A Brief Clinic-Based Intervention to Reduce Alcohol Misuse and Sexual Risk Behavior in Young Women: Results from An Exploratory Clinical Trial

Abstract

Background.

This exploratory trial determined the feasibility, acceptability, and preliminary efficacy of a brief intervention (BI), supplemented with text messaging and a curated website, on alcohol use and sexual risk behavior among young women.

Methods.

Young women seeking care at a reproductive health clinic were screened for alcohol misuse and sexual risk behavior. Those who screened positive and who agreed to participate (N = 48; M = 22.67 years) were randomized to either (a) a brief in-person session during which personalized feedback regarding alcohol use and sexual risking taking was provided and discussed, or (b) a control condition. Feasibility was assessed by recruitment and retention rates. Acceptability was assessed with participant ratings of their intervention. Efficacy was measured using self-reported alcohol use and sexual behavior at baseline and during a 3-month follow-up. We supplemented the quantitative data with qualitative data from semi-structured interviews.

Results.

Feasibility data indicated that 64% of eligible women agreed to participate, 74% of eligible women were enrolled, and 86% of enrolled women were retained through follow-up. Acceptability data showed that women who received the BI reported strong satisfaction with their intervention (M = 4.65 vs. 3.98 on a 5-point scale) and also reported that text messaging was helpful (M = 4.73 on a 7-point scale) and acceptable (M = 5.27 on a 7-point scale). Qualitative data provided additional support for BI feasibility and acceptability. Efficacy data showed that women in both conditions reduced alcohol use and sexual risk behavior over time; women who received the BI reduced their maximum daily alcohol intake more than controls (BI: 7.68 to 4.82 standard drinks vs. Control: 6.48 to 5.65; Wald x² = 4.93, p < .05). Women in the BI (Median = 2.50) reported fewer occasions of condomless sex than controls (Median = 5.00) at the follow-up, but this difference was not statistically significant (OR = 0.61, 95%CI [0.32, 1.15]).

Conclusions.

A brief intervention, supplemented with text messaging and a website, that targeted alcohol use and sexual behavior was feasible and acceptable to young women, and led to lower levels of alcohol misuse and sexual risk behavior.

Sexually transmitted infections (STIs) threaten the health of young women. Young women have the highest rates of STIs relative to similarly aged men and older women; recently, these rates have been increasing (CDC, 2019). In addition, women account for 19% of new HIV diagnoses, with the large majority attributed to heterosexual sex; low-income and racial and ethnic minority women are particularly vulnerable (CDC, 2018). Consequences of STIs include reproductive health problems, fetal and perinatal health problems, cancer, and facilitation of the sexual transmission of HIV (Green, Zarek, & Catherino, 2015; Nadeau, Fujii, Lentscher, Haney, & Burney, 2018).

Sexual risk reduction interventions designed for young women can increase condom use, decrease number of partners, and reduce STI incidence (Ruiz-Perez, Murphy, Pastor-Moreno, Rojas-García, & Rodríguez-Barranco, 2017). Enthusiasm for these interventions is tempered, however, due to relatively small effects that tend to decay over time. In addition, women who are the most at risk for STIs (e.g., those who have multiple partners and/or who misuse alcohol) can be difficult to reach. Further limiting the uptake of behavioral interventions is the concern that they can be burdensome to providers and patients (Wilson & Albarracín, 2015). Needed are more practical, efficacious, and enduring interventions that reach women at the greatest need in settings they frequently attend.

Building more efficacious interventions.

To optimize intervention efficacy, it is critical to identify and target the determinants of the risk behaviors. For young women, one of the strongest determinants of sexual risk behavior is alcohol misuse. Relative to older women, women aged 18 to 29 years are more likely to drink, to engage in heavy episodic drinking, and to meet criteria for an alcohol use disorder (Substance Abuse and Mental Health Services Administration, 2012). Nearly 50% of young women drink alcohol and 20% report heavy episodic drinking (≥ 4 drinks during a drinking day) in the last 30 days (Marchetta et al., 2012). Alcohol use is associated with having more sexual partners (Lee et al., 2018), sexual partner concurrency (Adimora, Schoenbach, Taylor, Khan, & Schwartz, 2011), non-use of condoms (Lee et al., 2018; Patrick, O’Malley, Johnston, Terry-McElrath, & Schulenberg, 2012), and more STIs (Aicken, Nardone, & Mercer, 2011; Seth, Wingood, DiClemente, & Robinson, 2011; Walsh, Fielder, Carey, & Carey, 2014). In one study, women who engaged in heavy episodic drinking had twice the rate of multiple partners and five times the rate of gonorrhea compared to abstainers (Hutton, McCaul, Santora, & Erbelding, 2008). Relative to other women, women who drink heavily are also more likely to report unintended pregnancy and/or emergency contraception use (Aicken et al., 2011). Notably, the association of alcohol use with risky sexual behavior is stronger for women than for men (Carey, Senn, Walsh, Scott-Sheldon, & Carey, 2016; Scott-Sheldon et al., 2009).

Alcohol use interacts with relationship factors to increase women’s risk (Carey et al., 2019). Gender-based power imbalances (e.g., differences in age, sex roles, physical strength, economic resources) contribute to risk, particularly for young women (Connell, 1987), who often must persuade a male partner to use a condom. Condomless sex is more prevalent with “main,” “steady,” or “regular” partners (Reid & Aiken, 2011) but research shows that condom use also varies among different types of “casual” partners (e.g., ex-boyfriend, friend with benefits) (Catallozzi et al., 2013). Sexual “hookups” (i.e., short-term relationships with no expectation of mutual commitment) are more likely to occur following alcohol use, and the likelihood of condomless sex during hookups increases with higher levels of alcohol use (Fielder & Carey, 2010; Fielder, Walsh, Carey & Carey, 2013; Haberland, 2015).

Given the associations among alcohol use, sexual risk taking, and relationship type, targeting all three factors in an integrated way is indicated. However, even though many sexual risk reduction interventions acknowledge the role of alcohol use, few clinic-based sexual risk reduction interventions target alcohol use directly. In a rare exception, Crawford et al. (2015) examined the efficacy of brief advice (i.e., feedback on alcohol and health, written information, and an offer of an appointment with an alcohol health worker) on alcohol consumption and unprotected sex among adults recruited from a sexual health clinic. However, in this study, the brief advice intervention lasted only 2 to 3 minutes and even the offered appointment with an alcohol health worker lasted only “up to 30 minutes.” This extremely brief intervention did not reduce either alcohol use or sexual risk behavior. In contrast, a large body of research from other settings shows that somewhat longer but still “brief” interventions (i.e., those lasting no more than one hour) can effectively reduce alcohol use by young adults if they target alcohol use in a direct but personalized way (e.g., using normative feedback) (Carey, Scott-Sheldon, Carey, & DeMartini, 2007; Reid & Carey, 2015).

Increasing uptake of interventions.

A second challenge to increasing the public health impact of sexual risk reduction interventions involves their poor uptake. Most empirically validated interventions have been “intensive,” that is, group-based and involving multiple sessions (Crepaz et al., 2009). These structural features undermine their appeal to participants and their feasibility. Even single-session interventions are poorly attended when they place too many demands on participants (e.g., time from work, need for childcare). Even a single-session 4-hour workshop drew only 55% of interested women, despite providing financial incentives and childcare (Carey, Senn, Vanable, Coury-Doniger, & Urban, 2010); low participation and poor attendance are even more common among multiple session, group-based programs.

