Cluster Randomized Trial of a College Health Center Sexual Violence Intervention

Abstract

Introduction

Sexual violence, particularly in the context of drinking, is prevalent on college campuses. This study tested a brief intervention to prevent sexual violence among students receiving care from college health centers.

Study design

Two-arm, unblinded cluster RCT.

Setting/participants

On 28 campuses with health/counseling centers (1:1 randomization allocation; 12 intervention and 16 control), from September 2015 to March 2018, a total of 2,291 students seeking care at college health centers completed surveys prior to the appointment, immediately after, 4 months later, and 12 months later.

Intervention

Intervention college health center staff received training on delivering sexual violence education to all students seeking care. Control sites provided information about drinking responsibly.

Main outcome measures

The primary outcome was students’ change in recognition of sexual violence. Additional outcomes included sexual violence disclosure and use of services among students with history of sexual violence at baseline. Generalized linear mixed models accounting for campus-level clustering assessed intervention effects. Data were analyzed from September 2018 to June 2019.

Results

Half (55%) of students seeking care at college health centers reported any history of sexual violence exposure. No between-group differences in primary (β=0.001, 95% CI= –0.04, 0.04) or secondary outcomes emerged between intervention and control students. Post-hoc analyses adjusting for intensity of intervention delivery (intensity adjusted) revealed an increase in self-efficacy to use harm reduction strategies (β=0.09, 95% CI=0.01, 0.18) among intervention participants. Among those who reported sexual violence at baseline, intervention students had increased odds of disclosing violence during the visit (AOR=4.47, 95% CI=2.25, 8.89) in intensity-adjusted analyses, compared with control. No between-group differences emerged for remaining outcomes.

Conclusions

Sexual violence exposure is high among students seeking care in college health centers. A brief provider-delivered sexual violence intervention, when implemented with fidelity, was associated with improved self-efficacy to use harm reduction and increased disclosure of sexual violence during clinical encounters, but no increased use of services. More interventions that are stronger in intensity are needed to connect students to sexual violence services.

Trial registration

This study is registered at www.clinicaltrials.gov NCT02355470.

INTRODUCTION

Sexual violence (SV; including sexual coercion, non-consensual sexual contact, and rape) on college campuses is prevalent: A quarter of women and one in 20 men report experiencing sexual contact involving force or incapacitation during college.^1,2 Sexual assaults often happen in the context of alcohol intoxication by the victim, perpetrator, or both, occurring in more than half of sexual assaults on college campuses.^3–5 Negative consequences of SV include repeat violence victimization, mental health symptoms, hazardous drinking, and sexual risk behaviors.^6–8

Despite the large number of students affected by SV, the majority do not report such experiences to university, health care, or law enforcement.^1,9,10 Under-reporting may be related to stigma, self-blame, fear of consequences from illegal alcohol consumption, and belief that nothing will be done.^11,12 In addition, although community-based victim services have been demonstrated to be helpful for SV survivors,^13,14 low levels of awareness and use of services for SV highlight the need to help students who have experienced SV connect with campus and community-based resources.^15,16

College health and counseling centers (CHCs) are ideal sites for introducing brief interventions and facilitating connection to services as they are confidential and provide care to a substantial proportion of students (43%–48%).¹⁷ As SV contributes to poor health, students seeking care for medical and behavioral health concerns in CHCs are more likely to have experienced SV compared with general population-based samples, similar to studies in other healthcare settings.^8,18–21 To date, the limited research on CHC-based brief interventions has focused on high-risk drinking,^22,23 without addressing intersections with SV.

The purpose of this study was to test the effectiveness of a brief, universal education, and harm reduction counseling intervention, titled “Giving Information for Trauma Support and Safety” (GIFTSS), on increasing awareness of SV (especially alcohol-related SV) and connection to SV services among college students. This is the first study in CHCs that integrates prevention (using a bystander approach) and intervention (counseling and connection to victim services as needed) to address SV. As the intervention was designed for integration into CHCs, which precluded randomizing at patient or provider level, the study used a cluster design.

The primary hypothesis was that, compared with controls, male and female students seeking care at CHCs implementing GIFTSS would have greater recognition of what constitutes alcoholrelated SV (primary outcome), knowledge of and self-efficacy to enact harm reduction strategies (behaviors that may help increase safety such as a buddy system or using an app to discretely connect to friends when feeling unsafe), intentions to intervene with peers engaging in behaviors that contribute to SV, and knowledge of and self-efficacy to use SV-related services, in the short term (4 months) and longer term (12 months). Additionally, among students who witness peer SV during the study period, those who received GIFTSS would be more likely to report enacting behaviors to interrupt peers’ harmful behaviors than students not receiving GIFTSS. Finally, as an exploratory hypothesis, students with SV victimization history who received GIFTSS would be more likely to disclose SV during their clinic visit, report greater use of SV-related services, and through connection to supports, would be less likely to be exposed to SV at follow-up compared with students on control campuses.

