Home | About | Journals | Submit Manuscript | Contact Us 
pmc logo imagepmc logo image
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptNIH Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Health Educ Res. Author manuscript; available in PMC Feb 10, 2009.
Published in final edited form as:
PMCID: PMC2639555
NIHMSID: NIHMS85349
A randomized control trial of Internet-delivered HIV prevention targeting rural MSM
Anne M. Bowen,1* Keith Horvath,2 and Mark L. Williams3
1Department of Psychology, University of Wyoming, Department 3415, 1000 E. University, Laramie, WY 82071, USA
2Program in Human Sexuality, Department of Family Medicine & Community Health, U. Minn. Medical School, 1300 S. Second ST. Minneapolis, MN 55454, USA
3Center for Health Promotion and Prevention Research, University of Texas-Houston, School of Public Health, 7000 Fannin, Suite 2516, Houston, Texas 77030, USA
* Correspondence to: A. M. Bowen. E-mail: abowen/at/uwyo.edu
Abstract
The Internet may be important for delivering human immunodeficiency virus (HIV) risk reduction to men who have sex with men (MSM) in rural areas. This randomized control trial (RCT) tested the acceptability and efficacy of an Internet-delivered HIV risk-reduction intervention. Two modules include a conversation between an HIV-negative man and an HIV-positive man, with interactive graphics. Ninety men were randomly assigned to intervention or wait-list control and 79% completed the study. An ‘intent-to-treat’ model was used. HIV/acquired immune deficiency syndrome (AIDS) knowledge, self-efficacy and outcome expectancies increased after participating in the intervention, and changes were maintained at 1-week follow-up. Participants said they would participate again. This RCT provides support for the acceptability and efficacy of the Internet for delivering HIV prevention messages to rural MSM.
Introduction
Among rural people, men who have sex with men (MSM) plus MSM who are injection drug users (MSM/IDU) account for 42% of acquired immune deficiency syndrome (AIDS) cases and 47% of rural human immunodeficiency virus (HIV) infections [1]. Rural MSM are often geographically isolated from centers of gay culture and lack gay-identified meeting places in which to interact with potential sex partners. Because of stigma associated with homosexuality in rural areas [2], MSM may feel uncomfortable discussing health concerns related to sexuality with health care providers. The resources of rural public health agencies to provide HIV/AIDS information and prevention services to rural MSM are limited. Therefore, confidential and cost-effective approaches to reach rural MSM are needed.
The World Wide Web expands the possibilities for delivering health-related information and risk-reduction messages. The Internet, with its growing population of users [3], may provide an affordable way [4] to rapidly reach large numbers of individuals who might be at risk of a health-threatening condition. Rural MSM regularly use the Internet to meet sex partners [5, 6], making them excellent candidates for Internet-based HIV/AIDS information and risk-reduction activities. Almost every model of risk reduction begins with the premise that knowledge is a necessary, but insufficient, condition for behavior change [7, 8]. Social cognitive theory assumes that self-efficacy is the direct precursor of behavior change, while knowledge, peer norms and outcome expectancies are predictors of self-efficacy [9, 10]. Using relatively simple matching programs, theoretical constructs included in risk-reduction messages can be tailored to an individual's level of risk [11] and motivations to change behaviors that might be contributing to risk.
Social cognitive theory [9] was used to guide the development of this pilot intervention that provides basic HIV/AIDS information in the context of conversations about living with HIV and methods for risk reduction between peers. Interactive graphics were provided regularly to emphasize the points and allow each participant to individualize his experience. It was thought that the peer delivery and individualization would allow participants to increase not only HIV-relevant knowledge but develop more positive outcome expectancies for risk reduction, and increase their risk-reduction self-efficacy.
