		2.00
Mother connectedness		2.12 (0.78)	0.724 (0.184)	0.122 (0.365)	1 –5
		2.00

Table 3.

Percentage Days Used Over Time By Substance

	Alcohol		Marijuana		Tobacco
	% at 0	% days of use	% at 0	% days of use	% at 0	%days of use
Baseline	94.9%	3.92 (3.98)	91.6%	28.89 (32.3)	91.6%	66.70 (42.9)
3 months	96.9%	4.78 (7.55)	96.1%	23.85 (28.29)	94.3%	70.53 (35.14)
6 months	93.8%	2.64 (3.43)	90.0%	39.53 (37.7)	91.3%	71.30 (43.08)
12 months	95.0%	5.54 (14.72)	91.8%	43.2 (43.4)	93.7%	89.36 (23.92)
18 months	96.9%	5.96 (4.28)	89.3%	42.64 (40.46)	89.3%	74.95 (40.20)

We tested both a linear model and quadratic model for all three outcomes to determine which fit the data the best. In all three cases the quadratic model fit better than the linear model.

In order to avoid collinearity between the linear slope and quadratic term, we “centered” time, so that the 6-month follow-up was set to 0 and was the intercept for the models. Pretreatment and the 3-month follow-up were coded as −0.5 and −0.25, respectively. The 12-month and 18-month follow-ups were coded as 0.25 and 0.5, respectively. The chi-square difference between the linear and quadratic models for alcohol use was 97.32 with 10 degrees of freedom (p < .001). The difference between the linear and quadratic models for tobacco use was Δχ² (10) = 66.29, (p < .001) and between the linear and quadratic models for marijuana use was Δχ² (10) = 272.72, (p < .001). Thus, we regressed the intercept, slope and quadratic terms onto the predictor variables in the three separate models. We report the standardized results from the Mplus analysis in Table 4 for all three sets of outcomes.

Table 4.

Standardized Estimates For Each Substance For The Count Portion And Inflated Portion of The Models

	Alcohol						Marijuana						Tobacco
	Count			Inflated			Count			Inflated			Count			Inflated
Predictor	Int	Slp	Quad	Int	Slp	Quad	Int	Slp	Quad	Int	Slp	Quad	Int	Slp	Quad	Int	Slp	Quad
Sex of Child	0.21 *	0.40 *	−0.16	0.31 *	0.45	−0.25	0.12	−0.43 *	0.24	0.13	−0.24	−0.13	0.36	−0.28 *	−0.40 *	0.20	−0.06	−0.44 *
Treatment Condition	−0.26 *	0.28 *	0.28 *	−0.24	0.32	0.22	−0.28	0.01	−0.29 *	0.16	−0.12	−0.32	−0.14	0.03	−0.40 *	−0.19	−0.15	−0.35
Sessions	−0.83 *	0.27 *	0.82 *	−0.81 *	0.17	0.88	−0.45 *	0.51 *	−0.44 *	−0.15	−0.46	−0.07	−0.71*	0.26	0.56 *	−0.19	0.06	0.53 *
Child Age	0.04	−0.24 *	0.20 *	0.07 *	−0.17	0.10	−0.10	−0.49 *	0.67 *	−0.39 *	0.01	0.61	0.51	−0.53 *	0.20 *	−0.14	−0.65 *	0.19
Child Separate	−0.35 *	0.36 *	0.03	−0.28	0.40	0.09	−0.42	0.07	0.30	0.18	−0.77	−0.17	−0.26	0.49 *	0.36 *	−0.05	0.42	0.22
Child Connect	0.09	0.17	−0.40 *	−0.02	0.14	−0.30	0.59*	0.33	−0.47	0.18	0.18	0.18	0.06	0.28	−0.57 *	0.27	0.52	−0.48
Mother Separate	−0.05	−0.63 *	0.16	−0.13	−0.60	0.21	−0.55	0.28	0.26	0.02	0.37	0.39	−0.14	0.24	0.48 *	0.00	0.12	0.48 *
Mother Connect	0.27 *	0.24 *	0.06	0.28	0.26	−0.08	0.40	−0.28 *	0.17	−0.45 *	−0.28	0.14	−0.00	0.07	0.03	−0.43	−0.16	0.25
Structure Change	−0.12 *	−0.14 *	−0.32 *	−0.23	−0.11	−0.26	−0.39	0.44 *	−0.37	−0.73 *	0.18	0.48	−0.42 *	0.47 *	0.34 *	−0.86 *	0.31	0.29

Note.

