A School Health Center Intervention for Abusive Adolescent Relationships: A Cluster RCT
BACKGROUND AND OBJECTIVES: Few evidence-based interventions address adolescent relationship abuse in clinical settings. This cluster randomized controlled trial tested the effectiveness of a brief relationship abuse education and counseling intervention in school health centers (SHCs).
METHODS: In 2012–2013, 11 SHCs (10 clusters) were randomized to intervention (SHC providers received training to implement) or standard-of-care control condition. Among 1062 eligible students ages 14 to 19 years at 8 SHCs who continued participation after randomization, 1011 completed computer-assisted surveys before a clinic visit; 939 completed surveys 3 months later (93% retention).
RESULTS: Intervention versus control adjusted mean differences (95% confidence interval) on changes in primary outcomes were not statistically significant: recognition of abuse = 0.10 (−0.02 to 0.22); intentions to intervene = 0.03 (−0.09 to 0.15); and knowledge of resources = 0.18 (−0.06 to 0.42). Intervention participants had improved recognition of sexual coercion compared with controls (adjusted mean difference = 0.10 [0.01 to 0.18]). In exploratory analyses adjusting for intensity of intervention uptake, intervention effects were significant for increased knowledge of relationship abuse resources and self-efficacy to use harm reduction behaviors. Among participants reporting relationship abuse at baseline, intervention participants were less likely to report such abuse at follow-up (mean risk difference = −0.17 [−0.21 to −0.12]). Adolescents in intervention clinics who reported ever being in an unhealthy relationship were more likely to report disclosing this during the SHC visit (adjusted odds ratio = 2.77 [1.29 to 5.95]).
CONCLUSIONS: This is the first evidence of the potential benefit of a SHC intervention to address abusive relationships among adolescents.
- adolescent relationship abuse
- cyber dating abuse
- intervention study
- primary prevention
- school health services
- sexual violence
What’s Known on This Subject:
Adolescent relationship abuse (ARA) is prevalent in confidential clinic settings such as school health centers (SHCs) and is associated with poor health outcomes. No evidence-based interventions target reduction of ARA in the SHC setting.
What This Study Adds:
This study provides the first evidence of the potential benefits of a brief provider-delivered universal education and counseling intervention in SHCs to address and prevent a major public health problem: ARA.
Adolescent relationship abuse (ARA; physical, sexual, or psychological abuse in the context of a past or present romantic relationship) is associated with injuries, poor academic performance,1 unintended pregnancy,2,3 sexually transmitted infections,4 depression,5,6 suicidality,5–7 and substance abuse.3,5 ARA is not uncommon among adolescents seeking care in confidential settings such as family planning and school health centers (SHCs; comprehensive clinics located within high schools) with lifetime prevalence estimates ranging from 40% to 53%.8–11 No evidence-based interventions target ARA in the SHC setting.
SHCs represent a unique setting for addressing ARA because of the prevalence of ARA, ease of access to clinical services,12–14 emphasis on psychosocial assessments during visits, and potential to engage students through peer leaders.15,16 With health reform legislation, SHCs are an authorized federal program essential for health services delivery for adolescents. SHCs offer the opportunity to reach adolescents experiencing ARA (targeted intervention), identify adolescents at risk for ARA (early intervention), and offer universal education about ARA (primary prevention).
Clinic-based ARA assessment can be a step toward helping youth to recognize abusive behaviors.17,18 Discussion of ARA during a clinical encounter may enhance adolescents’ ability to address such abuse. The School Health Center Healthy Adolescent Relationships Program (SHARP) is a provider-delivered intervention implemented within routine SHC visits. The intervention is universal, inclusive of all genders, sexual orientation, and clinic visit types, addressing a range of abusive behaviors, including cyber dating abuse (the use of social media to abuse a partner).19,20 This study uses a cluster randomized controlled trial (RCT) design to evaluate the effectiveness of this brief universal education and counseling intervention to improve adolescents’ recognition of ARA, use of resources, and attitudes and behaviors to reduce ARA.
