Table 1 provides brief descriptive characteristics of the 48 included interventions [30–77] and Table 2 provides a summary of the characteristics across interventions. Among 48 studies, the majority of the studies were conducted in earlier HAART era (1996 to 2003) and later HAART era (2004 to 2012). Forty-three studies (90%) were randomized control trials (RCT). Regionally, most interventions were carried out in the West, followed by the Northeast and South. The fewest interventions were conducted in the Midwest. Not surprisingly, most of the studies were conducted in urban settings, such as Atlanta, Baltimore, Boston, Chicago, Houston, Los Angeles, Miami, Milwaukee, New York City, Philadelphia, San Francisco, and Washington DC, except one study that was conducted by phone in rural areas of 27 states.

The most commonly targeted groups were clinic patients [30, 37–40, 43, 48, 54, 59, 60, 62, 64, 65, 68–70, 74], followed by men who have sex with men (MSM) [33, 34, 47, 52, 54, 61, 64–66, 68, 71, 72, 75] and PLWH who engaged in TRB [31, 35, 36, 49, 52, 56, 60, 61, 64, 65, 75]. Another frequently targeted group was substance-abusing PLWH, including injection drug users [50, 53, 58], general drug users [51, 63, 69], substance-using MSM [52, 71], methamphetamine users [52], cocaine users [57], crack users [73], and alcohol abusers [71]. Fewer studies specifically targeted the following subgroups of PLWH: women [38, 42, 70, 74, 77], African Americans [34, 36, 72, 73], persons with depression or a history of childhood abuse [46, 67, 77], younger age groups (13 to 29 years) [30, 62, 63], older adults (45 years and older) [34, 43, 49], newly HIV-diagnosed persons [32, 55, 68], male prison inmates [41], and persons who were homeless or at risk of homelessness [76].

The majority of the studies (88%) reported the theoretical principles used in designing the interventions. The most commonly used theories included: Social Cognitive Theory [78], Information Motivation and Behavioral Skills (IMB) model [79, 80], Theory of Reasoned Action, Social Action Theory [81], Motivational Interviewing [82], Transtheoretical Model of Stage of Change [83], and Theory of Gender and Power [84]. Half of the interventions were conducted in healthcare settings, such as HIV outpatient clinics, community health centers, hospitals, or methadone treatment clinics. More than half of the interventions were delivered by professionals such as healthcare providers (19%), counselors or health educators (40%). Some were delivered by peers (27%). Two were computer-delivered interventions using interactive video doctors [39, 48]. The majority of the interventions consisted of 3 to 10 sessions (63%) and lasted 1 to 3 months (69%). The median time per session was 90 minutes, ranging from 3–5 minutes to 3 hours.

The most commonly reported outcomes (see Table 3) were TRB (21 studies) and unprotected sex behavior (partner serostatus not reported, 33 studies). About half of the studies that reported these two outcomes showed significant positive intervention effects (12 studies and 16 studies, respectively). There were fewer studies that reported injection drug use or needle sharing behaviors (5 studies). Additional three studies combined with sex and drug behaviors in a risk index. About half of these studies showed significant positive intervention effects. All the sex and drug use behaviors were based on self-report. Regarding biologic outcomes, only one study out of six studies that measured STD (lab confirmed or doctor’s diagnosis) showed a significant positive intervention effect.

Of the 48 included studies, 24 interventions (50%) had low risk of bias. Among these, 14 were EBIs that also showed significant positive intervention effects [31, 36, 37, 39, 40, 44, 45, 54, 59, 62, 63, 67, 74, 75] and the other 10 interventions were Rigorous non-EBIs [42, 43, 55, 58, 61, 64, 69, 71, 73, 76]. Twenty-four interventions had high risk of bias, including 13 Positive non-EBIs [30, 33, 35, 41, 48–50, 52, 53, 57, 60, 70, 77] and 11 Other non-EBIs [32, 34, 38, 46, 47, 51, 56, 65, 66, 68, 72]. Eighteen Positive and Other non-EBIs (75%) had multiple sources of bias. The common sources of bias included: analytic sample sizes less than 40 per arm, less than a 60% retention rate of study participants per arm, greater than 10% differential attrition between arms, substantial missing data, not conducting intent-to-treat analysis, or significant negative findings. Another way of looking at the breakdown of 34 non-EBIs is that 21 interventions (61%) did not find any significant positive intervention effects.