Fortunately, “brief interventions” (BIs) can be efficacious for both alcohol misuse (Cronce & Larimer, 2011; Moyer, 2013) and for sexual risk reduction, especially when the BI is focused on the determinants of the risk behavior (Eaton et al., 2012). Indeed, BIs appear to be an optimal intervention for reducing alcohol misuse among young adults (Carey, Scott-Sheldon, Elliott, Garey, & Carey, 2012; Samson & Tanner-Smith, 2015).

Accessing “at-risk” young women.

The venue where an intervention is offered affects its public health “reach.” Young adult women are less likely to see a generalist provider than are older women (Irwin, 2010), yet more than 75% of young women visit a reproductive health and family planning (RHFP) clinic annually (Frost, 2008; Irwin, 2010). Young women attend RHFP clinics to obtain birth control, STI/HIV testing and treatment, and other gynecologic services (e.g., PAP smears) in addition to more comprehensive care. Young women who attend RHFP clinics are at higher risk for alcohol misuse than women in the general population (Cook et al., 2006; Hutton et al., 2008; Thorley, Hettiarachchi, Nightingale, & Radcliffe, 2012). In previous research at a RHFP clinic, 52% of young women reported heavy episodic drinking in the last 3 months, alcohol use was associated with increased sexual risk behavior (Carey, Scott-Sheldon, Senn, & Carey, 2013), and young women who engaged in heavy episodic drinking were more likely to report multiple partners (Morrison-Beedy, Carey, Crean, & Jones, 2011). Importantly, RHFP settings are community-based and lower cost (or free), providing services for low-income women. For these reasons, sexual health settings may be “an ideal setting in which to integrate screening and intervention strategies related to substance use problems among young persons” (Cook et al., 2006). Despite the promise of RHFP clinics as a venue for at-risk young women, no study has implemented an integrated alcohol-sexual risk intervention in a RHFP setting.

Increasing BI “dose” with technology.

A limitation of brief interventions is brevity limits the “dose” of the intervention that can be provided. Fortunately, technology (e.g., text messages, websites) can be used to extend BIs (Hall, Cole-Lewis, & Bernhardt, 2015) while enhancing intervention fidelity, often at lower cost. Although stand-alone, technology-only interventions have limited efficacy (e.g., Berman, Gajecki, Sinadinovic, & Andersson, 2016; Fowler, Holt, & Joshi, 2016; Head, Noar, Iannarino, & Grant Harrington, 2013), technology can be used to extend a face-to-face intervention (Mastroleo et al., 2018; Tahaney & Palfai, 2017). In one study, the efficacy of a BI for problem drinking was enhanced when it was supplemented with a curated website (Cunningham, 2012). Empirically-based content is easily accessed on websites maintained by trusted organizations (e.g., Centers for Disease Control and Prevention [CDC]). Videos, which can be viewed at any time, can teach skills (e.g., condom use, drink refusal) using actors who model skills to teach new behaviors (Tuong, Larsen, & Armstrong, 2014) while promoting self-efficacy (Bandura, 1985). Text messaging can be used to provide reminders, social support, and encouragement as well as to encourage interaction with additional (web-based) educational material (Schneider, van Osch, & de Vries, 2012).

Using technology to extend the dose and duration of BIs is especially promising for young adults. Nearly all young people in the U. S. own cell phones (100% of 18–29-year-olds; Pew Research Center, 2018), and texting is an inexpensive and acceptable way to reach them (Hall et al., 2015). Most young adults (84%) access the internet with their phones (Duggan, 2013), and 72% look online for health information, especially on sensitive health topics such as substance use and sex (Lenhart, Purcell, Smith, & Zickuhr, 2010; Wartella, Rideout, Montague, Beaudoin-Ryan, & Lauricella, 2016). Furthermore, 95% watch videos online, with “how to” videos watched by 78% of this age group (Duggan, 2013). Thus, texting and online videos have an unprecedented “reach” among young people, magnifying the potential public health impact.

Purposes of this research.

This exploratory trial was designed to gauge the feasibility and acceptability of a BI with technology extenders, and to obtain preliminary evidence of its efficacy. We began by conducting formative research with the target population to refine the BI so that it targeted the pathways through which alcohol use increases sexual risk (e.g., impaired decision making, lessened ability to enact protective strategies) (Carey et al., 2019). The BI addressed the risks associated with common relationship types. We supplemented the BI with text messages (provided information about sexual health, alcohol use, and STI prevention) and a website of curated content. These technology extenders allowed us to provide a greater dose of intervention. We compared this intervention to a control condition, with the expectation that the BI and extenders would be feasible, acceptable, and more efficacious.

Methods

Setting

The NIH-defined Stage 1B Exploratory Clinical Trial took place at an urban reproductive health and family planning clinic in the northeastern U. S. that has been providing contraceptive and reproductive health services for more than 50 years. The range of medical services provided includes gynecological visits, Pap tests, testing and treatment for sexually transmitted infections, birth control services and supplies, diagnosis and treatment of vaginal infections and cervical conditions, and pregnancy tests and options counseling. The clinic serves a diverse clientele with ~50% being people of color, ~60% aged 30 or younger, and ~75% patients with incomes below 200% of the U. S. Federal Poverty Level. Thus, the vast majority of patients are people with low-incomes, and many are uninsured and unable to afford the full cost of their care. The clinic charges patients according to a sliding fee scale based on income and family size. Most pay a fraction of what they would be charged by a private physician, and the poorest pay little or nothing for their care.

The study procedures were approved by The Miriam Hospital’s Institutional Review Board.

Procedures

Screening.

All women aged 18 to 29 who presented for care were invited to complete a brief (< 5 minute) screening survey on a tablet computer in a private room. The survey assessed a range of health behaviors (e.g., smoking, exercise, diet) and health conditions to mask the inclusion criteria: (a) female assigned at birth, (b) aged 18 to 29 years, (c) at-risk drinking (> 3 drinks on any day and/or > 7 drinks per week) in the last 3 months (National Institute of Alcohol Abuse and Alcoholism, 2004); (d) sexual risk behavior (i.e., vaginal or anal intercourse with >1 partner; vaginal or anal intercourse with a male partner who has other partners; inconsistent condom use) in the last 3 months; (e) English speaking; (f) absence of acute intoxication, depression, or suicidal ideation (using the PHQ-9; Kroenke, Spitzer, & Williams, 2001); and (g) no plans for relocation in the next 3 months. Once the survey was completed, a scoring program informed the female Research Assistant (RA) regarding a woman’s eligibility.

Recruitment.

In order to recruit a sample of 50 women, eligible women were invited by the RA to join the Women SHARE (Sexual Health and Relationships) Study. The RA explained that the study sought to evaluate two interventions intended to help women improve their health by addressing relationships, sexual behavior, and alcohol use. One intervention would rely on written materials whereas the other would involve health coaching, and assignment would be determined by chance. Before and after the intervention, and 3 months later, women would complete a survey; they would be paid $30 for the initial survey and $40 for the follow-up survey. After hearing about the study, patients were invited to ask any questions they had. The RA also asked them questions to be sure that they understood what participation would involve. Women who were interested signed a written consent form and provided contact information.

Baseline assessment.

Women completed an online baseline survey in a private room. Most women completed the survey within 45 minutes. The survey assessed a range of variables required for the BI feedback report, including alcohol use, sexual behavior and history, relationship and health values, and attitudes and norms using standardized items. Variables used in this report (demographics and primary outcome variables) are described below.

Women reported their age, racial/ethnic background, relationship status, educational and employment status, gender identity, sexual orientation, lifetime alcohol and drug use, and history of sexual assault using standard items. These variables were collected to describe the sample.