The study’s 3 prespecified primary outcomes for the full sample focused on ARA knowledge and attitudes: (1) recognition of abusive behaviors, (2) intentions to intervene, and (3) knowledge of ARA resources. Self-efficacy to use harm reduction behaviors was a secondary outcome. For the subset of patients reporting recent ARA at baseline, additional outcomes were as follows: (1) disclosure of ARA during the clinic visit, (2) recent ARA at follow-up, and (3) use of harm reduction behaviors and use of ARA resources.
Description of Intervention and Control Conditions
A team of practitioners, violence prevention advocates, and researchers developed the SHARP clinical guidelines,21 training slides, and a palm-size brochure that discusses healthy relationships, how to help a friend, and ARA resources. Clinicians and staff at intervention SHCs received a 3-hour training on the SHARP intervention about ARA impact on health and how to introduce the brochure, conduct ARA assessment, and make a warm referral to a victim service advocate (connecting a patient to an advocate via telephone or in person). Provider discussion of healthy and unhealthy relationships is integrated into each clinical encounter with the provision of the palm-size brochure to every patient regardless of reason for visit. Even in the absence of disclosure, patients are encouraged to take extra brochures for friends. SHC providers reported the time required to review the brochure with a student was typically less than a minute but could lead to longer discussions when ARA was disclosed. In addition to the provider-delivered intervention, each of the intervention SHCs involved their youth advisory boards to organize school-wide outreach events to provide ARA information and encourage students to come to the SHC.
Patients attending control SHCs received usual care (ie, standard social history assessments without specific prompts to assess for ARA). In the event of a disclosure, clinicians and staff followed clinic protocol, including referring to advocacy services. The control SHC youth advisory boards conducted other health- and wellness-related outreach events at their schools.
Sample and Setting
Eleven SHCs in Northern California already part of the California School-Based Health Alliance, in proximity to the California Adolescent Health Collaborative in Oakland, California, were recruited to participate in this study. Several of the SHC schools were participating in the federal Safe and Supportive Schools (S3) project, which involves school climate interventions to reduce bullying and violence, which was accounted for in the randomization via stratification. Two of the clinics shared providers, thus were treated as a single cluster. Cluster randomization was selected as the intervention was available to all patients served by a clinic. Ten clinic clusters were evenly randomized into intervention and control arms by using computer-generated randomization.22 After randomization and before participant enrollment, 3 schools withdrew when new school administrators determined they did not want the SHC participating in research, leaving 8 SHCs (7 clusters: 4 intervention, 3 control) as the final sample.
This longitudinal study was conducted via baseline and follow-up survey 3 months post intervention. Participants were recruited from September to December 2012; all follow-up surveys were completed by June 2013. All English- and Spanish-speaking patients ages 14 to 19 years seeking care for any reason in participating SHCs and who expected to be available for follow-up were eligible. Upon check-in for a clinic visit, patients were told about the study, and trained research staff at each SHC screened interested patients for eligibility. Eligible patients were escorted to a private area in clinic for consent and survey administration. The study received a waiver of parental permission.
Survey data were collected via computer with questions read through headphones (Audio Computer-Assisted Self-Interview software). Each participant received a $10 gift card upon baseline survey completion and $15 after follow-up. The survey was available in English or Spanish; all participants opted for English. All study procedures were approved by University of Pittsburgh and Public Health Institute’s institutional review boards and reviewed by administrators at respective schools and SHCs. Data were protected with a National Institute of Justice Privacy Certificate.
Clinic staff from all 8 participating SHCs referred patients (n = 1062) to research staff: 1012 consented (95% participation), 1011 completed the baseline survey, and 939 completed the follow-up survey (93% retention; Fig 1). Baseline surveys occurred at the time of clinic visit (and during the fall, before implementation of student outreach activities). Follow-up surveys were conducted at the SHCs at a time convenient for participants; 4% of participants were unable to return to clinic because of moving and completed the survey via e-mail.