METHODS

Study Population

Details of this trial have been described elsewhere.²⁴ Briefly, CHCs were approached to participate in an unblinded, cluster RCT to test a SV brief intervention. Those agreeing to participate were randomized to intervention or control after stratification based on size (large; i.e., ≥3,000 students served annually versus smaller). The study statistician created a computergenerated randomly permuted block allocation using Stata, version 12. Campuses with both a student health center and a counseling center were randomized together as a single CHC; colleges separated across different campuses were randomized as separate CHCs. The statistician did not reveal the allocation to the Principal Investigator or research team until after each individual CHC enrolled.

Study participants (N=2,291) in both conditions completed online surveys at baseline (prior to clinical visit and intervention) and after the clinical visit (exit survey completed within 2 weeks after the CHC visit). Follow-up surveys were collected at 4 months (T2) and 12 months (T3), primarily via e-mailing or texting participants the online survey link. All research procedures were approved by the University of Pittsburgh Human Subjects Research Protection Office as well as each college IRB.

To be eligible to participate, campuses had to have a health or counseling center for students staffed ≥5 days a week with health services. Twenty-eight Pennsylvania and West Virginia college student health and counseling centers participated, all located on co-ed campuses. No sites had existing protocols for educating patients about alcohol-related SV. Participant-level eligibility for this study included all students who met the following criteria: age 18–24 years, literate in English, seeking care at CHC for any reason, and having time prior to the clinic visit to complete a 20-minute survey.

Student recruitment followed randomization and training of the CHCs. Age-eligible students presenting to the CHC were identified by clinic staff at registration and informed about the study via flyers. Research assistants assessed interested students for eligibility; a waiver of signed consent was approved by the IRB that covered 27 colleges, with one college requiring written consent. At some sites (n=22), flyers included a number to text that sent students study information, an eligibility screen, and baseline survey. Students with scheduled appointments were informed of the study by CHC staff providing study team contact information to an interested student or giving interested students’ contact information (with their permission) to the study team, or students texting a number that sent a link to the study.

Students were enrolled in the study from September 2015 to March 2017. Follow-up data collection concluded in March 2018.

Intervention

The brief intervention, designed with stakeholder feedback (i.e., students, practitioners, and advocates), used a palm-sized educational card with information about SV and harm reduction given to all patients during clinical encounters.^a GIFTSS provides: (1) education regarding SV; (2) discussion of harm reduction behaviors to reduce risk of SV for self and peers (including bystander intervention); and (3) supported referrals to SV services, including how to connect to confidential services. The control intervention used National Institute of Alcohol Abuse and Alcoholism tools for brief alcohol risk reduction counseling (www.collegedrinkingprevention.gov), with a palm-sized educational card that offered information regarding drinking responsibly.

A local SV victim service agency, together with the Principal Investigator, led a 3-hour GIFTSS training at each site involving all CHC staff. Trainers also connected with staff by phone, in person, or video conference to review materials and ensure staff (including new staff) were comfortable with intervention components.

Control sites received a 3-hour training led by an expert on screening and brief intervention to assess for alcohol misuse as well as a short introduction to motivational interviewing. Providers then delivered this intervention to individual patients.

Owing to the design of this intervention and training, this was an unblinded trial, with the Principal Investigator, study team, CHC providers, and patients aware of the allocation.

Measures

Measures, described in detail elsewhere,²⁴ were assessed at the individual level.

Twelve items²⁵ assessed participants’ recognition of what constitutes SV and attitudes related to the role of alcohol in risk for SV on a 5-point scale, with higher mean score indicating greater recognition (primary outcome; baseline Cronbach α=0.74).

Respondents reported on a 5-point scale how confident they were to reduce risk for SV using seven harm reduction strategies (e.g., I am confident that I could download the Circle of 6 safety app to keep me or my friends safe), with an emphasis on strategies involving using the CHC for assistance. Women were asked two additional items specific to pregnancy risk reduction. A higher mean score corresponded to greater self-efficacy to use harm reduction strategies (not measured at baseline; Cronbach α=0.89 at T2 and 0.85 at T3).

