Meta-Analysis of Single-Session Behavioral Interventions to Prevent Sexually Transmitted Infections: Implications for Bundling Prevention Packages

Abstract

Evidence-based, single-session behavioral interventions are urgently needed for preventing the spread of HIV and other sexually transmitted infections (STIs).

To estimate the efficacy of single-session, behavioral interventions for STI prevention, we collected data from 29 single-session interventions (20 studies; n = 52 465) with an STI outcome. Infection with an STI was 35% less likely (odds ratio = 0.65; 95% confidence interval = 0.55–0.77) among intervention group participants than among control group participants. Single-session interventions offer considerable benefits in terms of disease prevention and create minimal burden for both the patient and the provider.

Brief and effective STI prevention interventions are a valuable tool and can be readily adapted to bolster the benefits of biomedical technologies focusing on the prevention of HIV and other STIs.

INTERVENTIONS TARGETING prevention of HIV and other sexually transmitted infections (STIs) during the course of routine clinical services must be succinct to be effective. Interventions intended to reduce sexual risk behaviors and HIV and other STIs have been tested in a variety of clinical and community settings. Although behavioral interventions have demonstrated significant reductions in risk behaviors and have offered evidence of disease prevention, many consist of multiple steps and sessions,^1–5 thus placing a considerable burden on patients and requiring substantial resources.⁴

Of particular concern is the feasibility of implementing multiple-session behavioral interventions in conjunction with currently available health care services. These services continue to face budget reductions that lead to staff shortages and limited means for retaining patients throughout the course of an extended intervention.⁶ Limited resources can render multiple-session interventions unusable or force service providers to substantially modify these interventions.

In addition to the need for brief behavioral interventions in the public health sector, there is a growing demand for feasible behavioral interventions that can be used in combination with biomedical prevention technologies. It is well recognized that no single prevention strategy, including behavioral interventions, male circumcision, preexposure and postexposure prophylaxis, vaccines, and vaginal or anal microbicides, will be completely effective in protecting individuals against infection with HIV and other STIs.^7–14 Furthermore, the effectiveness of biomedical prevention technologies can be undermined by changes in risk behaviors, such as risk compensation.¹⁵

Single-session behavioral interventions can potentially add value to the protective effects of biomedical interventions. There is growing recognition of the need for bundling multiple prevention strategies to gain cumulative effects.^16,17 Behavioral risk reduction interventions can play a critical role in comprehensive programs designed to prevent the spread of HIV and other STIs, particularly when they are designed to fit within current health care services.¹⁸

We conducted a meta-analysis to examine whether single-session risk reduction interventions targeting HIV and other STIs have positive effects on disease outcomes. We focused on STI outcomes because they are clinically meaningful indicators of intervention efficacy. Moreover, we chose single-session interventions because they are most likely to be successfully incorporated into existing services and meta-analyses have not, to date, focused on their effects. There are now sufficient numbers of STI trials with outcome data available to determine whether single-session interventions can lead to disease reductions relative to a standard of care. We also investigated moderators of STI outcomes to identify characteristics of single-session interventions that result in a reduced prevalence of disease. Finally, in a subset of studies that provided behavioral data related to sexual risk taking, we conducted an additional meta-analysis to determine whether single-session interventions improve condom use.

METHODS

We searched for studies (through May 2011) in the MEDLINE (PubMed), PsycINFO, CINAHL, ERIC, and Proquest electronic databases; all international subdatabases in the World Health Organization’s Global Health Library (LILACS, SEARO, EMRO, WPRO, WHOLIS, and AFRO); and the Syntheses of HIV/AIDS Risk Reduction Project’s database and document depository of interventions related to HIV and other STIs. We also searched the reference sections of obtained articles (databases included gray literature). No language or date restrictions were applied. We crossed the following key terms in our search: intervention, behavior, STI, STD, AIDS, HIV, brief, single session, one session, education, program, and counseling. Search terms were truncated to increase sensitivity. Unpublished papers (e.g., dissertations) were included to avoid the file-drawer effect (i.e., stronger effects reported in published than in unpublished studies).¹⁹

We included studies in the sample if they satisfied the following criteria: the intervention consisted of a single session, the study reported at least 1 STI outcome, and the study reported a control arm (further details are provided in Figure 1). We included nonrandomized controlled trials only when their study designs were consistent with approaches set forth by Cochrane review procedures for nonrandomized designs.²⁰ Two independent reviewers conducted literature searches. Search results were compared and discrepancies were addressed. Then 3 independent reviewers evaluated all literature results to determine studies to include in the analysis. In this second step, the 3 independent reviewers overlapped search results and, again, any discrepancies between reviewers were addressed. We excluded interventions solely focused on standard HIV testing and counseling, because these interventions have been reviewed and analyzed.^21,22 Search results yielded no studies focused exclusively on HIV-positive individuals.

