This study used data from the publicly available 2019 YRBS Middle School Combined Dataset. We pooled representative samples from students in public schools surveyed in the 13 (of 15) states that reported representative data including YRBS’s bullying (at-school and electronic) victimization questions (Table 1 footnote § lists these states). These 13 states also all included questions about the age that the student first tried alcohol and marijuana. Twelve of these 13 states also asked questions about having ever used an electronic vapor product (EVP; see Table 1 footnote ¶) or having ever misused prescription pain medicine (Table 1 footnote #). Nine of these 13 states asked a question about the age that a student first tried cigarette smoking (Table 1 footnote ∥).

For the predictors, we used 2 existing binary variables in the dataset reflecting “yes” responses to ever experiencing bullying victimization at school or electronically, and we created a combined measure for students answering “yes” to both questions. For the outcomes, we used existing binary variables about having ever used an EVP or having ever misused prescription pain medicine. We also created new binary variables for ever having tried cigarette smoking, alcohol, or marijuana, using the questions about the age that the student first tried these substances (ie, the “never” tried the substance answer was coded 0, and all age range answers for having tried the substance were coded 1). A 2-category sex (female, male) and a 7-category race/ethnicity (non-Hispanic [NH] American Indian/Alaska Native (AIAN), NH Asian, NH black or African American, Hispanic/Latino, NH Hawaiian/Pacific Islander (NHPI), NH white, NH Multiple Races) variable were already included in the dataset. Since substance use initiation rates increase during adolescence,²⁴ for an adjustment variable, we recoded the existing 7-category age variable into 3 categories with sufficient cell sizes: ages 11 and younger, 12–13, and 14 and older.

We used the primary sampling unit, stratum, and weighting variables from the dataset to calculate pooled prevalence estimates for the bullying victimization predictors and substance use outcomes, for all students and for sex and race/ethnicity subgroups. Logistic regression models were next used to generate prevalence ratios for each measure of bullying victimization with each substance use outcome, adjusted by age group, sex, and race/ethnicity.

We report significant associations when p < .05. All analyses were performed in SAS version 9.4, using SUDAAN version 11.0.3 to account for YRBS’s weighting and complex survey design. SUDAAN applied listwise deletion for missing data in the logistic regression models.

These findings have implications for prevention and intervention strategies. Although bullying rates are mostly staying the same among high school aged youth,⁵ our study results are consistent with past YRBS results finding higher rates of bullying behaviors reported by middle school students.²⁵ In addition, increasing rates of EVP use and prescription pain medicine misuse, misperceptions about the harms associated with substance use among youth,¹⁶ and a larger context that may be contributing to youth drug use,^17–20 make now an important time for prevention. Middle schools could consider implementing prevention strategies addressing bullying and substance use, jointly or separately, recognizing the association between these risk behaviors. Evidence informed approaches for violence prevention in school settings could be implemented, including increasing school connectedness and creating a protective culture.²⁶

In this study, the female and male groups demonstrated a nearly equal number of significant associations across measures of bullying victimization and substance use (Table 3), suggesting the importance of including both groups in prevention strategies. Although there were fewer associations for the Multiple Races, NH AIAN, and NH NHPI groups, the associations we report here add new evidence for these groups—as well as for the Asian group—which are largely not included in previous research.^4–12 To help equitably prevent bullying victimization and substance use, cultural tailoring of messaging and interventions may be needed, accounting for cultural differences in perceptions, attitudes, and values about bullying and substance use. Last, in this study, we found that many of the aPRs for the measures including electronic bullying victimization and substance use behaviors were often close to or exceeding 2 (Table 3). This strong association, along with frequent media use among youth,²⁷ and low awareness among parents for children’s bullying experiences online,²⁸ may underscore the importance of addressing bullying in the online setting.

Data from the 2021 Middle School YRBS are expected to be released in spring of 2023. Our 2019 results could provide a baseline for which to analyze post-pandemic trends. In addition, the 2021 YRBS data were likely affected by hybrid learning, with issues similar to those reported for the Adolescent Behavior and Experiences Survey (ABES), which used YRBS questions and was also conducted in 2021 (in the spring as opposed to in the fall when YRBS 2021 data were collected).²⁹ The associations between bullying experiences and substance use behaviors among middle or high school students have not yet been assessed using ABES or YRBS 2021 data. A study examining these data could help indicate if associations differed during the pandemic.

Nevertheless, with the public health emergency for the COVID-19 pandemic having ended on May 11, 2023,³⁰ we are entering a new era in public health prevention. Middle school administrators could use our results to plan for bullying and substance use prevention strategies, as schools return to a similar context as 2019 with regard to in-person learning. Also noteworthy in our study was that we found significant associations between bullying victimization and substance use behaviors for all 7 race/ethnicity categories examined. The COVID-19 pandemic showed a light on health risk behavior disparities among high school aged youth,²⁹ and bullying and substance use research would benefit from a more nuanced examination of how bullying is understood by and affects younger students from diverse racial and ethnic backgrounds, as well as how bullying is related to documented disparities in other forms of violence and substance use.

Although the data for our study are made up of representative state level YRBS samples, our pooled data could only include the 13 states that chose to administer the YRBS’s bullying victimization questions in 2019. Our pooling data across over a dozen states that reflect different regions of the country helps to strengthen the evidence base on this topic. In addition, the pooled data allowed us to examine more diverse categories for race/ethnicity. Other limitations of our study include the inability to assess causation or the direction of the associations, and potential overlap between the at-school and electronic bullying questions. We also could not examine bullying perpetration, which the Middle School YRBS does not measure; illicit drug use, due to small cell sizes in the dataset; or sexual identity or the current use of multiple substances, which were not as commonly measured across states. Future analyses examining bullying behaviors, polysubstance use, and sexual identity could help inform strategies for youth at a higher risk for adverse outcomes.