Psychosocial and genetic risk markers for longitudinal trends in delinquency: an empirical assessment and practical discussion

Abstract

The increased use of biosocial perspectives in criminological research has expanded the scope of factors considered in understanding the etiology of adolescent antisocial behavior. At the same time, its practical utility for preventive and remedial intervention has not been examined to the same degree. Using a large, nationally representative sample of youth (N = 2573) and a series of latent growth curve models, this study examines the relative utility of a psychosocial risk composite and genetic indicators (DRD2, DRD4, DAT1, 5-HTTLPR, MAO-A) in predicting the onset and later developmental patterns of adolescent and early adult delinquency and criminal behavior. The results show that the psychosocial risk composite measure has significant effects on the latent growth factors, while the main and interactive effects of the genetic indicators do not. The subsequent discussion considers the practical implications of these empirical findings in the context of extant research and pinpoints some possible future applications of this area of research. It also identifies some parallel cases of translational criminology that may serve as indications of how this research might inform policy and practice going forward.

Keywords:

• psychosocial risk
• genetic markers
• translational criminology
• developmental trajectories

Acknowledgments

This study uses data from Add Health, a program project designed by J. Richard Udry, Peter S. Bearman, and Kathleen Mullan Harris and funded by a Grant P01-HD31921 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development, with cooperative funding from 17 other agencies. Special acknowledgment is due to Ronald R. Rindfuss and Barbara Entwisle for assistance in the original design. Persons interested in obtaining data files from Add Health should contact Add Health, Carolina Population Center, 123 W. Franklin Street, Chapel Hill, NC 27516-2524 ([email protected]). No direct support was received from Grant P01-HD31921 for this analysis. An earlier version of this study was presented at the 2013 International Academy of Law and Mental Health Congress in Amsterdam, the Netherlands. The authors are grateful to the session participants, special issue editors, and an anonymous reviewer for their helpful comments on earlier versions of this work.

Notes

1. It is important to note the distinction between using risk assessment information for dispositional as opposed to treatment or intervention-related decisions. Regardless of the source of the information (i.e. genetic, self-report, interviewer-recorded), the former usage is controversial as it can move the system away from decision-making based on legally relevant factors toward scientifically grounded, but still error-prone, assessment information (see, e.g. Auerhahn, 1999). Use for screening and intervention may share some of these concerns, but it is generally aimed at rehabilitation as opposed to developing punitive sanctions. We focus on the latter risk assessment context as a point of departure for the current study. Heilbrun (1997: 352) defines this as a ‘management-oriented’ approach in that it seeks to ameliorate risk in the future and improve developmental prospects for individual youth through treatment/intervention as opposed to formulating sanctions for a particular offense.

2. For simplicity, we refer to the outcome measure as ‘delinquency’ throughout the study. Many of these items technically become criminal behavior as Add Health participants cross the threshold into adulthood in later waves of the study, however.

3. This led to a loss of 20 cases on that variable (.8% of the initial sample).

4. Given this modest positive skew, the cumulative risk measure was re-expressed using log 10 and square root functions to check the sensitivity of key relationships (McClendon, 2002). This had very limited effects on the covariance (and, by extension, correlation) matrix, which serves as a basis for latent growth curve (LGC) and regression modeling. For example, the relationship between the risk composite and delinquency at Wave I was reduced from Pearson’s r = .37 to r = .34 and r = .15 to r = .14 with delinquency at Wave IV using the square-root-transformed risk variable.

5. These markers are often associated with risk/vulnerability, but, more recently, they have been labeled as indicators of ‘plasticity’ (e.g. Belsky & Beaver, 2011). The notion of plasticity indicates the ability of the individual to change behavior in response to the environment. The measures are still important to this study because recent research has shown that one’s genotype does influence how he/she responds to treatment, which is often a practical follow-on to risk and needs assessment. For example, Bakermans-Kranenburg, Van IJzendoorn, Pijlman, Mesman, and Juffer (2008) found that children with the 7-repeat allele of DRD4 showed the largest improvement in externalizing behaviors when parents participated in the Video-feedback Intervention to Promote Positive Parenting and Sensitive Discipline. Beach, Brody, Philibert, and Lei (2010) found that the prevention program Strong African American Families was most effective among youths with the 7-repeat allele of DRD4.

6. Previous studies have identified sex differences in expression of genetic effects for disease based on X inactivation generally (e.g. Migeon, 2006) and MAO-A effects on aggression specifically (Meyer-Lindenberg et al., 2006). This makes pooling or separating the groups somewhat tenuous. As a check, we did examine the specified models with females and largely reached the same conclusions with respect to the pattern of findings described below.

7. Given the developmental time window and number of observation periods used, we also explored the possibility of adding a quadratic slope term to the estimated equation but that did not significantly improve model fit.

8. Due to space limitations, we do not present the main effect models (i.e. those without interaction terms) for the genetic marker indicators. Those results, which are available upon request, are entirely consistent with the models presented in the study in terms of their conclusions about the genetic marker variables.

9. As there was some potential for linear dependence among covariates, we investigated collinearity diagnostics for all main effect models. For the DRD2 models, it did appear that there might be some issue with multicollinearity for the genetic marker measures (Tolerance = .27–.99; variance inflation factors [VIF] = 1.01–3.70). We subsequently re-estimated the models reported here including only those covariates and reached similar conclusions. For all other multivariate models estimated here, the tolerance and VIF values fell within even the most conservative cutoff boundaries (see Allison, 1999). Supplementary analysis with a cumulative genetic marker index similar to that used by Belsky and Beaver (2011) also yielded nonsignificant regression estimates for both main effects and interaction terms.

10. See Laub (2011) for a discussion of this terminology and Sullivan (2013) for consideration of some important implementation challenges in the developmental crime prevention area.

11. See Piquero, Farrington, Welsh, Tremblay, and Jennings (2009) for a systematic review of the evidence on the effects of programs like NFP on later delinquency.

Additional information

Funding

Data collection for this study was funded by a grant P01-HD31921 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development. The study authors did not receive funding for this work.