Dating Violence Prevention for Juvenile-Justice Involved Females: A Hybrid Trial

This study was conducted between September 2014 and September 2019. It was approved by the [blinded institution] institutional review board, and human subjects’ protections oversight was provided by a data and safety monitoring board.

Adolescent females (14–18 years old) with an open family court petition were recruited from a family court in the Northeast United States. To enhance real-world conditions, females from all levels of the family court system (eg, intake, truancy court, formal calendar, group homes serving justice involved youth) were eligible, regardless of dating status or ADV history. As shown in Figure 1, participants were urn randomized into the Date SMART experimental intervention (DS; n = 120) or a knowledge-only comparison (KO; n = 120). Retention rates to the final 12-month assessment were excellent: 75% in DS and 80% in KO.

Adolescent were recruited surrounding appointments with court staff. Youth were also referred to the study by intake workers, judges, and magistrates. Adolescents placed in group homes serving justice involved youth were offered the opportunity to participate as well. Although participation in one of our intervention groups (DS or KO) could be mandated by a magistrate or judge, participation in the research portion was voluntary. Court staff were not involved in data collection, nor aware of whether an adolescent was participating in research.

After obtaining assent and parental consent for minors (14–17 years old) and informed consent for adults (age 18), participants completed a baseline audio computer-assisted self-interview survey on tablet computers, an interview, and paper-pencil questionnaires. Adolescents were then randomized in blocks of 8 to condition. Group homes were randomized as a unit. Assessments were completed every 3 months through the final 12-month follow-up. Participants received $40 for each assessment.

Both groups were matched for time, attention, and approach (eg, discussions, games). They included 6 weekly group sessions (2 hours each), followed by 1 booster session 6 weeks after the active phase. All sessions were implemented at the family court, aside from 6 of the 41 cohorts whose sessions were implemented in youth group homes.

Date SMART (DS) is a cognitive behavioral therapy group intervention focused on providing skills related to depression and aggression reduction, emotion regulation, and relationship communication, to reduce ADV and unprotected sex. A detailed summary of the intervention can be found in Rizzo et al.⁹ 

Knowledge Only (KO) is an active intervention group developed to provide interactive games and activities that educate youth about dating violence and sexual health, as well as other common health topics. The ADV and sexual health content aims to increase knowledge, as well as shift attitudes. Cognitive behavioral therapy skills are not presented.

All groups were led by 2 facilitators. To promote system integration, family court staff were trained as facilitators, in addition to masters and doctoral-level psychology trainees, using standardized training procedures and manuals. Facilitator assignments were made by drawing from this large pool of staff or trainees (n = 17), with 38% of groups led by a facilitator team that included at least 1 court staff. Intervention fidelity was achieved through intervention manuals and weekly supervision meetings. Sessions in both arms were audiotaped (83%) to ensure proper implementation. Facilitator-completed ratings revealed excellent (98%) manual adherence, as did doctoral-level project staff ratings (rated for 20% of sessions; 97% adherence). There were no significant differences in adherence to the treatment manual between intervention arms or on the basis of whether a court staff was serving as 1 of the facilitators.

Participants reported demographic information including age, sex, race, ethnicity, teen residing in a single parent household, and access to free or reduced price lunch.

For all surveys, the term ‘partner’ was defined as “a boyfriend or girlfriend, sexual partner, or someone you are going out with. You could be committed to this person (dating only them) or you could be in an open relationship where you are dating other people.”

The Conflict in Adolescent Relationships Inventory¹⁶  is a 35-item measure assessing ADV perpetration and victimization with demonstrated reliability and validity.¹⁶  It includes response options from “never” to “often” (happened more than 6 times). At baseline, participants reported on lifetime ADV and past 90-day ADV. Follow-ups assessed the past 3 months. Perpetration and victimization experiences for all forms of ADV were correlated (Pearson rs = 0.20–0.67), suggesting mutual aggression and consistent with the larger literature.^24–26 

The Digital Relationship Behaviors includes 20 items assessing cyber dating abuse perpetration and victimization (eg, “has a dating partner asked for a personal password to a social networking site?”) Internal consistency for this scale was excellent (α = 0.83).

The Timeline Followback-Dating Violence (TLFB-DV) is a semistructured calendar-based interview method for assessing recent relationship violence on the basis of the TLFB-Spousal Violence interview with excellent test-retest reliability and evidence for both concurrent and discriminant validity.¹⁷  The TLFB-DV was administered by a trained research staff to measure the number of days participants were engaged in a romantic relationship, as well as incidents of physical ADV (hit, slapped, punched) and sexual ADV (forced sex) with each partner.

