Predictors of Retention in an Alcohol and Risky Sex Prevention Program for Homeless Young Adults

Abstract

Homeless young adults are at risk for alcohol and other drug (AOD) use and risky sexual behavior. Interventions are needed to help these young people reduce their risky behavior, but this population is often difficult to engage and retain in services. We offered a four-session AOD and risky sex reduction program to 100 participants and examined if retention in the program was predicted by a number of factors: demographics, homelessness severity, other service use, AOD behaviors, mental health symptoms, sexual risk behaviors, and readiness to change AOD and condom use. Nearly half (48%) of participants completed all sessions. In bivariate analyses, participants were significantly less likely to be retained in the program if they had slept outdoors in the past month, engaged in more alcohol and marijuana use, experienced more alcohol-related consequences, and received the program in an urban drop-in center (as opposed to a drop-in center near the beach). When controlling for all significant bivariate relationships, only sleeping outdoors and receipt of the program in the urban setting predicted fewer sessions completed. The most endorsed reasons for program non-completion were being too busy to attend and inconvenient day/time of the program. Findings can help outreach staff and researchers better prepare methods to engage higher risk homeless youth and retain them in services. Finding unique ways to help youth overcome barriers related to location of services appears especially necessary, perhaps by bringing services to youth where they temporarily reside or offering meaningful incentives for program attendance.

Homelessness among adolescents and young adults is a national problem. Although the exact figures are not known, it was reported that in a single night in 2016 in the United States, 35,686 unaccompanied youth under age 25 were homeless (Henry, Watt, Rosenthal, & Shivji, 2016). These unaccompanied youth lack stable and safe housing and are living on their own, or with other youth their age, away from their families. Life on the street can often lead to problematic risky behaviors such as alcohol and other drug (AOD) use and risky sexual behavior. For example, about 95% of homeless adolescents and young adults report lifetime substance use and nearly three-quarters report current substance use (Wenzel, Tucker, Golinelli, Green, & Zhou, 2010). Illicit drug use such as needle use (and sharing), cocaine use, and prescription drug misuse are also common, with upwards of one-quarter to one-half of youth in both street and service seeking samples reporting illicit drug use (Al-Tayyib, Rice, Rhoades, & Riggs, 2014; Huba et al., 2000; Kipke, Montgomery, Simon, & Iverson, 1997; Nyamathi, Hudson, Greengold, & Leake, 2012; Rice, Milburn, Rotheram-Borus, Mallett, & Rosenthal, 2005; Tucker et al., 2012b). Most young homeless individuals are also embedded in social networks that include regular drug users, making avoidance of use difficult in the context of peer influence (Green, de la Haye, Tucker, & Golinelli, 2013; Koopman, Rosario, & Rotheramborus, 1994; Wenzel et al., 2010).

Risky sexual behavior is also common among homeless adolescents and young adults (De Rosa, Montgomery, Hyde, Iverson, & Kipke, 2001; Kennedy, Tucker, Green, Golinelli, & Ewing, 2012; Kipke et al., 1997; Nyamathi et al., 2012; Tucker et al., 2012a; Tucker et al., 2012b) and they therefore evidence relatively high rates of HIV, hepatitis, and other sexually transmitted diseases (STDs), with prevalence rates of these conditions ranging from 2% to upwards of 40% across samples (Medlow, Klineberg, & Steinbeck, 2014; Rotheram-Borus et al., 2003). Rates of unintended pregnancy are also substantially higher among homeless than housed young adults (Begun, 2015; Greene & Ringwalt, 1998; Tucker et al., 2012b). Given their high rates of AOD use and sexual risk behavior, unaccompanied homeless adolescents and young adults represent a population that could greatly benefit from risk reduction programs.

