We conducted an exploratory analysis of differences in substance use patterns among AI/AN and non-Hispanic white adolescents in the United States. Pairwise correlations overall showed similar patterns of substance use, particularly for heroin with injected drugs and cocaine with methamphetamines, although the strength of correlations was generally higher among AI/AN adolescents. Exploratory factor analysis identified a common factor for cigarettes, alcohol, and marijuana across both groups, but distinct factors also emerged in each group. Among AI/AN adolescents, injected drugs and steroids comprised a latent factor, whereas among non-Hispanic white adolescents, cocaine and ecstasy comprised a separate latent factor, and methamphetamine alone loaded onto its own factor.

Our study parallels well-known findings about substance use disparities across racial/ethnic groups in adults (15–22). It adds to this body of knowledge by describing differences among adolescents, in particular showing a higher prevalence of substance use among AI/AN adolescents than among non-Hispanic white adolescents. The primary focus of our study, however, extends beyond the prevalence of substance use by examining differences in correlations and patterns of polysubstance use between AI/AN and non-Hispanic white adolescents, about which little is known. Our findings that AI/AN and non-Hispanic white adolescents have different patterns of substance use is consistent in a broader sense with previous findings that risk factors for risky behavior differ between these 2 populations (4–6).

One pattern of substance use common among both AI/AN and non-Hispanic white adolescents was the combination of cocaine and ecstasy; these were strongly associated according to both the correlation analysis and the exploratory factor analysis. Other remarkable correlations in both populations were cigarettes with marijuana and alcohol. Both the heatmap and the factor analysis indicated that if an adolescent used cocaine, he or she would be more likely than not to have also used ecstasy, heroin, methamphetamines, or inhalants. Likewise, if an adolescent used cigarettes, he or she would be more likely than not to have used marijuana and alcohol. Our findings are broadly consistent with previous research on latent classes of substance use (7) in that a pattern emerged for “lighter” substance use (cigarettes, alcohol, and marijuana) and other patterns emerged for “heavier” substance use (substances that are illegal at any age).

We found notable differences between AI/AN and non-Hispanic white adolescents in illegal substance use. In particular, nonmedical steroid use and injected drugs were associated with both AI/AN and non-Hispanic white adolescents, but this pairing comprised its own unique factor among AI/AN adolescents. This pairing indicates that AI/AN adolescents who illicitly use steroids and injected drugs are a distinct subpopulation of AI/AN adolescents. A link between steroid use and injected drugs was reported previously (23), but to our knowledge, this link has not been reported among AI/AN adolescents. Distinct subpopulations of non-Hispanic white adolescents were uniquely characterized by methamphetamine use, indicating that non-Hispanic white methamphetamine users are a distinct subpopulation of non-Hispanic white adolescents. We found this unique characterization despite the overall lower prevalence of methamphetamine use among non-Hispanic white adolescents than among AI/AN adolescents. Similarly, cocaine and ecstasy use characterized another subpopulation of non-Hispanic white adolescents.

Future research should focus on replicating these patterns in other independent samples, examining temporality and possible causality and further refining the exact nature (eg, which drugs are injected and what type of steroids are being used) and timing (eg, whether cocaine and ecstasy are being taken together in a single episode) of these patterns. Validation of racial/ethnic differences in polysubstance use patterns would motivate the development of tailored interventions for subgroups to prevent or reduce the prevalence of polysubstance use.

Strengths of our study include the novel use of exploratory methods to assess polysubstance abuse among AI/AN and non-Hispanic white adolescents. Additionally, our study extends previous research on racial/ethnic disparities in the prevalence of use to examine patterns of polysubstance use. Finally, our study used a large, national data set, which increases the generalizability of the findings to the larger population of high school–aged students in the United States.

Our study had several limitations. First, it used cross-sectional data, which cannot be used to assess causality or rule out residual confounding. Second, because YRBS data are self-reported, they are subject to social desirability and recall biases. Third, the use of cigarettes and alcohol were measured in “use within the past 30 days,” while all other substances were measured in “lifetime use.” This difference results in imperfect direct comparisons between substances. Fourth, the YRBSS has a large degree of missing data on substance use, and these data may not be missing at random. For example, because the sample is limited to students who were at school on the day of the survey, truancy could have affected the findings, especially since truancy is known to be associated with substance use. The patterns of missing data may have affected our findings in unknown ways, and thus the findings should be considered exploratory only. Fifth, it was not possible to examine national YRBS public use data by state or region, and because AI/AN subpopulations can differ by state or region in important ways, further research is needed to explore these differences. Sixth, factor analysis should be considered an exploratory method, and the results may be sensitive to the particular sample or to the specific method of measuring the latent factors (eg, the wording of the questions). Additionally, factor analysis is somewhat subjective, particularly in the way factors are interpreted and the role of interpretability in selecting the number of factors. Finally, several of the substance use items, such as steroids, inhalants, and injected drugs, were broadly defined, and detailed information on the exact substance was not available.

The practicality of our study is noteworthy. The data show strong pairwise correlations between common substances of abuse and classify AI/AN and non-Hispanic white adolescents into well-structured latent factors, indicating that distinct patterns of polysubstance use exist in each subpopulation. A health care professional could use this information to inquire about substance use if a substance is known to be used and subsequently educate susceptible adolescents about substance use habits with the goal of preventing polysubstance abuse. Public health interventions and policy makers can also use these data to formulate strategies for reaching certain racial/ethnic groups to decrease the prevalence of substance abuse with more efficiency and effectiveness.