Eligible participants met the following criteria: (1) male-identifying; (2) Black/African American; (3) aged 15‐24 years; (4) a score >1 on the 4-item Serious Fighting, Friend Weapon Carrying, Community Environment, and Firearm Threats (SaFETy) clinical screening tool, indicating elevated risk for future firearm violence based on 4 domains: Serious Fighting, Friends’ Weapon Carrying, Community Environment, and Firearm Threats [43,44]; (5) English literate; and (6) able to assent/consent (with parental consent if under 18).

Exclusion criteria were (1) being detained or incarcerated; (2) inability to complete assent/consent and assessment forms due to language barriers, cognitive dysfunction, or injury; (3) presenting with an acute psychotic disorder, a suicide attempt as the mechanism of firearm injury, or a recent sexual assault requiring intensive psychosocial support; and (4) a score ≥6 on the SaFETy tool, as such individuals were referred to more intensive programs, as recommended by Goyal et al [45]. Recruitment was successfully conducted in collaboration with community partners, including local organizations, school-based health centers, and social media, with minimal loss to follow-up due to the intervention’s brief and single-session format.

Upon expressing interest in the study, participants completed a sociodemographic enrollment form on Research Electronic Data Capture (REDCap; Vanderbilt University), adapted from the biennial cross-sectional Youth Risk Behavior Surveillance System (YRBS) survey. All study procedures were reviewed and approved by the Rush University Institutional Review Board. Informed e-consent was obtained in accordance with federal and institutional guidelines for research involving minors using REDCap. Consent and assent forms were written in age-appropriate and culturally sensitive language, and all participants were informed of their right to decline or withdraw at any time without penalty.

Participants then completed a pretest survey battery comprising 63 questions assessing psychological distress (Kessler Psychological Distress Scale [K10]), depression (8-item Patient Health Questionnaire [PHQ-8]), attitudes toward guns and violence (Attitudes Toward Guns and Violence Questionnaire [AGVQ]), and aggression (Reactive-Proactive Aggression Questionnaire [RPQ]). All survey questions remained unchanged, as these instruments have been previously validated for use with children and adolescents. To ensure greater control over participant access, engagement tracking, and data capture during the pilot phase, the intervention was delivered via REDCap, with embedded YouTube-hosted videos simulating the app’s video modules. This approach enabled secure and streamlined testing of the core acceptance and commitment therapy (ACT)–based video modules before full app deployment, allowing us to evaluate the acceptability and preliminary effects of the behavioral modules while testing fidelity to content delivery and use styles, and engagement tracking.

Immediately following baseline survey completion, participants were given access to the BrotherlyACT modules and instructed to complete all video-based modules and in-app content either in 1 sitting or at their own pace within a 48-hour window. The posttest survey was administered electronically exactly 4 weeks (28 days) after the pretest, regardless of when participants completed the intervention content.

Additionally, we tracked the amount of time participants spent on the REDCap page and their engagement with the videos (ie, in-app quiz completion, likes and comments). Participants who completed the video content prematurely, either by not viewing all the materials or by skipping through them rapidly, were encouraged to reengage through various means, including indicating approval, providing comments, or submitting a summary of the knowledge acquired from the video content via email. Incentives for participating in the study were distributed after completion of the pretest and video modules (US $50) and again 4 weeks later, following completion of the posttest questionnaires (US $50 in the form of Target and Amazon e-gift cards).

Drawing from evidence-based behavior change models and community insights, the app incorporated elements that address the personal, social, and environmental factors contributing to youth firearm-related violence. Its development involved collaborative design with young Black males from the target age group, ensuring that the app’s content, imagery, language, and delivery methods were culturally attuned to their real-life experiences.

As at testing, the app comprised a set of 8 concise, video-based psychoeducational modules that depict realistic situations. These modules address topics such as refusing guns, resisting gang involvement, practicing gun safety, saying no to alcohol and drugs, resolving conflicts, avoiding retaliation, recognizing healthy and unhealthy relationships, practicing mindfulness, cultivating a positive outlook on the future, and using adaptive coping strategies. They incorporate behavioral skills training, emotional regulation techniques, and decision-making strategies to help reduce youth violence and risky firearm behavior. The app’s video modules are primarily based on ACT [46-49], a third-wave cognitive behavioral therapy that emphasizes accepting and navigating difficult thoughts and emotions rather than attempting to persistently change or avoid them [50-54]. ACT has demonstrated efficacy in improving emotional regulation and reducing aggression among trauma-exposed youth, who often cannot control their environments but can control how they respond to and navigate these chronic and intersecting issues [55]. Utilizing a micro-learning approach, complex topics were broken down into smaller, ready-to-use and life skill-based components to enhance comprehension, application, and retention, particularly for youth with various attention spans [56,57].

