Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2021 Aug 1.
Published in final edited form as: J Fam Psychol. 2020 Feb 13;34(5):587–597. doi: 10.1037/fam0000640

Constellations of Family Risk for Long-Term Adolescent Antisocial Behavior

Emily J LoBraico 1, Bethany C Bray 2,3, Mark E Feinberg 4, Gregory M Fosco 1,4
PMCID: PMC7375013  NIHMSID: NIHMS1069137  PMID: 32052984

Abstract

Adolescent antisocial behavior (ASB) can have long-term individual and societal consequences. Much of the research on the development of ASB considers risk and protective factors in isolation or as cumulative indices, likely overlooking the co-occurring and interacting nature of these factors. Guided by theories of ASB risk (i.e., coercive family process, disengagement), this study uses latent profile analysis to evaluate whether there are subgroups of families in the population that conform to specific constellations of risk factors prescribed by established theories of risk for ASB, and whether subgroup membership confers differential risk for different ASBs. We leveraged a large sample of adolescents in Fall, Grade 6 (N = 5,300; Mage = 11.8; 50.9% female) for subgroup analysis, and predicted aggression, antisocial peer behavior, and substance use in Spring, Grade 8.

Four family profiles were identified: Coercive (15%), characterized by high family conflict, low positive family climate, low parental involvement, low effective discipline, low adolescent positive engagement, and low parental knowledge; Disengaged (41%), characterized by low positive family climate, low parental involvement, low adolescent positive engagement, and low parental knowledge; Permissive (11%), characterized by high parental involvement, low effective discipline, high adolescent positive engagement, high parental knowledge, and high family conflict; and High Functioning (34% prevalence). In turn, group membership predicted long-term outcomes. Adolescents in Coercive families were at highest risk for ASB during Grade 8, followed by those in Disengaged and Permissive profiles; all three of which were at greater risk than adolescents in High Functioning families for every outcome.

Keywords: Antisocial Behavior, Family Functioning, Latent Profile Analysis, Coercion, Disengagement

Introduction

Antisocial behavior (ASB) is a broad term that refers to socially deviant behaviors, including aggression, violence, lying, delinquency, and substance use (Calkins & Keane, 2009; Hiatt & Dishion, 2008). Adolescent ASB is prognostic of problems into adulthood, including physical illness (Odgers et al., 2007), poverty (Samuelson, Hodgins, Larsson, Larm, & Tengström, 2010), and adult substance misuse and criminality (Dishion & Patterson, 2006). ASB onset prior to age 15 is particularly risky, as there is more time for adolescents to escalate in their behavior and to become ensnared in consequences of their ASB (Calkins & Keane, 2009; Craig, Morris, Piquero, & Farrington, 2015). Decades of work have identified several risk and protective factors within the family context that can contribute to adolescent problem behavior development. What is less known is if there are distinct subgroups of family functioning that reflect different patterns of the established risk and protective factors, and if these subgroups are related to the development of different ASBs. Extant research on family risk and protective factors for ASB development are typically studied as individual factors in isolation or cumulative indices. And, although risk and protective factors often co-occur and interact with each other (January et al., 2017; Lanza, Rhoades, Nix, & Greenberg, 2011), we have yet to identify patterns of co-occurring family-level risk and protective factors for ASB development. Once identified, typologies can be used to screen individuals for risk, target intervention strategies, and even develop appropriate intervention content (Lanza & Rhoades, 2013). The present study aims to extend knowledge of ASB etiology by identifying distinct subgroups of family functioning and evaluating their associations with adolescent ASB.

The Family’s Influence on Adolescent ASB: Risk and Protective Factors

Key risk and protective factors for adolescent ASB development have been identified across multiple dimensions of family functioning. Three family domains are particularly noteworthy in their influence on adolescent ASB development: family-level functioning, parent-child relationship quality, and parenting skills (Fosco & LoBraico, 2018). Family-level functioning (Jaycox & Repetti, 1993; Moos & Moos, 1994), which includes family conflict and positive family climate as two distinct (Forgatch & Degarmo, 1999) domains, is predictive of risk for adolescent ASB. Family conflict involves frequent expressions of anger, hostility, and resentment, and consistently predicts adolescent aggression in prospective longitudinal studies (Benson & Buehler, 2012; Fosco, Caruthers, & Dishion, 2012). Positive family climate, a measure of closeness, supportiveness, and warmth between family members, is negatively associated with adolescent ASB (Benson & Buehler, 2012; Fosco et al., 2012; Van Ryzin, Fosco, & Dishion, 2012).

Features of the parent-child relationship are also implicated in adolescent ASB development. Parental involvement – the degree to which parents spend time with their child and participate in joint activities together – predicts less adolescent substance use (Pilgrim, Schulenberg, O’Malley, Bachman, & Johnston, 2006), perhaps by promoting strong bonds and connectedness in the parent-child relationship which protect against ASB in the middle school years (Fosco, Stormshak, Dishion, & Winter, 2012). Adolescents’ positive engagement with the family, indicated by voluntarily spending time and expressing warmth and affection with their parents, is another indicator of relationship quality that reflects their access and receptiveness to parental guidance, thereby reducing their risk for substance use (Ackard et al., 2006). A strong parent-adolescent relationship, reflected by parental involvement and adolescent positive engagement in the family, is protective against the development of adolescent ASB.

Two parenting skills, parental knowledge and effective discipline, are key features of family life and important protective factors for adolescent ASB. Parental knowledge (sometimes referred to as parental monitoring), or the awareness of adolescents’ activities and whereabouts, is a robust protective factor for a host of adolescent problem behaviors including substance use and deviant peer involvement (Dishion, Nelson, & Kavanagh, 2003; Fosco et al., 2012; Kiesner, Dishion, Poulin, & Pastore, 2009). Effective discipline – referring to the consistent use of praise, contingent reinforcers for desired behavior, inductive reasoning, and the absence of harsh parenting – reduces risk for adolescent ASB (Halgunseth, Perkins, Lippold, & Nix, 2013; Surjadi, Lorenz, Conger, & Wickrama, 2013). Effective disciplinary strategies reduce risk for ASB by managing adolescent problem behaviors while promoting prosocial behaviors (Burnette, Oshri, Lax, Richards, & Ragbeer, 2012; Laible, Eye, & Carlo, 2008; Lindahl, 1998). Overall, skillful parenting (i.e., parental knowledge, effective discipline) may protect against adolescent ASB by supporting the development of self-regulatory skills, helping adolescents manage developmentally normative impulses to engage in risky behavior (Fosco et al., 2012).

Person-Centered Approaches to Distinguishing Risk

Considered separately, each of the key factors above can only tell part of the story of adolescent risk. By integrating these factors, we may glean new insights into family functioning and the consequences of different family contexts for later adolescent ASB. Recent studies leveraging person-centered approaches indicate that youth in families with lower overall functioning, disengaged or unattached families, enmeshed or anxious families, and families with a negative father-child relationship or sibling relationship are at risk for a variety of internalizing symptoms and substance use longitudinally (Simpson, Vannucci, & Ohannessian, 2018; Skinner & McHale, 2016; Sturge-Apple, Davies, & Cummings, 2010; Withers, McWey, & Lucier-Greer, 2016, Xia, Weymouth, Bray, Lippold, Feinberg, & Fosco, In Press). These studies complement the previous variable-centered findings, and advance our ability for both identification of distinguishable subgroups at risk for specific outcomes, and our ability to offer specific groups appropriate and tailored intervention content (Dishion, Mun, Ha, & Tein, 2019).

