Using the Internet to Recruit Rural MSM for HIV Risk Assessment: Sampling Issues

Abstract

The Internet is an emerging research tool that may be useful for contacting and working with rural men who have sex with men (MSM). Little is known about HIV risks for rural men and Internet methodological issues are only beginning to be examined. Internet versus conventionally recruited samples have shown both similarities and differences in their demographic characteristics. In this study, rural MSM from three sizes of town were recruited by two methods: conventional (e.g. face-to-face/snowball) or Internet. After stratifying for size of city, demographic characteristics of the two groups were similar. Both groups had ready access to the Internet. Patterns of sexual risk were similar across the city sizes but varied by recruitment approach, with the Internet group presenting a somewhat higher HIV sexual risk profile. Overall, these findings suggest the Internet provides a useful and low cost approach to recruiting and assessing HIV sexual risks for rural White MSM. Further research is needed on methods for recruiting rural minority MSM.

Introduction

Men who have sex with men (MSM) and live in rural areas have seldom been targeted specifically as participants in HIV prevention research. The first issue that may inhibit research with rural MSM includes the high costs associated with contacting relatively few individuals in areas of low population density. Second, contacting rural MSM for research may be difficult because of their preference for anonymity and fear of exposure (Williams et al., in press). Third, HIV prevalence has been relatively low overall in rural areas.

Internet use is rapidly increasing in the United States (Rainie and Packel, 2001) and the Internet has been shown to be an efficient format for collecting data on human sexual behavior (Birnbaum, 2000). It has been suggested that most, if not all, rural MSM use the Internet as a means for meeting other men (Rosser, personal communication). Additionally, Swedish studies indicate that MSM who reside in rural areas may be using the Internet as a means for accessing the larger gay community (Ross et al., 2000; Tikkanen and Ross, 2000). Thus, the Internet may provide an inexpensive and useful tool for recruiting rural MSM into HIV research and prevention programs.

Recent studies have raised concerns as to whether sampling via the web results in different outcomes than those of more conventionally recruited samples. For example, Bailey et al. (2000) compared the findings produced using a sample of college students, traditionally used in many studies on human sexuality, to those obtained using a sample that was recruited via the Internet. The Internet sample was more likely to report being White, older, and less well educated than a college-based sample. The Internet sample also reported more homosexual and bisexual behaviors, less negative attitudes toward homosexuality, and lower social desirability scores than the college-based sample.

Studies focused on MSM have found differences in demographic characteristics and risk behaviors reported by samples recruited through the Internet compared to those recruited by more conventional sampling procedures. Ross et al. (2000) and Tikkanen and Ross (2000) found greater diversity in a sample of Swedish MSM recruited using the Internet than in a sample recruited through conventional procedures. They found that MSM in the Internet constructed sample were younger, had less education, lived outside urban areas, were more likely to work in blue-collar jobs, and were more likely to self-identify as bisexual or heterosexual. In a similar study, Rhodes et al. (2002) compared the responses of MSM recruited using the Internet to those of MSM recruited in a bar settings. Rhodes et al. (2002) also found that the Internet sample of MSM were less well educated and were more likely to report being bisexual. They found few differences between the samples in sexual risk behaviors or use of the Internet for health information. Unlike the Swedish studies, Rhodes et al.'s Internet sample was older than the bar sample. It would seem from these few studies of mostly Urban and Swedish men that there is no clear pattern of similarities and differences.

The Internet is an attractive methodology for sampling rural populations of MSM. As the Swedish studies suggest, the profile of MSM recruited in rural areas using the Internet might be significantly different from that of MSM recruited in urban areas. To our knowledge, no studies have examined the differences in samples of MSM from rural areas based on sampling methods. It is unknown whether Internet samples differ from conventional samples on important demographic or sexual risk behaviors. HIV-related activities in areas of low population density have recently been given more attention due to the shift of the disease into rural areas. Lam and Liu (1994) noted that by the end of the 1980s the counties with the largest increase in HIV had shifted to counties with fewer than 75,000 people.

The majority of rural HIV related studies have focused on the Northeast or Southeastern parts of the United States, with little work in the Rocky Mountain region. Unlike other parts of the country, in the Rocky Mountains there are few satellite communities and urbanized areas (>50,000 people) quickly transition to sparsely populated rural regions. For example, Wyoming has an average statewide population density of 5.1 people/square mile with the two largest “cities” (Casper and Cheyenne) having approximately 55,000 people each. Cheyenne, at 2 hr, lies closest to an urban center (Denver, CO). Public socializing for MSM in Wyoming is difficult since there are no gay bars or other public establishments identified as “gay.” Internet chat rooms, as a result, have emerged as virtual meeting grounds for men living in small communities, suggesting that the Internet may greatly enhance researchers ability to contact MSM in rural areas.

