A social network perspective on heroin and cocaine use among adults: evidence of bidirectional influences

Abstract

Aims: While several studies have documented a relationship between initiation of drug use and social network drug use in youth, the direction of this association is not well understood, particularly among adults or for stages of drug involvement beyond initiation. The present study sought to examine two competing theories (social selection and social influence) in the longitudinal relationship between drug use (heroin and/or cocaine) and social network drug use among drug-experienced adults.

Design: Three waves of data came from a cohort of 1108 adults reporting a life-time history of heroin and/or cocaine use.

Setting: Low-income neighborhoods with high rates of drug use in Baltimore, Maryland.

Participants: Participants had weekly contact with drug users and were 18 years of age or older.

Measurements: Drug use data were self-report. Network drug use was assessed through a social network inventory. Close friends were individuals whom the participant reported seeing daily or rated as having the highest level of trust. Findings Structural equation modeling indicated significant bidirectional influences. The majority of change in network drug use over time was due to change in the composition of the network rather than change in friends' behavior. Drug use by close peers did not influence participant drug use beyond the total network.

Conclusions: There is evidence of both social selection and social influence processes in the association between drug use and network drug use among drug-experienced adults.

Figures

Figure 1

Standardized estimated coefficients from a structural equation model of the bidirectional relationship between drug use and network drug use, using data from waves 1, 2, and 4 of the SHIELD study. Note. “D. U.” refers to drug use defined as heroin and/or cocaine use in the prior 6 months. All estimates are standardized, and those in bold are statistically significant at p < .05. CFI = .997, TLI = .991, RMSEA = .034.

Figure 2

Structural equation model results for index drug use predicted by both total network drug use and if there are any drug users among those friends rated at the highest level of trust. Note. “D. U.” refers to the drug use defined as heroin and/or cocaine use in the prior 6 months. All estimates are standardized, and those in bold are statistically significant at p < .05. CFI = .994, TLI = .989, RMSEA = .026. Correlations between the residuals are not shown.

Figure 3

Structural equation model results for index drug use predicted by both total network drug use and if there are any drug users among those friends who the index sees daily. Note. “D. U.” refers to the drug use defined as heroin and/or cocaine use in the prior 6 months. All estimates are standardized, and those in bold are statistically significant at p < .05. CFI = .981, TLI = .972, RMSEA = .050. Correlations between the residuals are not shown.