Despite these contributions to the literature, much of the research on risk for ASB development continues to rely on cumulative risk indices or individual risk factors in a variable-centered framework to identify at-risk youth. Building on this strong evidence base, alternative approaches that can more closely capture the holistic, multidimensional qualities of family life may advance our understanding of these risk processes by capturing how risk and protective factors for ASB co-occur and cluster together in different subgroups of families. In doing so, we can more accurately characterize the experiences of adolescents within their family context, and determine which family contexts confer risk for which ASBs. Thus, we applied latent profile analysis ([LPA]; Collins & Lanza, 2010), a person-centered approach that is a powerful tool for identifying latent subgroups of co-occurring risk and protective factors that exist in a population and evaluating whether subgroup membership is a predictor of consequences.

Holistic approaches to conceptualizing families (e.g., LPA) may be uniquely suited to capture multiple elements identified in theories of risk for ASB. Key models of family risk, such as coercion (Patterson, 1982) and disengagement (Ary et al., 1999) processes, conceptualize vulnerability for ASB as involving multiple, overlapping elements of family functioning. Thus, coercion and disengagement processes may be reflected by specific patterns of co-occurring risk and protective factors in the family (i.e., family-level factors, parent-child relationships, and parenting skills). Perhaps the most established theory of family risk for ASB is coercive family process, which involves aversive, harsh, and escalating conflicts, that only terminate after one party (often the parent) acquiesces (Patterson, 1982, 2016). These coercive interactions effectively train ASB by shaping problematic behavioral patterns and can lead to youth aggression (e.g., social interaction learning; SIL; Patterson, 1982). Entrenched coercive interactional patterns can lead to declines in relationship quality between adolescents and their families, and/or poor whole-family functioning (Bullard et al., 2010; Patterson, 2016) Thus, coercive families may be identified by the combination of elevated family-level conflict, elevated harsh and ineffective parenting practices, low adolescent positive engagement in the family, and low positive family climate.

Disengagement refers to a developmental risk process in which parents and/or adolescents withdraw, leaving the adolescent to seek guidance and support from peers instead, who are inadequate surrogates for these relationships (Dishion, Nelson, & Bullock, 2004; Fosco & LoBraico, 2018). Disengaged adolescents spend an increased amount of unstructured time apart from the family, and are at risk for joining other unsupervised peers that may be engaged in ASBs and substance use (Fosco & LoBraico, 2019; Van Ryzin & Dishion, 2014). Thus, family disengagement theory may be characterized as a combination of: low levels of positive family climate, low parental knowledge, low parental involvement, and low adolescent positive engagement in the family.

Although coercion and disengagement are two influential theories of risk for adolescent ASB in the family context, they are not exhaustive of the potential types of subgroups of family functioning risk. There may be other patterns of co-occurring risk and protective factors across the domains of family functioning that reflect other less-discussed family risk processes. For instance, there may be a subgroup of families who are more inconsistent in their risk, such that they have poor parenting skills, but positive family-level and parent-adolescent relationships. Moreover, different subgroups of family functioning risk may be associated with different types and degrees of ASB (e.g, substance use, aggression), as is hypothesized to be the case for coercive and disengaged families. With LPA, we can examine risk and protective factors as indicators that reflect risk processes in our model and determine whether coercion, disengagement, and/or other family process profiles emerge as distinct subgroups. We can then compare their prediction of a variety of ASB outcomes to further examine their distinctiveness, offering new insights into the etiology of adolescent ASB.

Present Study

The aims of the present study were to (1) identify latent profiles of family functioning risk factors at the start of middle school that may (2) predict risk for ASB by the end of middle school. We incorporated aspects of family-level functioning, parent-adolescent relationships, and parenting practices to categorize families into subgroups of co-occurring risk and protective factors, that reflect theorized family risk processes in models of adolescent ASB. We expected to identify at least three family profiles reflecting: (1) Coercive family profile (characterized by high family conflict, low positive family climate, low adolescent positive engagement, and low effective discipline); (2) Disengaged family profile (characterized by low family conflict, low positive family climate, low parental involvement, low adolescent positive engagement, and low parental knowledge); and (3) High functioning family profile (characterized by low conflict and high values on all other variables). We also had an exploratory hypothesis for a fourth, inconsistent family profile (low family conflict, high positive family climate, high positive adolescent engagement, low parental knowledge, and low effective discipline).

In addressing our second study aim, we expected that there would be differences among the identified profiles in their prediction of ASB during Grade 8. Based on coercion and disengagement theories for ASB development, we expected that adolescents in coercive families would be trained to be aggressive, while those in disengaged families would engage with deviant peers and be exposed to substance use. Thus, we hypothesized the following: (H1) adolescents in high functioning families would demonstrate the lowest levels of all ASB outcomes, (H2) adolescents in coercive families would report the highest levels of aggression, and (H3) adolescents in disengaged families would have the most elevated antisocial peers and substance use. We also had an exploratory hypothesis about adolescents in the inconsistent families, such that we expected them to demonstrate higher levels of ASB than the high functioning families, and lower levels of ASB than the coercive and disengaged families.

Method

Participants and Procedures

Data were from the PROmoting School-Community-University Partnerships to Enhance Resilience (PROSPER) trial (Spoth, Greenberg, Bierman, & Redmond, 2004): a partnership-based delivery system for evidence-based preventive interventions. The PROSPER study included 28 community school districts recruited from rural and semi-rural areas in Iowa and Pennsylvania, which were randomized at the beginning of the study to one of two conditions. The intervention condition received the PROSPER-delivered family-based and school-based intervention programs, and the control condition received “programming as usual.” Community eligibility criteria were (a) school district enrollment between 1300 and 5200 students, and (b) at least 15% of students eligible for free or reduced cost lunches. Participants were two cohorts of sixth graders; Cohort 1 began the study during the 2002–2003 school year and Cohort 2 began the study during the 2003–2004 school year. All research was conducted under the supervision of a university IRB and was in compliance with all ethical standards in the field.

Data were collected in classrooms during the Fall and Spring of Grade 6, followed by yearly assessments every Spring through Grade 12. A total of 10,845 students (approximately 90% of those eligible) completed the baseline assessment in the Fall of Grade 6. The analytic sample (N=5,300) includes the control group only, and excludes any individuals who were grade retained and thus provided duplicate data (n=4) or had an unspecified community (n=30). Analyses include data collected at the Fall Grade 6 and the Spring Grade 8 assessments. Adolescent demographic information can be seen in Table 5 (online supplemental material).

Measures

Latent profile indicators.

All indicators were measured in the Fall of Grade 6.

Family conflict.

Whole family conflict was assessed with three items about the overall conflict and anger in their families, drawn from the Family Environment Scale (Moos & Moos, 1994). The response scale ranged from 1 (Strongly disagree) to 5 (Strongly agree). An example item is “Family members hardly ever lose their tempers.” Reliability for this scale was α=0.50.

Positive family climate.

Positive family climate was assessed using four items about the overall environment in their households, including both family cohesion and organization in the home, drawn from the Family Environment Scale (Moos & Moos, 1994). The response scale ranged from 1 (Strongly disagree) to 5 (Strongly agree). An example item is “Family members really help and support each other.” Reliability for this scale was α=0.60.