Adolescents reported whether they’d ever had vaginal or anal sex at baseline, the number vaginal and/or anal sex acts, and the number of condomless sex acts over the past 90 days. Oral sex was not examined, because the sexual health content in the intervention focused on vaginal and/or anal sex.

The Delinquency Activities Scale is a 40-item checklist based in part on the Self-Reported Delinquency Scale²⁰  and designed to assess delinquent offenses (eg, “Have you attacked someone with the idea of seriously hurting or killing him/her?”). Adolescents responded “yes or no.” Total counts were computed for lifetime at baseline and for the past 3 months at follow-ups.

Legal information regarding source of referral (eg, truancy, intake) and any history of incarceration was extracted from a statewide court database of all juveniles processed through the family court.

As a metric of system buy-in, we monitored whether judges or magistrates chose to mandate participation in a group intervention (research enrollment remained voluntary).

Pretest equivalence of conditions on demographics and outcome variables was assessed by using analyses of variance for continuously scaled variables and χ² tests of independence for categorical variables. The size of the difference was estimated by using standardized mean difference scores using tableone v0.12.0.

Missing data ranged from 1% to 7% at baseline and 23% to 31% at 12 months. For the TLFB, 77% completed all assessments. Participants who missed previous assessments were asked to complete an extended TLFB-DV interview that covered the missed assessments. Bias because of missing values was addressed by using multiple imputations with imputations generated through chained equations,²¹  with 100 imputations generated using the R package mice v3.13.

Generalized estimating equations were used to model change from baseline to the 12-month follow-up for all primary and secondary outcomes (primary prevention). This approach accounted for nesting of assessments (baseline and 12-months) within participant and for overdispersion in the count outcomes because of zero-inflation. We examined between cohort variation on the outcomes using the median rate ratio,²⁷  which provides an estimate of the strength of cohort context in the metric of rate ratios. The median rate ratio ranged from 1.00 to 1.32 with an average among outcomes of 1.07. Given the limited between cluster variation we chose to simplify the models by not modeling the nesting of participants within cohort. For outcomes distributed as counts, the models were fit with a Poisson distribution, log link function, and exchangeable working correlation structure. All models included recruitment source (juvenile intake department versus other) as a covariate. Effect sizes were estimated by using rate ratios. Self-report measures were not behavioral counts per se, but data followed a zero-inflated distribution and data were transformed to integers to run the models.

ADV, sexual risk behaviors, and delinquency followed zero-inflated distributions. For ADV and sexual risk behaviors the zero-inflation had 2 sources, those who were not in a dating relationship and those in a relationship but not reporting the behavior. For primary prevention analyses, we used a hurdle model evaluating no-ADV versus ADV with logit link and binomial distribution. Among those reporting any ADV (indicated sample), we used a model with log link and zero-truncated Poisson to model amount of ADV. The same analyses were run for sexual risk behaviors. Given the size of the indicated sample, perpetration and victimization scores were combined to reflect ADV involvement. This decision was supported by our data and is consistent with previous work.²⁴  For delinquency, a hurdle model was run by using the full sample instead of just those reporting a dating relationship.

We again used generalized estimating equations for behaviors that were able to be summed across the 6- to 12-month follow-up to assess primary prevention and indicated prevention effects. The models were largely the same as those used for the primary and secondary outcomes except that we included an offset term to account for individual differences in the amount of time between baseline and the cumulative sum.

Adolescent females (N = 240) had a mean age of 15.58 years. The sample was 33% White, 27% African American, 14% American Indian, 2% Asian, 1% Hawaiian or Pacific Islander, and 24% identified as other. Roughly 49% percent of youth ethnically identified as Hispanic. Just under one-half of juveniles resided in a single parent household (47%). A majority qualified for a free or reduced-price lunch (79%). Forty-three percent identified as bisexual, homosexual, or undecided. Seventy-six percent of youth were dating at baseline. There was no significant difference between conditions on baseline characteristics (Table 1). Regarding system buy-in, a large portion (41%) were mandated to participate in a group by a magistrate or judge.

Baseline and 12-month outcomes are shown in Tables 2 and 3. Primary prevention analyses among the full sample revealed decreases in rates of ADV in both treatment conditions, with larger decreases in Date SMART, although these differences do not reach statistical significance. Among the full sample, there were minimal changes in unprotected sex and delinquency. When examining cumulative outcomes, both conditions showed decreases in physical and/or sexual dating violence as measured by the TLFB (Table 3). There were minimal differences between conditions.