Single items assessed sex, race/ethnicity, grade level, and nativity. Each school’s proportion of free lunch was used as proxy for socioeconomic status. SHCs’ overall demographic characteristics were obtained from already available clinic data.
The recognition of ARA scale23 assessed perceptions of abusiveness of relationship behaviors by using a 5-point Likert response from “not abusive” to “extremely abusive” (Cronbach’s α = 0.85), modeled as a mean of responses to 10 items. An additional 8 behaviors specific to reproductive and sexual coercion were developed and piloted with the same Likert response as above to assess recognition of sexual coercion, modeled as a mean of 8 items (Cronbach’s α = 0.86).
To assess intentions to intervene, participants were presented with abusive behaviors and asked how likely they would be to stop the behavior if they witnessed a peer doing this,24,25 with 5 responses ranging from “very unlikely” to “very likely” (Cronbach’s α = 0.89), modeled as a mean of 9 items.
Knowledge and recent use of ARA resources items measured whether the participant knew about ARA resources (ie, Teen Dating Abuse Helpline, SHC clinical/counseling staff, other hotlines) and whether they had used any of these in the past 3 months, each modeled as a summary score.
The SHARP intervention included discussion of harm reduction to reduce reproductive coercion risk for female participants. Self-efficacy to use harm reduction behaviors was measured with 4 harm reduction behaviors by using a 5-point Likert response from “strongly disagree” to “strongly agree” (Cronbach’s α = 0.77). These items were asked of female participants only, and to reduce measurement effect, introduced only at follow-up.
General self-efficacy was measured at baseline by using an established scale, with responses ranging from “not true at all” to “exactly true,” scored by using the item means26 (Cronbach’s α = 0.89).
Intimate relationships were defined as “someone you were going out with or hooking up with.” Recent (past 3 months) physical and sexual violence were assessed by using 3 items modified from the Conflict Tactics Scales-227 and the Sexual Experiences Survey,28 1 for physical violence and 2 for sexual violence (with and without the use of force or threats). Any affirmative response was coded as recent physical or sexual violence (Cronbach’s α = 0.49).
Recent ARA was a combined variable of any physical or sexual violence or cyber dating abuse in the past 3 months (Cronbach’s α = 0.72).
The ARA disclosure outcome was assessed in an exit survey immediately after the SHC visit only in patients who answered affirmatively to a broader branching question “Have you ever experienced an unhealthy relationship or been hurt by a sexual partner?” resulting in a different sample than those who reported recent ARA in the baseline survey.
The study was designed to provide 80% power to detect standardized effects of 0.3, conservatively assuming that outcome-specific variance inflation factors from the cluster RCT design could reduce the planned actual enrollment of 1200 students from 10 clinics into effective sample sizes as low as 378.31,32
To account for the hierarchical arrangement of the data (up to 2 measurements per student nested within SHCs) and the cluster randomization, survey data analysis methods33 and multilevel mixed-effects models34 were employed for statistical inferences. Statistical analyses were conducted by using SAS version 9.3 (SAS Institute, Inc, Cary, NC).35
Except for the self-efficacy outcome, between-arm adjusted mean differences (AMDs) in over-time changes of continuous outcomes were used to estimate intervention effects (ie, the change in each outcome score attributed to the intervention). Race/ethnicity, sex, grade, US nativity, and school-level socioeconomic status and S3 program participation were included as covariates. All available data were used, with students analyzed according to SHC treatment assignment. Sample size varied slightly across outcomes based on small amounts of missing data (missingness across outcomes ranged from <1% to 4%). Because only a general self-efficacy score was assessed at baseline to reduce potential measurement effect related to querying about specific harm reduction behaviors, intervention effects on self-efficacy to use harm reduction behaviors were assessed by estimating AMD by using a mixed-effects model that included the baseline general self-efficacy score as an additional covariate.