Participants reported if they knew about five SV resources (e.g., National Sexual Assault Hotline); women also reported if they knew about emergency contraception for an unwanted sexual experience. Each resource endorsed was counted as 1 point for a summary score of knowledge of SV-related services.^20,26–28

Using the items presented for knowledge of SV services, participants reported likelihood that they would use those services using a 5-point scale. A higher mean score corresponded to greater self-efficacy to use those resources (baseline Cronbach α=0.89).^20,26–28

For each of ten abusive behaviors they might witness peers perpetrating, respondents rated their likelihood to intervene on a 5-point scale; higher mean score corresponded to greater likelihood to intervene (baseline Cronbach α=0.92).^24,29

Using same items from the intentions to intervene scale, participants reported whether they had actually witnessed each scenario in the past 4 months. For each endorsed scenario, they were asked how they responded, with positive (saying something) and negative (doing nothing or going along with it) options. Selecting at least one positive behavior per scenario was added to create a summary score for positive bystander behavior (range, 0–10).^24,29

Respondents’ recognition of sexual and reproductive coercion was assessed by rating eight situations (e.g., “pressuring someone to have sex when they’ve said no”) using a 5-point scale. A higher mean score across eight items indicated greater recognition (baseline Cronbach α=0.87).²⁷

Respondents’ self-efficacy to obtain consent for sex was measured using the perceived behavioral control subscale of the Sexual Consent Scale–Revised.³⁰ Respondents rated each item on a 5-point scale; after reverse coding negative items, a higher mean score indicated greater self-efficacy (baseline Cronbach α=0.83).

Six items³¹ assessed frequency of unwanted sexual experiences (when you indicated you didn’t want to): fondling, kissing, or touching; attempted sex; vaginal, oral, or anal sex; or penetration with finger or object. An additional item assessed incapacitated multi-person sexual experiences. Frequencies of each experience (zero, one, two, three, four or more) were dichotomized and added to create a summary score for SV victimization.

In addition to SV victimization (which could include SV from an intimate partner), two items assessed for physical violence from a dating partner and stalking.³² Yes to either of these items or any SV victimization was used for analyses restricted to respondents with histories of SV/intimate partner violence (IPV) at baseline.

Participants who endorsed any history of SV, IPV, or stalking were asked immediately post-visit and at T3 whether they had told their health provider about such experiences.

Respondents reported whether they used SV-related resources (e.g., Called the National Sexual Assault Hotline or visited their website). Women also reported whether they had taken EC after an unwanted sexual experience. Each resource used was added to calculate a summary score for use of SV-related services (possible ranges: women, 0–6; men, 0–5).

Sample size calculations³³ incorporated within-subject correlation and are described in detail elsewhere.²⁴ Using a range of within-subject intraclass correlation coefficients, assuming 80% retention, the enrollment of 2,291 students across 28 schools yielded 80% power to detect standardized effect sizes of 0.14 (intraclass correlation coefficient=0.60) to 0.185 (intraclass correlation coefficient=0.30) for the primary outcome.

Statistical Analysis

Data were analyzed from September 2018 to June 2019. Descriptive statistics were used to characterize study participants; comparisons by treatment arm were conducted using Wald loglinear chi-square tests. Outcome values were calculated for each survey timepoint as mean (SE) or proportion (n). Attrition analyses assessed differences between those who completed followup surveys to those who did not.

Primary assessments of intervention effects were based on intent-to-treat estimates, with astreated effects estimated in secondary analyses. Generalized linear mixed models with random effects for within school and within participant were used, specifying outcome distributions as binary (for use of any SV services, disclosure), Poisson (for knowledge of SV-related services, positive bystander behaviors), negative binomial (for recent SV), or continuous (all remaining outcomes). The negative binomial model was selected for the recent SV variable owing to overdispersion that the Poisson model suggested. The negative binomial model improved the Pearson chi-square/df values from 1.80 to 1.09 at T2 and 2.15 to 1.11 at T3. The baseline value for each outcome was included in corresponding models; CHC-level covariate (large or small clinic) was included in all models to account for stratified randomization design. The authors specified a priori to adjust for between-arm imbalances, resulting in inclusion of age and race/ethnicity in the primary models.

Although most outcomes were assessed among the entire sample, the following were analyzed among specific groups: positive bystander behaviors, among those who had recently witnessed abusive behavior; use of any SV-related services among those who reported SV/IPV at baseline; and SV/IPV disclosure among those who reported a history of SV/IPV. Additionally, measures that included items relevant to women only (i.e., emergency contraception and pregnancyrelated) were modeled twice: The first model used the entire sample with items asked of all participants, and the second model was restricted to women only and included these additional items.