A Black male narrator presents the modules with multimedia and interactive heuristics, offering skills for navigating and preventing firearm-related scenarios such as peer pressure to carry weapons, retaliatory violence, and rejecting gang involvement. The modules use culturally familiar language and visuals, with a tone that emphasizes respect, empowerment, motivational enhancement, and nonpunitive language to avoid pathologizing young people’s behavior. Table 1 summarizes the behavioral skill objectives and theoretical frameworks underpinning each BrotherlyACT module.

The broader BrotherlyACT app also includes additional features such as (1) a safety planning toolkit with interactive tools for mood tracking, risk assessment, mindfulness exercises, goal setting, and a zip code–enabled directory of local mental health services, community support resources, and violence prevention programs within a 50-mile (80.47 km) radius; and (2) service engagement and talk therapy via an artificial intelligence–powered chatbot named DEVON, a confidential and programmed resource for discussing stressors, emotional triggers, and challenges related to violence and substance use. DEVON was designed as a trauma-informed conversational agent to simulate peer dialogue and support help-seeking behaviors. Embedded within the app interface, the chatbot provides confidential, just-in-time emotional support accessible outside structured intervention sessions. Conversations with DEVON are anonymous, with user input logged only as anonymized session data. Participants are advised not to enter personal identifiers. App access was password-protected and intended for use on personal or study devices in private settings. These additional features were not evaluated in this study.

We adapted items from the Centers for Disease Control and Prevention’s biennial YRBS, which assesses health-related behaviors among a nationally representative sample of US students in grades 9‐12. Our analysis included sociodemographic variables such as age, gender, race/ethnicity, socioeconomic status (ie, household income), school and employment status, zip code or address, dating relationship status, and smartphone ownership. The questionnaire also assessed violence (physical and psychological violence toward a partner, 30-day weapon carriage, and community violence) and substance use experiences (ie, 30-day adolescent binge drinking, nonmedical drug use, and substance-related violence perpetration). Substance use behaviors were assessed due to their known association with youth aggression, trauma exposure, and risky weapon-related behaviors [8,27,58]. These variables were included to examine participants’ broader behavioral profiles and identify overlapping intervention targets. Table 2 summarizes the sociodemographic characteristics of participants who completed the pretest and posttest questionnaires and received the intervention (ie, the analytic sample).

The SaFETy score, a 4-item firearm risk assessment questionnaire included in our enrollment survey, was used to categorize participants into low-risk (0), medium-risk (1-5), and high-risk (≥6) groups for firearm violence based on weighted item responses. The questionnaire inquired about the frequency of serious physical fights, friends carrying weapons, hearing gunshots, and having guns pulled on them in the past 6 months. This clinically feasible tool has been shown to correlate with firearm violence, although further validation is needed [43]. The SaFETy score demonstrates potential in predicting future firearm violence risk, with each point increase associated with a 1.5-fold higher likelihood of firearm violence within 2 years (odds ratio 1.5), underscoring its value as a clinical screening instrument [44,59]. Administration of the tool takes less than 1 minute, and Goldstick et al [44] suggested referring patients with scores ≥6 to more intensive programs.

The 23-item AGVQ assesses individual attitudes toward using guns and violence for conflict resolution, power assertion, or safety across 4 scales: Aggressive Response to Shame (disrespect sensitivity; items 1, 2, 5, 10, 15, 16, 18, and 20), Comfort With Aggression (violence acceptance; items 8, 12, 13, 17, 21, and 22), Excitement (gun-related excitement; items 6, 7, 9, 11, and 14), and Power/Safety (power and safety from guns; items 3, 4, 19, and 23). Cronbach α coefficients were 0.83, 0.81, 0.79, and 0.72, respectively. Respondents rated their agreement on a scale from 0 (disagree) to 2 (agree). The summed scores for the AGVQ subscales ranged from 0 to 46, with higher scores indicating more favorable attitudes toward guns and violence. Juvenile firearm possession is linked to violence lethality and may reflect risky exposure to firearms or a proviolence viewpoint. Although the correlation between attitude and behavior change varies, modifying attitudes is crucial for changing behavior [60]. Youth typically complete the AGVQ in 5‐10 minutes. The AGVQ correlates with handgun ownership [31] and with various psychological and behavioral assessment instruments, supporting its construct validity and helping to understand how attitudes toward guns and violence relate to broader patterns of violent behavior, cognitions, and emotions. Although the AGVQ remains an insufficiently addressed aspect of pediatric firearm violence prevention, interventions such as the “Cradle to Grave” program have shown potential in reducing AGVQ scores, particularly among students with higher baseline tendencies toward violence [61].