In order to be comprehensive while addressing substance use, risky sex, and the interactions between the two risky behaviors, programs targeting these behaviors often consist of multiple sessions (e.g., Carmona, Slesnick, Guo, & Letcher, 2014; Rotheram-Borus et al., 2003; Slesnick, Prestopnik, Meyers, & Glassman, 2007), which require that individuals return for subsequent sessions after initiation in order to receive a full dose of care. Yet, despite their need for risk reduction services, it is often difficult to engage homeless youth in services. For example, once they initiate evidence-based programs that address AOD use and risky sex, retention of homeless adolescents and young adults in such programs is challenging. This is due to a variety of factors, including fluctuations in temporary residences, AOD use and mental health problems, no permanent or mailing address to reach them for reminders, or disinterest in subject matter perceived to be irrelevant or punitive (Becker, Berry, Orr, & Perlman, 2014; Lankenau et al., 2008). Yet completion of more AOD and sexual risk sessions has been linked to better long-term outcomes (Carmona et al., 2014; McKay, 2005; Slesnick, Kang, & Aukward, 2008), emphasizing the importance of retaining homeless youth in multisession interventions and programs.

Researchers have begun formative work to examine predictors of retention in AOD and risky sexual behavior programs, though very little is known about the characteristics of the homeless adolescents and young adults who may be less likely to be retained in risk reduction programs. Homelessness itself has been shown to be a predictor of dropout from mental health care among young people (Baruch, Vrouva, & Fearon, 2009), though other work has failed to find a link between homelessness and either initiation of or retention in substance use care (Lee et al., 2012). Predictors of substance use program retention may depend on type of care received; for example, one study found that homeless male adolescents were more likely to drop out from family therapy, non-African American adolescents were more likely to drop out from the Community Reinforcement Approach, and those with fewer runaway episodes were more likely to disengage from motivational enhancement therapy (Slesnick, Erdem, Collins, Bantchevska, & Katafiasz, 2011). Other work has shown that AOD use is negatively correlated with adherence to treatment regimens for homeless adults involved in HIV medication treatment (Friedman et al., 2009). For homeless adolescents involved in AOD treatment, having a history of attempted suicide or a history of physical or sexual abuse was associated with using more services (Slesnick et al., 2008). A better understanding is needed of what types of homeless youth may be more or less likely to complete risk reduction programs in order to better engage these individuals at the onset of services to encourage retention in programs and therefore reduce their overall risk.

The Present Study

For the present study, we explored factors that predicted retention in an AOD and risky sexual behavior prevention program offered at two drop-in centers in Los Angeles County (Tucker, D’Amico, Ewing, Miles, & Pedersen, 2017). Among all major cities, Los Angeles has the largest population of unaccompanied homeless adolescents and young adults (Henry et al., 2016), making it an ideal place for outreach efforts related to AOD use and risky sex. For example, over 54% of a sample of 200 homeless young adults in Los Angeles met criteria for a substance use disorder (Bender, Brown, Thompson, Ferguson, & Langenderfer, 2015). Past three month risky sex behavior among a sample of 419 homeless adolescents and young adults in Los Angeles also indicated that nearly two-thirds of sexual events were unprotected and about 50% of the sample reported any AOD use prior to sex (Martino et al., 2011).

As part of a larger study, we enrolled 200 young adults in an AOD and risky sexual behavior program using targeted recruitment methods, which proved to be successful in terms of enrollment and retention for follow-up assessments (Garvey, Pedersen, D’Amico, Ewing, & Tucker, 2018). Yet, once these young people are engaged in programs, it is not well known what factors may prevent them from finishing such programs. Although outreach specifically targeted toward increasing attendance at drop-in centers increases service utilization (Slesnick et al., 2016), more research is needed to determine what may increase or decrease retention in multisession programming for homeless youth at drop-in centers (see Pedersen et al., 2016). Thus, we examined whether retention in multisession programming was associated with a number of factors across several areas, including demographic characteristics, homelessness severity, other service use, AOD behaviors, mental health symptoms, sexual risk behaviors, and readiness to change AOD use and condom use (behaviors specifically targeted in the program). Insights from findings can help outreach staff and researchers think more creatively to both engage and retain youth in services.