Parental involvement.

Parental involvement was assessed using the Parent-Child Activities Scale. Adolescents responded to six items about how often they spent time doing activities with their parents in the past month. The response scale ranged from 1 (Not during the past month) to 6 (Everyday). An example item is “How often did you work on something together around the house?” Reliability for this scale was α=0.88.

Effective discipline.

Effective discipline was assessed using three subscales from the General Child Management Scale (Spoth, Redmond, & Shin, 1998). Adolescents rated their parents’ use of effective disciplinary strategies, including consistent discipline, harsh discipline, and inductive reasoning (Spoth, Redmond, & Shin, 1998). The response scale ranged from 1 (Always) to 5 (Never). An example item is “When my parents discipline me, the kind of discipline I get depends on their mood.” Reliability for this scale was α=0.84.

Parental knowledge.

Parental knowledge was assessed using the child monitoring subscale of the General Child Management Scale (Spoth, Redmond, & Shin, 1998). Adolescents responded to four items about their parents’ knowledge of the adolescent’s behaviors and activities. The response scale ranged from 1 (Never) to 5 (Always). An example item is “My parents know who I am with when I am away from home.” Reliability for this scale was α=0.80.

Adolescent positive engagement with the family.

Adolescents reported on their positive engagement in the family with three items adapted from the Affective Quality of the Relationship Scale (AQRS; Spoth, Redmond, & Shin, 1998). Adolescents responded to three items about their expression of affective quality toward each parent in the past year. The response scale ranged from 1 (Never or almost never) to 5 (Always or almost always). The six total items were averaged for one adolescent positive engagement score. An example item asks “During the past year, when you and your [father/mother] have spent time talking or doing things together, how often did you let [him/her] know you really care about [him/her]?” Reliability for this scale was α=0.79.

Outcome variables.

All outcome variables were measured in the Spring of Grade 8.

Hostile/aggressive behavior.

Adolescents reported on their hostile/aggressive behaviors with four items adapted from the widely-used National Youth Survey (Elliott, Huizinga, & Ageton, 1985). Adolescents responded to four items about how often they had engaged in specific hostile/aggressive behaviors. The response scale ranged from 1 (Never) to 5 (Five or more times). An example item is “Purposely damaged or destroyed property that did not belong to you.” Reliability for this scale was α=0.76.

Antisocial peer behavior.

In order to measure adolescent affiliation with antisocial peers, adolescents responded to three items about their closest friends’ ASB (Spoth & Molgaard, 1999). The response scale ranged from 1 (Strongly disagree) to 5 (Strongly agree). An example item is “These friends sometimes get into trouble with police.” Reliability for this scale was α=0.82.

Substance Use.

Adolescents responded to four individual items about their substance use, derived from the Monitoring the Future survey items (Johnston et al., 2017). Drunkenness frequency was measured with one item that asked adolescents how many times they got drunk from drinking wine, wine coolers, or other liquor in the past month. The response scale ranged from 1 (Not at all) to 5 (More than once a week). Inhalant use initiation was measured with a dichotomous item (1=Yes, 0=No) that asked adolescents if they had ever, “…sniffed glue, paint, gas, or other things you inhale to get high.” Marijuana use initiation was measured with one dichotomous item (1=Yes, 0=No) that asked adolescents if they had ever, “…smoked marijuana (grass, pot) or hashish (hash). Prescription opioid (PO) initiation was measured with one dichotomous item (1=Yes, 0=No) that asked adolescents if they had ever, “…used Vicodin, percocet, or Oxycontin.”

Demographic covariates.

Adolescents reported on their gender, eligibility for free/reduced price lunch (proxy for low income status), and household single-parent status.

Analysis Plan

Analysis proceeded in three phases. First, we identified and described latent profiles of Grade 6 family functioning. Second, we examined whether prevalence rates of profile membership differed based on demographic covariates in order to understand who belonged to the identified profiles. Third, we determined whether profile membership was related to Grade 8 adolescent hostile/aggressive behavior, deviant peer affiliation, and substance use behaviors.

LPA identifies patterns (i.e., profiles) across multiple characteristics within individuals instead of effects of single variables or interactions between variables across individuals. LPA is a type of finite mixture model that uses manifest items with continuous responses to divide a population into a set of mutually exclusive and exhaustive latent classes (i.e., profiles; Collins & Lanza, 2010). A standard LPA consists of two central sets of parameters. The first set are the latent profile membership probabilities, which describe the distribution of the profiles in the population. The second set are the item-response means (and variances), which describe the profile-specific item means (and variances). Profiles are interpreted and named based on the patterns of item means. A key assumption of LPA is that the latent profile indicators are normally distributed within classes (Bauer & Curran, 2003). However, two indicators in the current study, parental knowledge and adolescent positive engagement were severely negatively skewed, and attempts at transformation to achieve a normal distribution were unsuccessful. The skewed distributions each had an evident, clear, and theoretically supported point at which the indicator could be discretized to “high” and “low” levels. Thus, these scales were included as binary indicators in a “mixed” indicator LPA. This approach increased the plausibility of the model assumptions while permitting the inclusion of all indicators of interest.

Model selection was based on the Akaike information criterion (AIC; Akaike, 1974), Bayesian information criterion(BIC; Schwarz, 1978), sample-size adjusted BIC (a-BIC; Sclove, 1987), entropy (Celeux & Soromenho, 1996), and a bootstrapped likelihood ratio test (McLachlan, 1987; McLachlan & Peel, 2005), as well as model stability and interpretability. Lower values for the AIC, BIC, and a-BIC indicated better model fit; higher values for entropy indicated higher classification utility; significant bootstrapped likelihood ratio test p-values indicated better model fit compared to models with one fewer profiles. Emphasis was placed on the utility and theoretical interpretation of a solution. Model identification for all models was checked with 1,000 initial stage starts and 500 final stage starts; all models were estimated using Mplus version 8 (Muthén & Muthén, 1998–2017). Item-response variances were restricted to be equal across profiles by default to improve model identification.

After model selection, adolescent gender, low income status, and single parent status were added simultaneously to determine whether they were significant predictors of profile membership. Predictors were added using baseline-category multinomial logistic regression based on modal class assignment with classification-error correction (Vermunt, 2010; R3STEP command in Mplus). Effects of predictors on profile membership are expressed as odds ratios describing the increase in odds of membership in a particular profile (i.e., the target profile) compared to the reference profile. Any profile may be selected as the reference profile to facilitate interpretation. Missing data for covariate analyses were handled using listwise deletion.

Profile membership was then used to predict outcomes using linear and logistic regression based on modal class assignment with measurement-error weighting (Bakk & Vermunt, 2014; using BCH weights in Mplus). This approach is currently recommended for predicting continuous and binary distal outcomes from profile membership (Bakk & Vermunt, 2016; Dziak et al., 2016). Levels of aggression, deviant peer affiliation, and drunkenness frequency, as well as initiation of inhalants, marijuana, and PO during Grade 8 were regressed onto Grade 6 profile membership. Effects of profile membership on outcomes are expressed as the association with each outcome for each profile compared to the reference profile. Any profile may be selected as the reference profile to facilitate interpretation. Missingness on outcome variables was associated with demographic covariates (adolescent male gender, low income status, and single-parent status), thus, outcome analyses included these covariates as control variables; for each outcome model, corresponding Grade 6 adolescent behavior was also included as a control variable. Individuals who had already initiated inhalants (n=196), marijuana (n=65), and PO (n=34) by the Fall of Grade 6 were treated as missing in corresponding outcome models to preserve temporal precedence in prediction of substance use initiation.