FIGURE 1—

Selection process for inclusion of studies in the meta-analysis.

Note: STI = sexually transmitted infection. For reports of studies excluded, most studies met multiple exclusion criteria, and therefore, k are not reported separately for each exclusion criteria. A list of excluded studies is available upon request.

We coded session content using descriptions in the published articles as well as in manuals and session outlines when available. Common intervention components included didactic education, personalized feedback, communication skill building, safe sex discussions, eroticizing safe sex, activities designed to alter perceived social norms, and condom skill training offered through role-playing risk scenarios.

We calculated individual effect sizes in the form of odds ratios for STIs. For this outcome, we used the most distal time point after the intervention—in some instances, as long as 2 years—to serve as the index of STI acquisition. Therefore, we examined the most conservative data points. The presence of STIs was established via medical records, surveillance, or self-reports. The data used to generate odds ratios were entered such that values less than 1 indicated a reduction in STI prevalence for the intervention group relative to the control group. Consistent with meta-analytic conventions, we treated each intervention as an individual study in our analyses.²³ Variables were created that represented either intervention or overall study characteristics to prevent double counting. Asymmetries suggestive of publication bias were analyzed through 3 different strategies: trim and fill, Begg’s strategy, and Egger’s test.^24–26 We used Stata 11 with macros for meta-analysis in conducting our analyses.^23,27

A random-effects model with maximum-likelihood variance estimation was used to obtain average STI effect sizes. Effect size homogeneity values (Q and I²) were examined.²⁸ We conducted sensitivity analyses to detect any possible outliers affecting the results and analyzed study features as possible effect modifiers. Moderator variables were entered into a series of weighted least squares bivariate regression models; fixed-effect assumptions were followed for moderator analyses.²⁹ In a subset of studies that provided the relevant behavioral data, we examined condom use as a secondary endpoint. For this analysis, we calculated individual effect sizes—that is, standardized mean differences (d values³⁰)—for condom use outcomes and used a random-effects model with maximum-likelihood variance estimation. Most interventions reported only 1 STI and 1 condom use outcome; however, in the case of studies reporting multiple outcomes, we calculated individual effect sizes and then averaged these calculations.

Two independent raters coded each study for sample characteristics and risks, such as ethnicity, gender, and age; specific design and measurement features, such as length of session, methodological quality (measures based on those of Jadad et al.³¹; see the appendix, available as a supplement to this article at http://www.ajph.org), and STI and behavioral outcomes; and format and content of control and intervention conditions. Interrater reliability (Cohen’s κ) for categorical variables was calculated as 0.90.³² For continuous variables, we calculated the Spearman–Brown correlation value (r = 0.98).³³

RESULTS

In total, 52 465 participants from 29 single-session interventions (20 intervention trials^34–53) were included in our review (Table 1). Demographic characteristics of participants varied across interventions (12 interventions involved female participants, 4 involved male participants, and 13 included both male and female participants), and studies included both adolescents (k = 6 [k refers to number of interventions]) and adults (k = 23). Some studies focused on participants of varying races (k = 18), whereas others focused specifically on African Americans (k = 8), Whites (k = 1), Latinos (k = 1), and Asians (k = 1). Women constituted 37% of the total study sample, and percentages of participants by race/ethnicity were as follows: Whites, 36%; African Americans, 29%; Hispanics, 26%; Asians, 1%; and those of other racial/ethnic backgrounds, 8%. Participants’ average age was 29.5 years. A majority of the studies were conducted in the United States. The remainder were conducted in Singapore,³⁴ the United Kingdom,⁴¹ Puerto Rico,⁴⁶ Mexico,⁴⁹ and Malawi.⁵³