Methods

Participants

Participants were part of a larger intervention study testing the efficacy of a four-session AOD use and sexual risk reduction program, AWARE, designed for homeless young adults aged 18–25 (Tucker et al., 2017). During visits to the two drop-in center sites, research staff screened a total of 214 young adults for eligibility, of which two were ineligible, one refused participation, and 11 were identified as repeat participants. Thus, the final sample for the intervention study was 100 intervention (AWARE) participants and 100 control participants. For the purposes of this report, we examined data from AWARE participants only. These 100 participants had a mean age of 21.75 (SD=1.86) years and were 69% male. About three-quarters (74%) reported heterosexual orientation, and 28% identified as non-Hispanic White, with 23% African American, 26% Hispanic, and 23% multiracial/other.

Design and Procedures

The AWARE program is an innovative group-based Motivational Interviewing (MI; (Miller & Rollnick, 2012) intervention for homeless 18–25 year olds, with targeted outcomes of reducing AOD use behavior and reducing risky sexual behaviors (e.g., sex without a condom). Eligible participants were between the ages of 18 and 25 at one of two drop-in centers in Los Angeles County (one in a beach location and one in an urban/downtown location) recruited between January of 2014 and March of 2015. Participants were recruited by advertising the study at the drop-in centers and soliciting volunteers. At each visit to the drop-in center, we used a sign-up sheet to indicate interest in participating. We used a group randomization procedure where the two drop-in centers served alternatively as intervention or control sites, with each drop-in center having two intervention phases and two control phases over four 16-week phases. We advertised the study at the drop-in centers in advance and, at each recruitment visit, participants were recruited by soliciting volunteers and using a sign-up sheet. We then enrolled youth into the study until our recruitment goal for the visit was met, with the provision that we would randomly sample youth from the sign-up sheet if more youth expressed interest than we could accommodate on that visit. Individuals were then screened for eligibility, and eligible participants were asked to provide written consent and complete the baseline survey (for which they received $20).

Detail about each of four AWARE sessions can be found in our other work (Tucker et al., 2017). In general, each AWARE session focused on teaching cognitive behavioral skills to reduce substance use and increase condom use, providing personalized feedback about risky behaviors, and sharing educational information about STD transmission and effects of AOD use on development. Because each AWARE session was designed to be free standing, and participants could attend the sessions in any order, those in the intervention condition attended their first AWARE session immediately after completing the baseline survey. Participants were not required to attend all four weekly AWARE sessions in a row; if they missed a particular session, they could attend the next time that it was offered. AWARE was structured this way due to the unique population in the unique setting for care; that is, drop-in centers are designed to facilitate “dropping in” around the unpredictable schedules of homeless adolescents and young adults. Allowing youth several chances to attend each of the four sessions over a 16 week period was a format informed by drop-in center staff to help facilitate attendance and encourage retention across multiple sessions. We have written about this flexible format more extensively in our other work (Garvey et al., 2018).

AWARE groups lasted approximately 45 minutes each and contained an average of 5 (SD = 2.86, minimum = 2 maximum = 13) participants per session. Participants were offered complimentary snacks and condoms at each session and were given $5 for each session they attended. If they attended all 4 sessions, they received an additional $15. Three months after the 16 week period when groups were conducted at the drop-in center, participants were contacted to complete a follow-up survey, for which they received $30. The follow-up rate among AWARE participants was 95%. All study materials and procedures were approved by the institution’s Internal Review Board. Responses on both surveys were protected by a Certificate of Confidentiality from the National Institutes of Health. A more thorough description of the recruitment methods can be found in our other work (Garvey et al., 2018).

