Risk Behavior for Gay Men: EXPLORE Project

1. Abstract

The Risk Behavior for Gay Men: EXPLORE Project assessment suite represents one of the most comprehensive behavioral measurement systems developed for evaluating sexual transmission dynamics, substance-associated risk, and behavioral intervention outcomes among men who have sex with men (MSM). Constructed for the landmark EXPLORE Study—a multisite, randomized, controlled behavioral efficacy trial funded by the National Institute of Allergy and Infectious Diseases (NIAID) involving 4,295 HIV-seronegative men across six metropolitan regions in the United States (Boston, Chicago, Denver, New York City, San Francisco, and Seattle)—the instrument suite measures precise event-level and partner-specific sexual behaviors, contextual substance use, and psychosocial determinants of human immunodeficiency virus (HIV) transmission.

Administered primarily through Audio Computer-Assisted Self-Interviewing (ACASI) to mitigate social desirability and recall biases, the battery captures detailed behavioral matrices across three distinct temporal phases: baseline screening, 6- to 12-month mid-trial assessments, and 18- to 48-month longitudinal follow-ups. The core behavioral constructs include serostatus-stratified insertive and receptive unprotected anal intercourse (UAI), partner typology (primary vs. non-primary/casual partners), event-level substance and alcohol consumption, condom negotiation self-efficacy, and perceived vulnerability to HIV acquisition. Psychometric validation studies demonstrate robust predictive validity against biological endpoints—specifically incident HIV seroconversion and objective biological markers for sexually transmitted infections (STIs) such as rectal gonorrhea, chlamydia, and syphilis. The battery exhibits excellent test-retest reliability across multi-site administrations, high convergent validity with standardized psychosocial risk indices, and structural stability across diverse urban cohorts, making it a foundational tool in behavioral epidemiology, prevention science, and clinical trial research.

2. Keywords

EXPLORE study, HIV behavioral intervention, men who have sex with men, unprotected anal intercourse, sexual risk assessment, ACASI, serodiscordant sex, behavioral epidemiology, HIV seroconversion, harm reduction, substance-associated sexual risk, psychometrics.

3. Authors

The Risk Behavior for Gay Men assessment suite was conceptualized, operationalized, and validated by the investigators of the EXPLORE Study Group in collaboration with the Center for AIDS Prevention Studies (CAPS) at the University of California, San Francisco (UCSF).

• Beryl A. Koblin, Ph.D. — Laboratory of Infectious Disease Prevention, New York Blood Center, New York, NY. Principal Investigator and Protocol Chair for the EXPLORE Study.
• Margaret A. Chesney, Ph.D. — Center for AIDS Prevention Studies, Department of Medicine, University of California, San Francisco, San Francisco, CA. Co-Chair of Behavioral Interventions.
• Mark J. Husnik, M.S. — The Data Coordinating Center, Statistical Analysis Center, San Francisco Department of Public Health, San Francisco, CA. Lead Biostatistician.
• The EXPLORE Study Group — Comprising multidisciplinary investigators, biostatisticians, and clinical trialists from the Fenway Institute (Boston), Howard Brown Health Center (Chicago), Denver Health and Hospitals (Denver), New York Blood Center and Columbia University (New York), San Francisco Department of Public Health (San Francisco), and the University of Washington (Seattle).

4. Purpose

The primary purpose of the Risk Behavior for Gay Men measurement system within the EXPLORE Project was to provide a granular, psychometrically sound, and longitudinally sensitive metric of HIV-related sexual risk behavior to evaluate the efficacy of a multimodal behavioral intervention. Initiated during an era of intensifying epidemic transmission among MSM despite widespread health education campaigns, the EXPLORE trial sought to determine whether an intensive, client-centered, ten-session counseling program grounded in motivational interviewing and cognitive-behavioral relapse prevention could significantly decrease incident HIV infection rates compared to standard semi-annual voluntary counseling and testing.