TABLE 1—

Single-Session Behavioral Interventions Included in the Meta-Analysis

Study	Participants	Study Sample and Setting	Study Conditions/Design	Intervention Length in Minutes	Type of STI and Data Source	Main Intervention and Control Components
Archibald et al.34	442 female sex workers; 100% Asian	Brothel-based sex workers in Singapore	I: group; C: WLC	I: 180; C: 0	Gonorrhea; medical record	I: ED, SB; C: SC
Boekeloo et al.35	219 adolescents; 52% men; 19% White, 64% Black, 4% Hispanic, 13% other	Patients scheduled to see a physician at 1 of 5 participating sites in Washington, DC, area	I: audiotape; C: SC	I: NA; C: NA	Multiple STIs; self-report	I: ED; C: SC
Cohen et al.36	903 adults; 61% men; 5% White, 72% Black, 21% Hispanic, 3% Asian, 4% other	Patients in waiting rooms at 5 STI clinics in Los Angeles, CA	I1: individual; 12: individual; I3: individual; C: SC	I1: 15–20; I2: 15–20; I3: NA; C: NA	Multiple STIs; medical record	I1: CU, SB; I2: CU, ESS, SN; I3: CU; C: SC
Cohen et al.37	426 adults; 71% men; ∼90% Black	Patients at STI clinic in Los Angeles, CA	I: group; C: SC	I: NA C: NA	Multiple STIs; medical record	I: video, CU, RP, SB, SN; C: SC
Crosby et al.38	266 adult men; 100% Black	STI clinic patients in southern US city	I: individual; C: SC	I: 45–50; C: 5	Multiple STIs; medical record	I: SB, MI, CU; C: CU
Gollub et al.39	1591 adult women; 4% White, 91% Black, 3% Hispanic, 1% Asian, 1% other	Patients attending STI clinic in Philadelphia, PA	I: group; C: group	I1: 15–30; C: 15–30	Multiple STIs; medical record	I1: multiple sexual protection options; C: male or female condoms
Grimley and Hook40	456 patients; 44% men; 89% Black, 9% White, 2% other	Patients attending STI clinic in Birmingham, AL	I: computer delivered; C: computer delivered	I: 15; C: 15	Gonorrhea, chlamydia; medical record	I1: tailored CU; C: health risk assessment
James et al.41	492 adults; 51% men; 100% White	STI clinic patients in UK	I: individual; C: SC	I: 20; C: 0	Multiple STIs; medical record	I: SB, CU, SS, SN; C: SC
Jemmott et al.42	682 female adolescents; 68% Black, 32% Latina	Sexually experienced patients attending adolescent medicine clinic in Philadelphia, PA	I1: group; I2: group; C: group	I1: 250; I2: 250; C: 250	Gonorrhea, chlamydia, trichomoniasis; medical record	I1: SB; I2: ED; C: general health
Jemmott et al.43	564 adult women; 100% Black	Sexually experienced patients attending women’s health clinic in Newark, NJ	I1: individual; I2: individual; I3: group; I4: group; C: group	I1: 20; I2: 20; I3: 200; I4: 200; C: 200	Gonorrhea, chlamydia, trichomoniasis; medical record	I1: CU, SB, RP; I2: ED; I3: CU, SB, RP; I4: ED; C: general health
Kalichman et al.44	612 adults; 69% men; 9% White, 85% Black, 3% Hispanic, 3% other	Patients attending STI clinic in Milwaukee, WI	I1: individual; I2: individual; I3: individual; C: individual	I1: 90; I2: 90; I3: 90; C: 90	Multiple STIs; medical record	I1: ED, MI; I2: ED, SB; I3: ED, MI, SB; C: ED
Mansfield et al.45	90 adolescents; 7% men; 5% White, 82% Black, 7% Hispanic, 6% other	STI patients at clinic in children’s hospital in northeastern US	I: individual; C: individual	I: 20; C: 10	Multiple STIs; self-report	I: SC + CU, RP; C: SC
Neumann et al.46	3336 adults; 49% men; 1% White, 40% Black, 51% Hispanic, 8% other	Patients attending STI clinics in New York City and San Juan, Puerto Rico	I: group; C: SC	I: 70; C: 20	Multiple STIs; STI surveillance	I: ED, CU, SB; C: SC
O’Donnell et al.47	2004 adult men; 38% Hispanic, 62% Black	Patients attending STI clinic in South Bronx, NY	I1: video; I2: video + group; C: SC	I1: NA; I2: NA; C: NA	Multiple STIs; STI surveillance	I1: ED, CU, SB; I2: ED, CU; C: SC
Orr et al.48	209 female adolescents; 55% Black	Sexually active patients with positive C trachomatis test results attending family planning and STI clinic in US	I: individual; C: individual	I: 10–20; C: 10–20	Chlamydia; medical record	I: ED, CU, SB; C: SC
Patterson et al.49	924 adult female sex workers; 100% Mexican	Patients with reported high-risk behaviors attending private clinic in Tijuana and Ciudad Juarez, Mexico	I: individual; C: individual	I: 35; C: 35	Multiple STIs; medical record	I: ED, MI, CU; C: ED
Pedlow50	100 adult women; 76% White, 7% Black, 6% Hispanic, 5% other	Patients attending STI clinic	I: individual; C: individual	I: 90; C: 45	Chlamydia, gonorrhea; medical record	I: ED, MI, SB; C: ED
Smith et al.51	205 female adolescents; 10% White, 73% Black, 18% Hispanic	Patients diagnosed with STI at health clinic in US	I: group; C: SC	I: 37; C: 0	Multiple STIs; medical record	I: MI, CU; C: SC
Warner et al.52	38 635 adults; 70% men; 46% White, 18% Black, 25% Hispanic, 11% other	Patients in waiting room at STI clinic in Denver, CO; Long Beach, CA; or San Francisco, CA	I: video; C: SC	I: 23; C: NA	Multiple STIs; medical record/STI surveillance	I: ED, CU, SB; C: SC
Wynendaele et al.53	309 adults; 71% men; 100% African	STI clinic patients in Malawi	I: individual; C: SC	I: NA; C: 0	Multiple STIs; medical record	I: ED, CU, SB; C: SC

