Interparental Conflict and Long-Term Adolescent Substance Use Trajectories: The Role of Adolescent Threat Appraisals

Abstract

Although interparental conflict (IPC) has been linked directly and indirectly (via adolescents’ appraisals) with a wide range of adolescent outcomes, little is known about the implications of IPC and related adolescent threat appraisals for substance use. Drawing on the cognitive-contextual framework, we test competing hypotheses about how IPC may impact adolescent substance use outcomes, specifically testing whether: a) threat appraisals are directly related to escalation in alcohol and tobacco use over adolescence or b) threat appraisals are indirectly associated with substance use through their impact on adolescent internalizing problems. Family data from 768 two-caregiver families were analyzed for this study. Adolescents (53% female) were followed on 7 occasions starting in the fall of 6^th Grade (mean age 11.3) through the spring of 11^th grade. IPC and family demographic data were collected from parents. Youth provided data on their appraisals of conflict, internalizing problems, and substance use. Using longitudinal growth curve models, findings supported threat appraisals as a direct risk factor for escalating cigarette use, but not escalating alcohol use, during adolescence. In the alcohol trajectory model, IPC was a direct predictor of increases in alcohol use over time. These findings indicate that high levels of threat appraisals are a specific and direct risk for greater increases in cigarette use over the course of adolescence and that IPC confers risk for increasing rates of alcohol use over adolescence.

Substance use initiation and escalation are most prevalent during adolescence. Recent data from the National Survey on Drug Use and Health indicate that substance use initiation is rare for youth ages 12–13, yet by ages 16–17, 52% have used alcohol and 26% have smoked cigarettes (Center for Behavioral Health Statistics and Quality, 2015). Trajectories of escalating alcohol and/or tobacco use during adolescence forecast risk for a host of problems in later life, including substance use disorders, poor physical health, and mental health problems (DeWit, Adlaf, Offord, & Ogborne, 2000; Tucker, Ellickson, Orlando, Martino, & Klein, 2005; Van Ryzin, Fosco, & Dishion, 2012). In addition, there are different health risks associated with alcohol and tobacco use. Alcohol use can elevate adolescent risk for sexual victimization (Champion et al., 2004), riding in a car with an intoxicated driver (Miller, Naimi, Brewer, & Jones, 2007), sexual risk-taking behavior and STI’s (Tapert, Aarons, Sedlar, & Brown, 2001), and suicidal ideation (Miller et al., 2007). Tobacco use is concerning because of its addictive qualities. Relative to other substances, adolescents who start smoking are much more likely to continue into adulthood than adolescents who start using alcohol or marijuana (Van Ryzin et al., 2012). From a lifespan health perspective, tobacco remains among the leading global risks for mortality (World Health Organization, 2009), particularly due to cancer and heart disease risks (Go et al., 2014; World Health Organization, 2009). Developmentally, it is important to consider trajectories of escalation in adolescent alcohol and tobacco use over time. Adolescents who exhibit a pattern of increasing use are at highest risk for later life problems than those that experiment a little but do not escalate to patterns of frequent and intense use (Tucker et al., 2005). Thus, this study focuses on understanding antecedent risk factors that predict rates of escalation in alcohol and tobacco use over adolescence.

Despite considerable research on family risk factors for adolescent substance use, exposure to interparental conflict (IPC) has largely been overlooked. Yet, IPC is a well-documented and strong risk factor for a wide range of youth behavioral and emotional problems (Grych & Fincham, 2001). Work examining the underlying processes accounting for this link has identified youths’ subjective evaluations of conflict, or appraisals, as key mechanisms of risk for psychopathology outcomes in prospective longitudinal studies in childhood (Grych, Harold, & Miles, 2003) and adolescence (Fosco & Feinberg, 2015). In this study, we extend this work to explore threat appraisals as a key risk factor through which IPC may be associated with patterns of substance use over the course of adolescence.

Generally speaking, threat appraisals reflect adolescents’ evaluation of IPC as posing a risk to their own or their family’s well-being and stability. Adolescent threat appraisals include general fears or worries about the implications of IPC for the family or that IPC may lead to something bad. Threat appraisals also may include specific concerns that IPC may foreshadow divorce, escalation to violence, or that parents will become angry with the adolescent (Atkinson, Dadds, Chipuer, & Dawe, 2009; Fosco et al., 2007; Grych & Fincham, 1990; Grych, Seid, & Fincham, 1992). Over time, adolescents’ threat appraisals can lead to diminished beliefs that they are able to cope with IPC specifically (Fosco & Grych, 2010), declines in overall self-efficacy (Fosco & Feinberg, 2015), and increases in internalizing problems—all of which are risk factors for substance misuse (Brook, Ning, & Brook, 2006; Marmorstein, 2009; Witkiewitz & Marlatt, 2004). Youth in families where IPC is infrequent, of low intensity, or generally resolved effectively tend not to develop beliefs about IPC as threatening (Grych & Fincham, 1993). However, youth who have been exposed to chronic, poorly managed IPC, or have witnessed interparental violence, may become “sensitized” and tend to appraise later IPC—even low severity IPC—as threatening (Grych, 1998; Grych, Raynor, & Fosco, 2004). It is this tendency to appraise IPC as threatening that serves as a proximal risk factor for emotional and behavioral maladjustment (Rhoades, 2008), with longitudinal evidence emphasizing risk for internalizing problems in particular (Fosco & Feinberg, 2015; Grych et al., 2003).