^*p < .05

Int = Intercept; Slp = Linear Slope; Quad = quadratic.

Alcohol Use

For the count portion of the model, the average intercept (Mplus standardized estimate = 0.266; p = .438) and average quadratic term (standardized estimate = −0.757; p = .133) were not significantly different than zero but the linear portion was significantly different than zero and negative (standardized estimate = −0.605; p = .05). This is known as the instantaneous rate of change or the average rate of change from one time point to the next. This suggests a decrease in alcohol use from one follow-up point to the next. The inflation linear slope and quadratic terms were also significantly different than 0 (standardized estimates = −1.29 and −0.177; p’s < .001, respectively) and negative suggesting that the odds of being at 0 increased over the course of the study. Table 4 provides the standardized estimates for the regression equations.

For the count portion of the model, there were several significant associations for the intercept of alcohol use. The intercept was set at the 6-month follow-up, so treatment had been completed. The intercept was significantly associated with child’s sex (estimate = 0.216, p < .001), treatment condition (estimate = −0.260, p < .001), number of treatment sessions attended (estimate = −0.835, p < .001), child’s perceptions of healthy separateness with mother (estimate = 0.352, p < .001), mother’s perception of connectedness with child (estimate = 0.276, p < .001), and change in family structure (estimate = −0.125, p < .05). The coefficient for child’s sex was positive. Males were coded as 1 and females were coded at 2, this suggests that females had a higher intercept or higher use of alcohol at the 6-month follow-up in comparison to males. The coefficient for treatment condition was negative. The treatment condition variable was coded as 0 for attention control and 1 for family therapy. This suggests that alcohol use at the 6-month follow-up was lower for the family therapy group than for the attention control group. The coefficient for treatment sessions was also negative. This suggests that alcohol use at the 6-month follow-up was lower for those who attended more sessions. The coefficient for the child’s perception of healthy separateness was negative. This suggested that alcohol use at the 6-month follow-up was lower for children who perceived higher levels of healthy separateness. The coefficient for mother’s perception of connectedness was positive suggesting that alcohol use was higher at the 6-month follow-up when mothers perceived more connectedness between herself and her child.

The linear slope or the instantaneous rate of change for the count portion of alcohol use was significantly associated with all the predictor variables in the regression equation (see Table 4 for estimate). Treatment condition, number of sessions attended, age of child, child’s perception of connectedness to mother, and change in family structure, all significantly predicted the quadratic term for the count portion of the model. The change in alcohol use over time for the two treatment groups can be seen in Figure 1A. The coefficient for the slope was negative, and the coefficient for the quadratic term was also negative suggesting a convex or “U” shape.

Figure 1.

Change in Alcohol (A), Marijuana (B), and Tobacco use (C).

To interpret these results, the analyst has to take into account both the coefficient for the linear and quadratic terms (Atkins & Gallop, 2007; Liu, 2007). Child sex was positively associated with the linear slope and negatively associated with the quadratic term. This suggests that females’ slope is more negative or their rate of decrease in use is greater than it is for males and they do not increase their use as much post treatment. Treatment condition was positively related to the slope and quadratic term, suggesting that those in the family therapy condition decreased use more and did not increase use as much after treatment. This was also true for those who attended more sessions.

Children who perceived more healthy separateness and more connectedness with mother and children whose mothers perceived more connectedness showed a steeper decrease in use with more connectedness negatively related to the quadratic suggesting a less deep convex curve. Child age, and change in family structure were predictive of less decrease in use with a less deep convex curve. Mother’s perception of healthy separateness with child was predictive of less of a decrease in the slope but was not related to the quadratic term.