An intensity-adjusted analysis to reflect actual receipt of the intervention used a continuous score instead of the original binary intervention indicator (0 or 1). This score was based on post-visit survey questions, which asked whether they received the educational card, whether the provider reviewed this with them, and whether they discussed topics such as sexual assault and healthy relationships during the visit. For intervention arm participants, receiving the card and either reviewing the card or having at least one discussion with their provider was scored as 1 (“full intervention intensity”). Receiving the card or having at least one discussion point with their provider was scored as 0.5, to indicate having received approximately half the intervention. Participants who did not receive a card or any discussion with their providers were scored as 0. Intervention participants who did not complete the exit survey were excluded from this analysis. Control arm participants were assigned a score of 0, regardless of exit survey. In addition to a priori covariates, intensity-adjusted models included housing location and lifetime experience of SV as these differed by intensity levels. The authors also conducted a per-protocol analysis in which only intervention arm participants who received the full intervention (corresponding to an intensity score of 1) were compared to all control arm participants. SAS, version 9.4 was used for statistical analyses.

RESULTS

Of 36 CHCs approached for participation, 29 were eligible and agreed to participate (Figure 1). One CHC assigned to intervention withdrew following randomization but prior to enrolling students, resulting in 28 CHCs (12 intervention, 16 control). In the intervention arm, 1,362 students met eligibility criteria. Of those, 1,154 provided informed consent (85% of eligible) and 1,040 completed baseline surveys. In the control arm, 1,425 students were eligible. Of those, 1,330 provided informed consent (93%), and 1,251 completed baseline surveys. Though almost all colleges (96%) offered both health and counseling services to their students, half (50%) had only health services participate; the remainder were counseling centers only (11%) or both health and counseling centers (39%), which did not vary by treatment arm.

Figure 1.

CONSORT diagram.

Note: One study participant was found to have completed the baseline visit twice in error. As a result, the second set of measures was removed from the study dataset.

LTF, lost to follow up.

More Asian (10%) and black/African American (13%) and fewer white (58%) participants were in the intervention arm compared with control (1%, 7%, and 74%, respectively). Most identified as female (72%), with no difference by arm. The majority were aged 18–21 years; more control participants were younger with fewer years in college (Table 1).

Table 1.

Sample Characteristics

Characteristics	Overall (n=2,291)% (n)	Treatment arm		χ2p-valuea
		Intervention (n=1,040)% (n)	Control (n=1,251)% (n)
Race/ethnicity				0.0006
White	67.6 (1,549)	58.8 (611)	75.0 (938)
Hispanic or Latino	13.7 (313)	14.4 (150)	13.0 (163)
Black or African American	9.5 (218)	12.8 (133)	6.8 (85)
Asian	5.4 (123)	10.2 (106)	1.4 (17)
Multiracial or other	3.6 (83)	3.7 (38)	3.6 (45)
Gender				0.4393
Female-identified	72.3 (1,657)	70.9 (737)	73.5 (920)
Male-identified	26.9 (616)	28.3 (294)	25.7 (322)
Other gender	0.7 (15)	0.9 (9)	0.5 (6)
Age, years				0.0334
18–19	40.8 (935)	41.4 (430)	40.4 (505)
20–21	40.6 (931)	36.9 (384)	43.7 (547)
22–24	17.1 (392)	20.4 (212)	14.4 (180)
Year in school				0.0041
1st year undergraduate	26.9 (616)	28.0 (291)	26.0 (325)
2nd year undergraduate	25.1 (575)	22.7 (236)	27.1 (339)
3rd year undergraduate	20.7 (475)	19.7 (205)	21.6 (270)
4th or 5th year undergraduate	19.6 (449)	18.5 (192)	20.5 (257)
Graduate/professional school or other	7.0 (160)	11.1 (115)	3.6 (45)
Enrollment status				0.6632
Full-time	97.1 (2,224)	97.4 (1,013)	96.8 (1,211)
Part-time or other	2.1 (49)	2.3 (24)	2.0 (25)
Housing location				0.7871
Campus residence hall	50.7 (1,162)	49.1 (511)	52.0 (651)
Other off-campus housing	34.7 (796)	34.5 (359)	34.9 (437)
Other college housing	7.9 (180)	10.0 (104)	6.1 (76)
Parent/guardian’s home	4.1 (94)	3.9 (41)	4.2 (53)
Fraternity or sorority house	1.8 (42)	2.2 (23)	1.5 (19)
Lifetime history of sexual violence	55.0 (1,259)	53.7 (558)	56.0 (701)	0.2877
Fondled, kissed or sexually touchedb	45.0 (1,030)	43.8 (455)	46.0 (575)	0.2553
Attempted sex (but it did not happen)b	36.7 (841)	36.0 (374)	37.3 (467)	0.3070
Forced vaginal sexb	15.0 (343)	14.1 (147)	15.7 (196)	0.2442
Forced oral sexb	15.3 (351)	15.3 (159)	15.4 (192)	0.8615
Forced anal sexb	4.3 (99)	3.7 (38)	4.9 (61)	0.1492
Penetrated with a finger or objects (vaginally, orally, or anally)b	13.8 (317)	14.0 (146)	13.7 (171)	0.9357
Multi-person incapacitated sex	2.1 (47)	1.4 (14)	2.6 (33)	0.0603