Note. C = control; CU = condom use; ED = education; ESS = eroticizing safer sex; I = intervention; MI = motivational interviewing; NA = not available/applicable; RP = role play; SB = skill building; SC = standard care; SN = social norms; SS = safer sex; STI = sexually transmitted infection; WLC = wait list control.

All trials provided biological outcomes; most reported an aggregate measure of multiple STIs that included HIV (k = 19). Other trials reported specifically on Neisseria gonorrhea, Chlamydia trachomatis, or trichomoniasis (k = 10). STI outcomes were gathered through medical records (chart abstraction and laboratory results; k = 23), disease surveillance systems (k = 3), self-reported data (k = 2), and a combination of medical records and surveillance (k = 1). There were no asymmetries in effect sizes (Begg’s test: z = 0.39, P = .68; Egger’s test, t = −2.31, P = .03). The trim-and-fill analysis did not identify any added or omitted studies that were necessary to normalize the distribution, nor did it suggest that any bias was present.

Sensitivity analyses revealed no study variations significantly affecting the results. Specifically, we conducted 4 additional sets of analyses to test the effects of each of the following studies on overall STI effect sizes: the study with the largest sample size,⁵² the study with the largest effect size,⁵³ studies with self-reported data,^35,45 and studies with nonrandomized controlled designs^34,50,53 (1 study involved a controlled cohort study design, and 2 involved a controlled before-and-after study design with attention to matching). Overall, STI effect sizes remained significant even with the removal of these studies.

Summary of Intervention Characteristics

The reviewed interventions varied considerably in their design, session duration, and components. Although most of the intervention trials used 2-arm randomized designs, some tested multiple interventions against control conditions. Intervention formats included one-on-one counseling conducted face-to-face, computer-delivered counseling, small group workshops, and videotapes or DVDs. Single-session interventions ranged from 15 minutes⁴⁰ to 250 minutes⁴² in duration, averaging 79 minutes. The average length of time between intervention and follow-up was 58 weeks.

Although most of the interventions integrated multiple active components, their brief duration typically required emphasizing a particular component as a major feature of the intervention. The most frequently used intervention elements were educational and skill building strategies. Twenty-one interventions reported these components as major themes in the intervention.^{34,36–38,41–43,47,48,52,53} In a smaller number of interventions (k = 4),^38,44,49,51 motivational interviewing was the general framework used. Trials often included a risk education counseling session as a control condition, although many trials compared the experimental intervention with treatment as usual.

Intervention Effects on Sexually Transmitted Infections

On the whole, interventions succeeded in reducing STI incidence. The weighted mean risk reduction, expressed as an odds ratio, was 0.65 (95% confidence interval [CI] = 0.55, 0.77; Table 1, Figure 2). Although effect sizes exhibited heterogeneity (I² = 70; 95% CI = 59.42, 80.68; Q = 99.99), there were no trials for which the control group exhibited a significant reduction in STIs relative to the intervention group. Furthermore, 28 of the 29 control groups were considered active, meaning that control group participants received some form of risk reduction counseling. Thus, reductions in disease incidence resulting from single-session interventions occurred even in the context of relatively stringent controls.