The set of predictor variables were also used to predict the inflation part of the model or the probability of not using. To interpret this part, a positive coefficient suggests a greater likelihood of being at 0 and staying at 0 (for the slope and quadratic terms). A negative coefficient would suggest a decrease in that likelihood. The computations for the inflated part of the model were restricted because of the low number of users in the dataset, thus, many of the coefficients in this part of the model were fixed by Mplus and either not estimated or the random effect was fixed at 0. All of the coefficients for the inflated part of the slope and quadratic terms were fixed by Mplus. Only number of sessions attended, sex of child, and age of child were free to be estimated for the intercept of the inflated part of the model. These were significant in predicting the likelihood of being at 0 at the 6-month follow-up. Females were more likely to be at 0 compared to males, attending more sessions was associated with a lower likelihood of being at 0, and being older was associated with a higher likelihood of being at 0.

Marijuana Use

The same model was estimated for marijuana use. The standardized estimates for the model can be seen in Table 4. For the count portion of the model, the intercept was significantly different than 0 (estimate = 2.61 p < .001), the slope of use was positive but not significantly different than 0 (estimate = 1.96 p = .168), and the quadratic term was negative and significantly different than 0 (estimate = −9.88 p < .01). This suggests a convex curve. However, very few of the predictor variables were significantly associated with the intercept, linear slope, and quadratic terms. For the intercept, only mother’s perception of connectedness was positively associated (estimate = 0.116, p = .032), suggesting that with higher scores on mother’s connectedness the youth used more at the 6-month follow-up. The slope, or instantaneous rate of change was significantly associated with the child’s sex (estimate = −1.92, p = .032), number of treatment sessions attended (estimate = 0.131, p = .031), child age (estimate = −0.245, p = .004), and mother’s perception of connectedness to the child (estimate = −0.457, p = .026). Being female was associated with a lesser increase in use (since the slope was positive), attending more sessions increased the slope, being older decreased the increasing slope. Number of sessions attended (estimate = −0.249, p = .041) and child age (estimate = 0.731, p < .001) were predictive of the quadratic term in the count portion of the model. Attending more sessions was negatively associated with the quadratic term and positively associated with the slope term, suggesting that more sessions increased the “dip” in the convex curve; we see more decrease as well as more increase in use over time. The age of the child was positively associated with the quadratic term and negatively associated with the linear term which suggests that the older the child the less convex the curve, suggesting less decrease in use and less increase as well.

There were a few predictors that were related to the intercept of the inflated part of the model which represented the probability of being at 0 at the 6-month follow-up. The child’s age, mother’s perception of connectedness and change in family structure were significantly associated with this intercept. The older the child, the less likely they would be at 0 at the 6-month follow-up. If mother perceived more connectedness with the child, the less likely they would be at 0 at the 6-month follow-up, and if there was a change in family structure during the child’s lifetime, less likely the child would be at 0 at the 6-month follow-up (see Table 4 for estimates). Since treatment condition was not significantly associated with change in marijuana use over the course of the study, Figure 1B shows the average change in marijuana use across the follow-up times for all participants.

Tobacco Use

The model was estimated for tobacco use as well. The results of the model can be found in Table 4. For the count portion of the model, the intercept and linear terms were not significantly different than 0 (estimate = 1.51 and −0.983, respectively), but the quadratic term was (estimate = −1.44, p < .001). Both the linear term and the quadratic term were negative. This suggests a convex or “U” shaped trend with an initial decrease in use and then an increase over time. In the regression model, none of the predictors were significantly associated with the count intercept or use at the 6-month follow-up. Sex of the child (−0.284, p = .007), age of the child (−0.538, p = .004), child’s perception of healthy separateness (0.489, p < .001), and change in family structure (0.469, p < .001) were significantly associated with the linear slope of the count portion of tobacco use. Sex of the child, treatment condition, number of treatment sessions attended, child’s perception of connectedness to mother, and change in family structure were significantly associated with the quadratic term in the model (See Table 4 for estimates). Females showed a steeper instantaneous rate of change with a steeper convex curve as well. Those in the family therapy condition showed a deeper convex curve. This can be seen in Figure 1C. Those in the family therapy condition show an initial decrease in use with a steep increase at 6 months (after treatment had ended), then a decrease and increase. Those in the attention control condition show a similar initial decrease (the linear slope was not different), and a less deep convex curve with a leveling off of use at about the same rate as the initial rate of use. If the youth and/or mother attended more sessions, the convex nature of the curve was less deep, suggesting that with more sessions use of tobacco did not increase as steeply after treatment. When the child perceived more connection with mother, the convex nature of the curve deepened, suggesting that the child’s increase in use after treatment was steeper if they perceived more connectedness with mother. A change in family structure was associated with a flatter instantaneous rate of the change and the convex nature of the curve was also less deep.

