Observed Family and Friendship Dynamics in Adolescence: a Latent Profile Approach to Identifying "Mesosystem" Adaptation for Intervention Tailoring

Abstract

Nuanced understanding of adolescents' interpersonal relationships with family and peers is important for developing more personalized interventions that prevent problem behaviors and adjustment issues. We used latent profile analysis (LPA) to classify a community sample of 784 adolescents with respect to their observed relationship dynamics with friends and family using videotaped observations and five-minute audiotaped speech samples collected at ages 16-17. The resulting latent classes served to predict behavioral and emotional health in early adulthood. The LPA of the video- and audio-coded observational variables revealed a three-class model: (1) the healthy relationship group (n = 587), representing low levels of deviant and drug use talk with friends and positive, noncoercive relationship with parents; (2) the disaffected group (n = 90), representing high levels of drug use talk with friends and negativity about their parent(s) in the five-minute speech sample; and (3) the antisocial group (n = 107), representing high levels of deviant talk, drug use talk, coercive joining with friends, and coerciveness in family interactions. In contrast to the healthy relationship group, the disaffected group showed elevated risk for substance use problems and depression and the antisocial group showed higher risk for substance use problems and committing violent crimes in early adulthood. Outcome differences between disaffected and antisocial groups were mostly nonsignificant. We discuss the viability of applying these findings to tailoring and personalizing family-based interventions with adolescents to address key dynamics in the family and friendship relationships to prevent adult substance use problems, depression, and violence.

Keywords: Antisocial behaviors; Depression; Drug use; Intervention tailoring; Observation; Relationship dynamics.

Figure 1.

Results of the Latent Profile Analysis of observed peer and family relationship dynamics

Figure 2.

Summary of depression and anxiety symptoms as outcomes Note: Only significant unstandardized path estimates are shown here. Values in brackets are standard errors. 19 = Ages 18-19, 22 = Ages 21-22, 24 = Ages 23-24, 27 = Ages 26-27. Dep = Depression, Anx = Anxiety. Factor loadings of depression latent variable ranged from .37 to .68, anxiety from .49 to .72, and all were statistically significant. All predictors correlated to each other but we do not show paths here for visual parsimony.

Figure 3.

Summary of substance abuse and dependence as outcomes Note: Only significant unstandardized path estimates are shown here. Values in brackets are standard errors. Tob = Tobacco dependence, Alc = Alcohol abuse and dependence, MJ = Marijuana abuse and dependence. Factor loadings of tobacco dependence latent variable ranged from .65 to .90, alcohol abuse and dependence from .51 to .75, and marijuana abuse and dependence from .70. to .81, and all were statistically significant.

Figure 4.

Summary of violent offenses as an outcome Note: Only significant unstandardized path estimates are shown here. VO = Violent Offenses. Factor loadings of violent offenses latent variable ranged from .68 to .77 and all were statistically significant.