FIGURE 2—

Forest plot of sexually transmitted infection incidence effect sizes (odds ratios), ordered by magnitude.

Note: CI = confidence interval; ES = effect size. The figure presents data as odds ratios and corresponding confidence intervals for each intervention. Weights are from random effects analysis. CIs represent the relative weight of the intervention on estimates. Larger confidence intervals correspond to less relative weight and smaller samples size while smaller confidence intervals correspond to greater relative weight and larger sample sizes. Effect sizes < 1 are indicative of fewer infections in the intervention group relative to the control group.

Of particular note are the STI reduction outcomes observed among female sex workers in Mexico,⁴⁹ adolescents in the northeastern United States,^42,43 and STI patients in Malawi.⁵³ The studies involving these participants demonstrated feasibility under varying scenarios and risk groups. In addition, Warner et al.,⁵² whose study involved the largest number of participants, found that a 23-minute video intervention led to a reduction in STIs for intervention participants more so than for standard care participants. This study demonstrates the efficacy of an intervention engendering minimal burden yet resulting in clinically meaningful outcomes.

Moderating Factors Related to Intervention Efficacy

Our analyses examined how various intervention components and sample characteristics might affect STI effect size results (Table 2). Interventions achieved greater efficacy when they were conducted with non-White participants, they were conducted with exclusively African American participants, they were of longer duration, they were evaluated at intervals nearer to the completion of the intervention, they were compared with wait list and relevant content (vs standard-care) control groups, and they were conducted at the individual and group levels (vs media delivery). In no instances was the STI incidence lower in the control group than in the intervention group. Finally, age, gender, risk group, publication year, methodological quality, and time-matched intervention and control groups did not significantly moderate overall STI effect sizes.

TABLE 2—

Sexually Transmitted Infection (STI) Incidence Effect Sizes as a Function of Study Characteristics

Sample or Study Feature	ORa (95% CI)	B or Multiple Rb
Sample characteristics
Mean age, y (k = 28)		−0.04
13	0.82 (0.59, 0.90)
36	0.77 (0.66, 0.91)
Composition of sample: White participants (k = 29)		0.32**
None White	0.68 (0.62, 0.76)
All White	1.05 (0.85, 1.28)
Composition of sample: Black participants (k = 29)		−0.24*
None Black	0.82 (0.75, 0.91)
All Black	0.68 (0.59, 0.78)
Baseline sample composition by gender (k = 29)		−0.17
All men	0.83 (0.75, 0.93)
All women	0.76 (0.59, 0.81)
Age group		0.06
Adults (k = 23)	0.77 (0.72, 0.83)
Adolescents (k = 6)	0.85 (0.61, 1.19)
Study characteristics
Publication year range (k = 29)		0.17
1992	0.69 (0.59, 0.80)
2011	0.82 (0.75, 0.90)
Methodological quality score range (k = 29)		0.16
5	0.65 (0.52, 0.82)
16	0.87 (0.74, 1.03)
Total implementation time range, min (k = 25)		−0.33**
14	0.82 (0.76, 0.89)
250	0.64 (0.50, 0.80)
Control and intervention time matched		0.10
Yes (k = 11)	0.70 (0.58, 0.85)
No (k = 18)	0.78 (0.73, 0.84)
Time range between session completion and evaluation, wk (k = 29)		0.27*
8	0.66 (0.57, 0.75)
104	0.90 (0.79, 1.03)
Type of control group		0.37**
Wait list (k = 1)	0.36 (0.23, 0.56)
Standard care (k = 13)	0.80 (0.74, 0.86)
Relevant content (k = 15)	0.71 (0.59, 0.85)
Implementation level		0.33**
Individual (k = 15)	0.66 (0.57, 0.77)
Group (k = 8)	0.64 (0.52, 0.79)
Media (k = 6)	0.83 (0.76, 0.90)

Note. CI = confidence interval; OR = odds ratio. Lower ORs represent greater STI reductions among intervention participants than control participants. The lowest and highest values are used in presenting data for continuous variables.

^aValue under fixed effects assumption.

^bValues for variables with more than 2 categories are multiple Rs; otherwise, values are standardized parameter estimates.

*P < .05; **P < .01.

