Objectives
Design
Results
Conclusion
We consider gender and ethnic differences in the co-occurrence of adolescent behaviors related to health and well-being.
Using a nationally representative sample of adolescents in the National Longitudinal Survey of Youth (1997–2000), we examine behavior among students as well as school drop-outs. We use latent class models (LCMs) to identify subpopulations of adolescents with similar patterns of co-occurring behaviors. The generalizability of the findings for African American adolescents in the 1970s is considered using a sample of inner-city youth from the Pathways to Adulthood Survey.
For all ethnic groups, we find a subpopulation with ‘problem behavior’ characteristics (in which early sexual initiation, alcohol use, smoking, marijuana use, and truancy are all highly prevalent). This cluster is most common among European American adolescents and among young men. A subpopulation characterized by behaviors often leading to poor social outcomes (e.g. truancy, early sexual initiation and fighting) is most common for African American adolescents, especially young African American men.
Our findings suggest that multi-factorial interventions which address the interrelationships between all of the behaviors are relevant regardless of gender or ethnicity. However, the ethnic and gender differences in the likelihood of specific patterns of interrelationships highlight the importance of considering the ethnic and gender composition of a population when developing future research and interventions.
This research was supported by the National Institute of Child Health and Human Development grant No. R01 HD39018. We thank Saifuddin Ahmed for thoughtful comments on an earlier draft of this article, and for providing statistical consultation and an introduction to the LEM software.
Notes
1. Additional information about the study design, sampling, and survey methods is detailed in the NLS97 User's Guide 2002 (http://www.bls.gov/nls/97guide/nls97usg.htm).
2. Less than 1% of the total sample had missing information on risk behaviors, and only about 10% of the total sample had missing data on frequency. (The rate was slightly higher among African American men at 15%.) Less than 10% of the total sample had inconsistent data across the surveys on the reporting of marijuana use.
3. Additional information about the study design, sampling, and survey methods is detailed in the documentation associated with Hardy and Shapiro (1999).
4. The JHCPS is a component of the National Collaborative Perinatal Project (CCP) of the National Institute of Neurological and Communicative Disorders and Stroke. The CCP was a prospective study designed to identify the antecedents of infant and later child mortality (PAS 1997).
5. Of the 1,102 men who qualified for the survey, there is complete data for 636, and of the 1,098 women who qualified, there is complete data for 755.
6. Since all but one study assessing the interrelationships between the variables considered in these studies used continuous outcomes, direct equivalence of the indicators is not possible. Additionally, the data from the one study employing LCM is cross-sectional and is conducted among children (Fergusson et al. 1994), so direct equivalence with this study is also not appropriate.
7. The European Americans in the PAS include low income women as well as women experiencing complications, who would have delivered at Johns Hopkins Hospital where the JHCPS sample (and thus the PAS) was drawn.