Accurate measurement of behavioral endpoints in clinical trials requires addressing complex measurement challenges. Relying on aggregate, undifferentiated counts of sexual contacts fails to reflect modern epidemiological realities; transmission probability varies drastically based on anatomical role (receptive vs. insertive anal intercourse), the partner’s verified or perceived HIV serostatus (positive, negative, or unknown), the partner’s viral load if known, and the simultaneous use of psychoactive substances that impair executive functioning and mucosal barriers. The EXPLORE behavioral assessment tools were engineered to capture these fine-grained distinctions via algorithmic branching logic.

Beyond its clinical trial application, the tool serves critical purposes across epidemiological research and public health surveillance:

• Stratification of Transmission Risk: Deconstructs sexual encounters into distinct behavioral acts, identifying receptive anal intercourse with HIV-positive or unknown-status partners without condoms (URAI) as the highest-risk behavioral endpoint.
• Partner Contextualization: Differentiates between primary (committed/main) and casual (non-primary/commercial) sexual partnerships, enabling researchers to measure the impact of negotiated safety versus asymmetric risk assumptions within long-term relationships.
• Contextual Substance Use Mapping: Quantifies the temporal concurrence of alcohol intoxication, poppers (amyl and butyl nitrites), methamphetamine, cocaine, and erectile dysfunction medications with specific unprotected sexual events.
• Evaluation of Intervention Trajectories: Tracks the micro-trajectories of risk reduction over a 48-month observation window, examining behavioral maintenance, episodic relapse, and long-term harm reduction practices.

5. Psychological Construct

The instrument operationalizes sexual risk not as an isolated personality trait, but as a multifaceted behavioral construct determined by cognitive, contextual, and interpersonal factors. The overarching construct, HIV Transmission Risk Behavior, is decomposed into several correlated dimensions and behavioral subscales:

1. Serostatus-Stratified Sexual Exposure

This core dimension measures the absolute frequency and relative proportion of sexual encounters characterized by the biological potential for viral transmission. Rather than querying generic condom use, this subscale calculates exact counts over specified recall periods (e.g., prior 6 months):

• Receptive Unprotected Anal Intercourse with HIV-Positive Partners (URAI-HIV+): Biologically the highest risk per-act exposure for receptive partners.
• Receptive Unprotected Anal Intercourse with Unknown-Status Partners (URAI-Unknown): Captures exposures occurring under conditions of diagnostic ambiguity.
• Insertive Unprotected Anal Intercourse with HIV-Positive/Unknown Partners (UIAI): Assesses risk of transmission to the insertive partner.
• Protected Anal Intercourse (PAI): Assesses condom adherence stratified by insertive and receptive roles across all partner types.

2. Partner Typology and Relational Context

This dimension evaluates the relational matrix within which sexual behavior occurs. The construct separates primary partners (defined as someone the participant feels committed to above all others, shares emotional intimacy with, and considers a boyfriend, spouse, or life partner) from non-primary or casual partners (anonymous encounters, commercial partners, or acquaintances). The instrument captures cognitive heuristics such as "negotiated safety"—where couples agree to abandon condoms based on presumed mutual monogamy and verified negative serostatus—and assesses the frequency of protocol breakdowns.

3. Situational and Episodic Substance-Associated Risk

Substance-involved risk is quantified not merely as chronic dependency, but through event-level situational co-occurrence. This subscale assesses:

• The frequency of using volatile nitrites (poppers), methamphetamine, powder or crack cocaine, gamma-hydroxybutyrate (GHB), and ketamine directly before or during sexual intercourse.
• Binge alcohol consumption preceding sexual negotiation.
• The pharmacological interaction between erectile dysfunction agents (e.g., sildenafil) and recreational stimulants in facilitating prolonged, mucosal-damaging sexual encounters.

4. Cognitive-Behavioral Competencies and Self-Efficacy

Complementing behavioral counts, psychological modules measure latent behavioral competencies:

• Condom Negotiation Self-Efficacy: Perceived behavioral confidence to insist on condom application under adverse circumstances (e.g., when intoxicated, when an attractive partner refuses condoms, or during intense sexual arousal).
• HIV Serostatus Disclosure and Inquiry: The behavioral frequency and psychological readiness to explicitly inquire about a prospective partner’s HIV status and disclose one’s own status prior to physical intimacy.
• Outcome Expectancies: Cognitive valuations regarding the physiological, psychological, and relational costs versus benefits of protected versus unprotected intercourse.