Intervention Effects on Condom Use Outcomes

Overall, single-session behavioral interventions demonstrated a pattern of positive effects on sexual risk, including reductions in number of unprotected sex acts and increases in condom use (k = 20; Figure 3). Efficacy (expressed as a d value) was 0.22 (95% CI = 0.06, 0.37; I² = 82; 95% CI = 73.98, 88.19; Q = 108.38). This effect size can be interpreted as a small but significant effect of improved condom use among intervention group participants relative to control group participants. Heterogeneity was present; however, no study exhibited a significant reversal such that the treatment group exhibited less condom use than the control group.

FIGURE 3—

Forest plot of condom use effect sizes (d values), ordered by magnitude.

Note. CI = confidence interval; ES = effect size. The figure presents data as standard means differences d and corresponding confidence intervals for each intervention. Weights are from random effects analysis. CIs represent the relative weight of the intervention on estimates. Larger confidence intervals correspond to less relative weight and smaller sample sizes while smaller confidence intervals correspond to greater relative weight and larger sample sizes. Larger effect sizes are indicative of greater condom use among intervention participants than control group participants.

Most studies^{38,40,42–44,48,49,53} demonstrated reductions in sexual risk taking among intervention participants relative to controls. In all of the remaining studies,^{34,35,39,41,45} sexual risk reduction rates at follow-up were similar between groups; that is, participants in all arms of the trials reported less sexual risk taking. Although their study was not included in the condom use meta-analysis, Neumann et al.⁴⁶ reported that intervention participants were more likely than control participants to acquire condoms (by redeeming study-provided condom coupons at local stores) after the completion of the intervention (P < .05).

DISCUSSION

This meta-analysis is the first to our knowledge to focus on single-session behavioral interventions designed to reduce STI incidence. Our findings demonstrate that single-session interventions can have a substantial impact on clinically meaningful outcomes across a wide array of sample and intervention characteristics. The documented STI reduction among intervention group participants rivals the effects observed with biomedical technologies targeting disease prevention.^8–10,54 Investing in single-session interventions and delivering them during routine health care visits is practical and efficient. Doing so offers potential cost savings when taking into consideration the expense of treating patients infected with STIs. STI reductions are also consistent with improvements observed in condom use. This result suggests that interventions targeting reductions in sexual risk taking have an impact on clinically meaningful outcomes.

A number of our findings warrant further investigation. We observed that multiple moderators affected the efficacy of interventions with respect to preventing STIs. The strength of the interventions varied according to the demographic characteristics of the participants and suggested that interventions involving members of ethnic minority groups were most effective in reducing STI prevalence. With respect to long-term findings, we did observe a decrease in efficacy as time between intervention and follow-up increased. Behavioral and biomedical interventions generally require adherence to regimens or plans, and their effects are prone to dilution over time. It is possible that single-session interventions delivered during routine care will result in patients being exposed to their content repeatedly. This practice may, in effect, act as a booster and improve long-term efficacy.

Bundling Multiple HIV Prevention Packages

Single-session interventions can help resolve the issues of partial efficacy, risk compensation, and nonadherence associated with new and emerging biomedical HIV prevention technologies.^8,9 Mathematical modeling studies consistently show that combinations of HIV prevention interventions are required to reverse HIV epidemics.^55,56 Models that evaluate HIV prevention interventions demonstrate that even relatively small increases in transmission risk behaviors can reduce the protective benefits of lowered risk of transmission.^56,57 For example, it is now established that male circumcision reduces HIV transmission risks by as much as 55%,⁵⁸ microbicides by 39%,⁹ and preexposure prophylaxis by 44%.⁸ Single-session interventions could boost the disease reduction rates observed with these biomedical forms of prevention. Furthermore, in preexposure prophylaxis HIV prevention trials, nonadherence has been documented as a critical factor affecting efficacy.¹⁰ In summary, single-session behavioral interventions have established efficacy among at-risk populations, can directly address risk taking, and can potentially increase the beneficial effects of biomedical HIV prevention technologies.

