The study sample included 2,002 participants in the control arm of the Community Youth Development Study (CYDS), a community-randomized trial of the Communities That Care (CTC) prevention system in communities across 7 states (Colorado, Illinois, Kansas, Maine, Oregon, Utah, Washington).^32–34 All public school students in the 5^th grade in 2003–2004 were eligible to participate. Parental consent and student assent were obtained for 77% of the eligible population. The analyses in this study used survey responses from ten waves of data collected across two developmental periods: six waves in adolescence (from ages 12–18) with at least 89.8% retention through Age 18 and four waves in young adulthood (ages 19–26) with at least 86.4% retention through age 26. These retention rates refer to response rates at each study wave rather than cumulative retention. Participants were retained in the sample even if they did not respond in a particular survey wave. Most participants had fully populated survey response data across all study waves. For example, responses on handgun carrying were fully populated or only missing in one study wave for 79.7% of the sample (1595 of 2002). The sample was gender-balanced and ethnically and racially diverse; additional details about procedures and the demographic characteristics of participants are provided in the Appendix. The CYDS protocol and this study were reviewed and approved by the University of Washington Human Subjects Review Committee.

Participants were asked how many times they carried a handgun in the prior year from age 12–26. We used latent class growth analysis (LCGA), a person-centered, data-driven, probabilistic, full information maximum likelihood model to estimate handgun carrying trajectories from age 12–26 for all participants in the sample in Mplus software version 8.³¹ Six distinct trajectories of handgun carrying among all 2,002 study participants were identified (Figure 1).³¹ Each participant was assigned to a trajectory based on the highest posterior probability of group membership for the present analyses.^35,36 A never/low probability of carrying across the entire study period was the most common pattern (n=1,590;79.4%). LCGA grouped these participants with a zero probability of handgun carrying and very low probabilities of handgun carrying from age 12–26 into one trajectory, suggesting there are little to no meaningful differences in those who never or rarely carry. A small proportion of the sample (n=6;0.3%) had a high and persistent probability of carrying throughout both adolescence and young adulthood. About 12% of the sample had elevated probabilities of carrying during adolescence. Of these, one trajectory (n=53;2.6%) was characterized by moderately high probabilities of carrying a handgun during early adolescence that declined rapidly into young adulthood. A smaller group (n=24;1.2%) carried with high probabilities in adolescence and gradually declining probabilities into young adulthood. The largest adolescent group (n=163;8.1%) had carrying probabilities that steadily increased from adolescence into young adulthood. The final trajectory (n=166;8.3%) initiated handgun carrying later with positive probabilities of carrying limited to young adulthood.

We measured experiences and uses of different forms of violence: (1) experiencing bullying during adolescence, (2) using bullying during adolescence, (3) experiencing parental physical violence more than once before age 18, (4) using physical violence (e.g. beating or attacking with intent) during adolescence, (5), experiencing intimate partner physical violence during young adulthood, and (6) using physical violence during young adulthood, including the same forms during adoelsecence as well as intimate partner violence. All of these measures were coded as equal to one if a participant reported the behavior in any wave during the developmental period and zero otherwise. Additional details about the survey questions used to create these measures are available in the Appendix. These forms of violence have been noted in urban literature as associated with weapon carrying, particularly handgun carrying. These measures are developmentally appropriate for the age ranges of the study (e.g., bullying in adolescence and intimate partner violence in young adulthood), but do not fully describe all possible exposures to bullying or physical violence in both developmental periods.

All participants were included in all analyses. First, we used latent class growth analysis (LCGA) to estimate handgun carrying trajectories from age 12–26. Next, data on participants in each wave were assessed across two time periods, adolescence (ages 12–18) and young adulthood (ages 19–26), and each violence measure captured whether a participant ever endorsed the behavior in each developmental period. Although many questions remained the same across study waves, there were adjustments over time to reflect changing developmental salience. Details on the availability of survey questions for the measures used in this study are provided in Table A1. There was no missingness for exposure or outcome measures, but missing information for component questions is provided in Table A2. We described the proportion of each violence measure overall and among each handgun carrying trajectory. To characterize the strength of the association between the handgun carrying patterns (exposure) and six measures of experiencing and using physical violence (outcomes) we used logistic regressions. In all regressions, the never/low carrying trajectory was the reference trajectory to capture whether handgun carrying patterns are risk markers for each form of violence. We used single-level regressions because intraclass correlation (ICC) of handgun carrying at the community-level was low in LCGA (ICC = 0.000554, se = 0.0029168), and a higher degree of clustering increases the need for a multi-level regressions.³⁷ Covariates were not included because this study uses a within-person approach to examine patterns of behavior in a group of rural youth. There is no temporal order between the exposure and the outcome, and many potential confounders (e.g., peer behavior, parental monitoring) may be intermediaries on the various causal pathways between handgun carrying, bullying, and physical violence.³⁸