Missing data for outcome analyses were handled using listwise deletion. Although not recommended generally, alternative approaches to handling missing data (e.g., multiple imputation) are not yet available when BCH weighting is used in the outcome analysis phase, but BCH weighting is the currently recommended approach to reduce (or eliminate) bias from other sources in these types of analyses (Bakk & Vermunt, 2016; Dziak et al., 2016). We understand the potential biases incurred when using listwise deletion, but also the methodological challenges and potential biases of choosing other approaches at this time. Thus, we have carefully assessed the sample reduction at each stage of analysis and tested whether the reduced sample is as generalizable as the original full sample. For the first phase, the analysis sample included n=5,209 adolescents (of the original 5,300 adolescents) who provided data on at least one of the 6 latent profile indicators during Grade 6. For the second phase, the sample included n=4,378 (of the 5,209 at phase one) who provided data on all of the demographic covariates. For the third phase, the sample included n=3,409 adolescents (of the 4,378 at phase two) who provided any outcome data at Grade 8 (and the corresponding baseline data as a control). Latent profile membership was not associated with missing data on covariates or outcomes. The same four profile model emerged when tested in each reduced sample.

Results

Descriptive statistics for adolescent demographic characteristics, latent profile indicators, predictors of profile membership, and outcomes are shown in Tables 1 and 2. Model fit information and model selection criteria are shown in Table 2. Models with 1–7 profiles were considered; the AIC, BIC, and a-BIC were not minimized. Additionally, the bootstrapped likelihood ratio test remained significant for each model. However, practical decrements in the fit criteria stopped around the 4-profile model. Entropy ranged from .69 to .73, with values for larger models in the lower .70s. Therefore, we considered models with 4 or 5 profiles. Upon examination of the 4- and 5-profile models, the 5-profile model included an additional, small, redundant, and theoretically uninterpretable profile, suggesting overextraction. Thus, we selected the 4-profile model for theoretical interpretation and additional analysis.

Table 1.

Descriptive Statistics for Profile Indicators and Outcomes

Indicator Frequency (valid %) or mean (SD) Min Max Skewness Kurtosis
Low family conflicta 3.46 (0.89) 1 5 −.149 −.272
Positive family climatea 3.79 (0.77) 1 5 −.500 .111
Parental involvementa 4.33 (1.28) 1 6 −.812 −.051
Effective disciplinea 3.65 (0.88) 1 5 −.389 −.505
Adolescent positive engagementb 2,168 (41.8%) 0 1 -- --
Parental knowledgeb 2,032 (40.3%) 0 1 -- --
Outcome
Aggression 1.29 (0.59) 1 5 3.03 11.16
Deviant peer affiliation 1.97 (0.99) 1 5 0.92 0.07
Drunkenness frequency 1.17 (0.61) 1 5 4.13 18.28
Inhalant use initiation 723 (17.1%) 0 1 -- --
Marijuana use initiation 675 (16.3%) 0 1 -- --
PO use initiation 232 (5.7%) 0 1 -- --
Open in a new tab

Notes. Min = minimum; Max = maximum.

a

Indicators were standardized for analysis

b

Indicators were binary. Frequency values for binary indicators are the frequency of the “high” value.

Table 2.

Model Fit Information for Latent Profile Analyses

No. of profiles Log-likelihood No. of free parameters AIC BIC a-BIC Entropy BLRT
1 −34768.35 10 69556.70 69622.28 69590.50 -- --
2 −32552.39 17 65138.78 65250.26 65196.24 0.69 <.001
3 −31973.03 24 63994.06 64151.45 64075.19 0.69 <.001
4 −31675.61 31 63413.23 63616.53 63518.02 0.71 <.001
5 −31460.01 38 62996.02 63245.23 63124.48 0.73 <.001
6 −31261.65 45 62908.41 62765.41 62765.41 0.71 <.001
7 −31111.93 52 62327.85 62668.88 62503.64 0.72 <.001
Open in a new tab

Parameter estimates, including within-profile item means (for continuous indicators) and probabilities (for binary indicators), are presented in Table 3. Profile 1 (15% prevalence) was characterized by high family conflict, low positive family climate, low parental involvement, low effective discipline, low adolescent positive engagement, and low parental knowledge; we labeled them Coercive. Profile 2 (41%) was characterized by average family conflict, low positive family climate, low parental involvement, average discipline, low adolescent positive engagement, and low parental knowledge; we labeled them Disengaged. Profile 3 (11%) was characterized by high family conflict, average family climate, high parental involvement, low effective discipline, high adolescent positive engagement, and high parental knowledge; we labeled them Permissive. Profile 4 (34%) was characterized by low family conflict, positive family climate, high parental involvement, high effective discipline, high adolescent positive engagement, and high parental knowledge; we labeled them High Functioning.

Table 3.

Parameter Estimates for Four-Profile Model

Profile prevalence Item mean/probability Within-profile mean
1. Coercive .15 (n = 769) 2. Disengaged .41 (n = 2114) 3. Permissive .11 (n = 570) 4. High Functioning .34 (n = 1756)
Family conflict 0a −0.967c −0.052 −0.407c 0.647d
Positive family climate 0a −1.109c −0.118c −0.188 0.737d
Parental involvement 0a −1.487c −0.098c 0.647d 0.607d
Effective discipline 0a −1.143c 0.063 −1.191c 0.845d
Adolescent positive engagement 0.418b 0.101c 0.206c 0.639d 0.741d
Parental knowledge 0.403b 0.135c 0.173c 0.575d 0.742d
Open in a new tab

Notes.

a

Indicators were standardized for analysis

b

Indicators were binary.

c

Statistically significantly lower than the overall item mean at p < .05.

d

Statistically significantly higher than the overall item mean at p < .05. For family conflict, (−) values indicate higher family conflict, (+) values indicate lower family conflict.

We then examined demographic factors as predictors of profile membership. Pairwise comparisons between profiles were significant for all but two tests. Results showed the following patterns. Males were more likely than females to be in the Coercive (OR = 1.71), Disengaged (OR = 1.75), and Permissive (OR = 1.54) profiles than the High Functioning profile. Low-income families were more likely than higher-income families to be in the Coercive (OR = 1.77) and Permissive (OR = 2.73) profiles than the High Functioning profile. Adolescents in single-parent households were more likely than those in two-parent households to be in the Coercive (OR = .48) and Disengaged (OR = .73) profiles than the High Functioning profile.