Although IPC has a robust association with externalizing problems (Buehler, Krishnakumar, Stone, & Anthony, 1997), only a handful of studies have examined IPC as a risk factor for adolescent substance use. However, other studies, conceptualizing conflict in families more generally, have documented family conflict as a risk factor for adolescent substance use (Cordova et al., 2014; Hawkins, Catalano, & Miller, 1992). This literature points to different risk processes linking family conflict to substance use. One process focuses on disruptions to adolescents’ connections with their family (e.g., weakened bonds to parents) that lead them to befriend substance-using peers (Ary et al., 1999; Bahr, Maughan, Marcos, & Li, 1998) or to engage in delinquent behavior (e.g., Andrews, Foster, Capaldi, & Hops, 2000; Fosco, Caruthers, & Dishion, 2012), both of which increase their risk for substance use. A different perspective emphasizes intrapersonal processes, such as stress reactions, that also can place adolescents at risk for substance use (e.g., Bray, Adams, Getz, & Baer, 2001; Wills, 1986). A stress perspective has been found to inform both alcohol use risk (Bray et al., 2001) and tobacco use risk (Parrott, 1999).

It is important to increase specificity in our operationalization of family conflict and risk processes in families in order to better understand adolescent risk for substance use (Bahr et al., 1998). Across studies of family conflict and adolescent substance use, family conflict has been measured as marital conflict, parent-adolescent conflict, sibling conflict, or various combinations of these factors. These different operational definitions of family conflict make it difficult to compare findings across studies, and these various forms of conflict may expose adolescents to different types of risk for escalation in substance use. From a family systems perspective, interparental and family-level conflict should be conceptualized as distinct processes, and evidence suggests that they correlate with youth developmental outcomes in unique ways (Fosco & Grych, 2013). Thus, an important step forward for this body of research is to examine associations between specific types of conflict in families and specific risk mechanisms.

The Current Study: IPC, Threat, and Adolescent Alcohol and Cigarette Use

Guided by the cognitive-contextual framework, this study explored the possibility that adolescent threat appraisals about IPC may predict escalations in substance use trajectories across adolescence. Preliminary studies have documented a short-term association between exposure to IPC and adolescent substance use levels (Tschann et al., 2002). Building on this foundation, we explored two potential hypotheses regarding how threat appraisals may be a risk factor for escalating patterns of substance use over adolescence.

Threat Appraisals as a Direct Pathway to Substance Use Risk

Threat appraisals may represent a direct pathway linking IPC and adolescent substance use. Persistent threat appraisals about IPC can be distressing (Fosco & Grych, 2008; Grych & Fincham, 1990). Moreover, IPC is often outside of adolescents’ control, which may explain why threat appraisals are related to decreases in adolescents’ beliefs that they are able to cope with IPC in the future (Fosco & Grych, 2010) and decreases in their sense of self-efficacy (Fosco & Feinberg, 2015). Drawing on this empirical foundation, IPC, and the degree to which adolescents perceive conflict as threatening, represent a unique stress process in which adolescents are often not empowered to effect change for the better (Fosco & Feinberg, 2015).

This evidence base led us to propose the “threat as a direct pathway to substance use” hypothesis to test whether adolescents’ threat appraisals would serve as a mechanism by which IPC transfers risk for escalation in alcohol and cigarette use. Adolescents may engage in substance use as a means of managing their feelings of anxiety or distress (i.e., threat) that stems from exposure to IPC. Threat may have particular relevance for cigarettes, as some work underscores that there is a common perception that tobacco products contain anxiolytic properties which can help manage feelings of anxiety (Parrott, 1999). Thus, adolescents who feel threatened by IPC may find cigarette use to be particularly appealing for managing anxiety about conflicts they cannot otherwise control.