Only change in family structure was related to the intercept of the inflation part of the model (−0.858, p < .001). Those who did not experience a change in family structure were more likely to be at 0 at the 6-month follow-up than those who did. Child age was negatively associated with the linear slope term of the inflation part of the model, suggesting that younger children were more likely to stay at 0. Finally, child sex, number of treatment sessions attended, and mother’s perception of healthy separateness from child were associated with the quadratic term in the inflation part of the model (see Table 4 for estimates). Females were more likely to stay at 0, those who attended more sessions were more likely to stay at 0 (i.e., not begin using tobacco over the course of the study), and mothers’ who perceived more healthy separateness had children who were more likely to remain at 0 over the course of the study.

DISCUSSION

The results of this study suggest a complex response to treatment dependent on demographic factors, relational factors, type of treatment, and number of sessions attended. It should be noted again, that the children who participated in the family therapy sessions were not the identified patient or substance users, it was their mothers. Thus, any change in use on the part of the child was an additional benefit of treatment, not the target of treatment. The children in the sample were fairly young (8–16) and thus level of use of these three substances was low. This necessitated a complex data analysis that estimated the count portion of the data (percent days of use categorized) and the inflation part or 0’s in the data. We used predictor variables to explain the variance in the intercept, linear slope, and quadratic terms. Sex of the child was predictive of not using and continuing to not use, as well as change in use, with females more likely to not use and showing a steeper decrease in use if they had begun to use depending on the substance. Previous research suggests that male adolescents are more likely to have problems with substance use and tend to over-represent in all categories of adolescent substance use and early onset of substance use with one exception; early onset tobacco use (Griffin, Botvin, Scheier, Diaz, & Miller, 2000; Moss, Chen, & Yi, 2014).

Whether the child had experienced a change in family structure during their lifetime was also predictive of both not using, and change in substance use for all substances. For alcohol use, if the child had experienced a change in structure (either experienced a divorce or break-up, or experienced mother’s new relationship) they tended to continue use over the course of the study with a less convex curve. For marijuana use, a change in family structure was associated with a higher likelihood of using generally (lower likelihood of being at 0). For tobacco use, if there was no change in family structure, the child was less likely to use tobacco, and more likely to stay at 0 or not begin using tobacco during the study period. Evidence suggests that children in single-parent households are more likely to report substance use and initiate substance use early (Griffin et al., 2000; Hoffman, 1993; Turner, Irwin, & Millstein, 2014). This is believed to be related to children in single-parent families having fewer resources and less effective coping skills (Elder, Eccles, Ardelt, & Lord, 1995; Gabel, 1992; : Norton & Glick, 1986). Also children of single-parent families are more susceptible to peer pressure and are less likely to consult with their parents than other children (Steinberg, 1987; Dornbusch et al., 1985). These are theorized to play a role in why these youth are more at risk for adolescent substance use and early onset of substance use. Specifically, links between single-parent families and alcohol use, tobacco use, and marijuana use have been observed across several studies (Dornbusch et al., 1985; Saucier & Ambert, 1983; Stern, Northman, & Van Slyck, 1984; Flewelling & Bauman, 1990; Turner et al., 2014). However, less is known about how changes or disruptions in family structure change adolescent substance use over time. The results of this study would suggest that changes or disruptions in family structure are as influential as being from a single-parent household.

The family relationship variables were also related to use at 6-month follow-up (the intercept) and the change in use over the course of the study. Mothers and sometimes child’s perception of more connectedness was associated with “negative” outcomes, or more use, and less change in use over the course of the study. It could be that more connectedness is associated with a lack of boundaries and a lack of a sense of autonomy on the part of the child, which has been associated with more alcohol use in college-aged participants (Durkin, Wolfe, & Clark, 1999). Healthy separateness from both the mother’s perspective and child’s perspective was associated with more positive outcomes. With a sense of separateness from mother or from child, the child was less likely to have started using and more likely to stay a nonuser over the course the study. If they had started using, healthy separateness was predictive of steeper decreases in use and less steep increases after treatment. This provides evidence that a sense of individual boundaries along with a sense of belonging fosters an environment that decreases the likelihood of early substance use, even if mother is abusing substances.