We next predicted adolescent ASB in Grade 8 from latent profile membership in Grade 6, controlling for covariates and initial levels of ASB. First, we evaluated profile membership prediction using the High Functioning profile as the reference group to determine whether members of the other profiles were at higher relative risk for ASB outcomes. As expected, membership in the Coercive, Disengaged, and Permissive profiles were each associated with significantly higher hostile/aggressive behavior, antisocial peer behavior, drunkenness frequency, and higher rates of inhalant, marijuana, and PO initiation, than the High Functioning profile. Second, we rotated the reference group to the Disengaged profile in order to gain a more nuanced understanding of the differences in risk among the higher risk (i.e., Disengaged, Permissive, Coercive) profiles. Results are displayed in Table 4. Membership in the Disengaged profile was associated with significantly lower antisocial peer behavior and drunkenness frequency than membership in the Coercive profile. Effects can be interpreted as the difference in the within-profile average level of each outcome as compared to the average level of the outcome for the Disengaged profile. For example, the average antisocial peer behavior was .18 higher on the scale for adolescents in the Coercive profile compared to adolescents in the Disengaged profile. There were no significant differences between the Permissive and Disengaged profiles for hostile/aggressive behavior, antisocial peer behavior, or drunkenness frequency. Turning to the substance use initiation outcomes, odds of inhalant initiation and PO initiation were significantly lower for members of the Disengaged profile compared to members of the Coercive and Permissive profiles. Further, the odds of marijuana initiation were significantly lower for the Disengaged compared to Coercive profile (but not compared to the Permissive profile). Effects can be interpreted as, for example, the odds of inhalant initiation (vs. non-initiation) were 2.06 times higher for adolescents in the Coercive profile compared to those in the Disengaged profile.

Table 4.

Effects of Profile Membership on Grade 8 Outcomes

 Variable Hostile/Aggressive Behavior Antisocial Peer Behavior Drunkenness Frequency
B (SE) B (SE) B (SE)
 Intercept .84*** (.06) 1.72*** (.07) .94*** (.17)
 Baseline level .42*** (.05) .29*** (.03) .30 (.16)
 Profile membership
Coercive profile .04 (.05) .18* (.07) .14* (.06)
Disengaged profile ref ref ref
Permissive profile −.02 (.06) −.08 (.10) .04 (.06)
High Functioning profile −.13*** (.03) −.39*** (.05) −.12*** (.03)
 Control Variables
Adolescent gender (Male) .11*** (.02) .01 (.03) −.01 (.02)
SES (Low) .06* (.02) .12** (.04) .02 (.03)
Single-parent household status (No) −.07* (.03) −.11** (.04) −.06 (.03)
 Variable Inhalant Initiation Marijuana Initiation PO Initiation
OR [95% CI] OR [95% CI] OR [95% CI]
 Intercept .22*** [.17-.28] .20*** [ .16-.26] .06*** [.04-.09]
 Profile membership
Coercive profile 2.06***[1.55–2.73] 2.16*** [ 1.65–2.84] 1.77* [1.12–2.81]
Disengaged profile ref ref ref
Permissive profile 1.24 [.83–1.86] 1.35 [.93–1.96] 2.40** [1.40–4.13]
High Functioning profile .29*** [.20–.41] .38*** [.27–.52] .54+ [.32–.91]
 Control Variables
Adolescent gender (Male) .87 [.73–1.04] 1.02 [.86–1.20] 1.20 [.91–1.57]
SES (Low) 1.07 [.88–1.30] 1.38** [1.15–1.66] 1.15 [.86–1.54]
Single-parent household status (No) .82 [.67–1.01] .81 [.67–.99] .67* [.50-.91]
Open in a new tab

Notes.

***

p < .001

**

p < .01

*

p < .05

+

p = .05. Baseline levels of outcomes were included as controls in the linear regression models; analyzed samples for the logistic regression models included only those adolescents who had not yet initiated at baseline.

Discussion

Prior work has identified risk and protective factors for adolescent ASB; yet, despite established theories of ASB risk that describe interplay among factors, holistic approaches that incorporate constellations of risk and protective factors to characterize the family risk context remain understudied. We applied novel, person-centered methods to address this gap in the literature and we identified four distinct subgroups of adolescents’ family functioning that can differentiate risk for adolescent ASB outcomes. The findings from this study complement prior research on family risk and protective factors, provide additional evidence for key theories of risk for ASB, and hold important implications for adolescent ASB etiology and intervention.

We first applied person-centered methods to key family risk and protective factors to determine whether it was possible to identify distinct family subgroups that reflect theorized family risk and protective processes. We included dimensions we expected to find in coercive, disengaged, and high functioning families in our analyses. Four family subgroups emerged, largely confirming our hypotheses. The first subgroup, coercive families represented 15% of the sample and exhibited high family conflict and low positive family climate, parental involvement, effective discipline, adolescent positive engagement, and parental knowledge. This pattern of risk factors is reflective of a family experiencing coercive processes, which involve established patterns of frequent, aversive, and escalating conflicts and poor relationships among family members (Patterson, 1982, 2016). Another element of the coercive profile was low parental involvement and low adolescent positive engagement in the family, which may reflect escape conditioning aspects of established coercive dynamics in families that shape parent avoidance of engaging in challenging interactions with their adolescent (Patterson, 2016). It may be that these coercive families during adolescence have experienced a long-standing pattern of problematic family relations, resulting in a very low-functioning and high-risk family climate.

The second hypothesized subgroup – disengaged families – emerged in our models. Disengaged families often have low connectedness among family members and parents in these families tend to withdraw their parenting (i.e., low parental knowledge and discipline; Dishion et al., 2004; Fosco & LoBraico, 2018; Van Ryzin & Dishion, 2014). As such, the Disengaged subgroup (41% prevalence) exhibited low levels of positive family climate, parental involvement, adolescent positive engagement, and parental knowledge, and average levels of family conflict and effective discipline. The Disengaged subgroup was distinguished from the Coercive subgroup by lower levels of family conflict and higher levels of effective discipline. Thus, disengaged families is a relatively prevalent, at-risk group during the middle school period.

The third subgroup (“Permissive,” 11% prevalence) emerged differently than expected, disconfirming our exploratory hypothesis of an “Inconsistent” subgroup. The Permissive subgroup exhibited higher than average levels of family conflict, parental involvement, adolescent positive engagement, and parental knowledge; low levels of effective discipline; and average levels of family climate. Thus, Permissive families were characterized by generally warm, involved, and positive parent-child relationships, but also with low levels of effective parenting practices, analogous to conceptualizations of permissive parenting styles (Baumrind, 1971; Darling & Steinberg, 1993; Smetana, 1995; Steinberg, 2001). The High Functioning (34% prevalence) subgroup, as hypothesized, exhibited low levels of family conflict, and high levels of positive family climate, parental involvement, effective discipline, adolescent positive engagement, and parental knowledge.

The second goal of this study was to evaluate the relative risk associated with membership in these different family subgroups for adolescent ASB two years later. Across all outcomes, adolescents in the coercive, disengaged, and permissive families were at elevated risk, relative to those in high functioning families. These findings suggest that all three groups, although reflected by different constellations of family functioning, confer risk for hostile-aggressive behavior, antisocial peers, and substance use. We then explored comparisons among the risk groups to evaluate the nuances in risk, using the disengaged subgroup as a reference. Our analyses revealed a robust pattern of results which point to coercive families as conferring particularly high risk to youth, across all six ASB outcomes, validating and advancing prior work documenting developmental risk in coercive families (e.g., Patterson, 2016; Van Ryzin & Dishion, 2013). Coercion theory describes a social learning process by which children learn and develop coercive behavior that eventually manifests outside of the home as aggression (Patterson, 1982; 2016). Aggressive behavior can lead to rejection by prosocial peers, peer deviancy training from antisocial peers, and substance use. Accordingly, membership in the coercive family profile was either significantly more, or equally as, likely as other risk profiles to be engaged in all six ASB outcomes in Grade 8.