Notes: Boldface indicates statistical significance (p<0.05).

^aP-values calculated using Wald log-linear χ² testing to compare treatment arms, accounting for college-level clustering.

^bItems were asked in the context of “when you indicated you did not want to”.

Retention was 76.7% (n=1,757) at T2 and 75.5% (n=1,729) at T3. Attrition analyses indicated that compared with participants who did not complete each timepoint, participants who completed T2 and T3 were more likely to be non-Hispanic white and female. Completers, compared with non-completers, had higher scores at baseline on recognition of alcohol-related SV, recognition of sexual and reproductive coercion, intentions to intervene, self-efficacy to obtain sexual consent (T3 only), self-efficacy to use SV-related services, knowledge of SVrelated services, and positive bystander behaviors (T2 only). Among those who reported SV/IPV at baseline, completers reported less use of SV-related services compared with non-completers (T3, among women only); more T3 completers reported recent SV compared with noncompleters at baseline.

More than half of participants (55%) reported history of SV at baseline (59% when including physical IPV and stalking). The most common types of SV were unwanted fondling, kissing, or other sexual touching (45.0%; 53.1% women, 23.1% of men) and attempted sex (36.7%; 43.1% women, 19.6% men). About 15% (14.9% intervention; 15.2% control; 18.3% women, 6.3% men) reported any SV in the past 4 months; this increased to 23.3% and 22.7%, respectively, at 4-month follow up (25.3% women, 14.7% men), then decreased to 12.6% and 12.2% at 1 year (14.7% women, 4.5% men; no difference by treatment arm). Further, 34.5% of those who reported SV at baseline reported a perpetrator took advantage of them when they were unable to consent because of alcohol or drugs (33.3% intervention; 35.4% control). This tactic was reported by 23.9% of those who experienced recent SV at T2 (18.7% intervention, 28.2% control, p=0.052) and 23.8% of those at T3 (22.5% intervention, 25.0% control, p=0.645).

Outcomes at baseline, 4 months, and 12 months are summarized in Table 2. Mean scores for recognition of sexual and reproductive coercion, intentions to intervene, and self-efficacy to obtain sexual consent were >4 at baseline (range, 1–5) and remained high at follow-up in both arms. Self-efficacy to use SV services mean score was 3.42 among intervention participants and 3.36 among controls at baseline and increased to 3.86 and 3.87, respectively, at 12-month follow-up. Knowledge of SV services was high at baseline, with >90% in both arms knowing at least one service. On a scale from 1 to 5, participants had summary scores of 3.65 (intervention) and 3.63 (control) at baseline, which increased to 4.10 and 4.02 at 4-month follow-up.

Table 2.

