Re-Examining the Reciprocal Relationships of Adolescent Substance Use and Peer Substance Use within the Context of Child Maltreatment: Comparison of the Cross-Lagged Panel Model and Random-Intercept Cross-Lagged Panel Model

Background/Purpose: Several theories address how adolescent substance use may emerge and be perpetuated. Socialization theory posits that adolescents’ substance use is influenced by peers who engage in substance use, by mimicking their behaviors and values. However, the selection theory suggests that as adolescents increase their substance use, they select peers with more substance use. Given the conflicting directions that these theories suggest, previous studies have examined both directions and identified bidirectional relationships. However, these findings neglected within-person effects by solely focusing on between-person effects, which limits their ability to explain the causality. Furthermore, within- versus between-person effects have not yet investigated with an at-risk of maltreatment sample, despite the strong association between child maltreatment experiences, substance use, and peer relationships. To address these gaps, this study compared the results of the standard Cross-Lagged Panel Model (CLPM; based on between-group patterns) to the Random-Intercept Cross-Lagged Panel Model (RI-CLPM; taking into account both between-and within-group patterns).

Methods: This study included 1,093 adolescents (52.3% female, 54.8% Black, and 28.4% low income) drawn from the Longitudinal Studies of Child Abuse and Neglect (LONGSCAN) who completed the main study variables across ages 12, 14 and 16. Self-reported adolescent substance use was assessed using the number of alcohol, tobacco, and marijuana use within the past 12 months. Peer substance use was assessed using the modified version of the Youth Risk Behavior Survey. The standard CLPM and RI-CLPM were conducted using Mplus 8.3.

Results: The standard CLPM model found a reciprocal relationship between peer and adolescent substance use (βs range: .094 to .368), despite the poor model fit: CFI = .850, RMSEA = .105 (90% CI = .081 to .131), and SRMR = .071. The final RI-CLPM exhibited significantly better model fit: CFI = .988, RMSEA = .06 (90% CI = .016 to .117), and SRMR = .015. However, the reciprocal relationships between peer and adolescent substance use were attenuated in the final RI-CLPM. We also found a positive correlation at the between child level (r = .486, p < .001) in the final RI-CLPM, suggesting that the reciprocal relationship is partly attributable to stable, trait-like characteristics that are commonly related to both peer and adolescent substance use. These findings indicate that the standard CLPM over-estimated the bidirectional relations.

Conclusion: The present study first shed light on the reciprocal effects of peer and adolescent substance use after separating between and within-person effects. While the standard CLPM corroborated the reciprocal relations between peer and adolescent substance use, the RI-CLPM only partially supported the socialization effects. Despite this partial support, the significant role of peer substance use remains paramount in the RI-CLPM. Hence, practitioners could routinely monitor peer substance use and implementing preventive interventions for adolescent who associate with substance-using peers. Furthermore, while this rigorous model offers valuable insights, future research is warranted to conduct a more comprehensive examination to assist practitioners in developing policies and programs aimed at preventing peer and adolescent substance use.