For victimization outcomes, multinomial logistic regression models for clustered data were used to estimate mean risk difference (MRD) in victimization at follow-up within subgroups (based on recent victimization status at baseline).36 Only participants with baseline and follow-up data available for these outcomes were included (81% of the baseline sample); this approach was chosen to elucidate the pattern of abuse over time in each arm. For the ARA disclosure outcome, the clustered data multiple logistic regression model used a reduced set of covariates (S3 program participation, age and school-level socioeconomic status along with the binary intervention group indicator) because of small cell sizes.
Posthoc Intervention Intensity-Adjusted Analyses
The SHARP intervention involved providers discussing healthy relationships with patients and providing a brochure about ARA. Exit surveys completed immediately after the visit (by 95% of participants) demonstrated that in the intervention clinics, 69% (range, 63%–79%) reported having a discussion with a provider about relationships and 55% (range, 38%–74%) reported receiving the brochure. In posthoc analyses, intervention participants were assigned a “score” for intervention intensity, with 0.5 assigned if the participant reported no discussion with provider and no receipt of brochure, 0.75 for either provider discussion or brochure, and 1.0 if the participant received both. Intensity-adjusted intervention effects were estimated by substituting the intervention intensity score in place of the binary intervention indicator used in primary analyses.
Demographic Characteristics and Attrition Analyses
Seventy-six percent of the entire sample was female (Table 1). Almost all participants identified as nonwhite (5% white), with no significant differences between intervention and control clinics by race/ethnicity. Schools varied in students eligible for free lunch (37%–79%). Compared with each SHC’s demographics, participants were more likely to be female, in higher grades, and self-report as white, African-American, and multiracial/other.
Participants who did not complete the follow-up survey tended to be younger compared with those who completed (P = .09). Noncompleters were more likely to report recent ARA at baseline (63% vs 51%; P = .02). Attrition did not differ significantly between intervention (10%) and control (5%; P = .12).
Differences in Outcomes of Interest at Baseline
Control participants had lower baseline scores on recognition of abusive behaviors and were more likely to report recent physical or sexual abuse at baseline (16% vs 10%, P = .01) compared with intervention participants. Both arms were similar at baseline on recognition of sexual coercion, and knowledge and use of ARA-related resources (Table 2).
Female participants were more likely to report any recent ARA victimization at baseline (48% vs 35%, P < .01), as well as recent cyber dating abuse (45% vs 31%, P = .01) and physical or sexual violence victimization (14% vs 10%, P = .07) than male participants.
Overall Intervention Effects
No differences between intervention and control emerged in intentions to intervene, knowledge of and recent use of ARA-related resources, or self-efficacy to use harm reduction strategies (Table 2). Compared with controls, at follow-up, intervention participants demonstrated greater increases in recognition of sexual coercion (AMD = 0.10 [95% confidence interval (CI): 0.01 to 0.18]).
Intervention Effects by Baseline ARA
Among those reporting recent ARA at baseline, intervention participants demonstrated an increase in recognition of ARA (AMD = 0.14 [0.01 to 0.27]) and knowledge of ARA resources (AMD = 0.26 [0.09 to 0.43]) compared with controls (Table 3). Analyses of intervention effects by sex revealed no significant differences. Among those reporting any lifetime experience of an unhealthy relationship or being hurt by a sexual partner, 36% of intervention versus 22% of control patients reported disclosing to the provider (adjusted odds ratio [AOR] = 2.77 [1.29 to 5.95]).
Fewer intervention participants experiencing ARA at baseline reported ARA at follow-up compared with controls (65% vs 80%; MRD = −0.17 [−0.21 to −0.12]), including cyber dating abuse (62% vs 76%; MRD = −0.15 [−0.22 to −0.09]) and physical or sexual abuse (16% vs 24%, MRD = −0.07 [−0.12 to −0.01]; Table 4). Among participants not experiencing ARA at baseline, the intervention was associated with less likelihood of recent physical or sexual abuse at follow-up (7.3% vs 7.4%; MRD = −0.02 [−0.04 to −0.001]).