The goals of this randomized control trial (RCT) of an Internet-delivered intervention were to determine whether RCT was a viable approach on the Internet and to increase precursors of sexual risk reduction (e.g. HIV knowledge, sexual self-efficacy and outcome expectancies). The study design is a two-group RCT with a 1-week wait-list control group (Fig. 1). The data were analyzed using an ‘intention-to-treat’ model. MSM residing in rural areas of the United States were chosen as participants because of their risk for HIV infection, social isolation and high rates of Internet use. The research hypotheses included (i) HIV knowledge, self-efficacy and outcome expectancies would increase contingent upon participation in the intervention and (ii) increases in HIV knowledge, self-efficacy and outcome expectancies would be maintained for 1 week.
Fig. 1
Fig. 1
Diagram of the RCT with wait-list control. Includes the timing of the three assessments (T1, T2 and T3) and number of participants at each assessment.
Method
Participants
Rural, Internet-using MSM were recruited face-to-face (n = 6) or using Internet banners at a popular web site (n = 84) during April and May 2004. To participate in the study, men were required to be at least 18 years old, have had sex with another man in the last 12 months and live in a rural area. A rural area was defined as ‘living in town of 75 000 or less and more than 60 min drive time from an urban area [1214]’. Eligibility was determined from responses to a brief screening questionnaire completed prior to obtaining consent. Participants were offered gift certificates to an online shopping site after each of three assessments ($10, $15, $20, respectively). The project was approved by the University of Wyoming and the University of Texas Health Science Center at Houston Institutional Review Boards.
Materials
Sociodemographic
Sociodemographic characteristics included self-reported age, race/ethnicity, sexual orientation, education, income, place of residence and current living situation (Table I).
Table I
Table I
Comparison of demographic variables by intervention group
HIV/AIDS knowledge
The HIV/AIDS knowledge questionnaire included thirteen questions about facts identified in previous qualitative research and thought to be related to risk to rural MSM [6]. Eight of the questions were answered ‘true’, ‘false’ or ‘do not know’. The remaining five questions had multiple-choice responses.
Outcome expectancies
Nine outcome expectancy questions were rated on six-point Likert-type scales that ranged from very negative to very positive outcomes. The specific anchor varied according to question. Principle components analysis (PCA) with varimax rotation resulted in two factors: outcomes of condom use (OCU; four items) and outcomes of insisting on safe sex (OISS; five items). The OCU factor accounted for 28% of the variance and had a kappa of 0.78. An example of an item would be ‘When I use a condom, feelings of intimacy are…(‘lost completely’ to ‘just as good’)’. The OISS factor accounted for 26% of the variance and the kappa for this factor was 0.77. A sample item from this scale is ‘If I'm drunk, convincing my partner to use a condom for anal sex is…(‘extremely unpleasant’ to ‘extremely pleasant’)’.
Self-efficacy
Self-efficacy was assessed using 13 questions rated on a six-point Likert-type scale from ‘not at all confident’ to ‘very confident’ and PCA with varimax rotation yielded two factors: safe sex assertiveness (SSA; seven items) and safe sex communication (SSC; five items). The SSA factor accounted for 49% of the variance and had a kappa of 0.93. A sample item would be ‘How confident are you that you could stop at oral sex or mutual masturbation if you didn't have a condom?’ The SSC factor accounted for 27% of the variance and had a kappa of 0.87. A sample question would be, ‘How confident are you that you could talk to your partner about condoms before sex?’
Intervention
Intervention content was identified from focus groups conducted in 2001 and an Internet-based assessment conducted from January 2002 to January 2003. The format of the intervention was based on data obtained from two focus groups conducted in May 2003. Intervention content included HIV prevention information not generally known to MSM residing in rural areas and was presented as a conversation between an HIV-positive gay man who represented the ‘expert’ and an ‘inexperienced’, HIV-negative man who had recently engaged in high-risk sex. Dialogue in the two intervention modules is interspersed with interactive activities and graphics illustrating key points of information.
The intervention is delivered in two modules each of which takes ~20 min to complete. The first module focused on the inexperienced man's risky sexual encounter and the possibility of his having been infected with HIV. The men discussed topics including HIV testing, living with HIV, treatment issues and routes of infection. The second module was set 6 months later after the inexperienced man received a negative HIV test result. The content focuses on how the inexperienced man might maintain his HIV-negative status, including safer sex option, condom types and correct condom application.