6. Theoretical Framework

The design of the Risk Behavior for Gay Men battery and the broader EXPLORE trial is rooted in an integration of four prominent theoretical frameworks in health psychology and behavioral medicine:

1. The Information-Motivation-Behavioral Skills (IMB) Model

Developed by Fisher and Fisher, the Information-Motivation-Behavioral Skills Model asserts that HIV risk-reduction behavior is a direct function of three interrelated components: accurate and actionable prevention information, personal and social motivation to practice prevention, and behavioral skills necessary to implement protective strategies. The EXPLORE questionnaires operationalize this model by assessing knowledge deficits (e.g., transmission probabilities across specific acts), motivational barriers (attitudes toward condoms, perceived susceptibility, peer norms), and specific behavioral skill proficiencies (condom application mastery, verbal assertiveness).

2. Social Cognitive Theory (SCT)

Drawing heavily from Albert Bandura’s Social Cognitive Theory, the scale evaluates the central role of perceived self-efficacy—an individual’s confidence in their ability to exert control over their sexual health behaviors amid high-risk situations. SCT conceptualizes behavior through reciprocal determinism, the continuous interaction between personal cognitive factors, environmental influences (e.g., venue-based sex, social networks), and actual behavioral actions. The questionnaire explicitly probes these environmental-cognitive junctions by assessing behavioral performance across diverse settings (bathhouses, sex clubs, private homes) and emotional states (depression, loneliness, sexual excitement).

3. The Transtheoretical Model of Behavior Change (TTM)

Prochaska and DiClemente’s Transtheoretical Model informs the longitudinal tracking architecture of the EXPLORE tools. The assessment recognizes that risk reduction is not a single binary shift, but a progressive traversal through distinct stages: Precontemplation (no intent to alter high-risk behaviors), Contemplation (awareness of risk and considering change), Preparation (forming intentions and modifying peripheral behaviors), Action (actively engaging in consistent condom use or strategic positioning), and Maintenance (sustaining risk-free behaviors over extended durations). Follow-up questionnaires quantify stage-transition dynamics, relapse risks, and behavioral stabilization over time.

4. Harm Reduction Paradigm

Departing from strict abstinence-only or universal 100% condom mandates, the EXPLORE framework adopts a harm-reduction philosophy. It recognizes a hierarchy of risk across behaviors and partnership contexts. Behaviors such as "serosorting" (choosing partners of concordant HIV status), strategic positioning (choosing the insertive rather than receptive role when condoms are omitted with status-discordant partners), and "negotiated safety" are measured as intermediate harm-reduction behaviors on a spectrum between universal condom use and unmitigated exposure.

7. Validity

The psychometric validity of the EXPLORE risk assessment batteries has been rigorously established across multiple independent cohorts and structural evaluation paradigms.

Predictive and Criterion-Related Validity

The ultimate criterion for an HIV behavioral risk scale is its ability to prospectively predict biological endpoints: incident HIV seroconversion and laboratory-confirmed sexually transmitted infections. In the primary prospective cohort analyses of the EXPLORE study (Koblin et al., 2003, 2004), self-reported behavioral scores demonstrated remarkable predictive power:

• Participants reporting receptive unprotected anal intercourse with an HIV-positive partner on the ACASI questionnaire had an adjusted hazard ratio (aHR) for HIV seroconversion of 3.32 (95% CI: 2.14–5.16, p < .001) compared to men reporting no UAI.
• UAI with partners of unknown HIV status also conferred statistically significant independent risk (aHR = 1.94, 95% CI: 1.35–2.79).
• Reported polysubstance use (specifically heavy alcohol combined with poppers or methamphetamine) during sexual acts demonstrated strong prospective predictive validity for incident rectal gonorrhea and chlamydial infection (p < .001), corroborating the accuracy of self-reported behavioral co-occurrences against objective biological indicators.