Adolescents in the disengaged families profile also evidenced risk for ASB outcomes. Membership in the Disengaged profile was associated with elevated levels of hostile/aggressive behavior, deviant peer behavior, and substance use. Previous literature on disengagement describes and provides evidence for the premature autonomy hypothesis, a process by which adolescents disengage from their families, no longer receiving benefits from the family management practices that can protect them from engaging in deviant behavior (Dishion et al., 2004). They then use this unsupervised time outside of the home to socialize with other disengaged adolescents and engage in deviant behaviors together, learning from each other (Dishion et al., 1996). Thus, we had expected that membership in this group would be related to the highest risk for having antisocial peers and substance use, and also risk for aggression. However, our findings indicate that adolescents in disengaged families were at elevated risk for all six ASB outcomes, but at less risk than those in coercive families across most outcomes.

Often considered separately as theories of risk for ASB development, when compared in the same model it appears that both coercion and disengagement are important indicators of risk for ASB; however, coercive family environments confer significantly more risk for all ASB outcomes except hostile-aggressive behavior. Based on theories of coercion and disengagement, we expected more nuanced results. We would expect that adolescents in coercive families would be the most at risk for displaying aggressive behavior. However, it may be that adolescents from coercive families become aggressive earlier in life, through social learning processes within their family, while adolescents in disengaged or permissive families may develop aggressive behavior later, in response to deviant peer influence and poor family involvement. Thus, although the level of risk for aggression may not differ, the pathways to aggression may differ across groups based on their family risk environment. In addition, membership in the Coercive subgroup, not the Disengaged subgroup, was associated with the highest risk for deviant peer affiliation and substance use. Though we were not able to capture the longitudinal process of disengagement, it is possible that the pathway to aggression and substance use from disengagement was through deviant peer involvement (Dishion, Capaldi, Spracklen, & Li, 1995).

The adolescents in the permissive families evidenced comparable risk to those in the disengaged families. Specifically, membership in the Disengaged and Permissive subgroup predicted nearly equivalent risk across all six ASB outcomes included in this study. This finding implies that even in the context of a warm and positive parent-child relationship, poor parenting skills and a generally more negative family environment (climate, conflict) place the adolescent at just as much risk for ASB as if the parent and adolescent also had a poor relationship. Prior, variable-centered research has found that the quality of the parent-adolescent relationship matters above and beyond parenting styles (Bronte-Tinkew, Moore, & Carrano, 2006; Fosco et al., 2012). The divergent findings in the present study may be explained by the person-centered methodological approach, which does not ‘control’ for the other family risk factors, it instead subdivides the population into groups with similarities across the range of variables. Our models, which capture different family processes as reflected by different patterns of co-occurring factors, provide evidence that high connectedness and poor parenting can and do co-occur, and that adolescents in families where these factors co-occur experience risk despite a positive relationship with their parent. However, we note that our measures focused on behavioral engagement, rather than measures of connectedness that reflect emotional bonds, trust, and subjective feelings of closeness with parents. Future work might explicitly measure connectedness to determine which profile would include this process.

Implications and insights gained by using the person-centered approach

By implementing a person-centered approach, this study offers three key contributions to the literature. First, this study provides evidence that coercion and family disengagement are two unique risk processes which confer different levels of risk for long-term ASB outcomes. Second, the person-centered approach allowed for separation among risk groups and specificity in prediction of long-term adolescent outcomes. Traditional approaches to determining risk status through a tally-based or cumulative approach may result in overlooking families with more “moderate” levels of risk, because they have less severe overall risk status (i.e., disengaged) – and are even protective in some domains (i.e., permissive). In fact, the permissive families would likely go unnoticed in some other approaches, or in practice, as a risk group. This key finding was made possible by using a person-centered approach and assessing risk and protective factors simultaneously. Lastly, given the high prevalence of the ‘moderate’ risk groups (52%, disengaged and permissive families combined), the findings provide guidance for potential areas of focus in universal family-based prevention programming. Targeting the majority of a population (i.e., universal programs) rather than the severe minority is a more effective strategy toward having population level impacts on problem outcomes (i.e., the prevention paradox; Rose, 1981). By identifying the specific needs of the more moderate risk majority (e.g., effective discipline strategies, whole-family functioning), intervention content can be strategically focused toward this group, which may benefit from small changes and, in turn, account for a large proportion of the problem outcomes in the population.

Limitations

Some limitations of the present study warrant attention. Our sample was youth in largely homogeneously White and rural families, which may limit generalizability. Additionally, our single-method measurement of coercion may have limited our ability to capture the granularity intrinsic in this family process. Given the interactional nature of coercion, observational work may be the ideal way to capture it (e.g., Smith et al., 2014). Some of our measures of family risk and protective factors had reliability estimates that fell below ideal levels, and may have introduced error variance into model estimates. However, excluding these measures would have resulted in less informative profiles. Replication is needed, using multi-informant, multi-method measurement from families from other populations. This study focused on capturing subgroups that reflect specific theorized processes of family risk; however, other subgroups may emerge when including additional unmeasured factors which may further explain the risks for ASB. For example, interparental conflict (Bradford et al., 2008) and sibling deviancy and/or conflict (Fosco et al., 2012) are each related to risk for ASB development and could provide more complete family profiles for study. Moreover, although interpretations of subgroup differences in adolescent outcomes correspond to theories of coercion and disengagement, it is not possible to rule out an alternative interpretation that it is the number of risk factors in a subgroup, rather than the patterning of them, that corresponds to the subsequent intensity and severity of ASBs. Additionally, future work may be able to further distinguish the risk associated with different family profiles by extending the LPA longitudinally to evaluate change, and consequences of change, in profile membership over time.

Conclusion

This paper uses LPA to gain additional information about ASB etiology by holistically capturing family risk for ASB development and evaluating if there are differences in risk for specific types of ASB across identified subgroups. Findings offer insight into the multidimensional nature of family functioning and provide more evidence for the etiological differences and similarities across types of ASBs. Results indicate that (1) three subgroups of families (coercive, disengaged, and permissive) experience elevated risk compared to high functioning families, and (2) adolescents in coercive families experience the most robust risk across ASB outcomes. These person-centered analyses shed light on the distinctiveness of two family risk processes, coercion and disengagement, and a third risk process, permissive parenting. Future work should continue building on these findings by replicating with more diverse samples. Accumulating knowledge about subgroups can be used to develop prevention content and identify intervention targets in order to more effectively meet the needs of families with different patterns of co-occurring risk and protective factors.

Supplementary Material

Supplemental Material

Acknowledgments

Grant Support

This study was funded by grant R01-DA013709 from the National Institute on Drug Abuse, and co-funded by the National Institute on Alcohol Abuse and Alcoholism. Authors were supported by several funding sources: The National Institute on Drug Abuse (LoBraico: T32 DA017629; PI: L.M. Collins; LoBraico: F31-DA048522; Bray: P50 DA039838), the National Institute of Child Health and Human Development (Fosco & Feinberg: R01 HD092439). The content is solely the responsibility of the authors and does not necessarily represent the official views of the funding agencies.