Women completed the following measures to describe their recent alcohol use: (a) maximum number of drinks on a single day (“Considering all types of alcoholic beverages, what was the largest number of drinks you drank on any one day in the past 3 months?”), (b) typical number of drinks per week (“In the past 3 months, counting all types of alcohol combined how many drinks did you consume over a typical week?”), (c) number of heavy episodic drinking days in the last month “(“In the past 30 days, how many days did you drink 4 or more drinks on one occasion?”), and (d) alcohol-related consequences using the Short Inventory of Problems (SIP-2R; Feinn, Tennen, & Kranzler, 2003; Kiluk, Dreifuss, Weiss, Morgenstern, & Carroll, 2013). These variables were used as primary outcomes.

Women completed the following measures to describe their recent sexual behavior in the past three months: (a) their number of male sexual partners and (b) the number of times they had vaginal or anal sex with a man (count). Women also reported the number of times they had (c) used alcohol “before or during” vaginal/anal sex (count), and (d) used a condom (count). Finally, women answered an item about the quantity of alcohol consumed before or during sex, namely: “What was the largest number of drinks you consumed?” The sexual risk questions resulted in three summary variables: (1) number of male partners (count); (2) number of condomless sex acts (count); and (3) maximum number of alcohol drinks consumed before/during sex (count), which were used as primary outcomes.

Interventions

Control.

Women assigned to the control condition were given two brochures prior to leaving the clinic. One addressed “Alcohol Use and Your Health” (CDC, 2012) and the second provided “Information for Teens and Young Adults: Staying Healthy and Preventing STDs” (CDC, 2017). These brochures provided information regarding gender differences in alcohol intoxication, alcohol use and reproductive health, prevention of STIs, and STI testing and treatment. Women were encouraged to read these brochures, which served as a “standard care” control condition. There were no additional intervention sessions. Prior to exiting, they were scheduled for their 3-month follow-up.

Brief intervention (BI).

Women assigned to the BI condition received a face-to-face intervention delivered by a health coach/counselor. BI duration (≤ 60 minutes) was based on previous work showing the feasibility and efficacy of BIs of this length (Carey et al., 2010; Carey, Senn, Coury-Doniger, et al., 2013). Overall, the BI sought to reduce alcohol misuse and sexual risk taking by raising awareness of how alcohol use impairs decision making and sexual risking taking (e.g., condomless sex). The BI was guided by two health behavior theories.

First, with regard to sexual risk behavior, the BI was guided by the Information-Motivation-Behavioral Skills (IMB) model (Fisher & Fisher, 2000), which posits that information, motivation and behavior skills influence health protective behaviors. Informational variables include knowledge of STI transmission and protective behaviors. Motivational variables include attitudes, perceived norms, and self-efficacy. Skills variables include self-management, assertiveness, and condom use. Motivational variables, in particular, were expected to vary by partner type such that self-efficacy predicts condom use with casual partners whereas attitudes and partner norms predict condom use with regular partners (Reid & Aiken, 2011).

Second, with regard to alcohol use, the BI was guided by the dual-process model (Moss & Albery, 2009), which posits that alcohol use affects sexual behavior before (“pre-consumption”) and after (“post-consumption”) drinking. The pre-consumption phase recognizes that some individuals drink to facilitate sexual activity (e.g., “liquid courage”: Stoner, George, Peters, & Norris, 2007) and that drinking activates positive expectations of sexual activity (Goldman, Reich, & Darkes, 2006). The post-consumption phase recognizes the pharmacological effects of alcohol, namely, that intoxication reduces cognitive processing capacity, impairs executive function, and narrows attentional focus to the most salient cues (e.g. sexual arousal). Under conditions of inhibitory conflict (e.g., when the desire for sex is inhibited by concerns about consequences such as STI), intoxication increases the probability of risky sexual behavior.

The content and therapeutic goals of the BI are provided in Table 1. A key component of the BI was personalized normative feedback compared to age-matched women derived from the Behavioral Risk Factor Surveillance System (Centers for Disease Control and Prevention, 2015). This feedback was designed: (a) to prompt discussion of safer sex behaviors and drinking levels, (b) to correct exaggerated norms regarding alcohol use and sexual behavior, (c) to enhance self-efficacy regarding less risky alternatives to heavy episodic drinking and risky sex, and (d) to set goals consistent with self-protection. The topics were discussed in the context of specific relationships using a grid, anticipating potential barriers to safer sexual practices. Using a visual meter indicating risk for STIs, women were provided with personalized normative feedback on the risk inherent in each partnership based on frequency of condom use, alcohol use prior to sex, partner STI testing history, and partner’s engagement in concurrent sexual partnerships.

Table 1.

Brief Intervention Content and Therapeutic Goals

Content/Activity	Therapeutic Goals
Orientation Introductions Overview of session topics	Establish collaborative tone Demonstrate respect for autonomy
Health Values Provide feedback on health values Elicit participant’s definition of sexual health Discuss strategies used to take care of sexual health	Engage with self-affirmations Link sexual health with broader health concerns
Relationships Provide general feedback on reported number of partners, their HIV status, & incidence of condomless sex Discuss partner-specific feedback using risk gauges based on condom use, co-occurring alcohol use, testing status, concurrent partners, level of trust in relationship Elicit perceptions of why risk varies across partners and opportunities for reducing risk within each relationship	Engage participant with personalized information Promote risk sensitization (M) Identify links between drinking & sexual outcomes (I) Attend to role of alcohol and partner dynamics in engaging in condom protected sex (M) Identify partner-specific barriers (I) Prompt change talk (M, B)
Condom Use Provide personalized feedback on condom-protected sex acts Provide personalized feedback on assertive condom conversation strategies and areas for improvement	Enhance risk sensitization (M) Affirm positive behaviors (B) Elicit change talk (M, B)
STI testing Review STI testing history and testing recommendations Recommendations for reducing risk of STIs	Reinforce/provide guidelines for appropriate testing practices (I) Elicit behavioral intentions (B)
Alcohol Use Patterns Provide personalized feedback on average drinks per day, maximum drinks, and heavy episodic drinking frequency Compare to age-linked norms of women in this state Provide personalized feedback on BAC on typical and heaviest day, compared to legal driving limit (.08) and BAC associated with loss of consciousness (.30) Provide feedback on number of drinks prior to having sex, and relate back to heavy episodic drinking and BAC	Engage participant with personalized information Correct exaggerated peer norms (I) Develop self-other discrepancy (M) Raise awareness of alcohol-related impairment and sexual decision making (I)
Protective Strategies Discuss strategies that can be used to increase positive drinking outcomes and reduce negative outcomes Generate participant-specific strategies and partner-specific strategies if relevant	Raise awareness of strategies to moderate levels of intoxication (I) Elicit change talk (M, B)
Alcohol Consequences Provide feedback on participant-reported alcohol-related consequences across domains of health, mood, finances, decision making, relationships	Engage participant with personalized information Develop discrepancy between real and ideal selves (M) Generate motivation to reduce drinking in order to reduce consequences (M) Prompt change talk (M, B)
Change Plan Summarize areas of concern raised during session Invite participant to complete a change plan outlining goal, reason, specific steps, anticipated barriers, social supports, desired outcomes Elicit self-ratings of importance, confidence, and empowerment	Prompt public commitment to change and accountability (M) Generate SMART goals: specific, measurable, achievable, realistic, and time-targeted (M, B) Generate change talk and/or commitment language (M, B) Support self-efficacy (M)

I = information; M = motivation; B = behavioral intentions/skills. All feedback was based on baseline assessments covering the past 3 months.