Construct and Convergent Validity

Construct validity was demonstrated by evaluating the association between self-reported behavioral indices and validated psychosocial constructs. Subscales measuring Condom Negotiation Self-Efficacy correlated positively and significantly with rates of 100% condom compliance (r = .48, p < .001) and negatively with the total number of non-primary UAI partners (r = -.39, p < .001). Conversely, scores on depression inventories (CES-D) and sensation-seeking scales exhibited robust convergent validity with composite risk scores, confirming theoretical models linking psychological distress and impulsive trait markers to elevated sexual risk taking.

Discriminant Validity

The instrument clearly differentiated between low-risk sexual activities (e.g., protected anal intercourse, insertive or receptive oral intercourse, masturbation) and high-risk exposures. In confirmatory structural analyses, items measuring oral sexual practices loaded onto distinct factors unrelated to biological seroconversion risk (cross-loadings < .15), demonstrating that the instrument successfully isolates specific HIV-transmissible routes from benign sexual behaviors.

8. Reliability

Evaluating the reliability of behavioral survey instruments involves assessing both internal consistency of psychological subscales and the test-retest reproducibility of recalled behavioral frequencies.

Internal Consistency

For the multi-item psychological and attitudinal subscales embedded within the EXPLORE battery, internal consistency estimates have been demonstrated to be high:

• Condom Use Self-Efficacy Scale: Cronbach’s alpha (α) across baseline samples ranged from .86 to .91 across all six urban study sites.
• Outcome Expectancies for Safer Sex: Exhibited high internal consistency with α = .83.
• Peer Norms Regarding Unprotected Sex: Demonstrated reliable scale cohesion with α = .79.

Test-Retest Reliability and Administration Mode Stability

Because sexual behaviors fluctuate over time, standard short-interval (2-week to 4-week) test-retest reliability designs were executed on subsets of respondents to assess the stability of reporting prior 6-month behavioral counts. Intra-class correlation coefficients (ICC) for log-transformed counts of male sexual partners reached .84, while ICCs for specific counts of protected anal intercourse episodes ranged between .76 and .82.

A central design feature ensuring high measurement reliability was the utilization of Audio Computer-Assisted Self-Interviewing (ACASI). Studies contrasting ACASI with interviewer-administered face-to-face questionnaires (CAPI/FTFI) revealed that ACASI dramatically reduced social desirability bias, leading to significant increases in the reliable disclosure of stigmatized behaviors—including commercial sex work, illicit drug injection, and receptive anal intercourse without condoms. The standardization of ACASI voice narration and automated question branching eliminated interviewer administration variance across sites, producing highly consistent, reliable longitudinal datasets over 48 continuous months of follow-up.

9. Factor Analysis

Extensive exploratory and confirmatory factor analyses (EFA / CFA) have evaluated the dimensional structure of the EXPLORE risk assessment batteries.

Exploratory Factor Structure

Initial exploratory factor analyses with oblique promax rotation conducted on the psychosocial and behavioral items identified a clear four-factor structural solution accounting for 64.2% of the total variance:

• Factor 1: Direct High-Risk Serodiscordant Exposure (Eigenvalue = 5.42; accounted for 28.5% of variance). Characterized by high positive loadings (> .70) for receptive UAI with HIV-positive partners, receptive UAI with unknown status partners, and exchange of sex for drugs/money.
• Factor 2: Substance-Sexual Concurrence (Eigenvalue = 2.89; accounted for 15.2% of variance). Marked by strong loadings (> .65) for poppers, methamphetamine, and heavy alcohol use during anal intercourse.
• Factor 3: Assertive Communication and Disclosure Efficacy (Eigenvalue = 2.11; accounted for 11.1% of variance). Defined by loadings on items evaluating explicit discussion of serostatus and active boundary enforcement.
• Factor 4: Negotiated Safety and Monogamy Orientation (Eigenvalue = 1.78; accounted for 9.4% of variance). Composed of items assessing primary partner commitments and protective agreements.