Primary and Secondary Outcomes

Variable	Baseline		4-month follow up		12-month follow up
	Intervention Mean (SE)	Control Mean (SE)	Intervention Mean (SE)	Control Mean (SE)	Intervention Mean (SE)	Control Mean (SE)
All participants	n=1,040	n=1,251	n=786	n=971	n=780	n=949
Recognition of alcohol-related sexual violence and sexual risk	3.99 (0.06)	3.98 (0.03)	4.02 (0.03)	4.00 (0.02)	4.06 (0.04)	4.04 (0.03)
Recognition of sexual and reproductive coercion	4.35 (0.04)	4.33 (0.02)	4.44 (0.03)	4.41 (0.02)	4.48 (0.04)	4.46 (0.03)
Intentions to intervene	4.04 (0.05)	4.10 (0.03)	4.12 (0.04)	4.12 (0.04)	4.17 (0.04)	4.19 (0.02)
Self-efficacy to obtain sexual consent	4.10 (0.03)	4.12 (0.02)	4.11 (0.03)	4.15 (0.02)	4.05 (0.03)	4.09 (0.01)
Self-efficacy to use harm reduction strategies	–	–	4.20 (0.03)	4.22 (0.03)	4.34 (0.05)	4.41 (0.03)
Among women onlya	–	–	4.23 (0.03)	4.27 (0.03)	4.37 (0.05)	4.42 (0.03)
Self-efficacy to use SV related services	3.42 (0.06)	3.36 (0.05)	3.60 (0.03)	3.59 (0.03)	3.86 (0.03)	3.87 (0.03)
Among women onlya	3.62 (0.07)	3.56 (0.03)	3.79 (0.05)	3.75 (0.03)	4.03 (0.04)	4.03 (0.02)
Recent sexual violence (scale, 1–7)	n/ab	n/ab	0.39 (0.04)	0.41 (0.03)	0.25 (0.02)	0.21 (0.02)
Any recent SV, % (n)	14.9 (155)	15.2 (190)	23.2 (182)	22.7 (220)	12.6 (98)	12.2 (116)
Knowledge of SV related servicesa (scale, 0–5)	3.65 (0.08)	3.63 (0.06)	4.10 (0.07)	4.02 (0.04)	3.98 (0.06)	4.08 (0.04)
Among women onlya (scale, 0–6)	4.56 (0.10)	4.53 (0.08)	5.04 (0.07)	4.96 (0.04)	4.94 (0.08)	5.01 (0.04)
Knowledge of at least one SV related services,a % (n)	93.9 (976)	91.1 (1,139)	94.8 (745)	94.3 (916)	96.9 (756)	98.2 (932)
Among women onlya	96.3 (724)	93.2 (866)	96.9 (590)	95.8 (722)	97.7 (581)	98.5 (723)
Among those who witnessed at least one abusive behavior by a peer at baseline	n=858	n=1,041	n=651	n=813	n=649	n=798
Positive bystander behaviors (scale, 1–10)	1.64 (0.04)	1.64 (0.06)	1.17 (0.05)	1.19 (0.05)	0.98 (0.04)	1.07 (0.07)
At least one positive bystander behavior, % (n)	79.8 (685)	79.9 (832)	63.0 (410)	64.6 (525)	56.7 (368)	57.4 (458)
Among those who reported SV or IPV at baseline	n=597	n=748	n=466	n=590	n=451	n=582
Use of any SV-related services, % (n) 10.7 (64)	10.7 (64)	9.1 (68)	9.4 (44)	8.8 (52)	18.2 (82)	13.9 (81)
Among women onlya	12.2 (60)	13.2 (82)	12.6 (50)	10.6 (53)	21.8 (84)	18.2 (90)
Among those who reported a history of SV or IPV			n=557c	n=690c	n=500d	n=645d
SV/IPV disclosure to CHC provider, % (n)	–	–	12.4 (69)c	9.6 (66)c	21.2 (106)d	24.2 (156)d

^aOutcome was first calculated restricted to items asked of all participants and then calculated restricted to participants assigned female at birth and including the additional item(s) these participants were asked (e.g., after an unwanted sexual experience, you can take emergency contraception).

^bPast 4 month SV was assessed at baseline as yes or no; a summary score is not available.

^cImmediately post-visit.

^d12-month follow up.

SV, sexual violence; IPV, intimate partner violence; CHC, college health center; – indicates that outcome was not asked at that time point.

Intent-to-treat analyses did not find significant differences between intervention and control arms on outcomes of interest at either 4- or 12-month follow-up (Table 3). Intervention effects for the primary outcome, recognition of alcohol-related SV and sexual risk, were 0.001 (95% CI= –0.04, 0.04) at 4 months and –0.003 (95% CI= –0.05, 0.04) at 12 months.

Table 3.

Intent-to-Treat and Intensity-adjusted Intervention Effect Analyses at 4- and 12-Month Follow Up