The BI culminated in a change plan tailored to each woman’s circumstances. This plan addressed alcohol use, condom use, and partner reduction using a menu of options. For example, women might set goals (a) to reduce the quantity of drinking and/or drinking before sex; (b) to increase condom use; (c) to discuss and seek STI testing prior to engaging in a new sexual relationship; (d) to reduce the number of partners; and (e) to prepare for situations in which alcohol will be used (i.e., protective behavioral strategies; cf. Prince, Carey, & Maisto, 2013).

The health coach followed principles of motivational interviewing (MI) (Miller & Rollnick, 2013), a collaborative, goal-oriented form of communication that strengthens motivation and commitment toward specific goals. Coaches elicited a woman’s reasons for change and respected her autonomy (Miller & Rollnick, 2013), taking a collaborative stance. Technical therapeutic skills (e.g., creating discrepancy to motivate change) were used. The health coaches were mental health professionals (MA- or PhD-level) who received training and supervision from the investigators.

We used two technology extenders. After receiving the BI, women were sent text messages daily for 12 weeks. Message content and frequency were determined by formative work. As detailed in Table 2, the texts provided facts about alcohol use, STIs, contraception, and related issues. Once a week, a text reminded women of the website by noting new content and ending with a link to the website.

Table 2.

Samples of the Content of the Daily Texts

Topics	Sample Text
General Women’s Health	Adults need at least 7–8 hours of sleep every night. Sufficient sleep is an essential part of staying mentally and physically healthy. Women should start getting regular Pap tests at age 21. Women ages 21–29 should get a Pap test every 3 years. Eating a balanced diet of fruit, vegetables, and protein and getting a regular dose of physical activity can decrease fatigue.
Alcohol/Drugs	There are 5 standard drinks in a 750ml bottle of wine, and over 8 standard drinks in a pint of hard liquor. Even after you stop drinking, your BAC may still be rising as alcohol is absorbed into your bloodstream. Most women who drink find that staying in the “green zone” (under a BAC of .08) brings more positive than negative effects.
Contraception	Use water-based and silicone-based lubricants to help prevent condoms from breaking or slipping. Avoid oils, lotions, or Vaseline. The birth control implant is a tiny rod inserted into your arm that releases hormones that prevent pregnancy for up to 4 years. Condoms 101: Check the expiration date and make sure there are no tears or defects before using a condom.
STIs	Condoms may cost a little, but medical care and diapers cost a whole lot more. An ounce of prevention is worth a pound of cure. Having multiple sex partners or sex partners you don’t know well can increase your risk for STDs. It is important to get tested for STIs; treatments, like antibiotics, can cure some STIs, while others can keep the disease from worsening.
New blog post alert	Check out the new blog post: Testing Before Sexing Check out the new blog post: Drinking Decisions: Limits and Tips Check out the new blog post: The Best Defense is a Good Offense

The website complied with best practices in website design (Schneider et al., 2012); thus, we built an easy-to-navigate website tailored to young women. The website was accessible only to participants and study staff. We drew upon our formative work and website usability principles (Devine, Broderick, Harris, Wu, & Hilfiker, 2016) to make the website appealing, intuitive, and informative. The content targeted the antecedents of alcohol use and sexual behavior, with engaging textual, graphic, and video materials. To address information and motivation, we developed blog posts based on themes raised by women in our formative work (e.g., “How to bring up using condoms”) and provided a section to allow women to ask questions. The blogs provided information in an engaging way on topics of interest. We also provided links to websites with information about reducing alcohol use and sexual risk.

To promote behavioral skills, we used professionally produced videos. These videos depicted compelling vignettes (Tuong et al., 2014), using Edutainment principles (Moyer-Guse, 2008) to model self-management and inter-personal skills and to address barriers in the natural environment. We selected content to meet women’s needs as identified during our formative work. Information about the videos is summarized in Table 3.

Table 3.

Skills Domains, Video Content and Duration of Website Videos

Use a condom properly so you are protected from beginning to end Condom application tips from the lady you wish was your health teacher (0:59) (Be the Change, 2012; Carey et al., 2015) Condom use 101 (2:29) (Planned Parenthood, 2011) Condom tips for the ladies: Getting past the double standard (3:10) (Green, 2015a) What is a female condom? (0:42) (Planned Parenthood, 2015b)

Introduce condoms into a relationship and effectively deal with resistance from your partner Let’s try something new – Turning a no to a yes by eroticizing condom use (1:40) (Select Media, 2009) Responding to 5 common excuses for not using condoms (2:03) (Meld, 2014) Holding the line – Introducing condoms with a long-term partner (2:55) (Be The Change, 2012; Carey et al., 2015) Talk condoms to me BBY: assertive and sex positive come-back lines (3:15) (Green, 2015b)

Communicate with partners What is consent? And how do you ask for it? (4:43) (Brown University, 2012) How to start the conversation (4:02) (Crain, 2014) Skills for healthy romantic relationships (15:52) (Davila, 2015)

Know when, where, and why to get tested for STI When to get tested and what to expect, with enthusiasm! (2:57) (Green, 2015c) Telling your partner you have an STI – Demystifying the process and staying positive (3:57) (Green, 2013b) Adventures in HIV testing – Good reasons to know your status (3:45) (Green, 2012a) Demystifying pre-exposure prophylaxis: When would PrEP be right for you? (4:42) (YourekaScience, 2014)

Sort through what type of contraception is right for you Birth control basics: What you might want to know about the pill. Hear a woman’s perspective (2:26) (Green, 2012c) What is the easiest birth control ever? Learn about LARCs (3:02) (Green, 2017) IUD basics from Planned Parenthood (1:13) (Planned Parenthood, 2017) Birth control basics: Are implants right for you? Hear a woman’s perspective (3:06) (Green, 2012b) Learn about the sponge from Planned Parenthood (1:26) (Planned Parenthood, 2010b) Non-hormonal barrier methods – Pros and cons (1:37) (Planned Parenthood, 2010a) What you need to know about “pulling out” (hint: 1 in 3) (3:40) (Green, 2013a) Condoms: Double protection from pregnancy and STDs (0:45) (Planned Parenthood, 2015a)

Turn down a drink (or anything else) if you want to Resistance skills: How many of these 6 types of offers have you experienced? Watch this for strategies to use in the face of drink offers. (7:34) (Flipp, 2013) Refusing a drink: how to deal with the pressure (5:03) (Charisma on Command, 2015) Prepare yourself with ways to say no to a drink (2:04) (Peters, 2016)

Post-intervention assessment.

Feasibility was assessed by recruitment and retention rates. Acceptability was assessed by participant ratings (of the BI session and the coach) using items from the Session Evaluation Questionnaire (Stiles et al., 1994), as modified for use with BIs (Carey, Carey, Maisto, & Henson, 2006). Using 6-point bipolar scales, women rated their perception of their health coach on 7 dimensions (e.g., likable-not likable, caring-not caring, warm-cold), and their perception of the session on 7 dimensions (e.g., difficult-easy, valuable-worthless, comfortable-uncomfortable). The women also rated their satisfaction with the session (e.g., personal relevance, interest value, their overall impression, and whether they would recommend the intervention to others) on 5-point scales.

Follow-up assessment.

Three months after baseline, women returned to complete the follow-up, which obtained additional evidence of BI acceptability during a semi-structured interview and BI efficacy using a web-based survey.