Measures

Demographics, Homelessness Severity, and Service Usage

The baseline survey contained demographic items of age, biological sex, sexual orientation, and race/ethnicity. Participants also completed items about current homelessness severity, which included length of their most recent period of homelessness (in months) and an item about which places they spent the night in the past 30 days (e.g., outdoors, the street, or park; a car, van, or camper; an emergency shelter; a motel or hotel). We coded this latter item for analyses as “slept outdoors” in the past 30 days versus not. Participants indicated their use of various services (other than the AWARE program) in the past 3 months, including job training services, mental health counseling, AOD counseling, and medical/dental services. It was also noted which of the two drop-in centers the participant was recruited and received the AWARE sessions from during the study (beach location versus urban location).

AOD Behaviors

AOD use was assessed with two items regarding frequency of alcohol use and frequency of marijuana use in the past three months, with response options ranging from 0 (never) to 7 (everyday). Participants also completed the 5-item Global Appraisal of Individual Needs – Short Screener (GAIN-SS; Dennis, Chan, & Funk, 2006)) to assess AOD severity related to past year alcohol use disorder or cannabis use disorder (α = 0.73). Negative consequences resulting from alcohol use were assessed with the 24-item Brief Young Adult Alcohol Consequences Questionnaire (B-YAACQ; Kahler, Hustad, Barnett, Strong, & Borsari, 2008; Kahler, Strong, & Read, 2005)). A summary score for the B-YAACQ was created by summing the number of negative consequences participants had experienced in the past 3 months (α = 0.90).

Mental Health Symptoms

To assess depression symptoms, participants completed the 2-item Patient Health Questionnaire (PHQ-2; Kroenke, Spitzer, & Williams, 2003) with items of feeling down, depressed, or hopeless and finding little interest or pleasure in doing things. Items were rated based on the past two weeks from 0 (not at all) to 4 (nearly every day) (α = 0.74). To assess generalized anxiety symptoms, participants completed the 7-item Generalized Anxiety Disorder – 7 item scale (GAD-7; Spitzer, Kroenke, Williams, & Lowe, 2006)), with items such as “feeling nervous, anxious, or on edge” and “not being able to stop or control worrying.” Items were rated based on the past two weeks from 0 (not at all) to 4 (nearly every day) (α = 0.92).

Sexual Risk Factors

Participants indicated their total number of sexual partners in the past 3 months, which ranged from 0 to 15. This variable was log transformed in order to reduce the positive skew of the distribution and bring the distribution of the variable closer to normal. In addition, as an indicator of risky sexual behavior, we created a measure of having unprotected sex with a casual partner in the past month. This was derived from multiple survey items: how many casual partners they had vaginal or anal sex with in the past 30 days, and for those with at least one casual partner, the percentage of times that they used a condom during vaginal or anal sex with a casual partner in the past 30 days. Those who did not report a casual partner or who always used condoms with casual partners were coded as 0, whereas those who reported condom use less than 100% of the time with a casual partner were coded as 1.

Readiness to Change Targeted AOD Use and Condom Use

Participants completed a series of four readiness to change rulers in which they rated how ready they were to cut down or stop their alcohol, marijuana, and other drug use, and how ready they were to increase their condom use. Rulers ranged from 0 (not at all) to 10 (extremely) and were modified from prior work (Boudreaux et al., 2012).

Follow-up Survey

The follow-up survey assessed AOD and risky sexual behavior outcomes pertinent to the larger study (Tucker et al., 2017). For the purposes of this study, we examined follow-up data regarding participants’ self-reported reasons for failing to attend all AWARE sessions. These items were asked of the 52 AWARE participants who did not complete all four sessions, with 47 of these participants responding to items and indicating yes/no to the reasons presented in Table 2. They had the option to fill in “other reasons” as well. These open- ended responses were recoded to fit into established categories or combined to create new categories (see Table 2).

Table 2.