Variable	4-month follow up		12-month follow up
	Intervention effecta	Intensity adjusteda,b	Intervention effecta	Intensity adjusteda,b
All participants, β (95% CI)
Recognition of alcohol-related sexual violence and sexual risk	0.001 (−0.04, 0.04)	−0.01 (−0.06, 0.05)	−0.003 (−0.05, 0.04)	−0.001 (−0.06, 0.06)
Recognition of sexual and reproductive coercion	−0.01 (−0.07, 0.04)	−0.03 (−0.10, 0.04)	−0.02 (−0.07, 0.04)	−0.06 (−0.13, 0.02)
Intentions to intervene	0.04 (−0.05, 0.13)	0.04 (−0.08, 0.16)	0.02 (−0.07, 0.11)	0.10 (−0.02, 0.21)
Self-efficacy to obtain sexual consent	−0.03 (−0.09, 0.04)	0.03 (−0.06, 0.11)	−0.03 (−0.10, 0.04)	−0.004 (−0.10, 0.09)
Self-efficacy to use harm reduction strategiesc,d	−0.01 (−0.10, 0.08)	0.07 (−0.01, 0.16)	−0.04 (−0.14, 0.07)	0.10 (0.01, 0.19)
Among women only	−0.05 (−0.15, 0.05)	0.06 (−0.04, 0.16)	−0.05 (−0.16, 0.06)	0.07 (−0.03, 0.17)
Self-efficacy to use SV related servicesd	−0.07 (−0.17, 0.03)	0.02 (−0.11, 0.15)	−0.07 (−0.17, 0.02)	0.03 (−0.10, 0.15)
Among women only	−0.04 (−0.15, 0.06)	0.01 (−0.13, 0.15)	−0.08 (−0.18, 0.02)	−0.04 (−0.17, 0.10)
Recent sexual violence, RR (95% CI)	0.94 (0.73, 1.21)	0.82 (0.60, 1.13)	1.18 (0.85, 1.65)	1.02e (0.66, 1.59)
Knowledge of sexual violence related services,d RR (95% CI)	1.01 (0.95, 1.08)	1.03 (0.95, 1.12)	0.97 (0.91, 1.03)	0.98 (0.90, 1.07)
Among women only, RR (95% CI)	1.01 (0.94, 1.08)	1.00 (0.92, 1.09)	0.98 (0.91, 1.04)	0.98 (0.90, 1.07)
Among those who witnessed at least one abusive behavior by a peer at baseline, RR (95% CI)
Number of positive bystander behaviors	0.98 (0.84, 1.13)	1.20 (1.00, 1.45)	1.10e (0.94, 1.28)	1.12 (0.92, 1.36)
Among those who reported SV/IPV at baseline, OR (95% CI)
Use of any SV-related servicesd	0.95 (0.54, 1.68)	1.29e (0.65, 2.58)	1.22 (0.74, 2.02)	1.36 (0.73, 2.54)
Among women only	1.41 (0.80, 2.46)	1.61 (0.81, 3.23)	1.44 (0.87, 2.37)	1.39 (0.74, 2.61)
Among those who reported a history of SV/IPV, OR (95% CI)
SV/IPV disclosure to CHC provider	1.42 (0.55, 3.64)f	4.60 (2.29, 9.24)f	0.83 (0.46, 1.49)g	1.34 (0.81, 2.20)g

Notes: Boldface indicates statistical significance (p<0.05). ICCs for the primary outcome (recognition alcohol-related sexual violence and sexual risk) in the intervention effect (intention-to-treat) models were 0.58 (within-person) and 0.04 (within-school) at 4-month follow up and 0.53 (within-person) and 0.04 (within-school) at 12-month follow up.

^aBeta coefficients (β), relative risks (RR), and ORs adjusted for baseline outcome value (where available), year in school, race/ethnicity, randomization design covariate (large/small school), and college-level clustering.

^bIntensity score models replaced the binary treatment arm variable with a continuous score of intervention implementation and additionally adjusted for gender (except for women-only models), housing location, and lifetime experience of SV (except for use of any SV services as restricted to a subsample), which significantly differed by intensity score. All control arm participants were assigned an intensity score value of zero.

^cModel adjusted for baseline general self-efficacy in place of self-efficacy to use harm reduction strategies.

^dOutcome was first modeled as restricted to items asked of all participants and then a secondary model was run, restricted to participants assigned female at birth and including the additional item(s) these participants were asked (e.g., after an unwanted sexual experience, you can take emergency contraception).

^eNot adjusted for year in school due to model non-convergence with full set of covariates.

^fImmediately post-visit.

^g12-month follow up.

Intervention implementation varied significantly by site from 17% to 93% reporting receiving the card, with 41.6% of intervention arm participants receiving the intervention as intended (card and at least one discussion point; 40.7% of women and 44.8% of men, p=0.564). In interventionintensity adjusted analyses, intervention participants reported greater self-efficacy to use harm reduction strategies (β=0.09, 95% CI=0.01, 0.18). No other differences between arms emerged among participants overall.

Among those who had witnessed SV behaviors among their peers at baseline (83%), participants in both arms reported some positive bystander behaviors at baseline (summary score=1.64; range, 1–10), which decreased over time in both arms. Positive bystander behaviors did not differ by arm in either intent-to-treat or intensity-adjusted analyses.