Behavioral Interventions and Biomedical Prevention Technologies

To establish a frame of reference for our results, we conducted a nonsystematic review to identify STI or HIV outcomes in other meta-analyses that focused on multisession behavioral interventions^59–68 as well as in biomedical trials (involving male circumcision, preexposure prophylaxis, microbicide, vaccines, diaphragms, and herpes simplex virus treatment) aimed at prevention of HIV and other STIs.^{8–10,14,69–71} Using odds ratios to compare the results of these studies with our study’s biomedical outcomes, we found that single-session interventions rivaled the effects of both multisession behavioral interventions and biomedical prevention trials (Table A, available as a supplement to this article at http://www.ajph.org). However, further analyses that account for heterogeneity, trial length, settings, and so forth are needed to precisely compare interventions, and such analyses should be the focus of future research.

Single-Session Interventions and the Teachable Moment

One mechanism that may play an important role in determining the efficacy of brief, single-session interventions is their delivery, among populations at risk for HIV and other STIs, during a “teachable moment.”⁷² Health behavior interventions often take advantage of periods of heightened awareness and increased perceived vulnerability to enhance health promotion outcomes. The teachable moment has been critically important in interventions, such as those focusing on smoking cessation, treatment of alcoholism, cardiovascular risk reduction, and cancer prevention, in which significant changes in lifestyle are demanded to ward off life-threatening conditions.

It is likely that bundled HIV prevention measures will target individuals at risk for HIV and other STIs; therefore, we must capitalize on these teachable moments. For example, the openness to behavior change that can occur after a diagnosis may explain at least some of the observed effects in single-session interventions with clinic patients. Receiving an STI diagnosis is itself a potent motivator for many people to change their sexual risk behaviors; thus, this time period is an important window of opportunity for intervention.

Limitations

Our findings should be considered in light of their limitations. We are unable to offer evidence on differences and similarities in the effectiveness of single-session interventions in preventing individual STIs because an insufficient number of studies reported results separately for STIs. We were unable to identify any single-session interventions with sufficient power to allow HIV to be used as an outcome. Likewise, condom use was reported differently by different studies (e.g., as numbers of unprotected acts, percentages of unprotected acts, or event-level condom use); therefore, reported risks may have varied across studies. Similarly, some studies did not report on behavioral outcomes, and thus we were unable to include all studies in our analysis investigating condom use.

Our results also do not speak to the critical active ingredients that are necessary to motivate protective behaviors or reduce risky behaviors, thereby decreasing STI prevalence. The most common approach emphasized education and skill building, but there were insufficient numbers of studies to determine whether use of these elements is associated with the greatest risk reduction. Moreover, the definition of standard of care appeared to vary greatly across studies. Many studies cited treatment and prevention guidelines as their basis for constructing control conditions; however, there was clearly variability in what the control condition provided.

Implementation level and age group were 2 moderating variables determined post hoc, and therefore the analyses involving these variables were prone to bias.²⁰ In some cases, the unit of analysis did not match the unit of assignment; this limitation should be the focus of future studies. Furthermore, results demonstrated heterogeneity of tested interventions. The presence of heterogeneity creates limitations with respect to making conclusive statements regarding meta-analytic findings. However, we used random effects models to address heterogeneity,^73–75 and we conducted moderator analyses to account for variability in study results.

To create a comparison across other meta-analytic results and biological studies, we completed statistical approximations and transformations from standardized mean differences, proportions of risk, and risk ratios to odds ratios. As a result, some small deviations are possible relative to completing these procedures with raw data.⁷⁶

Conclusions

Single-session risk reduction interventions are effective in reducing STI prevalence. On the whole, brief interventions require considerably lower levels of services, resources, and attendance motivation from participants than multisession interventions. The low cost of single-session interventions offers high public health utility (i.e., a greater likelihood of being implemented in community-based organizations), as well as their potential to achieve widespread coverage. Moreover, brief interventions appear to succeed in behavior change as well as or even better than longer interventions. This finding is consistent with the results of a recent meta-synthesis of health promotion meta-analyses⁷⁷ and a meta-analysis focusing on behavioral interventions targeting prevention of STIs among female African Americans.⁶⁴

The promise of single-session interventions rests mainly in their ability to reach large numbers of people in the populations at greatest risk for STIs. Single-session interventions can readily fit into existing services. Attaching interventions to routine care eliminates the need for additional infrastructure and staffing. Single-session behavioral interventions require capitalizing on teachable moments while condensing effective intervention elements. Focused behavioral interventions will prove necessary for the success of many biomedical prevention technologies. Given their potential for reducing high-risk behaviors, effective single-session interventions should be made more available. Particularly in resource-constrained settings, they may be the most viable option in terms of behavioral change interventions. Future research should focus on establishing the potential benefits of biomedical technology in combination with single-session interventions.