Reasons for not completing the program among participants who did not complete all 4 AWARE sessions (n = 47) and mean (standard deviation) number of sessions of completed

	Number and percent endorsing reason
Why didn’t you attend all 4 sessions
I was too busy to attend the programa	n = 16; 34%
The day/time of the program was not convenient for meb	n = 10; 21%
Not in town when group was heldc	n = 6; 12%
The location of the program was not convenient for me	n = 5; 11%
Other (i.e., personal conflict about group bias, got into a fight at session, did not know the date of the session)c	n = 5; 11%
The $5 incentive for attending each session was not enough	n = 3; 6%
I felt the issues/topics covered in AWARE did not apply to me and my life	n = 3; 6%
Otherwise unavailable (i.e., too sick, too high on drugs, in jail)c	n = 3; 6%
My partner wasn’t attending the program and I didn’t want to go without him/her	n = 2; 4%
I thought the program was boringd	n = 1; 2%
I didn’t want others in the group to know personal things about me	n = 1; 2%
I was concerned that center staff would view me differently if I shared something personal	n = 1; 2%
I didn’t think the information provided in AWARE was helpful	n = 0; 0%

^atwo open-ended “other” responses added to this category (“working two jobs,” “I had stuff to do”),

^bone open-ended “other” response added to this category (“bad timing”),

^cauthor-created category based on open-ended “other” responses,

^done open-ended “other” response added to this category (“they talk about what most youth are aware about”)

Analytic Plan

Analyses consisted of descriptive statistics and regression analyses. We examined the following factors as correlates of program retention: demographic characteristics (age, biological sex, sexual orientation, race/ethnicity), homelessness severity (slept outdoors in the past month, period of most recent homeless), other service use (site where AWARE was received, overall service use at drop-in centers, job training service use, mental health counseling use, AOD counseling use, and medical/dental service use), AOD behaviors and mental health, (past year drug use severity, past month alcohol frequency, past month marijuana frequency, past month alcohol consequences, depressive symptoms, anxiety symptoms), sexual risk (number of partners, condom use with casual partners), and readiness to change AOD use and use of condoms as correlates of retention. Each of these factors was examined individually in separate bivariate models predicting number of treatment sessions. Because very few participants reported completion of three sessions only (n = 4), we compared those who attended one session (n = 21) to those who attended two or three sessions (n = 31) to those who attended all four sessions (n = 48). We treated our outcome variable, number of sessions, as a three-category outcome variable and therefore conducted ordinal logistic regression models. Once it was determined which factors independently associated with retention in the program, all the factors statistically significant at p < 0.05 in bivariate models were entered into a single multivariate model predicting the ordinal number of sessions completed.

Results

Predicting Retention in the AWARE Program

Overall, 48% of participants completed all four sessions, 31% completed two or three sessions, and 21% completed only one session of AWARE. None of the demographic factors (i.e., age, biological sex, sexual orientation, race/ethnicity) were associated with retention in the program (Table 1). There was a main effect for the homeless severity factor of sleeping outdoors in the past 30 days (estimate = −1.14, SE = 0.48, p = 0.017), with those who slept outdoors less likely to be retained in the program. However, number of months homeless was not significantly associated with retention. For service-level factors, there was a main effect for location of the drop-in center (estimate = 0.95, SE = 0.39, p = 0.014), such that those who received AWARE at the drop-in center in the beach location were more likely to attend AWARE sessions after their first session compared to participants at the urban drop-in center. Use of specific drop-in center services was not significantly associated with retention in AWARE.

Table 1.

Regression analyses of each independent factor predicting number of sessions (bivariate models) and multivariate models with significant bivariate indicators