Posthoc Intervention Intensity-Adjusted Analyses
Intensity-adjusted intervention effects were associated with increased knowledge (AMD = 0.25 [0.11 to 0.39]) of ARA resources and increased self-efficacy to use harm reduction strategies (AMD = 0.33 [0.06 to 0.60]) among intervention participants compared with controls (Table 2). For ARA disclosure during the clinic visit, the intervention-intensity adjusted odds ratio for the intervention was 9.30 (2.44 to 35.51; results not shown).
Findings suggest the potential utility of a brief SHC provider-delivered intervention, which discusses healthy relationships, integrating education, and connection to resources as part of routine care. Changes in prespecified outcomes of ARA knowledge and attitudes for the entire sample were not significant. Exposure to the SHARP intervention was associated with improvements in recognition of sexual coercion, and among youth recently experiencing ARA, improvements in recognition of ARA and knowledge of ARA resources. Disclosure to SHC providers about unhealthy relationships was greater among participants in the intervention clinics. Although the intervention did not have significant effects on use of harm reduction strategies, relative reductions in overall ARA, as well as cyber dating abuse and physical/sexual violence victimization are promising.
The SHCs in this study serve mostly female patients, who reported greater ARA victimization than their male counterparts, consistent with studies that include sexual violence victimization in the measurement of ARA.37–39 Intervention effects by sex did not differ significantly. Continued attention to similarities and differences in ARA experiences by sex is needed to enhance ARA prevention.
Brief interventions such as SHARP embedded in clinical settings are a promising strategy for prevention and intervention, yet uptake by providers remains an anticipated challenge. Even with focused provider training, only two-thirds of patients in intervention sites reported having a conversation with a provider about relationships and even fewer reported receiving the brochure. Posthoc (as-treated) analyses suggest that intervention receipt was associated with increases in knowledge and use of ARA resources, self-efficacy to use harm reduction strategies, and disclosures about unhealthy relationships to SHC staff. Given multiple pressures on providers and time limitations, scaling up provider-based interventions such as SHARP will require attention to system-level changes.40 Practice-based supports such as electronic health record prompts and involvement of nonclinical staff may facilitate intervention implementation. Additionally, testing the effectiveness of text messaging, computerized interventions, and related strategies is needed to enhance clinic-based assessment and counseling for ARA.
The SHARP intervention also introduced strategies youth could use to help a friend in an unhealthy relationship. Each of the intervention sites had student leaders presenting information during lunches and assemblies to encourage students to come to the SHC to learn about ARA. Although youth did report increases in knowledge and use of resources in the intervention arm, no changes in intentions to intervene were observed. Given the effectiveness of bystander approaches in preventing interpersonal violence,41 further studies are needed to identify how to encourage positive helping behaviors among high school students.
Findings should be interpreted in light of several limitations. As a cluster RCT with a small number of clusters, the study may have been underpowered to assess outcomes. Interval to follow-up was short, precluding examining longer-term clinical outcomes. Differences in demographics across clinics at baseline, while adjusted for in analyses, may reflect other unmeasured clinic differences left unaccounted. Sites were not well balanced at baseline, with a higher prevalence of ARA in control sites. Intervention implementation varied across sites, and the degree to which participants were exposed to any of the school-wide outreach events is not known. Findings from 8 SHCs in 1 region also cannot be generalized to all US high schools. A larger cluster RCT with a greater number of clusters, more geographically diverse clinics, and longer-term follow-up with assessment of health outcomes is needed.
These limitations notwithstanding, this trial provides the first evidence of the potential benefits of a brief universal education and counseling intervention in SHCs to address and prevent a major public health problem: ARA.
We thank the staff of the SHCs in Balboa, Edison, Logan, Oakland, Oakland Technical, San Lorenzo, Skyline, and Tennyson high schools for their invaluable support with this study. Monica Beas, Katarina Blagojevic, Alex Coleman, Kyla Lew, Deanna Staggs, Cassie Stiftl, Alexandra Taylor, and Sami Newlan provided invaluable research assistance. Alison Chopel, Public Health Institute, provided critical review of the manuscript.
- Accepted October 10, 2014.