A female team member conducted the semi-structured interview, using open-ended questions to elicit feedback on women’s experiences with the study, including intervention components and assessments. Women in the BI condition also reported on unique intervention components (e.g., what they remembered about the personalized feedback sheets; most and least helpful parts of the sheets; impressions about the change plan); website usability (e.g., their impressions of each section; any difficulties accessing the site; favorite and least favorite parts); and text messaging (e.g., overall impressions as well as frequency, time of day, ideal number).

The survey was identical to the baseline survey (e.g., assessing alcohol use and sexual behavior) except that the women in the BI condition completed additional items rating the text messages and website (Larsen, Attkisson, Hargreaves, & Nguyen, 1979). The website usability questions were drawn from a validated measure (Sindhuja & Dastidar, 2009) designed to assess five factors: information content, design, format, ease of use, and likelihood of returning. From these items, we calculated an overall rating reflecting website usability and user satisfaction.

Finally, website usability statistics were harvested to complement women’s self-reported use of and satisfaction with the website. Anonymized Google analytics provided additional objective measurement of website use.

Data Analyses

For the feasibility data, we produced a Consort flow figure and calculated summary statistics to report the findings. For the acceptability data, we calculated summary statistics and identified illustrative quotes from the semi-structured interviews. Prior to analyzing the efficacy data, we used t-test and chi-square analyses to determine if randomization produced equivalent groups with respect to demographic characteristics and outcome measures.

For the efficacy data, we used inferential statistics to compare the two conditions to one another. Specifically, we used generalized estimating equations (GEE; Hardin & Hilbe, 2013) to examine the impact of the intervention on primary outcomes (i.e., typical drinks per week, maximum drinks per day, number of heavy episodic drinking days, alcohol-related consequences, number of sexual partners, and alcohol use before sex) as well as to model change over time. We used an intent-to-treat approach such that data from all women were used. We hypothesized an intervention-by-time interaction such that participants in the BI, but not the control, condition would reduce alcohol use and sex risk behavior over time. All analyses examined the impact of time (baseline vs. follow-up), intervention condition (control vs. BI), and the time-by-condition interaction controlling for age. When interactions were not significant, models were run without the interaction terms. Count variables (i.e., drinks per week, maximum drinks per day, heavy episodic drinking days, number of partners, drinks before sex) were analyzed using a Poisson distribution unless otherwise indicated. Alcohol consequences were analyzed using linear models.

For condomless sex, a GEE model could not be fit due to significant over-dispersion bounded by a large number of zeroes. Instead, a zero-inflated negative binomial model was chosen based on the (a) lower AIC and BIC values in the negative binomial versus Poisson models, and zero-inflated versus non-inflated models; and the (b) likelihood ratio test comparing zero-inflated Poisson and negative binomial models (lr test: x² (1) = 60.76, P < .001) (Long & Freese, 2006). Further, a zero-inflated model is considered particularly appropriate for sexual health research, such as condomless sex (He, Tang, Wang, & Crits-Cristoph, 2014), given that some mechanisms responsible for zero counts in the data differ from the traditional regression model. For example, women who are not engaging in any sexual activity will produce zero counts in a condomless sex outcome, no matter how they vary on other predictors, and are therefore considered to be structurally different (in statistical modeling terms) than women who are at-risk of engaging in condomless sex but do not do so. Outliers (z scores ≥ 3.29; 3 cases at baseline and 2 at follow-up) were re-expressed to the next highest value (Tabachnick & Fidell, 2007).

Results

Feasibility

Figure 1 shows that, of the 591 women screened, 106 (18%) met eligibility criteria. Of these eligible women, approximately two-thirds (n = 68; 64%) agreed to participate in the study. Of the 38 women who declined to participate, most did so due to time restrictions (n = 23; 61% reported they were too busy). Other decliners cited lack of interest in participating in research (n = 11), transportation difficulties (n = 2), lack of childcare (n = 1), or concerns about privacy/confidentiality (n = 1) as their reason for declining.

Figure 1.

CONSORT diagram for study enrollment and procedures

We enrolled 50 of the 68 women who agreed to participate. The 18 women who did not enroll could not be scheduled within 2 weeks of their screening date. Of the 50 enrolled women, two withdrew prior to randomization, leaving an analytic sample of 48 women (described next). After consenting, one woman completed the baseline survey but did not return any contacts thereafter and did not participate in the intervention; nonetheless, her baseline data were included in the analyses as part of the intent-to-treat sample. Two women (one in each condition) were excluded due to highly inconsistent self-reports (e.g., incompatible reports on sexual behavior and alcohol use within the same survey). Therefore, there were 46 women available for follow-up. Forty-two of the 46 women participated in the 3-month follow-up; one woman could not return in person but she was able to complete the follow-up survey remotely.

Participant Characteristics

The intent-to-treat sample comprised 48 young (18 to 29 years; M = 22.67), single women. Twenty-eight identified as White (58%), five as African American (10%), six as multiracial (13%), four as Asian-American (8%), and five as “other” (10%); 16 women (33%) identified as Latina/Hispanic. Forty-six identified as female (96%) and two as “genderqueer/gender nonconforming” (4%); 33 were heterosexual (69%) and 15 were bisexual (31%). Women identified the reason for their visit as: STI or pregnancy testing (62%); birth control (32%); general reproductive health (e.g., pelvic exam) (17%); or general health screening (e.g., routine physical exam; breast exam) (15%) [women could select multiple reasons].

Most women (81%) reported a history of forced or coerced sexual contact or attempted penetrative assault (lifetime), with 35% reporting childhood sexual abuse and 52% reporting a completed unwanted penetrative (i.e., oral, vaginal, or anal) sex event. Many women reported a lifetime history of substance use, including marijuana (98%), hallucinogens (37%), and cocaine/crack (35%) as well as various tobacco products (hookah 78%, combustible cigarettes or cigars 61%, and electronic cigarettes / vaping 37%).

Preliminary Analyses

Table 4 summarizes the participant characteristics at baseline, overall and by group. Preliminary analyses determined whether randomization produced equivalent groups. There were differences between conditions with respect to age (p < .001) and student status (p < .01). These two variables covaried (r_pb = −0.60, p < .001), such that women who were younger were more likely to be full-time students. Therefore, women assigned to the BI condition were younger, on average, and were more likely to be full-time students (Table 4). Given that these variables covaried, subsequent analyses controlled for age only. With respect to the outcome variables, women in the BI condition reported a larger maximum number of alcoholic drinks before sex at baseline (M = 6.10) than women in the control condition (M = 4.21; B = 0.37, IRR = 1.45, 95%CI [1.09, 1.92], p = .01).

Table 4.

Baseline demographic variables in the full sample and by condition

	Full Sample (N = 46)	Control (n = 21)	Brief Intervention (n = 25)	t	χ2	P
Age (M, SD)	22.67 (3.13)	24.33 (2.92)	21.28 (2.61)	3.75		< .001
Race					1.88	0.82a
Caucasian	28 (61%)	12 (57%)	16 (64%)
African American	5 (11%)	3 (14%)	2 (8%)
Asian	4 (9%)	2 (10%)	2 (8%)
Multiracial	5 (11%)	3 (14%)	2 (8%)
Otherb	4 (9%)	1 (5%)	3 (12%)
Ethnicity					0.06	1.00
Non-Hispanic	32 (70%)	15 (71%)	17 (68%)
Hispanic	14 (30%)	6 (29%)	8 (32%)
Gender Identity					0.02	.90
Female	44 (96%)	20 (95%)	24 (96%)
Genderqueer	2 (4%)	1 (5%)	1 (4%)
Sexual Orientation					1.07	.30
Heterosexual	32 (70%)	13 (62%)	19 (76%)
Bisexual	14 (30%)	8 (38%)	6 (24%)
Employment Status					2.08	.35
Unemployed	12 (26%)	4 (19%)	8 (32%)
Part-Time	19 (41%)	8 (38%)	11 (44%)
Full-Time	15 (33%)	9 (43%)	6 (24%)
Relationship Status
Single, never married	46 (100%)	21 (100%)	25 (100%)
Student Status					7.10	< .01c
Non-Student	17 (37%)	9 (43%)	8 (32%)
Part-Time Student	6 (13%)	6 (29%)	0 (0%)
Full-Time Student	23 (50%)	6 (29%)	17 (68%)

^aChi-square comparison of White participants to those who endorsed African American, Asian, Multiracial, or Other identities.