Threat Appraisals as an Indirect Risk Pathway via Internalizing Problems

A second possibility builds on established findings that threat appraisals are consistently related to internalizing problems and other work suggesting that internalizing problems mediate the association between family factors (e.g., family conflict, maltreatment) and substance use (Lewis et al., 2011; Wu, Lu, Sterling, & Weisner, 2004). Associations also have been found for cigarette use (Gehricke et al., 2007) and alcohol use (White, Xie, Thompson, Loeber, & Stouthamer-Loeber, 2001) specifically. Thus, established developmental models linking IPC and internalizing problems may be extended to predict adolescent substance use risk. Preliminary evidence is provided in a short-term longitudinal study of Mexican-American early adolescents, which found that appraisals of IPC, analyzed as a composite of threat and self-blame, were indirectly associated with substance use 6 months later via internalizing symptoms (Tschann et al., 2002). This study faced two key limitations. First, it combined measurement of threat and self-blame appraisals, which may have unique associations with outcomes. Second, it was limited to two measurement occasions, making it impossible to test prospective mediational pathways from appraisals to internalizing problems to substance use in a temporal sequence. In the current study, we evaluated internalizing problems as an indirect pathway by which threat appraisals may be linked with substance use trajectories.

We draw on prior work demonstrating a developmental model in which IPC impacts adolescent internalizing problems through a threat appraisal mechanism (Fosco & Feinberg, 2015). Our “threat as an indirect risk pathway” hypothesis extends prior findings by testing whether threat appraisals place adolescents at risk for escalations in alcohol and cigarette use to the extent they are related to increased internalizing symptoms. To test this hypothesis, this study applied an autoregressive, longitudinal test of the developmental model in which IPC would be related to increases in threat appraisals, threat appraisals would be related to increases in internalizing problems, and internalizing problems would predict escalations in alcohol and cigarette use. Substance use was modeled as latent growth curves (separately for alcohol and cigarettes) and the latent slope term was regressed on IPC, threat appraisals, and internalizing problems. Thus, we were able to simultaneously evaluate the direct and indirect associations between threat and escalations in substance use over time. Finally, to provide a conservative test of the hypothesized developmental model of IPC and substance use, we included established family risk factors – parenting practices and parental substance use – as covariates in our models.

Method

Participants

This study analyzed a subsample drawn from PROSPER (PROmoting School-community-university Partnerships to Enhance Resilience): a large-scale effectiveness trial aimed at reducing substance use initiation starting in 6^th Grade. A random sample of 2,267 families from the original sample were invited to participate in family assessments; 979 (43%) completed the in-home surveys completed independently by the adolescent, mother, and, if present, father. Comparisons of youth in the in-home sample to the original sample at baseline indicated no differences in substance use initiation, but youth in the in-home sample were slightly less likely to engage in delinquent behavior than those in the original sample (F [1, 27] = 18.32, p<.01, d =.15). Although similar in most dimensions to the general population of cases, the in-home subsample may be at slightly lower risk for problem behavior.

The focus of this study was IPC; thus, two-parent families were selected for analyses (n = 768). We defined two-parent families by parents’ responses to a marital status question as either: a) married and living with their spouse, or b) living with someone in a steady, marital-like relationship. No data were collected about the length of relationship. This two-parent sample included 768 families at Wave 1. At W1, 47% of adolescents were male and participant ages were: adolescents (M = 11.3 years, SD = .49); female caregivers (M = 38.7, SD = 6.05); and male caregivers (M = 41.2, SD = 7.14). Female caregivers included mothers (94.9%), stepmothers (1.3%), and other parental figures (3.8%; e.g., parents’ significant other, foster parent). Male caregivers included fathers (75.3%), stepfathers (16.9%), and other parental figures (7.8%). Sixty-one percent of families resided in Iowa and 39% lived in Pennsylvania. The median household income was $52,000 (in 2003) and 64% of adolescents had parents with some post-secondary education (more detail provided in measures). Adolescents identified their race as White (89%), Hispanic (6%), African American (1%), Asian (1%), or Other (3%).

Procedure

The larger PROSPER trial was an evaluation of a community-level implementation of family-based and universal programs during 6^th and 7^th grades, respectively, that were intended to prevent substance use (see: Spoth, Guyll, Lillehoj, Redmond, & Greenberg, 2007). This larger trial randomized 28 rural communities and small towns in Iowa and Pennsylvania to either receive the PROSPER intervention delivery system or not (controls). It is worth noting that the PROSPER trial examined a universal population, thus, all adolescents and families in participating communities were invited to participate.

Measurement occasions were the Fall of 6^th grade (W1), 6 months later in Spring of 6^th grade (W2), and annually thereafter in 7^th (W3), 8^th (W4), 9^th (W5), 10^th (W6), and 11^th (W7) grades. Sample retention rates were: 97% at W2, 93% at W3, 80% at W4, 74% at W5, 67% at W6, and 61% at W7. Data on IPC, family demographic information, adolescent threat appraisals, and internalizing problems were collected in the in-home family assessments in the Fall of 6^th grade, Spring of 6^th grade, and Spring of 7^th grade. Of families retained in the study, 92% of couples remained together over the course of W1-3 family assessments. Substance use data were drawn from in-school assessments for the larger PROSPER study during the Spring of 6^th through 11^th grades. Youth were assured that their information would be kept private and that their answers would not be shared with school staff, teachers, peers, or their families. The University Institutional Review Board approved this study.