Intervention acceptability
Questions measuring acceptability of the intervention included (i) ‘Overall, how interesting was it?’ (ii) ‘Overall, how useful was the intervention?’ (iii) ‘Would you do the intervention again?’ (iv) ‘Would you recommend the intervention to a friend?’ rated on six-point Likert-type scales and (v) ‘How was the time it took for pictures to load?’, rated just right, to short or too long.
Procedure
The men first completed the screening questionnaire and those who matched eligibility requirements viewed an online informed consent form, including a study overview, risks and benefits and confidentiality issues. Links to more in-depth discussions of each point were provided. Consent to participate was indicated when a participant devised his own password and provided online contact information.
Participants then completed the pre-test questionnaire (T1) and were randomly assigned by the computer to the ‘intervention’ group or to the ‘wait-list control’ group (Fig. 1). After assignment, participants were shown a schedule of intervention activities, expected completion dates and the reimbursement schedule. Participants in the intervention group were allowed to begin the intervention modules immediately and were encouraged to complete them within 7 days. The minimum interval between intervention modules was 24 hours. The interval between the pre-test (T1) and the first post-test was a minimum of 7 days and a maximum of 14 days. One-week follow-up assessment (T3) was scheduled for 7 days after the first post-test (T2) and participants were dropped if they did not take it within 14 days after T2. Participants assigned to the wait-list group completed the pre-test assessment (T1) and a second pre-test (T2) 7 days later. After the T2, wait-list participants had the same constraints on completing the intervention as the experimental group with T3 scheduled between 7 and 14 days after T2. Participants who exceeded a maximum of 14 days between any two assessments were dropped from the study.
Analyses
An intention-to-treat model was used to analyze intervention effects. This approach is a conservative method of examining treatment outcome because, rather than excluding dropouts from the analyses, their pre-test scores are used as follow-up scores. The assumption is that we ‘intended’ to treat them, so they should be included in the analyses. Therefore, T1 scores are pasted in at T2 and T3 for participants who did not complete their intervention post-test within 14 days of their intervention pre-test. Preliminary sample size was set at 100 or when one group reached 50 participants. Group differences at T2 and T3 for knowledge, outcome expectancies and self-efficacy scores were examined using two (group) by two (time) repeated measures analysis of covariance, with T1 scores as the covariate. Changes due to participation in the intervention for each group were examined using repeated measures analysis of variance.
Results
Participant characteristics
Ninety men completed the pre-test assessment and the computer randomly assigned them to intervention or wait-list control. There were no significant differences in demographic variables between groups (Table I). The men were from 29 states, and 40% were from towns smaller than 20 000 people. Participants ranged in age from 19 to 59 years, with a mean of 29.02 (SD = 9.61, median = 25.50).
Recruitment, attrition and acceptability
Four hundred and seventy-four individuals completed the study screener, 81% (N = 384) of whom did not meet study eligibility because they were from an urban area (Fig. 1). Of the 91 men who qualified for the study, one did not complete the pre-test questionnaire (T1) and was dropped prior to randomization. Overall study completion was 78.9% with no participants removed for adverse events. Twenty percent of the intervention group and 21% of the wait-list group dropped out before completing all activities with no significant difference between groups. The completers versus the dropout were not different in age, ethnicity, city size, income or sexual orientation.
Seventy-four men responded to the question, ‘Overall, how interesting was the intervention?’ with a mean score of 5.19 (SD = 1.54) on a Scale of 1–6. Eighty-nine percent of the 74 said that they would participate in the intervention again, and 93% said that they would recommend the intervention to a friend. Sixty-nine percent rated the time to load pictures as ‘just right’. Men with dial-up connections were significantly more likely to rate time to load pictures as ‘too long’ (Mann–Whitney U; Z = −3.30, P = 0.001) compared with men with high-speed connections.