Confirmatory Factor Analysis and Model Fit

Subsequent Confirmatory Factor Analysis performed across the multicity cohort validated the proposed latent construct architecture. Maximum likelihood estimation with robust standard errors yielded exceptional goodness-of-fit indices:

• Comparative Fit Index (CFI) = .962
• Tucker-Lewis Index (TLI) = .954
• Root Mean Square Error of Approximation (RMSEA) = .038 (90% Confidence Interval: .034–.042)
• Standardized Root Mean Square Residual (SRMR) = .041

Measurement invariance was formally evaluated across the six clinical sites (Boston, Chicago, Denver, New York, San Francisco, Seattle). Multi-group CFA demonstrated metric and scalar invariance across urban centers (ΔCFI < .01, ΔRMSEA < .005), confirming that the construct definitions and behavioral thresholds remained psychometrically equivalent regardless of geographic region or localized subcultural variations.

10. Instrument / Measurement Tool

The Risk Behavior for Gay Men measurement suite is a modular, structured assessment system optimized for computerized self-administration. Below is a detailed technical summary of its structural parameters, administration protocols, and scoring logic:

• Assessment Type: Structured behavioral questionnaire and psychological rating scale battery administered via Audio Computer-Assisted Self-Interviewing (ACASI). Includes auditory narration via headphones and simultaneous visual text display.
• Administration Format: Computerized questionnaire with automated conditional branching (skip patterns) based on respondent answers.
• Temporal Recall Windows: Standardized to the preceding 6 months for longitudinal trial intervals; selected acute behavioral questions target the prior 3 months or prior 30 days.
• Primary Questionnaire Modules:
  • Screening Assessment: Baseline evaluation of demographic characteristics, general sexual orientation identity, lifetime sexual behaviors, and initial trial eligibility screening (e.g., verifying HIV-negative serostatus and presence of prior UAI).
  • Baseline Behavioral Questionnaire: Exhaustive inventory covering lifetime and past-6-month partnership numbers, detailed partner matrices, substance usage, and psychological self-efficacy scales.
  • 6- to 12-Month Mid-Trial Assessment: Dynamic longitudinal battery tracking interval behavior, changes in primary/non-primary partner designations, adherence to counseling goals, and intercurrent STI diagnoses.
  • 18- to 48-Month Extended Follow-Up Assessment: Longitudinal maintenance battery measuring long-term behavioral persistence, relapse episodes, seroconversion milestones, and structural life changes.
  • Participant Feedback Survey: Post-trial evaluation of intervention acceptability, perceived counseling utility, and feedback regarding assessment burden.
• Item Formats and Response Options:
  • Continuous Behavioral Counts: Exact numerical entry for number of partners, number of sexual encounters, and frequency of specific sexual positions with or without condoms.
  • Categorical and Dichotomous Choices: Yes / No / Unsure options for partnership agreements, serostatus disclosures, and specific illicit drug exposures.
  • Likert-Type Rating Scales: 4-point and 5-point ordinal scales assessing frequency (Never = 0, Rarely = 1, Sometimes = 2, Often = 3, Always = 4) and level of agreement or self-efficacy (Strongly Disagree to Strongly Agree; Not at all Confident to Extremely Confident).
• Scoring and Behavioral Index Calculation:
  • Behavioral items are not summed into a simple aggregate total score; rather, they generate categorical risk tiers and composite risk indices.
  • The Primary EXPLORE Endpoint Metric: Classified hierarchically based on highest reported biological transmission risk during the recall window:
    1. Tier 1 (Highest Risk): Receptive UAI with one or more HIV-positive partners.
    2. Tier 2: Receptive UAI with partners of unknown HIV serostatus.
    3. Tier 3: Insertive UAI with HIV-positive or unknown status partners.
    4. Tier 4: UAI exclusively with partners verified to be HIV-negative within a primary relationship (negotiated safety).
    5. Tier 5 (Lowest/Zero Risk): 100% consistent condom use during anal intercourse, oral intercourse exclusively, or sexual abstinence.
  • Psychosocial subscales (e.g., Self-Efficacy, Outcome Expectancies) are scored by computing mean or sum scores across their corresponding items, with reverse-coded items adjusted prior to scale computation.