Measures

Interparental conflict

Mothers and fathers responded to seven items about the frequency of conflict behaviors over the past month from a measure that was adapted from the Iowa Youth and Families Project (Conger, 1989) at W1. These items had a 7-point scale, ranging from always (1), almost always (2), fairly often (3), about half the time (4), not too often (5), almost never (6), to never (7). Items were reverse-coded so that higher values reflect more frequent conflict. Parents reported on their own behavior and their partners’ behavior, including yelling, criticism, and hitting, shoving, pushing, shoving. Prior work has found that this scale is correlated with marital dissatisfaction and marital distress (Cui & Conger, 2008) and with observer ratings of marital conflict (Harold, Fincham, Osborne, & Conger, 1997). Reliabilities were acceptable for mother and father reports (.84–.89). Adolescent report was drawn from one item at W1: “thinking about your parents or guardians, how often would you say they argue or disagree with each other?” Responses ranged from always (1) to never (5). Scores were recoded so that higher values reflected more frequent IPC. Correlations of mother, father, and adolescent reports ranged from .41 to .59 (p’s<.01). Seven hundred sixty-one mothers (99.1%), 641 fathers (83.5%), and 737 youth (96.0%) provided data on IPC; when these indicators are combined in a latent variable, all 768 families provided data.

Threat Appraisals

Adolescents completed four items adapted from the Children’s Perceptions of Interparental Conflict Scale (CPIC; Grych et al., 1992) at W1 and W2. Items included: “When my parents argue, I’m afraid that something bad will happen”, “When my parents argue I worry that one of them will get hurt”, “When my parents argue I’m afraid that they will yell at me too”, and “When my parents argue I worry that they might get divorced.” Youth rated items on a 5-point scale: strongly agree (1), agree (2), neutral or mixed (3), disagree (4), or strongly disagree (5). This scale was computed as an item average and scaled so that higher values reflected greater threat appraisals (reliabilities .86–.87).

Internalizing Problems

Adolescent reports were selected over parent reports, because adolescents are more accurate informants of internalizing symptoms (Lagattuta, Sayfan, & Bamford, 2012). Adolescents completed the depressed/anxious subscale of the Child Behavior Checklist, Youth Self-Report (YSR) (Achenbach, 1991) at W1 and W3. Adolescents rated how true each item was for them “now or within the past six months” on a scale ranging from not true (0) to very true or often true (2). Reliability was acceptable at W1 (.85) and W3 (.87). Items were averaged and higher values reflected more internalizing symptoms.

Cigarette and Alcohol Use

During Waves 2–7, adolescents reported on their past month frequency of cigarette and alcohol use. Items were: “During the past month, how many times have you smoked any cigarettes?” and “During the past month, how many times have you had beer, wine, wine coolers, or other liquor?” Item response options included: not at all (1), one time (2), a few times (3), about once a week (4), more than once a week (5). As described below, these items were analyzed separately to evaluate hypotheses for alcohol and cigarette use. Single-item measures of past-month substance use are commonly used and tend to fit well in latent trajectory models (e.g., Buu, Dabrowska, Heinze, Hsieh, & Zimmerman, 2015; Nelson, Van Ryzin, & Dishion, 2015; Passarotti, Crane, Hedeker, & Mermelstein, 2015). In addition, there is evidence indicating that single-item measures of substance use are reliable and valid measures (Dollinger & Malmquist, 2009). In the current sample, the proportions of youth who indicated any past-month alcohol use were 7.6% (W2), 11.0% (W3), 19.0% (W4), 24.8% (W5), 30.0% (W6), and 33.6% (W7). The proportions of youth who indicated any past-month cigarette use (compared to non-use) were 2.5% (W2), 3.9% (W3), 7.8% (W4), 13.2% (W5), 13.2% (W6), and 18.0% (W7).

Covariates

Parents reported on their education, income, and parent substance use at the initial measurement occasion (W1). Parent education was scored 0–20, such that 0–16 reflected each year completed through a Bachelor’s degree, 17 = beyond Bachelor’s degree, 18 = Master’s degree, 19 = beyond Master’s, and 20 = doctoral degree. The average education level in this study was 13.25 (SD = 2.19). Family income was rated on a 1–11 scale from 1 (up to $10,000/year) to 11 ($100,000+/year). Mean family income was 6.09 (SD = 2.63), falling in the $50,000–59,999 range. Parents were asked a series of yes/no questions as to whether they used tobacco (cigarettes, pipe, cigars, or chewing tobacco), alcohol, or illegal drugs. Parent substance use was recorded as any use by a parent, such that 0 = no parental use, 1 = one parent use, and 2 = both parents used substances (M = 1.13, SD = .60). Parenting practices were assessed by youth report on the general child management scale (Spoth, Redmond, & Shin, 1998), which includes ratings of parental monitoring, consistent parenting practices, and harsh parenting practices (1 = almost never to 5 = almost always). Items were averaged and scored so that higher values reflected more effective parenting (α = .79; M = 5.64, SD = .78).