Intervention effects
Knowledge
Participants' means and standard deviations for their knowledge scores, adjusted for the intention-to-treat model, at T1, T2 and T3 are shown in Table II. Analysis revealed a significant interaction effect between group and time [F (1,87) = 46.84, P < 0.000, η2 = 0.35]. The two groups' knowledge scores were significantly different at T2 (B = 2.08, 95% CI = 1.15–2.71, η2 = 0.33), but not at Time 3 (B = −0.40, 95% CI = −1.15 to 0.36, η2 = 0.01). One-way repeated measures indicated that the intervention group showed a significant main effect for time [F (2,76) = 42.77, P < 0.000, η2 = 0.53], with a significant increase in knowledge [F (1,38) = 48.82, P < 0.000, η2 = 0.56] from T1 to T3 but not from T2 to T3 [F (1,38) = 1.52, P = 0.22, η2 = 0.04]. The wait-list control group also showed a significant main effect for time [F (2,100) = 42.77, P < 0.000, η2 = 0.53]. There was a significant change in mean knowledge score from T1 to T3 [F (1,50) = 82.90, P < 0.000, η2 = 0.62] and from T2 to T3 [F (1,50) = 62.34, P < 0.000, η2 = 0.56].
Table II
Table II
Means and standard deviations across groups and time for knowledge, outcome expectancies and self-efficacy
Outcome expectancies
Participants' means and standard deviations for the OCU and OISS factors are shown in Table II. The scores on the OISS factor did not show a significant group by time interaction [F (1,87) = 0.77, P = 0.38, η2 = 0.01]. The two groups' OISS scores were significantly different at T2 (B = 0.30, 95% CI = 0.09–0.51, η2 = 0.08), but not at Time 3 (B = −0.01, 95% CI = −25 to 0.23, η2 = 0.00). One-way repeated measures indicated that intervention group showed a significant main effect for time [F (2,76) = 22.34, P < 0.000, η2 = 0.37], with a significant positive increase in the OISS from T1 to T3 [F (1,38) = 39.56, P < 0.000, η2 = 0.51] and from T2 to T3 [F (1,38) = 5.96, P = 0.02, η2 = 0.14]. The wait-list control group also showed a significant main effect for time [F (2,100) = 8.23, P = 0.001, η2 = 0.14]. There was a significant increase in mean OISS score from T1 to T3 [F (1,50) = 11.96, P = 0.001, η2 = 0.19] and from T2 to T3 [F (1,50) = 8.08, P = 0.006, η2 = 0.14].
The scores on the OCU factor did not show a significant group by time interaction [F (1,87) = 0.36, P = 0.55, η2 = 0.00]. The two groups' OCU scores were significantly different at T2 (B = 0.34, 95% CI = 0.06–0.43, η2 = 0.04), but not at Time 3 (B = 0.18, 95% CI = −05 to 0.41, η2 = 0.03). One-way repeated measures indicated that the intervention group had a significant main effect for time [F (2,76) = 14.57, P < 0.000, η2 = 0.28], with a significant positive increase in OCU from T1 to T3 [F (1,38) = 21.66, P < 0.000, η2 = 0.36], but not from T2 to T3 [F (1,38) = 3.53, P = 0.07, η2 = 0.14]. The wait-list control group also showed a significant main effect for time [F (2,100) = 8.06, P = 0.001, η2 = 0.14]. There was a significant increase in mean OCU score from T1 to T3 [F (1,50) = 10.64, P = 0.002, η2 = 0.18] and from T2 to T3 [F (1,50) = 8.63, P = 0.005, η2 = 0.15].