Table 2 and Table 3 show comparisons between EBIs and rigorous non-EBIs. The two groups were similar in terms of target populations (i.e., MSM, those who engaged in HIV transmission risk), sample characteristics (i.e., race/ethnicity, gender, age), reporting of power analysis and theories, use of ACASI for data collection, several intervention components (i.e., self-efficacy, skills building, serostatus disclosure, social support, personalized risk reduction plan, personal responsibility, normative influence, intimate partner violence), and type of outcomes reported (e.g., TRB, unprotected sex behavior, injection drug use or needle sharing).

Despite the similarities, EBIs and Rigorous non-EBIs differed in a few ways. More EBIs than Rigorous non-EBIs were evaluated in the earlier HAART era (1996–2003), carried out at multiple study locations, targeted clinic patients, delivered to individuals, conducted in healthcare settings, used standard of care or non-HIV attention controls (defined as receiving non-HIV intervention such as general health promotion that matched length and doses of HIV-intervention) as comparison groups. There was a higher percentage of EBIs than non-EBIs that addressed the following intervention components: discussing HIV risk-reduction, promoting motivation for behavioral change, addressing misperception about HIV, reducing negative affect such as depression or anxiety, enhancing medication adherence, and conducting risk screening to guide prevention messages. In contrast, more Rigorous non-EBIs were conducted in the later-ART era, targeted substance users, used HIV demand controls as comparisons (defined as participants in the comparison group are aware of the intervention they received were intended to change their sex or drug use risk), and had 3 to 12 intervention sessions over a period of 1 to 3 months

There were several similarities between EBIs and Positive non-EBIs. Comparable percentages of EBIs and positive non-EBIs were observed on the following: intervention level (i.e., individual, group, couple), intervention intensity and time span, and some intervention components (e.g., building skills and self-efficacy, conducting risk screening to guide personalized prevention messages, working with participants on personalized risk-reduction plans, emphasizing personal responsibility to take care of one’s and partner’s health, and addressing medication adherence issues). However, Positive non-EBIs were more likely to be small-scale studies conducted with subgroups of PLWH (e.g., rural areas, depressed MSM, meth-using MSM, injection drug users, substance users, women) compared to EBIs. Positive non-EBIs were less often conducted as multi-site studies and in health care settings or community-based establishments in contrast to EBIs. Positive non-EBIs were also less likely than EBIs to use non-HIV attention controls, use ACASI for data collection, report theories, and report and show positive intervention effects on TRB.

While the comparisons between EBIs and two non-EBIs groups (i.e., Rigorous and Positive) can inform research gaps and future intervention development, a closer examination of the EBIs can provide helpful direction for providers in healthcare and non-healthcare settings in selecting model programs suitable for their target populations. More than half of EBIs targeted HIV clinic patients [37, 39, 40, 54, 59, 62, 74], but none targeted newly diagnosed PLWH. Six EBIs targeted specific subgroups of PLWH: MSM [54, 75], heterosexual African American discordant couples [36], substance using youth and young adults [63], women [74], and PLWH with a history of childhood sexual abuse [67]. All 14 EBIs had greater than 50% ethnic minority participants (range: 53% to 100%), seven of which included a majority of African Americans. The key components of EBIs included addressing HIV risk reduction behaviors, motivation for behavioral change, misconception about HIV, and issues related to mental health, medication adherence, and HIV transmission risk behavior.

A variety of intervention delivery methods, ranging from brief prevention messages delivered during regular HIV care visits to intensive multi-session interventions over several weeks or months, were shown to be successful in reducing TRB as well as unprotected sex with any sex partners. Three EBIs were brief interventions. In one intervention, the healthcare provider delivered 3- to 5-minute prevention messages that focused on self-protection, partner protection, and disclosure. Posters and patient education brochures in the clinics reinforced provider-delivered prevention messages [59]. Another intervention used clinicians to deliver the 5- to 10-minute tailored prevention message based on risk screening information that patients provided [37]. In the third intervention, patients completed a computer-based risk assessment and then viewed a 24-minute video clip in which an actor-portrayed physician delivered risk-reduction messages tailored to the patient’s unique risks. At the conclusion of the video section, patients received an educational worksheet for self-reflection, harm reduction tips, and local resources and clinicians received a cueing sheet that summarized patient’s risk profile and suggested counseling statements [39].

