Covariation Among Adolescent Problem Behaviors: Cross-Country Perspectives

 

Background/Purpose: To the extent adolescent problem behaviors (e.g. alcohol use, cigarette smoking, drug use) are impacted by common causes, they will be highly correlated with one another. To the extent they are impacted by unique, behavior-specific determinants, the correlations will be low. Documenting the degree of correlation among problem behaviors is important because the magnitude of the correlation provides information about the ceiling effects we can expect for different interventions, such as those emphasizing youth development (common causes) versus behavior-specific programs (unique causes). Adolescent problem behaviors are traditionally thought to be highly correlated. The present study documented the correlation among adolescent problem behaviors in each of 104 countries in youth ages 13-15. World systems theory was used to make predictions of country-level predictors of correlation magnitudes, including Gini Index, GNI per capita, and health expenditures (% of GDP).

Methods: Data were analyzed on adolescent problem behaviors from the most recent Global School-based Student Health Survey and the Health Behavior in School-aged Children survey. Data on nation-wide indicators were obtained from the World Bank DataBank, Standardized World Income Inequality Database, and world system classifications. The study focused on recent problem behaviors (e.g., alcohol use, cigarette smoking, drug use, sexual behavior, physical fight, and suicidal attempt). Problem behaviors were quantified both as dichotomous variables and count variables. We used multiple effect size indices (e.g., phi coefficient) for dichotomous variables and Pearson correlation for count variables. Meta-regression for calculating across-country average correlations and analyzing country-level predictors were performed using robust variance estimation methods assuming a random effects model.

Results: The sample included 358,141 adolescents (mean age 14.2, 52% female). The average correlation for engaging in any two problem behaviors among the six problem behaviors was 0.194, with the highest correlation of 0.365 between alcohol use and cigarette smoking and the lowest correlation of 0.082 between sexual behavior and suicidal attempt, and. The average correlation for counts between any two problem behaviors was 0.233 with the highest between alcohol use and cigarette smoking (0.376) and lowest between physical fight and suicidal attempt (0.154). Health expenditure and GNI per capita were positively associated with the magnitude of the correlations between problem behaviors. Periphery countries had lower correlations than core countries when evaluated in the context of world systems theory. Income inequality was negatively associated with the correlations, but the coefficient was non-significant when controlling for GNI per capita.

Conclusions and Implications: Correlations between problem behaviors exist but the magnitude is modest in 104 countries across different cultural, economic, and political environments, suggesting that problem behaviors have far more unique than common variance. The correlation varies depending on the specific problem behavior pairs. GNI per capita and health expenditure are robust nation-level predictors of the magnitude of correlations. Programs that focus on broadly defined common causes likely have a somewhat low ceiling effect in terms of change potential. It is necessary to develop both general youth development and problem behavior-specific prevention programs to address the common and unique determinants of each problem behavior.