Footnotes

Findings from this work were presented the Pennsylvania State University Prevention Research Center in April 2019 and the Society for Prevention Research Annual Conference in San Francisco, C.A. in May 2019.

References

  1. Akaike H (1974). A new look at the statistical model identification. IEEE Transactionson Automatic Control, 19(6), 716–723. 10.1109/TAC.1974.1100705 [DOI] [Google Scholar]
  2. Ary DV, Duncan TE, Biglan A, Metzler CW, Noell JW, & Smolkowski K (1999). Development of Adolescent Problem Behavior. Journal of Abnormal Child Psychology, 27(2), 141–150. 10.1023/A:1021963531607 [DOI] [PubMed] [Google Scholar]
  3. Bakk Z, & Vermunt JK (2016). Robustness of stepwise latent class modeling with continuous distal outcomes. Structural Equation Modeling, 23(1), 20–31. [Google Scholar]
  4. Bauer DJ, & Curran PJ (2003). Distributional Assumptions of Growth Mixture Models: Implications for Overextraction of Latent Trajectory Classes. Psychological Methods, 8(3), 338–363. 10.1037/1082-989X.8.3.338 [DOI] [PubMed] [Google Scholar]
  5. Baumrind D (1971). Current patterns of parental authority. Developmental Psychology, 4(1 PART 2), 1–103. 10.1037/h0030372 [DOI] [Google Scholar]
  6. Benson MJ, & Buehler C (2012). Family Process and Peer Deviance Influences on Adolescent Aggression: Longitudinal Effects Across Early and Middle Adolescence. Child Development, 83(4), 1213–1228. 10.1111/j.1467-8624.2012.01763.x [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Bloom BL (1985). A Factor Analysis of Self-Report Measures of Family Functioning. Family Process, 24(2), 225–239. 10.1111/j.1545-5300.1985.00225.x [DOI] [PubMed] [Google Scholar]
  8. Bradford K, Vaughn LB, & Barber BK (2008). When There Is Conflict. Journal of Family Issues, 29(6), 780–805. [Google Scholar]
  9. Bronte-Tinkew J, Moore K, & Carrano J (2006). The Father-Child Risk Behaviors in Intact Families. Journal of Family Issues, 27(6), 850–881. [Google Scholar]
  10. Bullard L, Wachlarowicz M, DeLeeuw J, Snyder J, Low S, Forgatch M, & DeGarmo D (2010). Effects of the oregon model of parent management training (PMTO) on marital adjustment in new stepfamilies: A randomized trial. Journal of Family Psychology, 24(4), 485–496. 10.1037/a0020267 [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Burnette ML, Oshri A, Lax R, Richards D, & Ragbeer SN (2012). Pathways from harsh parenting to adolescent antisocial behavior: A multidomain test of gender moderation. Development and Psychopathology, 24(3), 857–870. [DOI] [PubMed] [Google Scholar]
  12. Calkins SD, & Keane SP (2009). Developmental origins of early antisocial behavior. Development and Psychopathology, 21(4), 1095–1109. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Celeux G, & Soromenho G (1996). An entropy criterion for assessing the number of clusters in a mixture model. Journal of Classification, 13, 195–212. [Google Scholar]
  14. Collins LM, & Lanza ST (2010). Latent Class and Latent Transition Analysis: With Applications in the Social, Behavioral, and Health Sciences. Hoboken, New Jersey: Wiley. [Google Scholar]
  15. Craig JM, Morris RG, Piquero AR, & Farrington DP (2015). Heavy drinking ensnares adolescents into crime in early adulthood. Journal of Criminal Justice, 43(2), 142–151. [Google Scholar]
  16. Darling N, & Steinberg L (1993). Parenting Style as Context: An integrative model. Psychological Bulletin, 113(3), 487–496. 10.1037/0033-2909.113.3.487 [DOI] [Google Scholar]
  17. Dishion TJ, Mun CJ, Ha T, & Tein JY (2019). Observed Family and Friendship Dynamics in Adolescence: a Latent Profile Approach to Identifying “Mesosystem” Adaptation for Intervention Tailoring. Prevention Science, 20(1), 41–55. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Dishion TJ, Nelson SE, & Bullock BM (2004). Premature adolescent autonomy: Parent disengagement and deviant peer process in the amplification of problem behaviour. Journal of Adolescence. 10.1016/j.adolescence.2004.06.005 [DOI] [PubMed] [Google Scholar]
  19. Dishion TJ, Nelson SE, & Kavanagh K (2003). The Family Check-Up with high-risk young adolescents: Preventing early-onset substance use by parent monitoring. Behavior Therapy, 34(4), 553–571. 10.1016/S0005-7894(03)80035-7 [DOI] [Google Scholar]
  20. Dishion T, & Patterson G (2006). The development and ecology of antisoacial behavior in children and adolescents In Cicchetti D & Cohen DJ (Eds.), Developmental Psychpathology, Volume 3: Risk, Disorder, and Adaptation (2nd ed., pp. 503–541). Hoboken, New Jersey: John Wiley & Sons, Inc. [Google Scholar]
  21. Dishion T, Spracklen K, Andrews D, & Patterson G (1996). Deviancy training in male adolescent friendships. Behavior Therapy, 27(3), 373–390. [Google Scholar]
  22. Elliott DS, Huizinga D, & Ageton SS (1985). Explaining Delinquency and Drug Use. Beverly Hills, CA: Sage. [Google Scholar]
  23. Fosco GM, & LoBraico EJ (2018). A Family Systems Framework for Adolescent Antisocial Behavior: State of the Science and Suggestions for the Future. In Celano M, Deater-Deckard K, Jouriles E, & Whisman M (Eds.), APA Handbook of Contemporary Family Psychology. [Google Scholar]
  24. Fosco GM, Stormshak EA, Dishion TJ, & Winter C (2012). Family relationships and parental monitoring during middle school as predictors of early adolescent problem behavior. J Clin Child Adolesc Psychol, 41(2), 202–213. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Fosco GM, Caruthers A, & Dishion TJ (2012). A six-year predictive test of adolescent family relationship quality and effortful control pathways to emerging adult social and emotional health. Journal of Family Psychology, 26, 565–575. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Fosco Gregory M., & LoBraico EJ (2019). Elaborating on premature adolescent autonomy: Linking variation in daily family processes to developmental risk. Development and Psychopathology, Advanced o. 10.1017/s0954579419001032 [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Halgunseth LC, Perkins DF, Lippold MA, & Nix RL (2013). Delinquent-Oriented attitudes mediate the relation between parental inconsistent discipline and early adolescent behavior. Journal of Family Psychology, 27(2), 293–302. 10.1037/a0031962 [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Hiatt KD, & Dishion TJ (2008). Antisocial personality development. Child and Adolescent Psychopathology., (JANUARY 2008), 370–404. [Google Scholar]
  29. January SAA, Mason WA, Savolainen J, Solomon S, Chmelka MB, Miettunen J, … Järvelin MR (2017). Longitudinal Pathways from Cumulative Contextual Risk at Birth to School Functioning in Adolescence: Analysis of Mediation Effects and Gender Moderation. Journal of Youth and Adolescence, 46(1), 180–196. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Jaycox LH, & Repetti RL (1993). Conflict in Families and the Psychological Adjustment of Preadolescent Children. Journal of Family Psychology, 7(3), 344–355. [Google Scholar]