Results

Missing Data

Little’s MCAR tests were conducted for the model predicting past month cigarette use (χ² (973) = 1256.63, p<.01) and for the model predicting past month alcohol use (χ² (967) = 1180.892, p<.01) and suggested that data were not missing completely at random. We examined whether W1 IPC, threat appraisals, or internalizing problems were correlated with missingness at one or more waves. IPC was not associated with attrition; however, adolescents with higher threat appraisals and internalizing problems were slightly more likely to drop out at one or more later waves. Among covariates, parent education, family income, parent substance use were all correlated with missingness at one or more waves and thus they were included as covariates. Analyses were conducted using Full Information Maximum Likelihood (FIML) procedures to reduce bias due to attrition (Widaman, 2006).

Preliminary Analyses

Descriptive statistics and correlations are presented in Table 1. Longitudinal models were conducted using Mplus 7.3 (Muthén & Muthén, 2013), with Robust Maximum Likelihood (RML) estimation to avoid assumptions of normal distribution of the variables (i.e., skew in substance use). RML was deemed most appropriate because measures of substance use were not count variables. RML employs a FIML treatment of missing data to minimize potential bias incurred due to missing data (Widaman, 2006). Models were evaluated across three criteria for acceptable fit: comparative fit index (CFI) and Tucker-Lewis index (TLI) > .90, and Root Mean Square Error of Approximation (RMSEA) < .08. Excellent model fit was indicated by values for CFI/TLI > .95 and RMSEA < .05 (Hu & Bentler, 1999).

Table 1.

Correlations, Means, and Standard Deviations of Study Variables

	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17	18	19	20	21
1. IPC 1	–
2. Thr 1	.37**	–
3. Thr 2	.30**	.55**	–
4. Int 1	.21**	.33**	.27**	–
5. Int3	.16**	.26**	.29**	.48**	–
6. Cigs2	.03	.04	.07	.06	.09*	–
7. Cigs3	.12**	.13**	.13**	.01	.03	−.02	–
8. Cigs4	.04	.07	.10*	.06	.02	.06	.40**	–
9. Cigs5	.08	.06	.10*	.13**	−.03	.03	.26**	.67**	–
10. Cigs6	.11*	.09*	.07	.07	.06	.07	.17**	.46**	.59**	–
11. Cigs7	.15**	.00	.07	.08	.09	.16**	.03	.37**	.54**	.63**	–
12. Alc2	−.04	.04	.07	.07	−.03	.20**	.01	.12**	.13**	.06	.12*	–
13. Alc3	.16**	.13**	.15**	.06	.07	.00	.44**	.27**	.20**	.16**	.11*	.34**	–
14. Alc4	.12**	.08*	.05	.03	−.01	.10*	.28**	.44**	.39**	.27**	.22**	.17**	.33**	–
15. Alc5	.06	.03	−.01	.02	−.03	−.01	.24**	.29**	.44**	.45**	.30**	.15**	.31**	.47**	–
16. Alc6	.11*	.02	.00	−.05	−.01	.08	.09*	.27**	.32**	.59**	.46**	.09*	.21**	.28**	.51**	–
17. Alc7	.10*	−.03	−.00	−.02	−.01	.10*	−.02	.17**	.27**	.36**	.55**	.23**	.13**	.18**	.43**	.48**	–
18. P Edu 1	−.05	−.20**	−.15**	−.02	−.04	.05	−.10**	−.11**	−.16**	−.14**	−.11*	.08*	−.05	−.04	−.08	−.10*	−.07	–
19. P−SU 1	.15**	.11**	.13**	.03	.02	−.03	.05	.04	.18**	.09*	.16**	.09*	.09*	.08*	.15**	.15**	.16**	−.01	–
20. Inc. 1	−.09*	−.21**	−.13**	−.08*	−.10*	−.01	−.09*	−.11**	−.14**	−.15**	−.05	.00	−.02	−.02	−.06	−.04	.02	.50**	−.06	–
21. Par 1	−.29**	−.32**	−.26**	−.19**	−.11**	.01	−.04	−.12**	−.15**	−.12**	−.10*	−.08*	−.16**	−.10*	−.08	−.02	−.05	.21**	−.03	.17**	–
N	768	737	636	767	615	685	634	617	570	517	471	685	634	617	569	516	470	755	768	702	765
M	.00	2.25	2.17	.22	.20	1.02	1.08	1.18	1.31	1.38	1.49	1.09	1.16	1.30	1.43	1.57	1.67	13.25	1.13	6.1	5.65
SD	.84	1.08	1.08	.26	.28	.18	.45	.74	.91	1.06	1.18	.36	.50	.70	.86	1.01	1.09	2.19	0.6	2.63	.78
Skew	1.21	.66	.75	2.07	2.05	9.50	7.27	4.43	3.24	2.77	2.33	4.42	3.46	2.64	2.17	1.75	1.48	−0.34	0.1	0.26	.09

