Antisocial personality disorder as a predictor of polydrug use: a longitudinal study of the dual mediating roles of deviant peer association and self-control among juvenile offenders

Abstract

Background

Antisocial personality disorder (APD) has been linked to increased risk for polydrug use (PU). However, there is a dearth of research focusing on the mechanisms underlying this risk.

Aims

To examine the relationship between APD and PU; to test for the mediating effects of deviant peer association (DVA) and self-control in this relationship; to determine if the magnitudes of observed mediation effects are equivalent.

Methods

The Pathways to Desistance data were analyzed. Generalized structural equation modeling tested for direct and mediating relationships. Mediation effect equivalence was tested using a Clogg Z test.

Results

Antisocial personality disorder was associated with increased drug use heterogeneity (coefficient = 0.479; CI = 0.107–0.851). This was significantly mediated by self-control and DVA. The magnitude of the mediation was significantly stronger for the self-control path (self-control coefficient = 0.233; CI = 0.044–0.422; DVA coefficient = 0.134; CI = 0.028–0.241).

Conclusion

While both self-control and DVA were found to be important for explaining PU in APD, it may be beneficial to focus on self-control to address this problem.

Keywords:

• Antisocial Personality disorder; polydrug use; self-control; deviant peer association; mediation

Disclosure statement

There are no conflicts of interest to report.

Notes

1 This count measure included all substances that were illicit at the time of measurement, thus, eschewing alcohol and tobacco (or if these were illegal because of age). The drugs included in the measure then were: marijuana/hashish; stimulants/amphetamines; sedatives/tranquilizers; opiates; cocaine; hallucinogens; ecstasy; amyl nitrate/odorizers/rush; inhalants; abused prescription medications; and other drugs (specify).

2 Wave 10 was chosen as the observation period of measurement for key independent variables in this study because this was the only observation period at which APD status was measured. Wave 11 was chosen for the dependent variable because this allowed for parsing of temporal ordering that is a necessary component for identifying causal relationships.

3 Only the key relationships concerning APD, self-control, deviant peer association, and PU are described in Figures 1 and 2 for the sake of clarity of presentation.

4 Stata does not provide traditional fit statistics when using GSEM. As such, the traditional linear models were estimated to provide root mean squared error of approximation (RMSEA) and confidence fit index (CFI) statistics. Model 2 RMSEA = 0.103, RMSEA 90% confidence interval = 0.094, CFI = 0.410). These statistics should be interpreted with extreme caution given the extreme departure from normality in data distribution on the dependent variable.