^bAll women who selected “other” for the racial identification question self-identified as Hispanic in a separate text field entry.

^cChi-square comparison of non- and part-time students versus full-time students, given the zero count for part-time students.

Acceptability Analyses

Table 5 displays the summary statistics for the acceptability data.

Table 5.

Acceptability Measures

	Range	Control	Brief Intervention	t	p	Cohen’s d
Intervention session
Evaluation of Health Coach	1–6	5.77 (0.45)	5.86 (0.25)	−0.83	.41	.25
Evaluation of Session	1–6	5.44 (0.57)	5.50 (0.66)	−0.31	.76	.10
Satisfaction with Session	1–5	3.98 (0.96)	4.65 (0.36)	−3.19	.003	.92
Personal relevance	1–5	3.52 (1.17)	4.58 (0.65)	−3.69	.001	1.12
Interest	1–5	3.90 (1.21)	4.30 (0.77)	−1.29	.21	.39
Overall impression	1–5	4.10 (0.85)	4.74 (0.54)	−2.89	.01	.90
Recommend to others	1–5	4.21 (1.08)	4.78 (0.42)	−2.17	.04	.70
Texta
Helpful	1–7	--	4.73 (1.98)
Good vehicle	1–7	--	5.27 (1.93)
Frequency of messages
Wanted more		--	1 (5%)
Just right		--	12 (55%)
Too many		--	9 (41%)
Websitea
Information Content	1–5	--	3.73 (1.06)
Design	1–5	--	3.57 (0.95)
Format	1–5	--	3.78 (1.02)
Ease of Use	1–5	--	3.92 (1.07)
Likely to Return to Website	1–5	--	3.26 (1.29)
Overall Mean Score	1–5	--	3.67 (0.95)

^aOnly brief intervention (BI) participants participated in the text message and website extenders.

Clinic-based interventions.

Ratings of the health coaches and the intervention session were highly favorable for both conditions and did not differ. However, women’s mean satisfaction with the session differed by condition, as indicated by a significant t-test and large effect size of .92. Compared to controls, women in the BI condition reported greater perceived personal relevance of the material; BI women also described their overall impression as more positive and they reported a greater likelihood of recommending the intervention to others.

Women’s reports during the debrief interviews echoed the quantitative results. When asked what aspect of the research experience they liked best, 16 women in the BI condition named the coaching (one woman named the text messages, and the others did not identify a single aspect). Women noted that the coaching session prompted them to think about their sexual behavior in new ways. For example, one woman stated: “Putting a conversation with my partner about monogamy in the context of my health was really interesting, because I think I don’t necessarily otherwise really link that.” Another woman reported: “My favorite part was talking about how drinking alcohol, and STIs, and unprotected sex can come hand in hand, because it’s not something we think about a lot.” In contrast, women in the control condition frequently spoke about the survey being the most impactful part of the study (e.g., “[favorite part] was maybe the survey”). Many women in the control condition also noted that the survey prompted them to reflect on their behaviors. For example, one woman stated, “just being asked the questions made you think about some things,” with another woman noting “the first few questions, like, allowed me to think about my life in ways that I wouldn’t think about on purpose, you know?” Similarly, one woman noted how the survey prompted her to reflect on her number of sexual partners: “I like to be honest with my partners. And when I did [the baseline survey], I was kinda like ‘Oof’”.

Many BI women 12/21 (57%) accurately recalled their change plan goal(s) and made changes to reach their goals. For example, one woman described declining potential sexual partners’ advances as follows: “I think that this whole Women SHARE project has given me the confidence to improve my skills with saying no.” She clarified, “It was so much easier than I ever thought.” Another woman said, “because I talked about it with someone, I was able to refer to that stuff with my goals and everything that we set together.” Further, most (16/21; 76%) would have liked a second intervention session to revisit goals or problem-solve around barriers.

Text messages.

On average, women described the text messages as helpful (M = 4.73 on a 7-point scale) and rated text messages as a good mechanism to deliver health information (M = 5.27 / 7.00). In the debrief interviews, some women described the text messages as ongoing boosters (“It was just a good reminder that would keep popping up on my phone. And like reminding me that I’m doing this thing and trying to make myself better in the next 3 months”).

Approximately half of women (12/22; 55%) perceived the frequency of daily messages to be “just about right,” although 41% (9/22) thought there were “too many text messages.” During the follow-up qualitative interviews, half of the BI women (11/21; 52%) stated that daily text messages were enjoyable (e.g., “I personally loved them”) with four (4/21; 19%) women reporting that they wanted to continue to receive messages after their participation in the study ended. Of the women who reported the messages were too frequent, seven stated that the messages were limited in that they (a) already knew much of this information, (b) wanted more context (e.g., protective behavioral strategies rather than just statistics about STIs), (c) wanted to personalize the topics, and/or (d) wanted to interact with the content they received.

Website.

Most women reported that they visited the website (16/21), with 13 women able to describe aspects of the website they visited and preferred (e.g., blog postings; fact sheets; informational videos). Website analytics corroborated these reports. Website usability statistics were collected in a de-identified manner, so they cannot be directly linked to women’s self-reports. According to website analytics, 29 unique devices accessed the site during the study period, with an average of 1.29 sessions per device. The discrepancy between number of unique visitors according to web analytics (n = 29) and both the number of women in the BI condition (n = 25) and the number of women who self-reported visiting the site (n = 16) is likely due to: (1) women logging into the study site at the end of the BI session with their health coach but not returning or engaging meaningfully; (2) several women who reported they logged in with multiple devices (i.e., tablet, laptop, cell phone); and (3) two women who reported sharing the log-in information with friends.

Most sessions (17/29) consisted only of visiting the login and home page. Therefore, 12/29 women (41%) interacted in some way with the site other than logging into the main page, which is consistent with the 13 women who self-reported visiting particular pages of the website. Regarding their website use, women reported that they did not remember the website (“I think I just forgot about it, honestly”) or they believed the text messages and in-person session contained enough information (“I felt like all the other information I was provided with was sufficient, so I didn’t need to [use the website]”). For example, six women (24%) reported using the website immediately after the intervention session but forgot about it as time went on. Other women reached the website from links in the text messages, prompted by interest in the text information. Some women (n = 6; 24%) reported they sought out information when they had questions or to inform changes they wished to make (e.g., condom negotiation with a partner).

On average, women rated the website as having the information they desired, and that it was easy to use (all subscale scores > the neutral median). Women reported the highest scores for the ease of use and content scales. The return scale, which assessed women’s plans to visit the site in the future and recommend the site to friends, had the lowest score (M = 3.26).

Efficacy Analyses

Table 6 contains summary statistics for the outcome variables at baseline and follow-up, by condition. Tables 7 and 8 summarize the analytic models for alcohol use and sexual behavior outcomes, respectively. Interaction terms, where not significant, are reported in the text, as the tables represent the final, reduced models.