Most participants reported experiencing (73.2%) and using (52.4%) bullying during adolescence (Figure 2). The majority of youth who experienced bullying also bullied others during adolescence (59.3%; 869 of 1,466), and these behaviors were persistent over time. Of those who experienced bullying, 71.4% (1046 of 1466) reported this in multiple waves. Of those who used bullying, 59.5% (624 of 1049) reported doing so in multiple waves. Around 36.0% of youth reported using physical violence during adolescence, and 36.7% reported using physical violence during young adulthood (Figure 2). The use of physical violence co-occurred less frequently with experiencing physical violence, and physical violence behaviors were less persistent than bullying. Most participants who reported physical violence in young adulthood did so only once (experiencing: 64.8%, 406 of 627; using: 58.0%, 426 of 734). One exception was many who experienced physical violence by an intimate partner during young adulthood also used physical violence during young adulthood (82.3%; 516 of 627). Most participants never reported carrying a handgun (70%; 1401 of 2,002), but among those who did, some handgun carrying trajectories had high estimated probabilities of carrying in adolescence and/or young adulthood (Figure 1). Overall, 20.7% of participants reported carrying a handgun at least once during adolescence and 17.5% reported handgun carrying during young adulthood (Figure 2).

The most striking differences in bullying and physical violence across handgun carrying trajectories were related to the uses of these behaviors during adolescence and young adulthood (Figure 3). Trajectories with elevated probabilities of carrying during adolescence (e.g., declining, adolescent) had the highest proportions (91.7%, and 79.2%, respectively) of bullying others during adolescence. These patterns with high probabilities of carrying during adolescence were also more likely to report using physical violence during adolescence (range:59.5–87.5%) compared to the never/low carrying trajectory (30.6%), and the group with carrying limited to young adulthood (40.4%). Similarly, there were sizeable proportions of young adults who reported using physical violence between ages 19 and 26 among those with a pattern of high/persistent carrying (66.7%), declining carrying (70.8%), and steadily increasing carrying (49.7%) compared to the never/low trajectory (34.2%).

Almost all handgun carrying trajectories had higher odds of experiencing physical violence in both adolescence (between 1.5 and 2 times) and young adulthood (between 1.5 and 2.5) compared to the never/low trajectory though only some of these differences were precisely estimated (Table 1). The trajectory with handgun carrying limited to adolescence had 1.93 (95%CI=[1.07,3.48]) and 2.19 (95%CI=[1.26,3.79]) times the odds of experiencing physical violence during adolescence and young adulthood, respectively. Similarly, the trajectory with handgun carrying during young adulthood had around 1.5 times the odds of experiencing physical violence during adolescence and young adulthood (OR=1.47,95%CI=[1.02,2.12] and OR=1.50,95%CI=[1.07,2.09], respectively). The steadily increasing trajectory with increasing probabilities of handgun carrying through young adulthood also had 1.62 times the odds of experiencing physical violence during young adulthood (95%CI=[1.17,2.26]). Notably, all six individuals assigned to the high/persistent trajectory reported experiencing bullying during adolescence.

Trajectories with high estimated probabilities of handgun carrying during adolescence (declining, adolescent, steadily increasing) were associated with 11.22 (95%CI=[2.63,47.89]), 3.89 (95%CI=[1.99,7.62]), and 1.95 (95%CI=[1.39,2.73]) times higher odds, respectively, of using bullying during adolescence (Table 1). While the estimated association between the declining trajectory and using bullying during adolescence is imprecisely estimated, these results show there is a sizeable and strong association between these trajectories and bullying. While only 50% of those in the never/low carrying trajectory ever reported bullying someone, 92% of those in the declining trajectory did.

All five handgun carrying trajectories were associated with higher odds of using physical violence during adolescence compared to the never/low trajectory, and several were also associated with using physical violence during young adulthood (Table 1). The adolescent (9.74, 95%CI=[4.85,19.54]) and declining trajectories (15.85, 95%CI=[4.71,53.40]) had the largest associations with using physical violence during adolescence. While both estimates are imprecisely estimated due to small sample sizes, both trajectories were associated with large increases in the odds of using physical violence when the probability of handgun carrying was elevated. Similarly, the steadily increasing trajectory was associated with over three times the odds (OR=3.33, 95%CI=[2.39,4.63]) of using physical violence during adolescence and almost twice the odds (OR=1.90, 95%CI=[1.37,2.63]) of using physical violence during young adulthood. Overall, these three trajectories with higher probabilities of handgun carrying during adolescence were associated with various measures of violence perpetration across both adolescence and young adulthood. The findings described here are generally robust to regressions adjusting for male gender which varies across trajectories and is associated with both bullying and physical violence (Table A3 and Figure A1).