^**p < .01,

^*p < .05

P Edu = Parent Education, P-SU = Parent Substance Use, Income = Family Income, Par = General Parenting Practices; IPC =

Interparental Conflict, Thr = Threat, Int = Internalizing Problems, Cigs = Past Month Cigarette Use Frequency, Alc = Past Month Alcohol Use Frequency. Numbers following labels reflect the assessment wave (e.g., Thr1 refers to Threat at Wave 1). Interparental conflict (IPC) in this table is a composite variable, derived of an average of z-score transformed mother, father, and youth reports used in the latent model.

First, unconditional growth models were fit for cigarette and alcohol use separately. For both, the best fitting models included a quadratic term and both models yielded a good fit with the data (Table 2). The average alcohol use trajectory started with few youth reporting alcohol use at W1, increasing rates of use over time, but leveling off by W7. Intercept, slope, and quadratic terms had statistically significant variance, indicating meaningful individual differences from the average alcohol use trajectory. Similar to alcohol use, the cigarette use model indicated that the average trajectory started low, increased over time, and leveled off. Only the slope and quadratic terms had statistically significant variance, indicating meaningful individual differences in the rates of change (slope, quadratic) over time. However, there was not meaningful variation from the mean level at the intercept. Thus, the conditional model predicting adolescent cigarette use was estimated as a fixed intercept model.

Table 2.

Summary of Unconditional Substance Use Growth Models

Unconditional Model:	χ2(12)	CFI	TLI	RMSEA	Intercept		Slope		Quadratic
					M	VAR	M	VAR	M	VAR
1. Alcohol Use	29.993*	.97	.97	.03	1.09**	11**	07**	.12**	.01**	.01**
2. Cigarette Use	68.336*	.97	.95	.03	1.02**	.00ns	.06**	.13**	.01**	.01**

Note.

^*p < .05

^**p < .01

M = Mean, VAR = Variance

Next, we estimated conditional models to examine how IPC, threat appraisals, and internalizing problems might predict individual differences in trajectories of alcohol use and cigarette use. Our primary hypotheses focused on whether threat and/or internalizing problems predicted linear rates of change over time (slope); however, we retained quadratic terms in the predictive models to provide the cleanest possible estimate of linear change over time by disentangling linear (slope) and curvilinear (quadratic) aspects of the trajectories. Although we did not have specific hypotheses related to predicting a curvilinear form, we regressed the quadratic terms on covariates and predictors for exploratory analyses. In both models, all W1 predictors and covariates were allowed to co-vary. In addition, because of temporal overlap, W2 threat appraisals and W3 internalizing problems were allowed to correlate with the intercept (in the alcohol model) to avoid bias in other parameters caused by covariance among these constructs.

Predicting Alcohol use Trajectories

The first model, presented in Figure 1, fit the data well. Findings related to IPC predicting increases in threat appraisals (β = .11), and threat appraisals predicting increases in internalizing problems (β = .19), have been reported in a prior publication (Fosco & Feinberg, 2015). Among covariates, parental substance use predicted slightly higher initial levels of alcohol use (β = .12) and effective parenting practices predicted lower initial levels (β = −.14). Relevant to hypotheses in this study, higher levels of IPC were related to slightly below-average initial levels of alcohol use at W2 (intercept), but were related to higher than average rates of escalation in alcohol use over time (β = .20). Neither threat appraisals nor internalizing symptoms explained variance in the slope term; therefore, our results do not support the “threat appraisals as a direct risk for substance use” and the “indirect risk via internalizing problems” hypotheses for alcohol use. None of the predictors were associated with curvilinear change in alcohol use (quadratic term). Thus, only IPC predicted increases in alcohol use over time.

Figure 1.

A model predicting latent trajectories of adolescent alcohol use.

Model fit: χ²(74) = 150.72, p = .00; CFI = .96; TLI = .93; RMSEA = .032 (90%: .023–.041)

Statistically significant paths (p < .05) are represented with solid lines; dashed lines were not statistically significant. In the growth model, IPC = Interparental Conflict, Icept = Intercept, Parent Edu = Parent Education, Parent Subst. = Parent Substance Use. Numbers after labels reflect measurement occasions: 1 = Fall 6^th Grade; 2–7 = Spring 6^th–11^th Grades.

All W1 predictors and covariates were modeled with a saturated covariance structure.