Among those that reported violence, we examined how much ADV, vaginal and/or anal sex, unprotected sex, and delinquency was reported (Table 4). Those in the Date SMART condition reported lower rates of physical or sexual and cyber ADV involvement over follow-up.

Our type 1 hybrid trial of Date SMART⁹  suggest that a skills-based program for adolescent females in the justice system may lead to reductions in physical and sexual dating violence, as well as cyber dating abuse involvement, for those females with histories of ADV exposure. These findings emerged under real world conditions whereby trained family court staff coled many groups, and all adolescent females (ages 14–18) were eligible to participate regardless of their specific legal charge (eg, truant, delinquent, etc.). Through our trial design, we have demonstrated that it is feasible to integrate Date SMART into the structure of an existing family court system and, thus, hope to expedite the transition from intervention testing to standard clinical practice within the JJ setting.

The primary objective of this trial was to reduce adolescent dating violence (ADV) among juvenile-justice involved females over 1 year. First, we tested the primary prevention impact of Date SMART on ADV behaviors (victimization and perpetration) among the full sample. Date SMART was not found superior to psychoeducational programming as a primary prevention program. Rather, both interventions produced reductions in some ADV experiences (emotional or verbal ADV, physical or sexual ADV, and cyber dating abuse). Neither group reduced physical or sexual ADV perpetration rates at 12 months, whereas cumulative counts of physical and/or sexual ADV involvement were reduced. Importantly, both groups offered an opportunity for justice-involved females who share common experiences to discuss relationship related topics led by trained facilitators. Within-group change in ADV may have been fostered by gains in social support, which has been shown to reduce ADV experiences²²  and may be a powerful tool for primary prevention of ADV.

Next, we tested indicated prevention effects of Date SMART on ADV. The presence or absence of ADV was not related to intervention group. However, as expected, we found that ADV-exposed females in Date SMART reported fewer total instances in physical and/or sexual and cyber ADV experiences relative to the control condition. Reductions in the amount of emotional or verbal abuse were observed, although between-group effects were nonsignificant. Thus, Date SMART did not influence the presence of violent relationships but did reduce the amount of ADV taking place in those relationships. These findings are consistent with our previous work showing the benefit of Date SMART for adolescent females with serious ADV exposure.⁹  It also suggests that those with current ADV involvement are applying the intervention skills learned to their current dating relationship.

Our second aim was to examine the impact of Date SMART on unprotected sex and delinquency. No primary prevention effects were observed. Indeed, previous evaluations of sexual risk prevention programs for adolescents in the general population have proven to be less effective at reducing sexual risk in justice settings.²³  With regard to delinquency, many participants had some level of court monitoring during the 12-month follow-up period; thus, a longer follow-up period is needed to understand program impact on delinquency once court involvement ceases. In our indicated sample, we observed reductions in the amount of vaginal and/or anal sex acts reported over the 12-month follow-up period. Reduction in sex acts was not a target of the intervention; however, Date SMART teaches skills related to resisting pressured sexual experiences. The overall reductions in sex acts observed in our indicated sample may reflect reductions in pressured sex, but further work is needed to understand this finding.

Our final goal was to examine system buy-in by monitoring the decision by judges or magistrates to mandate youth to participate our groups. We did observe a large proportion of youth being mandated to treatment. This suggests that court staff perceived potential benefits from our program.

First, data collection relied on participant self-report. Future research that captures partner and naturalistic data of the dyadic experiences is needed. Second, although court staff coled many groups, we do not know whether translation to other justice settings could be hampered by the need for court staff to serve as the sole facilitators of these groups. Finally, given our use of an active control, identifying statistically significant differences between groups was more challenging than if we had implemented a no treatment comparison.

Our findings demonstrate that Date SMART is comparable to psychoeducational programming as a primary prevention tool but is superior as an intervention program to reduce ADV among females with ongoing ADV involvement. These findings emerged in the context of a hybrid trial, where we aimed to replicate real-world conditions for implementation. Thus, Date SMART has the potential to help reduce ADV involvement among justice involved females most at-risk for ongoing dating violence experiences and is ready to be rapidly evaluated and disseminated to other family court systems.

ADV

adolescent dating violence

CI

confidence interval

DS

Date SMART

JJ

juvenile justice

KO

knowledge only

SMART

Skills to Manage Aggression in Relationships for Teens

TLFB

Timeline Followback

TLFB-DV

Timeline Followback-Dating Violence