	Bivariate models				Multivariate model
	Estimate	SE	t-value	p-value	Estimate	SE	t-value	p-value
Demographics
Age	0.00	0.10	0.04	0.970
Male	−0.39	0.41	−0.94	0.350
Straight	−0.34	0.44	−0.77	0.442
Race/Ethnicity
Hispanic	−0.20	0.54	−0.37	0.709
Multi/Other	−0.08	0.56	−0.14	0.891
White	−0.40	0.53	−0.75	0.452
Black
Homelessness Severity
Slept Outdoors	−1.14	0.48	−2.39	0.017	−1.31	0.54	−2.43	0.015
Number of Months Homeless	0.00	0.01	0.23	0.815
Service Use
Beach location of program site1	0.95	0.39	2.45	0.014	1.41	0.45	3.11	0.002
Drop-in Center Use	−0.20	0.53	−0.38	0.704
Job Training Use	0.19	0.38	0.51	0.612
Alcohol or Drug Counseling Use	−0.26	0.42	−0.62	0.534
Mental Health Counseling Use	0.24	0.41	0.60	0.549
Medical or Dental Care Use	0.05	0.38	0.12	0.901
AOD Behaviors
Alcohol Frequency, Past month	−0.06	0.03	−2.38	0.017	0.00	0.03	−0.11	0.911
Marijuana Frequency, Past month	−0.04	0.02	−2.30	0.022	−0.03	0.02	−1.79	0.073
Drug Severity, Past 3 months	−0.01	0.01	−0.89	0.372
Consequences (BYAACQ)	−0.07	0.03	−2.09	0.036	−0.07	0.04	−1.84	0.066
Mental Health Symptoms
PHQ-2	−0.05	0.11	−0.45	0.654
GAD-7	0.02	0.03	0.56	0.573
Sexual Behavior
Total Number of Partners (log+1)	−0.18	0.26	−0.69	0.493
Unprotected sex with casual partner, Past 3 months	−0.36	0.42	−0.87	0.384
Readiness Rulers
Alcohol	0.09	0.05	1.82	0.068
Marijuana	0.09	0.05	1.82	0.069
Other Drug	0.00	0.04	−0.10	0.920
Condom	−0.02	0.05	−0.51	0.607

¹Beach location coded as 1, urban setting coded as 0. P-values < .05 are bolded.

Regarding AOD use, there were main effects for past month alcohol use frequency (estimate = −0.06, SE = 0.03, p = 0.017) and marijuana use frequency (estimate = −0.04, SE = 0.02, p = 0.022), such that those who engaged in more alcohol and marijuana use were less likely to attend AWARE sessions after their first session. There was also a main effect for alcohol-related consequences (estimate = −0.07, SE = 0.03, p = 0.036), such that those who attended only one AWARE session reported more alcohol-related consequences (M = 8.95, SD = 5.86) compared to those who attended 2 or 3 sessions (M = 6.23, SD = 5.33) and those who attended all 4 sessions (M = 5.58, SD = 5.67). However, program retention was not significantly associated with drug use severity, symptoms of depression or anxiety, or any of the four readiness to change rulers. Program retention was also not associated with number of sexual partners or with frequency of unprotected sex with a casual partner.

In the multivariate model, we entered all factors significant at the bivariate level (i.e., slept outdoors, location of drop-in center, alcohol use frequency, marijuana use frequency, alcohol consequences) and predicted number of sessions completed. Table 1 contains the estimates, standard errors, t-values, and significance levels for these factors. After controlling for other factors, we found that slept outdoors (estimate = −1.31, SE = 0.54, p = 0.015) and urban location of the program (estimate = 1.41, SE = 0.45, p = 0.002) were associated with less retention in the program. Past month alcohol frequency, past month marijuana frequency, and past month alcohol consequences were no longer associated with attending fewer sessions in the full model.

Reasons for Not Completing AWARE Program

Table 2 details reasons participants endorsed for not completing all 4 sessions of the AWARE program. The most endorsed reason was “I was too busy to attend the program,” mentioned by about 1 in 3 non-completers. About 1 in 5 non-completers endorsed “The day/time of the program was not convenient for me.” About 1 in 10 non-completers endorsed reasons related to being out of town, inconvenient location of the group, or “other reasons,” where one participant stated “personal conflict about group bias;” another stated they had gotten into a fight at the last group, and another stated they did not know the date of the next session. Other reasons endorsed less frequently related to feeling the incentive was not worth the effort, topics covered were not relevant, participant was unavailable (e.g., sick), and not wanting to attend a session if one’s partner decided not to attend that week. One participant each reported that the program was boring, not wanting others to know personal issues, and concern about staff viewing them differently if they shared personal information. No participant indicated that the information presented in the program was unhelpful.