Self-efficacy
Participants' self-efficacy means and standard deviations for SSC and SSA are shown in Table II. The scores on SSC showed a significant group by time interaction [F (1,87) = 6.30, P = 0.01, η2 = 0.07]. The two groups' SSC scores were significantly different at T2 (B = 0.31, 95% CI = 0.08– 0.53, η2 = 0.08), but not at Time 3 (B = 0.00, 95% CI = −22 to 0.23, η2 = 0.00). One-way repeated measures analyses indicated that the intervention group showed a significant main effect for time [F (2,76) = 5.03, P = 0.009, η2 = 0.12], with a significant positive increase in the SSC means from T1 to T3 [F (1,38) = 4.76, P = 0.035, η2 = 0.11], but not from T2 to T3 [F (1,38) = 0.98, P = 0.33, η2 = 0.02]. The wait-list control group also showed a significant main effect for time [F (2,100) = 7.12, P = 0.001, η2 = 0.12]. There was a significant increase in mean SSC scores from T1 to T3 [F (1,50) = 9.84, P = 0.003, η2 = 0.16] and from T2 to T3 [F (1,50) = 7.00, P = 0.011, η2 = 0.12].
The scores on the SSA factor showed a significant group by time interaction [F (1,87) = 7.45, P = 0.008, η2 = 0.08]. The two groups' SSA scores were significantly different at T2 (B = 0.30, 95% CI = 0.09–0.51, η2 = 0.08), but not at Time 3 (B = −0.01, 95% CI = −25 to 0.23, η2 = 0.00). One-way repeated measures analyses indicated that the intervention group had a significant main effect for time [F (2,76) = 4.10, P < 0.02, η2 = 0.10], with a significant positive increase in SSA from T1 to T3 [F (1,38) = 4.04, P = 0.05, η2 = 0.10], but not from T2 to T3 [F (1,38) = 0.60, P = 0.44, η2 = 0.02]. The wait-list control group also showed a significant main effect for time [F (2,100) = 8.58, P < 0.000, η2 = 0.15]. There was a significant increase in mean SSA score from T1 to T3 [F (1,50) = 5.76, P = 0.004, η2 = 0.16] and from T2 to T3 [F (1,50) = 9.97, P = 0.006, η2 = 0.17].
Discussion
The goal of this RCT was to examine short-term changes in HIV risk-reduction knowledge, outcome expectancies and self-efficacy among rural MSM. The most promising outcomes of this study were the significant increases in HIV/AIDS-related knowledge and safer sex attitudes. There is little doubt that participation in the intervention was the effective component, as both groups' knowledge, self-efficacy and outcome expectancies increased contingent upon participation in the intervention. It is also encouraging that the intervention group maintained their positive changes in cognitive variables and knowledge for 1 week. Although this pilot did not allow sufficient time to examine behavior change, both social cognitive theory [9] and the theory of reasoned action [15] would suggest that behavior follows change in attitudes.
A second focus of this Internet study was to examine the acceptability of RCT on the Internet. Most importantly, Internet interventions that maintain a person's interest and thus, their participation are critical to internal validity. The Internet intervention tested in this study was composed of a story line presented as text bubbles, interspersed with interactive graphics. The completion rate was very good and responses to acceptability questions were very positive. However, Internet connection speed may limit the sophistication of interventions. Although most men rated the loading time for graphics as just right, more men with dial-up connections rated them as too long. The development of Internet interventions including sound and animation may be attractive; however, researchers must strike a balance between exciting graphics and connection speeds. There may be a curvilinear relationship in which interesting graphics retain some men who might not otherwise complete an intervention, but eventually losing men as time to load increases. As the Internet evolves and higher connection speeds are available in rural areas, interventionist will have greater freedom to develop more sophisticated programs that match participants' interests.
This pilot study has a number of limitations, the strongest of which is the lack of sufficient time to examine behavior change. The sample size is small and participants were mostly gay identified, white MSM residing in rural areas of the United States. The intervention emphasized HIV/AIDS knowledge that was salient for rural MSM, so it may be less effective with MSM who live in urban areas or with differing levels of knowledge. Additionally, we did not evaluate the relative effects of the interactive versus peer story aspects of the intervention presentation, so the specific active ingredient is unknown. Finally, the two different recruiting methods may have an effect on outcomes. Participants recruited in a face-to-face manner may be more ‘out’ and thus risk behaviors may be different than Internet recruited men. Additionally, face-to-face recruits may have higher retention rates due to wanting to ‘please’ (i.e. social desirability) the recruiter. Given that only six were recruited face-to-face in this study, it was impossible to examine these factors, but future studies might look at these and other potential difference.