Table 6.

Means (SD) on alcohol and sexual health variables by condition.Median presented for condomless sex due to zero inflated nature.

	Full Sample		Control		Brief Intervention
	Baseline (n = 46)	Follow-up (n = 42)	Baseline (n = 21)	Follow-up (n = 20)	Baseline (n = 25)	Follow-up (n = 22)
Drinks per week	6.26 (5.34)	3.60 (3.14)	6.33 (5.56)	4.00 (3.08)	6.20 (5.26)	3.23 (3.22)
Heavy episodic drinking days	3.35 (3.96)	2.19 (2.72)	3.71 (4.85)	2.00 (2.00)	3.04 (3.09)	2.36 (3.27)
Maximum drinks/day	7.13 (3.34)	5.21 (2.12)	6.48 (2.99)	5.65 (2.70)	7.68 (3.57)	4.82 (1.37)
Alcohol consequences	5.67 (6.31)	2.60 (3.34)	6.47 (6.11)	3.05 (3.19)	5.00 (6.51)	2.15 (3.51)
Male sexual partners (3 months)	2.37 (2.02)	1.48 (0.99)	1.86 (0.79)	1.35 (0.75)	2.80 (2.58)	1.59 (1.18)
Condomless sex (number of unprotected events)	6.00	4.00	6.00	5.00	5.00	2.50
Maximum drinks before sex*	5.18 (2.11)	3.94 (1.63)	4.21 (1.90)	4.00 (1.97)	6.10 (1.92)	3.88 (1.26)

^*p < .05. Conditions differed significantly at baseline.

Table 7.

GEE models for alcohol outcomes

	Drinks per week			Maximum drinks			Number of heavy episodic drinking days			Alcohol consequences
	Wald χ2	B	IRR	Wald χ2	B	IRR	Wald χ2	B	IRR	Wald χ2	B
		(SE)	95% CI		(SE)	95% CI		(SE)	95% CI		SE
Time (ref: baseline)	10.44***	−0.53	0.59	1.97	−0.03	0.88	4.93**	−0.40	0.67	15.06***	−2.62
		(0.16)	0.43, 0.81		(0.15)	0.74, 1.05		(0.18)	0.47, 0.95		0.68
Condition (ref: control)	1.61	−0.24	0.78	0.20	0.33	1.06	3.79*	−0.47	0.31	1.20	−1.32
		(0.19)	0.54, 1.14		(0.17)	0.82, 1.37		(0.24)	0.39, 1.00		1.20
Time x Condition	--	--	--	4.93**	−0.33	0.72	--	--	--	--	--
					(0.15)	0.54, 0.96
Age	2.47	−0.06	0.95	6.84*	0.01	0.96	3.47	−0.12	0.91	0.14	−0.09
		(0.04)	0.88, 1.01		(0.03)	0.94, 0.99		(0.07)	0.80, 1.03		0.21

^*− p ≤ .05.

^**p < .01.

^***p < .001.

Table 8.

Models for sexual health outcomes

	number of partners			Maximum drinks before sex
			IRR			IRR
GEE	Wald χ2	B (SE)	95% CI	Wald χ2	B (SE)	95% CI
Time (ref: baseline)	15.89***	−0.44 (0.11)	0.64	0.03	−0.03 (0.15)	0.97
			0.52, 0.80			0.72, 1.31
Condition (ref: control)	2.11	0.28 (0.19)	1.32	3.75*	0.33 (0.17)	1.39
			0.91, 1.93			1.00, 1.94
Time x Condition	--	--	--	4.09*	−0.40 (0.20)	0.67
						0.44, 0.99
Age	1.17	−0.03 (0.03)	0.97	0.29	0.01 (0.03)	1.01
			0.93, 1.02			0.96, 1.07
	Condomless Sex Acts at Follow-Up
	Zero-Inflated			Negative Binomial
			OR			IRR
	z	B (SE)	95% CI	z	B (SE)	95% CI
Condition	1.45	2.22 (1.58)	9.19	−1.53	−0.50 (0.33)	0.61
			0.41,204.36			0.32, 1.15
Baseline Count of Condomless	−1.921†	−0.58 (0.30)	0.56	−0.77	−0.01 (0.01)	0.99
Sex			0.31, 1.12			0.97, 1.01
Follow-up Count of Total Sex	−1.92*	−0.19 (0.10)	0.83	6.10***	0.05 (0.01)	1.05
Acts			0.68, 1.00			1.03, 1.07
Age	1.28	0.33 (0.26)	1.39	1.86	0.09 (0.05)	1.09
			0.84, 2.32			0.99, 1.20

^† p = .055.

^*p ≤ .05.

^***p < .001.

Drinks per week.

The time-by-condition interaction was not significant (B = −0.19, IRR = 0.83, 95% CI [0.44, 1.59], Wald x² = 0.31, p = .58) but there was a main effect of time such that, regardless of condition, women reported consuming fewer drinks in a typical week at follow-up (M = 3.60) than at baseline (M = 6.26).

Maximum drinks per day.

The time-by-condition interaction was significant (B = −0.33, IRR = 0.72, 95% CI [0.54, 0.96], Wald x² = 4.93, p < .05) indicating that women who received the BI evidenced a greater decrease in maximum drinks at follow-up (7.68 to 4.82, a 2.86 drink decrease) compared to women in the control condition (6.48 to 5.65, a 0.83 drink decrease).

Number of heavy episodic drinking days.

The time-by-condition interaction was not significant (B = 0.27, IRR = 1.31, 95% CI [0.66, 2.60], Wald x² = 0.58, p = .45). There was a main effect of both condition and time (Table 7); thus, women in the BI condition reported fewer heavy episodic drinking days than women in the control condition (Table 6). Regardless of condition, women reported fewer heavy episodic drinking days at follow-up (M = 2.19) than at baseline (M = 3.35).

Alcohol consequences.

The time-by-condition interaction was not significant (B = 0.85, SE = 1.34, Wald x² = 0.40, p = .53). There was a main effect of time (Table 7); women reported fewer alcohol-related consequences at follow-up (M = 2.60) compared to baseline (M = 5.67).

Number of partners.

The time-by-condition interaction was not significant (B = −0.18, IRR = 0.84, 95% CI [0.56, 1.25], Wald x² = 0.76, p = .38). There was a main effect of time (Table 8), indicating that women reported fewer male partners over the follow-up period (M = 1.48) than in the 3 months prior to the study (M = 2.37), regardless of condition.

Condomless sex.

A zero-inflated approach models (1) the probability of being a zero as well as (2) nonzero counts using a negative binomial model. These analyses examined the total count of condomless sex acts, controlling for baseline count of condomless sex acts as well as total count of vaginal/anal sexual acts at follow-up. As expected, women who reported a greater number of sexual encounters also had a greater likelihood of engaging in condomless sex (i.e., negative coefficient in the zero-inflated portion of model) and to a greater degree (count portion) (Table 8). Similarly, there was a trend by which women who engaged in condomless sex at baseline were also more likely to do so at follow-up (zero-inflated portion).

There was no effect of condition on condomless sex in the odds of being a zero or in the count portion of the model (Table 8). However, both effects were in the expected direction, such that women in the BI condition were less likely to have engaged in condomless sex (i.e., they were more likely to be zeroes in the zero-inflated portion) and to do so less often if they did (i.e., smaller IRR in the count portion; see Table 6 for summary statistics). Further, the odds ratio of being a zero was in the large effect size range (Chen, Cohen, & Chen, 2010), indicating that women in the BI condition were less likely to report condomless sex at follow-up compared to women in the control condition. However, this should be interpreted in line with the small number of women in each cell.