Intensive behavioral risk-reduction interventions (defined as multiple sessions over weeks and months with a median of 90 minutes per session) in healthcare [40, 54, 62, 74] and non-healthcare settings [31, 36, 44, 45, 63, 67, 75] can also lead to reductions in risky sexual behaviors among PLWH. In 11 EBIs, peer-educators [44, 54, 74, 75] or health educators/counselors [31, 36, 40, 45, 62, 63, 67] provided multi-session interventions to individual PLWH [31, 40, 54, 63], small groups of adult PLWH [44, 45, 62, 67, 74, 75], or discordant couples with HIV [36] with varied demographic characteristics. Interactive sessions focused on many topics such as coping with an HIV diagnosis, addressing serostatus disclosure, building condom use skills, negotiating safe sex behaviors, avoiding risky drug use, or medication adherence. Participants of all 11 EBIs were significantly less likely to report TRB or unprotected sex with any partners at some point within 3 to 12 months after interventions.

The healthcare setting affords a great opportunity to integrate behavioral prevention with routine medical care and address behavior change over time. Consistent with a previous meta-analysis [8], we found two EBIs that showed brief prevention counseling messages (e.g., 3 to 10 minutes) delivered by healthcare providers during routine HIV care visits can lead to significant reductions in HIV transmission risk among PLWH. This brief provider-delivered risk-reduction approach has been implemented and evaluated in two large-scale demonstration projects [86, 87]. Both studies showed the feasibility of conducting brief provider-delivered risk-reduction interventions in busy healthcare settings and the effectiveness of this approach in reducing HIV transmission risk behaviors of PLWH. Based on the body of evidence, the brief provider-delivered risk-reduction intervention during HIV patient’s routine care visits has been recommended and currently promoted to be standard of HIV clinic care by CDC (i.e., Prevention IS Care [88]).

The importance of clinic provider’s role in facilitating healthier behaviors of patients is not new. However, evidence suggests that providers are not consistently talking to patients about safer sex, injection drug use, and HIV prevention methods. Approximately 23% to 29% of HIV patients reported that their providers have never talked to them about safer sex [89, 90]. The data on provider-patient communication in the most recent HIV primary care visit showed that 65% of HIV-seropositive injection drug users reported having discussed HIV prevention with their provider [91] and 53% of Ryan White CARE Act patients from 9 states reported having discussed safer sex and HIV prevention methods with their providers [92]. These percentages are similar to the percentages reported by healthcare providers [93]. Studies also showed that providers were more likely to provide prevention counseling to new patients rather than established patients [93, 94]. These findings highlight considerable room to increase the delivery of brief prevention counseling by providers during routine HIV care visits, especially among returning patients.

Published studies in the literature indicate several common barriers to providing risk screening and risk reduction interventions in healthcare setting, including lack of time, competing priorities, limited staffing, providers’ lack of risk-reduction counseling skills, discomfort with talking about risk behaviors, and the belief that interventions will not change behavior [92, 93, 95–99]. However, evidence suggests that training on brief risk screening methods requires minimal time and training on brief risk-reduction interventions enhances the provider’s comfort, skill, efficiency, and motivation [60, 94, 96, 99]. Providers are more likely to engage in risk-reduction prevention counseling if other providers in the same clinics are also providing prevention counseling. Additionally, providers who agree that risk-reduction prevention is part of the clinic’s mission are more likely to conduct counseling to HIV patients [94]. Innovative approaches are needed to prepare and support providers for delivering more consistent risk-reduction interventions to their HIV patients. Some approaches to consider include integrating behavioral prevention into the clinic mission, providing training to enhance the provider’s ability, motivation, and comfort to deliver brief preventions, reimbursing counseling time, and educating medical students about HIV prevention.

Our systematic review also found that longer and multi-session HIV interventions are efficacious in changing HIV transmission risk behaviors of PLWH. The feasibility of delivering interventions with multiple sessions over time is not clear, especially in busy clinic facilities. However, these interventions are not without merit because some PLWH require additional help to address multiple interconnected factors (e.g., substance use, depression, childhood sexual abuse, interpersonal and partner dynamics) underlying their risk behaviors. Several intensive EBIs are successful in reducing risk behaviors among PLWH at high risk of transmitting HIV. Referrals for evidence-based, multi-session risk-reduction interventions for PLWH who report high levels of risk or continue risk behaviors may be a beneficial component of comprehensive HIV prevention efforts at the clinic/agency, state, and federal levels. Creating directories of local clinics or agencies that offer evidence-based, intensive risk-reduction interventions, facilitating the collaboration between clinic and community providers, and establishing policies and procedures regarding patient’s referrals to intensive interventions can also be helpful to ensure the services are in place as needed [100].

