Estimating the prevalence and frequency of adolescent substance use using zero-inflated models and variables associated with social learning, social bond, and opportunity

Abstract

Purpose

To use zero-inflated regression models to examine the differences in prevalence and frequency of adolescent substance use. This addresses several limitations of prior research that uses dichotomous measures of substance use and does not estimate separate models depending on type of substance.

Methods

This study used data from the Arizona Youth Survey and included a variety of independent variables derived from criminological theory and several types of substances.

Results

Many of the independent variables were associated with substance use when logistic regression was used. Variables derived from social learning theory and the opportunity perspective had among the largest effect sizes. When zero-inflated models were used, findings were more nuanced, suggesting that there are differences in prevalence and frequency of use. These explanations were more consistent for models examining alcohol, tobacco, and marijuana, suggesting that social learning, social bonds, and opportunity may not explain street drug use and prescription drug misuse.

Conclusions

While the findings suggest that peer-based prevention and intervention programs should be used, differences between prevalence and frequency should be addressed. Additionally, while these programs may work for alcohol, tobacco, and marijuana, they may not be beneficial for less commonly used substances.