These limitations notwithstanding, the results of this study indicated that an RCT testing the efficacy of Internet-delivered interventions targeting psychosocial precursors to AIDS-related behavior change are feasible and potentially effective. This is even more important if we consider that MSM residing in rural areas or who do not identify as gay or bisexual are likely to be missed by HIV risk-reduction interventions delivered in urban gay-identified settings. For these reasons, the future of Internet-delivered HIV risk-reduction interventions appears promising.
Acknowledgments
This project was partially supported by a grant from the National Institutes of Health/National Institute of Mental Health (MH-63667).
Footnotes
Conflict of interest statement: The opinions expressed herein are solely those of the authors and do not reflect those of the Institute.
References
1. CDC. HIV Surveillance Report, 2002. Atlanta, GA: US Department of Health and Human Services, Centers for Disease Control and Prevention; 2002.
2. Preston DB, D'Aguelli AR, Kassab CD, et al. The influence of stigma on the sexual risk behavior of rural men who have sex with men. AIDS Educ Prev. 2004;16:291–303. [PubMed]
3. Madden M, Rainie L. America's Online Pursuits: The Changing Picture of Who's Online and What They Do. Washington, DC: PEW Internet and American Life Project; 2003.
4. Cooper A. Sexuality and the Internet: surfing into the new millennium. Cyberpsychol Behav. 1998;1:187–93.
5. Horvath K, Bowen A, Williams M. Virtual and physical venues as contexts for HIV-risk behaviors among rural men who have sex with men. Health Psychol. 2006;25:237–42. [PubMed]
6. Williams M, Bowen A, Horvath K. The social/sexual environment of gay men residing in a rural frontier state: implications for the development of HIV prevention programs. J Rural Health. 2005;21:48–55. [PMC free article] [PubMed]
7. Albarracin D, McNatt PS, Klein CTF, et al. Persuasive communications to change actions: an analysis of behavioral and cognitive impact on HIV prevention. Health Psychol. 2003;22:166–77. [PubMed]
8. Denning P. Practicing Harm Reduction Psychotherapy: an Alternative Approach to Addictions. New York: Guilford; 2000.
9. Bandura A. Social cognitive theory and exercise of control over HIV infection. In: DiClemente R, Peterson J, editors. Preventing AIDS: Theories and Methods of Behavioral Interventions. New York: Plenum Press; 1994. pp. 25–59.
10. Bandura A. Social Foundations of Thought and Action: a Social Cognitive Theory. Englewood Cliffs, NJ: Prentice-Hall; 1986.
11. Binik YM. Sexuality and the Internet: lots of hyp(otheses)—only a little data. J Sex Res. 2001;38:281–2.
12. Bowen AM, Williams MW, Horvath K. Using the Internet to recruit rural MSM for HIV risk assessment: sampling issues. AIDS Behav. 2004;8:311–9. [PMC free article] [PubMed]
13. Gesler WM, Ricketts TC, editors. Health in Rural North America: the Geography of Health Care Services and Delivery. New Brunswick, NJ: London: Rutgers University Press; 1992.
14. Ricketts TC, Savitz LA, Gesler WM, Osborn DN. Using Geographic Methods to Understand Health Issues: AHCPR Publication No. 97-N013, March 1997. Agency for Health Care Policy and Research, Rockville, MD http://www.ahrq.gov/research/geomap/geomap1.htm.
15. Fishbein M, Ajzen I. Belief, Attitude, Intention, and Behavior: an Introduction to Theories and Research. Reading, MA: Addison-Wesley; 1975.