Post-hoc tests were conducted to determine whether the magnitude of path coefficients in the model differed for a) families in intervention and control groups and b) male or female adolescents. Multiple-group model invariance tests were conducted by comparing CFI values for models in which the paths were freely estimated across groups to models in which path coefficients were constrained to be equal across groups. Change in the CFI (ΔCFI) of greater than .01 indicated meaningful differences in models across groups (Cheung & Rensvold, 2002). Model fit did not differ for the invariance test comparing intervention and control groups (ΔCFI = .009). The comparison of constrained and unconstrained models suggested that findings were not different for boys and girls (ΔCFI = .002). We also conducted an additional post-hoc test to evaluate the possibility that mother-reported youth internalizing problems might predict alcohol use in this model; however, the pattern of results remained the same.

Predicting Past Month Cigarette Use

A fixed intercept conditional model was estimated for cigarette use following the same procedures as the model for alcohol use, and this model yielded a good fit with the data (Figure 2). Among the covariates, only one was a significant predictor of cigarette use. Adolescents from families with higher income exhibited fewer increases in cigarette use over time (β = −.13). Related to the study hypotheses, only threat appraisals were associated with cigarette use. Specifically, adolescents who felt more threatened by IPC exhibited higher than average rates of escalation in cigarette use over time (β = .13). Internalizing problems were not correlated with cigarette use over time. None of the study variables predicted the curvilinear rate of change in cigarette use.

Figure 2.

A model predicting latent trajectories of adolescent cigarette use.

Model fit: χ²(84) = 204.85, p = .00; CFI = .95; TLI = .93; RMSEA = .032 (90%: .024–.041)

Statistically significant paths (p < .05) are represented with solid lines; dashed lines were not statistically significant. IPC = Interparental Conflict, Icept = Intercept, Parent Edu = Parent Education, Parent Subst. = Parent Substance Use. Numbers after labels reflect measurement occasions: 1 = Fall 6^th Grade; 2–7 = Spring 6^th–11^th Grades

All W1 predictors and covariates were modeled with a saturated covariance structure.

Post-hoc multiple group invariance tests indicated that the model did not differ for intervention and control groups (ΔCFI = .005) or for boys and girls (ΔCFI = .005). These comparisons suggest that the overall model was representative of the full sample. As with the alcohol use model, we re-estimated the model using mother-reported youth internalizing problems, but the pattern of results remained the same.

Discussion

Little attention has been given to IPC as a risk factor for adolescent substance use, despite well-documented evidence of IPC as a risk factor for emotional and behavioral problems (Fosco & Feinberg, 2015; Fosco & Grych, 2008; Grych & Fincham, 2001; Grych et al., 2003). In this study, we evaluated adolescents’ threat appraisals of IPC as a proximal risk factor for escalations in adolescent substance use. Prior work examining the cognitive contextual framework, applying autoregressive longitudinal models, has found that IPC is related to increases in threat appraisals, which in turn is related to increases in internalizing symptoms (Fosco & Feinberg, 2015; Grych et al., 2003). This study extended prior work by examining how IPC, threat appraisals, and internalizing problems were associated with patterns of escalating alcohol and cigarette use over a 5-year period. Two alternative hypotheses were tested: a) threat appraisals would represent a direct risk pathway to escalation in substance and b) internalizing problems would mediate the association between threat appraisals and escalation in substance use over time. The results revealed that there were distinct risk pathways for different substances. IPC predicted escalating alcohol use over time and threat appraisals predicted escalating cigarette use over time. Internalizing problems were not associated with alcohol or cigarette use in either model. These findings held even when accounting for parenting quality in the home, parent substance use, family income, and parent education.

Threat as a Cigarette Use-Specific Risk Mechanism

Threat appraisals were associated with increases in cigarette but not alcohol use, lending partial support to the “threat appraisals as a direct pathway to substance use” hypothesis. Specifically, adolescents who experienced higher levels of threat appraisals tended to have relatively more rapid escalations in cigarette use over time. This direct effect was found even when accounting for internalizing problems, which have been identified as a risk factor for cigarette use in prior studies (Lewis et al., 2011; Wu et al., 2004). These findings suggest that there is a robust association between adolescents’ threat appraisals and their engagement and escalation in cigarette use.

The findings that threat appraisals are associated with cigarette but not alcohol use suggests that adolescents who perceive parental conflicts as threatening may have a specific vulnerability to cigarette use. To the extent that IPC, and threat appraisals in particular, undermine adolescents’ beliefs that they can cope with worries about interparental conflicts (Fosco & Grych, 2010), adolescents may specifically seek out cigarettes because of common perceptions that nicotine has anxiolytic (anxiety reducing) properties (Parrott, 1999). Although both alcohol and tobacco may be used for self-medication purposes, adolescents may use cigarettes specifically to manage fears that feel beyond their sphere of control (Gehricke et al., 2007). The desire to feel in control or vigilant amidst potential family conflict situations may also explain why alcohol, which includes cognitive blunting and diminished inhibition, is not appealing as a means of coping with adolescents’ fears related to IPC. In this way, cigarettes may offer unique appeal for their anxiolytic properties that is not shared with alcohol. Overall, these findings underscore threat appraisals as a new risk factor for cigarette use and offers a specific avenue for prevention programs.