- Address correspondence to Elizabeth Miller, MD, PhD, Chief, Division of Adolescent and Young Adult Medicine, Children’s Hospital of Pittsburgh, University of Pittsburgh Medical Center, 3420 Fifth Ave, Pittsburgh, PA 15213. E-mail:
Dr Miller conceptualized and designed the study, designed the data collection instruments, supervised data analyses, interpreted the results, and drafted the initial manuscript; Ms Goldstein assisted in designing the data collection instruments, coordinated and supervised data collection, and critically reviewed the manuscript; Dr McCauley assisted in drafting the manuscript, conducting the analyses, and interpreting the results; Ms Jones conducted the data analysis, assisted in interpreting the results and drafting the manuscript, and reviewed and revised the manuscript; Ms Dick assisted in designing the data collection instruments, supervised data management, and assisted in drafting the manuscript; Ms Jetton coordinated data collection and critically reviewed the manuscript; Dr Silverman assisted in designing the data collection instruments and reviewed and revised the manuscript; Ms Blackburn recruited school health centers into the study and critically reviewed the manuscript; Ms Monasterio trained providers on the intervention and provided technical assistance to intervention school health centers, and she critically reviewed the manuscript; Ms James assisted in design of the research study as well as facilitating recruitment and retention of the school health centers in the study, and she critically reviewed the manuscript; Dr Tancredi assisted in the design of the research study and data collection instruments, and Dr Tancredi supervised the data analyses and critically reviewed the manuscript; and all authors approved the final manuscript as submitted.
The opinions, findings, and conclusions or recommendations expressed in this publication/program/exhibition are those of the author(s) and do not necessarily reflect those of the Department of Justice.
Ms Goldstein is now retired.
FINANCIAL DISCLOSURE: Ms Monasterio received consulting fees and travel support from Futures Without Violence; the other authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: This project was supported by award number 2011-MU-MU-0023, awarded by the National Institute of Justice, Office of Justice Programs, US Department of Justice.
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
- Foshee VA,
- Reyes HL,
- Gottfredson NC,
- Chang LY,
- Ennett ST
- Exner-Cortens D,
- Eckenrode J,
- Rothman E
- Keeling J,
- Birch L
- ↵Dick RN, McCauley HL, Jones KA, et al. Cyber dating abuse among teens using school-based health centers [published online ahead of print November 17, 2014]. Pediatrics. doi: 10.1542/peds.2014-0537
- ↵Miller E, Levenson R. Hanging Out or Hooking Up: Clinical Guidelines on Responding to Adolescent Relationship Abuse—An Integrated Approach to Prevention and Intervention. 2012; Development of these clinical guidelines funded by the Department of Justice Office of Juvenile Justice and Delinquency Prevention, the Administration of Children and Families, and the Office on Women’s Health. Available at: www.futureswithoutviolence.org/userfiles/file/HealthCare/Hanging-Out-or-Hooking-Up_Low_Res_Cropped_FINAL.pdf. Accessed December 13, 2012
- ↵Stata Statistical Software: Release 12 [computer program]. College Station, TX: StataCorp LP; 2011
- Rothman E,
- Decker M,
- Silverman JG
- ↵Schwarzer R, Jerusalem M. Generalized Self-Efficacy scale. In J. Weinman, S. Wright, & M. Johnston, Measures in health psychology: A user's portfolio. Causal and control beliefs. Windsor, United Kingdom: NFER-NELSON; 1995: 35–37
- Straus MA,
- Hamby SL,
- Boney-McCoy S,
- Sugarman DB
- Hsieh FY,
- Lavori PW,
- Cohen HJ,
- Feussner JR
- ↵SAS version 9.3 [computer program]. SAS System for Windows. Cary, NC: SAS Institute, Inc; 2002–2010
- Spiegelman D,
- Hertzmark E
- Mulford C, Giordano PC. Teen Dating Violence: A Closer Look at Adolescent Romantic Relationships. Washington, DC: NIJ Journal; 2008:261
- Copyright © 2015 by the American Academy of Pediatrics