Alcohol use before sex.

The time-by-condition interaction for maximum number of drinks before sex was significant (B = −0.40, IRR = 0.67, 95% CI [0.45, 0.99], Wald x² = 4.09, p < .05), such that women in the BI condition reduced their drinking more over the follow-up period (from 6.10 to 3.88 maximum drinks before sex; a 2.22 drink decrease) compared to women in the control condition (4.21 to 4.00; a 0.21 drink decrease).

Discussion

This exploratory clinical trial evaluated the feasibility, acceptability, and efficacy of a BI that integrated alcohol and sexual risk content for young women reporting alcohol misuse and sexual risk behavior. The BI was designed to be practical in the context of a busy clinic setting; therefore, it was structured so that it could be completed in less than one hour; however, the BI was supplemented with two technology extenders (i.e., daily text messages, a website) to increase the “dose” of the intervention and to boost motivation in the natural environment.

The results showed that the study processes and the BI itself were feasible. Approximately 83% of women presenting for care were screened prior to their discharge. Only a minority of women (7%) declined screening. Of those screened, one in five women reported both risky alcohol use and sexual behavior. Two-thirds of the women who screened eligible enrolled in the study. Reasons for declining study participation were related to limited time or lack of interest in being in a research study and not to the BI itself. Overall, these data confirm that a substantial minority of women were engaging in risk behaviors, demonstrating that there is a need for an integrated alcohol and sexual risk reduction intervention, and that most at-risk women wanted to learn more about their alcohol use and sexual behavior.

The acceptability analyses showed a high level of satisfaction with the BI itself. Women found the content relevant and interesting; they rated their health coaches and the coaching session very favorably, and they were enthusiastic about recommending the BI to others. Overall, the results suggest that delivering a BI in the context of a busy clinic is well-accepted by young women reporting at-risk drinking and sexual behavior.

The acceptability analyses, website analytics, and qualitative results for the technology extenders showed that the text messages received more favorable comments and use than the website. This is despite the fact that the website had rich content, including material designed to improve behavioral skills needed to reduce risk. The preference for text messages is consistent with research on young adults’ use of technology. More than half (60%) of all digital media consumption is now spent on mobile apps (e.g., Snapchat) as emerging and young adults spend a decreasing proportion of their time on desktops or browsers in general (comScore, 2017).

Results of the efficacy analyses showed, for alcohol use and sexual behavior, reductions in risk from baseline to the three-month follow-up. Women in both groups reported reductions in the number of drinks per week, the number of heavy episodic drinking days, alcohol consequences, and the number of sexual partners in the follow-up interval. The most parsimonious explanation for the reductions observed in both groups is assessment reactivity. Prior research has shown that completing detailed assessments can reduce alcohol use in clinical (Clifford, Maisto, & Davis, 2007) and non-clinical samples of young adults (Walters, Vader, Harris, & Jouriles, 2009). Assessment reactivity also influences sexual behavior, partly because of increased risk sensitization (Weinhardt, Carey, & Carey, 2000). Furthermore, after completing sexual behavior surveys, women report increased intentions to reduce their risk behaviors (Kalichman, Kelly, & Stevenson, 1997). Thus, the process of reflecting on one’s behavior can prompt risk reduction even in the absence of an intervention. This is a likely mechanism given that most women in the study (62%) were recruited from a clinic where they presented for STI screening or pregnancy tests, which likely increased their receptivity to this information. Women’s reports in the interviews support the assessment reactivity explanation; that is, women reported increased awareness of their risk behaviors as well as associated self-reflection and change.

Reductions observed in control conditions make it harder to detect intervention effects, a well-documented phenomenon in trials of alcohol interventions (Jenkins, McAlaney, & McCambridge, 2009). Nevertheless, despite the reductions observed for the control condition, as well as the limited power inherent in a small sample, we did observe effects for the BI. For two outcomes, we found a significant condition-by-time interaction, reflecting additional risk reduction for women who received the BI. Thus, relative to women in the control condition, women who received the BI reported a larger reduction in the maximum number of drinks per day (−2.86 vs. −0.83 drinks) as well as in the number of drinks before having sex (−2.22 vs. −0.21 drinks); however, it should be noted that the absolute number of drinks in each condition at the follow-up occasion was nearly equivalent.

Finding that the BI impacted alcohol use aligns with previous studies. In a recent systematic review, O’Connor et al. (2018) reported that across 68 studies with a variety of populations and settings (e.g., primary care, college students), brief interventions focusing on alcohol use only were associated with reductions in alcohol use (e.g., decreases in drinks per week, heavy episodic drinking). The magnitude of changes in alcohol use observed in this study are better than those obtained in the one study conducted in a sexual health clinic (i.e., Crawford et al., 2015) and similar to results obtained with other populations and in other settings (e.g., Ingersoll et al., 2005; Lewis et al., 2014). O’Connor et al. (2018) also observed that other health behaviors (which included sexual risk behavior) were rarely assessed or reported, and when they were, there was no evidence of efficacy. Our results with respect to sexual risk behavior are more encouraging and suggest that an integrated alcohol and sexual risk reduction intervention is promising. This finding aligns with research using college student samples, which also suggests that changing alcohol-related sexual behavior requires an integrated intervention that addresses alcohol-related sex risk (cf. Ingersoll et al., 2005; Lewis et al., 2014).

To determine if the effects we observed are enduring, it will be necessary to conduct larger trials with more participants and longer follow-ups. The effects of assessment reactivity are typically fleeting whereas intervention effects need to be sustainable to have public health importance. Thus, research is needed to evaluate the long-term effects of integrated alcohol- and sexual risk reduction interventions. Qualitative findings from this study suggest that technology “boosters,” particularly text messaging, are an acceptable way to extend an intervention into the natural environment and to promote behavior change maintenance. Text messaging is inexpensive, easily adapted to changing circumstances, and potentially interactive.

This exploratory trial had several strengths. First, we targeted both alcohol use and sexual risk behavior with an integrated intervention. Typically, these two behavioral domains are not addressed concurrently nor in an integrated way. Second, we developed the BI based upon empirical precedents and behavioral science theory. Third, we used formative research to optimize the integration of the intervention into the operations of a busy clinic. The BI did not disrupt clinic flow, earning the support of staff and administration. Fourth, we assessed partner type and tailored the intervention to match each woman’s partner-specific risk. Fifth, we recruited at-risk women from a community-based RHFP clinic where clients are more likely to be lower income, have relatively lower literacy skills and educational levels, lack access to routine medical care, and are vulnerable to STIs, sexual violence, and unintended pregnancy (Alam et al., 2015). Thus, the public health benefit is enhanced relative to research conducted with lower risk participants (e.g., college students). Sixth, to increase the dose of the BI, we used technology to extend the BI into women’s natural environment.

The primary weaknesses of this study include its small sample, a single recruitment site, the brief follow-up interval, and the reliance upon self-report. Future trials can improve upon these limitations by recruiting larger samples from multiple sites, following women for longer intervals, and using biomarkers to corroborate efficacy evaluations. Such improvements will also permit investigation of potential mediators or moderators of intervention efficacy.

Compliance with Ethical Standards:

All procedures performed in studies involving human participants were in accordance were in accordance with the ethical standards of The Miriam Hospital’s Institutional Review Board and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. This article does not contain any studies with animals performed by any of the authors.

Informed consent:

Informed consent was obtained from all individual participants included in the study.