While our review identified several EBIs for healthcare and non-healthcare settings, the dissemination of the interventions remains limited as these settings often do not have the human or financial resources to devote to interventions. One potential solution is the use of new technologies. Computer-based interventions are shown to be efficacious in increasing condom use and reducing sexual activity, numbers of sexual partners, and incident STD [101]. The advantages include greater intervention fidelity, lower delivery costs, and greater flexibility in dissemination channels such as in person, by mail, on the webs, through cell phones [101]. Several computer-based interventions that address HIV risk behavior as well as mediation adherence and other issues (e.g., retention, treatment readiness) for PLWH are on the way (e.g., CDC and NIH funded comprehensive prevention with positive project, CDC funded computer-based Interactive Screening and Counseling Tool, CBISCT). More research is needed to explore the best way to incorporate new technologies to deliver HIV behavioral interventions that address the prevention needs of PLWH.

Several identified differences and similarities between EBIs and the two groups of non-EBIs (i.e., Rigorous and Positive) can inform the design and testing of future behavioral interventions for PLWH. EBIs tended to target HIV clinic patients whereas more of the two non-EBIs groups targeted specific high-risk populations (e.g., substance-using MSM, IDUs, substance abusers, sexual actively older adults, inmates, homeless) or understudied populations (e.g., rural residents, newly diagnosed).

When comparing EBIs and Rigorous non-EBIs, there were many more similarities than differences in study design, implementation and analysis, outcome measures, and intervention components. One unique difference is that EBIs were more likely to use standard of care or non-HIV attention control and less likely to use HIV demand controls. For HIV-related comparison groups, using variations of the interventions as comparison groups may greatly reduce the ability to detect intervention effects [102]. Using a standardized comparison arm that the HIV prevention field could agree upon as a prevention standard can facilitate comparing intervention effects across studies.

Unsurprisingly, when comparing EBIs and Positive non-EBIs, there were obvious differences in the sources of bias. More positive non-EBIs than EBIs were small sample size studies intended as pilot studies to test the feasibility of the intervention implementation. Several of non-EBIs also suffered from substantial attrition, differential retention, and missing data issues. While positive non-EBIs showed at least one significant positive finding on sex or injection drug use outcomes, few reported and found significant positive findings on TRB. Positive non-EBIs are good candidates for further evaluation and they should be evaluated with more rigorous methods that reduce the risk of bias in study design, implementation and analysis and that measure HIV transmission risk behavior.

The findings of our review must be viewed within the context of the limitations of the available evidence. First, although the majority of the studies were RCTs, many were un-blinded and relied on self-reported sexual behavior, which may open to social desirability bias [103]. Given that blinded trials are not feasible in HIV behavioral prevention research, future intervention trials should consider complementing self-reported behavioral measures with biologic outcomes such as STD to assess intervention efficacy [104]. There were very few studies in the current literature that measured both behavioral and biologic outcomes. Second, while the majority of EBIs demonstrated significant positive intervention effects on reducing unprotected sex with HIV-negative and serostatus unknown partners, it is unclear how the observed behavioral changes may translate into averted new infections. Many factors such as individual’s viral load level, type of sex acts, and present of other STD may affect the probability of new HIV transmission. Although the complexity of the multiple influencing factors that could affect HIV transmission potential makes it impossible to estimate the number of new infections averted by the interventions reviewed here, our findings showed that some interventions are more successful than others in promoting positive behavior changes that are important factors in HIV transmission risk.

In addition, there are several limitations specific to this review that merit consideration when interpreting the findings. First, we classified all the U.S.-based interventions for PLWH into EBIs vs. non-EBIs based on the risk of bias and evidence of positive intervention effects. While this classification approach is intended to identify model programs, it differs from meta-analytic approaches, which provide the overall estimate of the intervention effects. Second, although we contacted primary investigators to confirm our evaluation of the risk of bias and intervention effects, the coding of study and intervention characteristics are based on published reports that may not provide complete information about the intervention. Third, our review relied on the published literature and our findings might be susceptible to for publication bias. Fourth, while we observed some patterns that may explain differences among EBIs and non-EBIs, there are multiple factors that may contribute to an intervention’s lack of evidence and it is difficult to disentangle a specific reason or combination of reasons.