Discussion

The present study adds to the sparse literature assessing retention in risk reduction programs aimed at reducing AOD use and risky sex among the in-need but often difficult to engage population of homeless young adults. We explored whether a number of factors relating to demographic characteristics, homelessness severity, service usage, AOD behaviors, mental health, sexual risk behavior, and readiness to change AOD and sexual risk behaviors assessed prior to the first AWARE session predicted retention in the program at later sessions. This information is important for providers and researchers seeking to engage homeless youth in multisession programs as it offers insight into what types of youth may need to be targeted more directly in retention efforts.

One important finding is that failure to attend all four sessions of the AWARE program was not significantly associated with age, gender, sexual orientation, or race/ethnicity. This suggests that in addition to attracting a demographically diverse group of young adults (72% non-White, 31% female, 26% reporting sexual orientation other than heterosexual; see also Tucker et al., 2017), we were able to retain young people in the program over time, as no particular demographic was more or less likely to drop out from the multisession program. This is promising for future adoption of AWARE and other programs discussing AOD use and sexual behaviors, as even though the program addressed sensitive topics such as sexual behavior, there was no evidence that certain types of young adults (e.g., women, sexual minorities) were less likely to return.

Failure to attend all four AWARE sessions was also not significantly associated with mental health symptom severity (i.e., depression, anxiety) or use of other drop-in center services, suggesting that we were not just retaining those who were not experiencing any mental health distress or those who were highly motivated to seek services. Along these lines, retention did not vary as a function of readiness to change the risk behaviors targeted by the intervention. As the AWARE intervention was conducted in the style of MI, this suggests that participants did not stop attending due to limited readiness to change their behavior; thus, the intervention was able to engage those at all stages of change, highlighting the relevance of the program content even though some might be in the precontemplation versus action stage (Prochaska & Velicer, 1997).

Findings also identified several factors that may increase the likelihood of homeless young adults failing to attend all sessions of a risk reduction program. One important finding is that those who used alcohol and marijuana more frequently, and experienced more alcohol- related problems, were less likely to complete the program; however these significant bivariate associations were not retained in multivariate analyses. Rather, after sleeping outdoors in the prior 30 days and location of the drop-in center where the program was delivered were included in the model, substance use no longer associated with retention. In addition, retention was not a function of risky sexual behavior reported, which was an additional focus of the group intervention. Thus, whereas those who were riskier in terms of AOD use were less likely to complete the program, this was not the case for those who were riskier in terms of sexual behavior.

The two factors that remained significant in the multivariate model were sleeping outdoors and location of where AWARE was delivered. Participants who slept outdoors were less likely to return for subsequent sessions of the program bivariate analyses and in multivariate analyses after controlling for substance use frequency, alcohol consequences, and location of the drop-in center where the program was received. Those who sleep outdoors – in parks, alleyways, and so forth – tend to have a higher risk profile compared to other homeless young people. For example, they are more likely to report engaging in risky sexual behavior (Rice et al., 2005), including sex trade (Tucker et al., 2012b). They also have higher rates of AOD use (Gomez, Thompson, & Barczyk, 2010), which could help to explain why the bivariate effects for substance use no longer remained significant after controlling for this factor. They are in particular need of risk reduction programs, especially relevant given the overlap between AOD use severity and sleeping outdoors as an indicator of homelessness severity. Yet, it may be particularly difficult to retain these young people in services, and more intensive outreach and engagement efforts may be needed. One method, if practical, could be to either reduce the length of time between sessions (e.g., AWARE sessions were one week apart but sessions could be offered twice a week over a 2-week period) or condense multiple brief sessions into a lengthier one or two-session format to make it as convenient as possible for those at risk of not returning.