The RPQ is a 23-item validated self-report instrument designed to differentiate between reactive aggression (11 items; eg, “Reacted angrily when provoked by others”) and proactive aggression (12 items; eg, “Had fights to show who was on top”). Respondents reflect on these behaviors over the past 30 days, rating each item on a 3-point Likert scale (0=never, 1=sometimes, and 2=often). Subscale scores were summed to generate separate reactive and proactive aggression scores, as well as a combined total aggression score. Higher scores indicate a greater frequency of aggressive behaviors [29,62]. The RPQ has demonstrated good reliability and construct validity in adolescent populations and has been extensively used in research on aggression, violence, and trauma exposure. Reactive aggression is more strongly linked to internalizing symptoms, such as anxiety and depression, whereas proactive aggression is associated with externalizing and antisocial traits [29,62].

The K10 [63] is a widely used self-report diagnostic tool for identifying individuals who may require further evaluation for anxiety and depression. It is applicable in both the general population and clinical settings. The K10 comprises 10 questions rated on a 5-point scale (5=all the time to 1=none of the time), reflecting experiences over the past 4 weeks. Scores range from 10 (no distress) to 50 (severe distress), with lower scores indicating less distress. The scale includes 2 subscales: depression (items 1, 4, 7, 8, 9, and 10) and anxiety (items 2, 3, 5, and 6). Items 3 and 6 are omitted if the previous response is “none of the time” [64].

The PHQ-8 is extensively used to evaluate depression severity in clinical and research settings [65,66]. It comprises 8 items aligned with the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition criteria for major depressive disorder and serves as a reliable tool for screening and monitoring depressive symptoms. The PHQ-8 is a shortened version of the PHQ-9, omitting the item on suicidal ideation. It assesses mood and feelings over the past 2 weeks on a 4-point scale from 0 (not at all) to 3 (nearly every day), with total scores ranging from 0 to 24. Scores above 10 indicate moderate depression, while scores above 20 indicate severe major depression.

Nearly half of the participants (32/70, 46%) reported engaging in a serious physical fight within the past 6 months. During the same period, 31 (44%) indicated that a friend had carried a weapon (knife, razor, or gun), and 31 (44%) reported having been directly threatened with a gun at least once. Additionally, 41 (59%) had heard gunshots in their neighborhood in the previous 6 months. Exposure to community violence was highly prevalent, with 49 (70%) participants endorsing direct or indirect involvement in violence (eg, fighting, threats, or gang activity) in the past year. Furthermore, 53 (76%) respondents reported witnessing someone being attacked, beaten, stabbed, or shot in their neighborhood. Regarding relationship aggression, 41 (59%) reported psychological aggression toward a romantic or dating partner, and 31 (44%) disclosed having engaged in physical aggression in the past year. Notably, 31 (44%) reported carrying a weapon at least once in the past 30 days.

Table 3 presents item-level responses and scoring contributions for each domain of the SaFETy score: Serious Fighting, Friend Carrying Weapons, Community Environment, and Firearm Threats. The most commonly endorsed level for Serious Fighting was “3‐5 times in the past 6 months,” reported by 19 of 70 (27%) participants, contributing 2 points to the SaFETy score. In the Friend Carrying Weapons domain, 38 (54%) participants reported that at least some of their friends had carried a weapon in the past 6 months, with 7 (10%) participants reporting that many, most, or all of their friends had done so, contributing 1 point.

Community environment exposure was notably high: 57 (81%) participants reported hearing gunshots at least “a few times” in the past 6 months, with 26 (37%) hearing them “many times,” contributing 3 points. The highest-weighted item was Firearm Threats: 17 (24%) participants reported being threatened with a gun 2 or more times, contributing 4 points. Notably, 22 (31%) respondents scored 0 across all categories of firearm threat. These distributions reflect a high prevalence of direct and indirect exposure to firearm violence in this sample, contributing to higher cumulative SaFETy scores and elevated future firearm violence risk.