Among those who reported SV or IPV at baseline (59%), use of any SV services increased among intervention participants from 10.7% to 18.2% and among control participants from 9.1% to 13.9% at 12-month follow-up (OR=1.22, 95% CI=0.74, 2.02). In both intent-to-treat and intervention adjusted analyses, differences between arms were not statistically significant. Disclosure to a health professional during the baseline CHC about history of violence exposure (reported via exit survey after the clinical encounter) was greater in the intervention arm (12.4%) compared with the control arm (9.6%). The difference between arms was significant only in intensity-adjusted analyses (OR=4.47, 95% CI=2.25, 8.89). At 1-year follow up, disclosure about SV to a healthcare provider in those 12 months did not differ by treatment arm. The authors conducted post-hoc analyses to assess referral to services in the context of a disclosure. On the exit survey, 4.5% of participants reported a referral for SV/IPV services overall, which was higher in the treatment arm (8.7% vs 1.7% of all participants, p<0.001; 9.2% vs 1.7%, p<0.001, among those with history of SV/IPV).

DISCUSSION

To the authors’ knowledge, this is the first large-scale study in CHCs to evaluate the effectiveness of a brief harm reduction intervention to reduce risk for alcohol-related SV. Strengths of this study include the number and heterogeneity of CHCs involved (by size, geography, services), stakeholder engagement (practitioners, advocates, and students involved in development of intervention), and rigorous design. Overall, no differences in intervention effects emerged comparing GIFTSS to an alcohol-focused intervention. Implementation of GIFTSS varied across campuses. In post-hoc analyses, when GIFTSS was implemented as intended, only self-efficacy to use harm reduction strategies was greater among intervention participants compared to controls. Students with histories of SV or IPV had increased odds of disclosing such experiences during their CHC visit when receiving the intervention as intended. Such disclosure appeared to increase referrals to, but not reported use of, SV related services.

More than half of participants reported a history of SV or IPV, underscoring how important CHCs are for connecting survivors to relevant services. Guided by stakeholders, the authors learned that the most salient aspect of communicating SV-related risk with patients was ensuring that all patients received information about SV (especially involving alcohol), regardless of disclosure. Findings suggest, however, that a brief psychoeducational intervention alone is not sufficient to increase uptake of SV services among survivors even though disclosures to providers about SV appeared to increase in GIFTSS sites. Post-hoc analyses suggest that providers in the intervention arm offered more referrals to services compared with control providers; further exploration about factors that increase actual uptake of services is needed.

Limitations

The limited implementation of this intervention across sites is a significant limitation, as it remains unclear the extent to which GIFTSS could be efficacious under ideal conditions. Although describing implementation barriers are beyond the scope of this manuscript and are addressed in detail elsewhere,³⁴ providers’ concerns about insufficient time, forgetting to offer the card, and type of visit not being relevant to SV (such as a student presenting with a cold) were common barriers to implementation. Though trainers conducted follow-up with CHC staff after the initial training to support implementation, training alone does not promote change in practices, and changes to protocols and CHC flow are needed to support intervention delivery. Studies to evaluate training modalities and to optimize implementation of such brief counseling interventions for CHC settings are needed.

Findings should be considered in context of additional limitations. First, this study relied on selfreport of sensitive topics. Surveys were administered online or in a private area in each clinic on a tablet to maximize privacy.^35–37 Second, colleges from Western Pennsylvania and West Virginia are not representative of college students across the nation; proportions of underrepresented racial/ethnic groups are well below national estimates at degree-granting postsecondary institutions. Third, although the authors tracked alcohol and SV prevention programs available on campuses, they were not able to discern differences in intervention effects across campuses that could be attributable to other campus-wide policies and programs. Additional analyses are needed to examine heterogeneity of intervention effects by setting (health versus counseling centers), exposure to other programs, demographic characteristics, and type of SV experiences including severity and alcohol-related incidents. Finally, reducing SV on campuses requires focus on stopping SV perpetration, which was not the focus of this intervention. The GIFTSS intervention had a dual focus on prevention (universal education for all students seeking care about reducing SV on campuses with an emphasis on bystander interventions and helping friends who have experienced SV) combined with increasing safety and support for survivors of SV. More research is needed to identify strategies that reduce SV perpetration on campuses.

CONCLUSIONS

Among students seeking care in CHCs, SV victimization is common. Although GIFTSS may be one strategy for disseminating information about SV and services for students already exposed to SV, such brief interventions are likely insufficient for addressing the complex needs of these students. More in-depth interventions designed with input from students, providers, administrators, and advocates combined with comprehensive prevention strategies are needed to improve connection of vulnerable students to relevant and effective services that can help reduce SV and the serious sequelae of such violence. Simultaneously, implementation studies that evaluate strategies to optimize delivery of brief counseling interventions in CHCs are needed.