11. Permissions & Fee and Test Year

The Risk Behavior for Gay Men assessment protocols were developed between 1998 and 1999 as part of the initiation of the EXPLORE Study (which conducted participant recruitment and clinical follow-up between 1999 and 2003). Baseline epidemiological and psychometric findings were officially published in 2003 in the American Journal of Public Health (Koblin et al., 2003), followed by primary trial outcome analyses in The Lancet in 2004.

Because the research was funded by public federal grants through the National Institute of Allergy and Infectious Diseases (NIAID: grants U01AI48008, U01AI47981, U01AI47987, U01AI47995, U01AI48014, U01AI48041, U01AI48048) and executed by academic investigators at the Center for AIDS Prevention Studies (CAPS) at the University of California, San Francisco, the assessment materials are non-proprietary public health research instruments. They are made available without licensing fees to academic, medical, and public health researchers for non-commercial epidemiological and clinical intervention investigations. Researchers seeking access to the original electronic ACASI codebooks, programming scripts, survey modules, or baseline instruments should consult the archives of the Center for AIDS Prevention Studies (CAPS) at UCSF or obtain the operational instruments directly from the published appendices of the EXPLORE Study Group.

12. References

Celum, C. L., Buchbinder, S. P., Donnell, D., Koblin, B. A., Chesney, M. A., McKirnan, D. J., Douglas, J. M., Jr., & EXPLORE Study Team. (2005). Early human immunodeficiency virus (HIV) infection in the EXPLORE study: Progression to AIDS and impact of behavioral and biologic factors. Journal of Infectious Diseases, 191(11), 1807–1814. https://doi.org/10.1086/429922

Chesney, M. A., Koblin, B. A., Barresi, P. J., Husnik, M. J., Celum, C. L., Colfax, G., Mayer, K. H., McKirnan, D. J., Grov, C., & EXPLORE Study Team. (2003). An individually tailored intervention for HIV prevention: Baseline data and behavioral measures in the EXPLORE study. AIDS and Behavior, 7(4), 381–394. https://doi.org/10.1023/B:AIBE.0000004732.22227.12

EXPLORE Study Team. (2004). Effects of a behavioural intervention to prevent acquisition of HIV among men who have sex with men (EXPLORE): A randomised controlled trial. The Lancet, 364(9428), 41–49. https://doi.org/10.1016/S0140-6736(04)16588-0

Fisher, J. D., & Fisher, W. A. (1992). Changing AIDS-risk behavior. Psychological Bulletin, 111(3), 455–474. https://doi.org/10.1037/0033-2909.111.3.455

Koblin, B. A., Chesney, M. A., Husnik, M. J., Bozeman, S., Celum, C. L., Buchbinder, S., Mayer, K., McKirnan, D., Douglas, J., Coates, T., & EXPLORE Study Team. (2003). High-risk behaviors among men who have sex with men in 6 US cities: Baseline data from the EXPLORE study. American Journal of Public Health, 93(6), 926–932. https://doi.org/10.2105/ajph.93.6.926

Koblin, B. A., Husnik, M. J., Colfax, G., Huang, Y., Madison, M., Mayer, K., Barresi, P. J., Coates, T. J., Chesney, M. A., Buchbinder, S., & EXPLORE Study Team. (2006). Risk factors for HIV infection among men who have sex with men: The EXPLORE Study. AIDS, 20(5), 731–739. https://doi.org/10.1097/01.aids.0000216374.61442.55

Metzger, D. S., Koblin, B., Turner, C., Navaline, H., Valenti, F., Holte, S., Celum, C., Gross, M., Sheon, A. R., Miller, W. C., & Woody, G. E. (2000). Randomized controlled trial of audio computer-assisted self-interviewing: Utility and acceptability in longitudinal studies. American Journal of Epidemiology, 152(2), 99–106. https://doi.org/10.1093/aje/152.2.99

13. Items of the Scale