The distribution of total SaFETy scores among participants reveals a wide range of firearm risk exposure. While a notable portion of the sample scored 0, indicating minimal or no reported exposure to serious fighting, weapon-carrying peers, gunfire in the community, or direct firearm threats, the majority reported at least one source of exposure. A significant subset scored between 4 and 7, reflecting high cumulative exposure, particularly in the domains of community gun violence and direct firearm threats. This right-skewed distribution suggests heterogeneity in risk profiles, with a sizable group facing elevated vulnerability to firearm-related harm, an important consideration for intervention targeting and future risk stratification.

We assessed multiple outcomes (summarized in Table 4), such as attitudes toward guns and violence, reactive and proactive aggression, and mental health outcomes (eg, psychological distress).

A significant reduction in AGVQ scores was observed among participants (n=43) from pretest (mean 29.8, SD 8.7) to posttest (mean 26.1, SD 9.0), with a mean difference of 3.69 (SE 1.05, t₄₂=3.522, P<.001), and a moderate effect size (Cohen d=0.53). Subscale analysis revealed the most considerable improvement in the “Aggressive Response to Shame” subscale (28% reduction), followed by the “Excitement Toward Guns and Violence” subscale (14.8%), the “Feelings of Power/Safety” subscale (8%), and the “Comfort with Aggression” subscale (2.3%). These crucial findings indicate that participants experienced reduced emotional reactivity to perceived threats and decreased excitement and reliance on firearms for power and safety. For example, Goldberg et al [61] found that small but statistically significant decreases in AGVQ scores were predictive of lower intentions to carry firearms and decreased willingness to use violence in future peer conflict situations.

Mean scores indicated that reactive aggression significantly decreased from 10.48 (SD 5.39) preintervention to 8.67 (SD 4.92) postintervention (mean difference 1.82, SE 0.66, t₅₃=2.74, P=.008), with a small-to-moderate effect size (Cohen d=0.37), suggesting a notable improvement in managing impulsive, emotionally driven responses to perceived threats following exposure to the intervention. By contrast, no significant change was detected in proactive aggression scores between the pre- and posttest (t₅₃=0.82, P=.42), with a small effect size (Cohen d=0.11), indicating the need for longer-term or more intensive and targeted interventions to address the underlying cognitive patterns and motivations associated with planned aggression.

A paired-samples t test was used to evaluate changes in psychological distress, focusing on subscales (depression and anxiety) and overall psychological distress (ie, total scores) from pre- to posttest. No significant changes were found in depression symptoms from pretest (mean 15.85, SD 6.28) to posttest (mean 15.97, SD 5.85, t₅₈=–0.20, P=.84, Cohen d=0.03). Likewise, anxiety scores did not significantly differ between pretest (mean 10.32, SD 3.93) and posttest (mean 10.02, SD 3.88, t₅₈=0.76, P=.45, d=0.10). In addition, overall psychological distress scores remained stable from pretest (mean 26.17, SD 10.02) to posttest (mean 25.98, SD 9.54, t₅₈=0.20, P=.84), with a negligible effect size (d=0.03). At baseline, 21 of 70 (30%) respondents reported potentially moderate to severe psychological distress (K10 scores>16), highlighting a substantial mental health burden within this sample.

A change of 5 points or more on the K10 is considered clinically meaningful for psychological distress [67,68]. Based on our sample mean difference (mean difference 0.186), the absence of both statistically and clinically significant changes suggests that the intervention may have had a limited short-term impact on psychological distress, underscoring the need for enhanced or sustained mental health support in the long term.

The Shapiro-Wilk test revealed a significant deviation from normality in the pretest scores (W=0.939, P=.008), indicating a need for a nonparametric approach. A Wilcoxon signed-rank test was conducted to compare pretest (mean 9.82, SD 6.41) and posttest (mean 9.33, SD 6.26) scores. The results showed no statistically significant difference (W=597.50, z=0.903, P=.37) in depression, corroborating earlier findings. Although descriptive statistics suggested a slight reduction in the mean PHQ-8 scores, this change was not significant. For the PHQ-8, a reduction of 5 points or more is widely accepted as the minimal clinically important difference in both primary care and research settings [66].

Backend analytics from REDCap were used to monitor time stamps and session durations. Almost all participants (68/70, 97%) completed the video modules in a single session, with 47 (67%) finishing within an hour, suggesting high feasibility of BrotherlyACT as a digital intervention. Participants were advised to complete 1-2 modules daily, with no upper limit per session. This flexible approach accommodated varying schedules and comfort levels while ensuring consistent content delivery. Future iterations will aim to explore sustained engagement and assess feasibility in school and clinical settings.