Of most importance, however, was the evidence that supported the use of family therapy over an attention control condition for children with mothers who abuse substances. Regardless of the change in mothers’ use, their children’s use of alcohol and tobacco changed as a result of participating in treatment with their mother. Those children who participated in family therapy with their mothers showed steeper declines in use and less steep increases in use of alcohol over the course of the study than those in the attention control group. Those in the family therapy condition also showed a steeper decline in tobacco use during treatment, but their use increased again after treatment and stayed higher than those in the control group.

Although many children of parents with a substance use disorder (SUD) show significant resilience, and do not develop problem substance use, research indicates that children of parents with a SUD have a 2–5 times higher-fold risk of developing problem substance use (Ali, Dean, & Hedden, 2016; Li, Pentz, & Chou, 2002). As women substance users are more likely to have minor children in their care than male substance users, this study examined the potential benefits of engaging children in family therapy as an adjunct to their mother’s substance use treatment. Children are not often included in their parents’ treatment plan, yet, given their increased risk for substance use, this exclusion represents a lost opportunity to engage in prevention and intervention with these youth. Findings from this study showed that when children were included in the family therapy of their substance-using mothers, significant benefits to the child were observed. Specifically, when children had begun to use alcohol or tobacco at the start of the study, their use decreased if they were in the family therapy condition versus the attention control condition.

Decreasing alcohol use has the potential to reduce the risks of engagement in other drugs, as well as co-occurring problem behaviors. Since many of the children in this study were not using these substances at the start of treatment, and substance use was tracked within a relatively short period of time (18 months), future research should examine the role of family therapy on delaying substance use onset over a longer period of time, along with the preventive benefits on other risk behaviors.

Although the family systems therapy was identical when delivered in the home or office, some studies suggest that home-based family therapy facilitates engagement, especially for multiproblem families (Henggeler, Borduin, Melton, & Mann, 1991). The number of users in this study was so small that it was not possible to compare the impact of office- versus home-based family therapy. However, it should be noted that the number of sessions attended made a difference in outcomes for the children in this study. Those who were in the home-based therapy condition participated in more sessions (7.68) in comparison to the office-based condition (5.18) and the attention control (2.89) (F(2) = 16.49, p < .001). This provides some evidence that home-based family therapy was more effective than office based because of the evidence that if the family attended more sessions, there were better outcomes for the children. It will be an important next step in this research to compare office- versus home-based therapy. It appears that mothers and children are more likely to engage with the therapy process and the therapist when the therapy is conducted in a context in which they feel more comfortable.

Limitations

The study is limited by a relatively small sample. A larger sample would offer greater power to examine predictive socio-demographic factors associated with substance use outcomes. The 18-month time frame was relatively short, and a longer follow-up might provide more detailed information on substance use outcomes for the children, as well as the long-term preventive effects associated with family therapy. The reliability of the separatedness subscale in this study was low for both mothers and children. It could be that the items included from the healthy separation subscale of the Separation Individuation Test for Adolescents (Levine et al., 1986) are not relevant for mothers and their younger children, or that a five-item scale with these particular items did not provide enough internal consistency. More work is needed to develop measures of family distance regulation that are sensitive to change, and appropriate for children and adolescents. Future work needs to determine how best to assess these family dynamics.

CONCLUSIONS

Some of the shortcomings of many of the clinical trials conducted to test the effectiveness of family therapy for substance use and other symptoms are that rarely is more than one member of the family assessed, and only the symptoms of the identified patient are followed over time. This study fills this gap and shows that family members other than the identified patient can benefit from family therapy. That is, to the best of our knowledge, this is the first study to document that children of women who have been diagnosed with a substance use disorder and who participated in family therapy appear to reap some benefits from that treatment. Youth in the family therapy treatment group showed decreasing use of alcohol and tobacco. The next steps in this line of research would be to assess what mechanisms influenced these changes in substance use. We used distance regulation at baseline to predict levels of use and change in use with some significant associations. The next step would be to determine if changes in distance regulation are associated with changes in substance use. It may be that distance regulation changed over the course of therapy, which may have been the mechanism of change that lead to decreasing use for children in the family therapy condition.