  31. Johnston LD, Miech RA, O’malley PM, Bachman JG, Schulenberg JE, & Patrick ME (2017). Monitoring the Future National Survey Results on Drug Use 1975–2017. [Google Scholar]
  32. Kiesner J, Dishion T, Poulin F, & Pastore M (2009). Temporal dynamics linking aspects of parent monitoring with early adolescent antisocial behavior. Social Development, 18(4), 765–784. 10.1111/j.1467-9507.2008.00525.x [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Laible D, Eye J, & Carlo G (2008). Dimensions of conscience in mid-adolescence: Links with social behavior, parenting, and temperament. Journal of Youth and Adolescence, 37(7), 875–887. 10.1007/s10964-008-9277-8 [DOI] [Google Scholar]
  34. Lanza ST, & Rhoades BL (2013). Latent Class Analysis: An Alternative Perspective on Subgroup Analysis in Prevention and Treatment. Prevention Science, 14(2), 157–168. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Lanza ST, Rhoades BL, Nix RL, & Greenberg MT (2011). Modeling the interplay of multilevel risk factors for future academic and behavior problems: A person-centered appraoch (Vol. 22). 10.1017/S0954579410000088.Modeling [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Lindahl KM (1998). Family process variables and children’s disruptive behavior problems. Journal of Family Psychology, 12(3), 420–436. 10.1037/0893-3200.12.3.420 [DOI] [Google Scholar]
  37. McLachlan G (1987). On bootstrapping the likelihood ratio test statistic for the number of components in a normal mixture. Applied Statistics, 36(3), 318–324. [Google Scholar]
  38. McLachlan G, & Peel D (2005). Finite Mixture Models. New York, NY: John Wiley & Sons. [Google Scholar]
  39. Moos R, & Moos B (1994). Family environment scale manual Muthén LK, & Muthén BO (n.d.). Mplus User’s Guide Eighth Edition. Los Angeles, C.A.: Muthén & Muthén. [Google Scholar]
  40. Odgers CL, Caspi A, Broadbent JM, Dickson N, Thomson WM, & Moffitt TE (2007). Prediction of Differential Adult Health Burden by Conduct Problem Subtypes in Males, 64. [DOI] [PubMed] [Google Scholar]
  41. Patterson GR (2016). Coercion theory: The study of change In Dishion TJ & Snyder JJ (Eds.), The Oxford Handbook of Coercive Relationship Dynamics (pp. 7–22). New York, NY: Oxford University Press. [Google Scholar]
  42. Patterson GR, & Stouthamer-Loeber M (1984). The correlation of family management practices and delinquency. Child Development, 55(4), 1299–1307. [PubMed] [Google Scholar]
  43. Rose G (1981). Strategy of prevention: Lessons from cardiovascular disease. British Medical Journal (Clinical Research Ed.), 282(6282), 2136. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Samuelson YM, Hodgins S, Larsson A, Larm P, & Tengström A (2010). Adolescent antisocial behavior as predictor of adverse outcomes to age 50: A follow-up study of 1,947 individuals. Criminal Justice and Behavior, 37(2), 158–174. [Google Scholar]
  45. Schwarz G (1978). Estimating the Dimension of a Model. The Annals of Statistics, 6(2), 461–464. 10.1214/aos/1176344136 [DOI] [Google Scholar]
  46. Sclove LS (1987). Application of model selection criteria to some problems in multivariate analysis. Psychometrika, 52(3), 333–343. [Google Scholar]
  47. Simpson EG, Vannucci A, & Ohannessian CMC (2018). Family functioning and adolescent internalizing symptoms: A latent profile analysis. Journal of Adolescence, 64(September 2017), 136–145. 10.1016/j.adolescence.2018.02.004 [DOI] [PubMed] [Google Scholar]
  48. Skinner OD, & McHale SM (2016). Parent–Adolescent Conflict in African American Families. Journal of Youth and Adolescence, 45(10), 2080–2093. [DOI] [PubMed] [Google Scholar]
  49. Smetana J (1995). Parenting styles and conceptions of parental authority during adolescence. Child Development, 66(2), 299–316. [DOI] [PubMed] [Google Scholar]
  50. Spoth R, Greenberg MT, Bierman KL, & Redmond C (2004). PROSPER community-university partnership model for public education systems: Capacity-building for evidence-based, competence-building prevention. Prevention Science, 5(1), 31–39. [DOI] [PubMed] [Google Scholar]
  51. Spoth R, & Molgaard V (1999). Project Family: A partnership in integrating rsearch with the practice of promoting family and youth competencies In Chibucos TR & Lerner R (Eds.), Serving children and families through community-university partnerships: Sucess sotires (pp. 127–137). Boston: Kluwer Academic. [Google Scholar]
  52. Spoth Richard, Redmond C, & Shin C (1998). Direct and Indirect Latent-Variable Parenting Outcomes of Two Universal Family-Focused Preventive Interventions: Extending a Public Health- Oriented Research Base. Journal of Consulting and Clinical Psychology, 66(2), 385–399. [DOI] [PubMed] [Google Scholar]
  53. Steinberg L (2001). We know some things: Parent-adolescent relationships in retrospect and prospect. Journal of Research on Adolescence, 11(1), 1–19. [Google Scholar]
  54. Sturge-Apple ML, Davies PT, & Cummings EM (2010). Typologies of family functioning and children’s adjustment during the early school years. Child Development, 81(4), 1320–1335. 10.1111/j.1467-8624.2010.01471.x [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Surjadi FF, Lorenz FO, Conger RD, & Wickrama KAS (2013). Harsh, inconsistent parental discipline and romantic relationships: Mediating processes of behavioral problems and ambivalence. Journal of Family Psychology, 27(5), 762–772. [DOI] [PubMed] [Google Scholar]
  56. Van Ryzin MJ, & Dishion TJ (2013). From antisocial behavior to violence: A model for the amplifying role of coercive joining in adolescent friendships. Journal of Child Psychology and Psychiatry and Allied Disciplines. 10.1111/jcpp.12017 [DOI] [PMC free article] [PubMed] [Google Scholar]
  57. Van Ryzin MJ, & Dishion TJ (2014). Adolescent deviant peer clustering as an amplifying mechanism underlying the progression from early substance use to late adolescent dependence. Journal of Child Psychology and Psychiatry and Allied Disciplines, 55(10), 1153–1161. 10.1111/jcpp.12211 [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. Van Ryzin MJ, Fosco GM, & Dishion TJ (2012). Family and peer predictors of substance use from early adolescence to early adulthood: An 11-year prospective analysis. Addictive Behaviors, 37(12), 1314–1324. 10.1016/j.addbeh.2012.06.020 [DOI] [PMC free article] [PubMed] [Google Scholar]
  59. Withers MC, McWey LM, & Lucier-Greer M (2016). Parent–Adolescent Relationship Factors and Adolescent Outcomes Among High-Risk Families. Family Relations, 65(5), 661–672. 10.1111/fare.12220 [DOI] [Google Scholar]
  60. Xia M, Weymouth B, Bray BC, Lippold MA, Feinberg ME, & Fosco GM (In Press). Exploring Influence of Triadic Family Relationship Profiles on Adolescents’ Early Substance Initiation. Prevention Science. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplemental Material
NIHMS1069137-supplement-Supplemental_Material.pdf (38.4KB, pdf)

RESOURCES