The location where participants in our study received the AWARE program was also important, such that those at the beach drop-in center were more likely to return for subsequent sessions compared to those at the urban drop-in center. This finding was consistent across the bivariate model and the multivariate model controlling for substance use frequency, consequences, and sleeping outdoors in the past 30 days. Although prior work has shown that homeless adolescents and young adults in different parts of Los Angeles County have distinct demographic profiles (Golinelli, Tucker, Ryan, & Wenzel, 2014), this is unlikely to account for site differences given that demographic factors were not significantly associated with program drop-out. Further, it should be noted that the same facilitator conducted all sessions at both the urban and beach sites and there were no differences in main outcomes of the intervention by site (Tucker et al., 2017). The higher completion rate at the beach drop-in center may have been due to its more limited hours of operation, and fewer competing agencies serving homeless adolescents and young adults in the area, compared to the urban drop-in center. These factors may have made it more likely that individuals would already be at the beach drop-in center when the program was being offered, thus making program attendance more salient and convenient for these youth. It may also be the case that those at the beach site were more likely to attend the program with friends. Many homeless young adults may desire to meet up with friends at drop-in centers and therefore peer involvement may have played a role in program attendance. Future work should evaluate how peers can be used to facilitate initiation of and retention in risk reduction programs at drop-in centers.

In order to understand why drop out from the program may have occurred, we asked AWARE participants who did not attend all four sessions to report their reasons for not completing the program. The most endorsed reasons were due to being too busy or due to the group being held at an inconvenient time. Logistically, we had to schedule the groups around the drop-in centers’ schedules and one of the drop-in centers was only open 2 days per week for a limited number of hours per day; both drop-in centers closed prior to nighttime. Participants may have had other obligations or commitments and were not able to complete groups. Of note, very few participants reported failure to attend group sessions due to boredom or irrelevant content, fears staff would view them differently if they shared, or concerns related to who else was present during groups (e.g., conflict with other group members, wanting a partner to attend as well, not wanting peers to know personal information), and no participants reported that the groups were unhelpful. Thus, it appears that the group content was relevant, and the use of MI strategies during the sessions helped keep young people engage in the offered services.

Limitations

Findings from this study should be interpreted with consideration of limitations. First, although results can help providers, researchers, and outreach workers better understand what factors influence retention in risk reduction programs, these may not be generalizable to other programs that differ in length or content. A clear example of this is the finding for location of the drop-in center where AWARE was received. Future work outside of the Los Angeles area should examine how location of the programs influences attendance. In addition, data collected were all self-report with the exception of attendance records observed by the research staff. Also, as noted, we did not have a measure of traveler status, which may have been a reason why this typically transient group did not return for subsequent sessions. To limit the impact of transience, youth needed to report that they believed they would be in the local area for at least the next month when enrolling. When we contacted participants in the larger study for the follow-up assessments three months later, we found that of the 181 of the 200 intervention and control participants we located, most of them (80%) were still local in the Los Angeles area, yet 20% were located outside the area, including participants in 12 states other than California (Garvey et al., 2018). However, we do not know if the participants moved out of town during the period when they were eligible to complete the program or in the three month period after.

Conclusions

Overall, our findings provide some insight for services providers, outreach workers, and researchers at what may affect homeless youths’ retention in risk reduction programming. Overall, most factors did not affect retention, which is encouraging. As one might expect, greater use of alcohol and marijuana, experiencing alcohol consequences, and homelessness severity were all associated with a lower likelihood of returning for subsequent sessions. Thus, services need to be geared toward helping these higher risk youth both access services and become engaged in programming. We also found that urban location was associated with a lower likelihood of attending multiple sessions, which may have been due to more options for services being offered in the downtown location. In light of research suggesting that group attendance is predictive of positive long-term outcomes (Carmona et al., 2014; McKay, 2005; Slesnick et al., 2008), is it important to increase our efforts to engage and retain homeless adolescents and young adults in multisession interventions.