IPC as a Risk Factor for Alcohol Use

Interestingly, IPC was directly associated with higher than average increases in adolescent alcohol use, even when threat appraisals and internalizing problems were included in the model. This direct association calls for further exploration of what other processes may explain the link between IPC and alcohol use. Drawing from the cognitive-contextual framework, self-blaming attributions – adolescents’ beliefs that they are responsible for causing or resolving parental disagreements – has been demonstrated to be a key risk factor for externalizing problems (Fosco & Grych, 2008; Grych et al., 2003). Perhaps self-blame is a unique pathway related to alcohol use. Another possibility is that IPC may be related to alcohol use through disruptions to effective parenting practices during adolescence. Although we included effective parenting practices as a covariate, other work emphasizes spillover from IPC to diminished parental monitoring, behavior management, and parental warmth (Buehler, Benson, & Gerard, 2008; Krishnakumar & Buehler, 2000). These parenting factors are proximal factors for substance use risk (Fosco, Stormshak, Dishion, & Winter, 2012; Lippold, Fosco, Ram, & Feinberg, 2015). Future work is needed to explicate how IPC places adolescents at risk for escalating patterns of alcohol use.

Finally, it was surprising that internalizing problems did not predict alcohol or cigarette use in this study. This finding runs contrary to other work that identifies internalizing problems as a risk factor for substance use (e.g., Brook et al., 2008; Marmorstein, 2009). However, some have noted that the evidence is inconsistent with regard to the link between internalizing problems and substance use, particularly when using broad-band assessments (e.g., YSR) compared to more specific assessments of depressive symptoms (Stone, Becker, Huber, & Catalano, 2012). Thus, our findings may fit with a growing body of evidence that fails to find a link between broad-band assessments of internalizing problems and adolescent substance use.

Limitations and Future Directions

The findings of this study should be considered within the context of its limitations. The sample was drawn from a rural, semi-rural, and small town community population that primarily comprised of White families. Future work should seek to replicate these findings in samples that offer more racial, geographical (rural/urban), and socioeconomic diversity. Another limitation of this study is the fact that there was meaningful, non-random attrition in our sample. Although sample attrition was addressed through FIML analytic methods, the issue that at risk youth dropped out at later waves may introduce persisting bias in the sample despite analytic efforts to minimize it.

In addition, assessments in this study did not include other cognitive appraisals, such as self-blame, which may be important risk processes for substance use (Tschann et al., 2002). In particular, future research might examine whether particular combinations of appraisals may offer new insights into substance use risk (Fosco & Bray, 2016). Measurement was limited to survey data. Replication of these findings with more objective assessments of IPC would be valuable. In addition, this study used single-item measures of alcohol and cigarette use. Although this approach is commonly used and considered valid, future work might consider more expansive measures of substance use in order to offer a more complete assessment of alcohol and tobacco use and potentially enhance the strength of findings. In addition, it may also be helpful o consider the role of IPC and threat appraisals for early substance use initiation, given the importance of this issue, as well (Odgers et al., 2008). Looking forward, it would be valuable to examine other family factors as a context for adolescent appraisals. Prior work has found that parenting quality and parent-adolescent connectedness are related to substance use risk, but also may temper adolescent threat appraisals (e.g., DeBoard-Lucas et al., 2010; Fosco, Stormshak, et al., 2012; Lucas-Thompson & George, In Press). The results would offer a valuable integration of family risk models that focus on externalizing or deviance pathways with those that examine internalizing pathways of risk for adolescent substance use.

Conclusion and Implications for Interventions

IPC is a robust risk factor for poor youth adjustment that has been largely overlooked in relation to adolescent substance use risk. Guided by the cognitive-contextual framework, this study provides support that threat appraisals are a proximal risk mechanism linking IPC and adolescent escalations in cigarette use over the course of adolescence. In addition, IPC was directly correlated with escalations in alcohol use over adolescence, suggesting that further exploration of risk mechanisms for alcohol use is needed. From a developmental psychopathology perspective, the current study offers insights into unique pathways of risk for cigarette and alcohol use, flowing out of IPC. These findings provide the ability to differentiate risk for different types of substance use, which is often not the case in family research (Van Ryzin et al., 2012). Further, these findings offer guidance to interventionists who may seek to incorporate IPC and threat appraisals into family assessments when determining risk for adolescent substance use. Further, most family-based programs aimed at preventing or reducing adolescent substance use do not address IPC in families (Van Ryzin, Roseth, Fosco, Lee, & Chen, 2016). The present findings suggest that adding intervention components that reduce IPC may bolster effects for programs that have focused on parenting practices.