The purpose of this study was to compare the demographic characteristics, Internet use, and HIV sexual risk behaviors of rural MSM living in the Rocky Mountain region and several adjacent states, recruited either via the Internet or conventional sampling methods. One sample of rural MSM was recruited via the Internet and a second through conventional sampling procedures. Since little is known about MSM in rural areas, especially in states in the “north central region” (i.e. northern Rocky Mountains and adjacent states), and definitions of rural vary (Ricketts et al., 1997; Ricketts and Gesler, 1992), we stratified the two samples by city size to determine whether behaviors in frontier areas differ from those in rural areas.

Methods

Participants

Two sampling methods were used to recruit rural MSM from the “north central region” of the United States between June 2001 and December 2002. The “conventional” sample was recruited through face-to-face contacts with project staff (47.5%), referral by friends in the study (36.5%), or advertisements displayed in community gathering places (6.6%) or at a club or community event (7.7%). Recruiting men for the conventional sample was conducted primarily in Wyoming (n = 118) and Idaho (n = 47), with a few men from adjacent states (Colorado, n = 1; Montana, n = 3; Nebraska, n = 4; South Dakota; n = 7; and Utah, n = 4). Men who met the study criteria had the option of completing a self-administered paper/pencil form (n = 26), a self-administered questionnaire on a notebook computer (n = 19), or by accessing the survey on the Internet (n = 141).

The second group, the “Internet” sample, was composed of men recruited solely on the Internet using primarily banner ads placed on the home page of Gay.com. Men in the Internet sample were limited to those residing in rural areas of the Northern Rocky Mountains and several adjacent states: Colorado (n = 13), Idaho (n = 24), Montana (n = 23), Nebraska (n = 5), North Dakota (n = 12), South Dakota (n = 18), Utah (n = 8), and Wyoming (n = 40). Men who indicated interest in the project by clicking on the banner were directed to the screening questionnaire. Men who completed the screening questionnaire and met eligibility, were provided with the consent form and the complete survey after indicating consent.

Eligibility criteria for the study were: male; 18 years of age or older; self-identified as gay or bisexual or reported having had sex with a man in the 12 months prior to responding to the survey; and living in a rural area. “Rural” may be defined in a variety of ways that include size of city, distance from an urban area and/or mean density per square mile and may vary by the purpose of the research (Ricketts et al., 1997; Ricketts and Gesler, 1992). For the purposes of this research project, we defined rural as “living in town of 75,000 or less and more than 60 min from an urban area. Eligibility for the purpose of taking the questionnaire was verified by self-reported size of city, and zip code was used for purposes of data analysis.

Measures

The questionnaire used in this study was constructed specifically for this study and included sociodemographic characteristics such as age, race/ethnicity, sexual orientation, education, income, place of residence, and current living situation. Sexual orientation was measured as heterosexual, gay, or bisexual. Transgender was not included as a separate category because we were interested in sexual orientation regardless of gender identification. Men reported their HIV status by checking one of the following: having tested negative more than 6 months prior to responding to the survey, testing negative less than 6 months before responding to the survey, and having tested HIV-positive, or never having had an HIV test.

Internet Use

Questions about Internet use included one question about the location of the computer used to surf the Internet to meet men (e.g. home, work/school, or other) and yes/no questions about reasons for using the Internet (e.g., for information, to chat with men, etc.). Frequency of Internet use to meet other MSM was measured using five responses categories; never, less than once/month, 2–3 times/month, once per week, or more than once per week. Men were asked to report the number of Internet sex partners they had had in the last 12 months and this was recoded into four categories: 0, 1, 2–5, and more than 5. The last Internet question was a yes/no question about whether the men had developed a long-term relationship with any of their Internet sex partners.

Measures of Sexual Behavior

Participants responded to questions about recent sex partners as well as the last partner with whom they had anal sex. The number of male sex partners in the last 12 months was reported with an open-ended question and then recoded into 0, 1, 2–5, and more than 5. Descriptions of current male sex partner(s) were elicited by checking one of five choices: none, one steady, one casual, more than one steady, or more than one casual. To examine the context of sexual risk behavior, men were asked to check “yes” or “no” to whether they had unprotected anal sex in any of five situations: with a steady partner, in love, turned on, drinking alcohol, and using drugs. Questions regarding the last partner with whom they had anal sex included a description of that partner (e.g. new, casual, monogamous), where they met him (e.g. Internet, bar, face-to-face), how long they knew him prior to having anal sex (e.g. 1 day, less than a month, 1–6 months, or more than 6 months) and whether they used a condom.

Analyses

All data were analyzed using descriptive contingency tables and chi-square statistics. Since some cells contained small sample sizes, p-values were generated using the Monte Carlo procedure (10,000 samples; 99% confidence interval). Size of residence stratification was done by three city sizes (e.g. <20,000 people; 20,000–49,999 people; and 50,000–75,000 people). We used three city sizes because we felt that dichotomizing by either the USDA, ERS Rural–Urban Continuum Code of greater or less than 20,000 (Butler and Beale, 1994), or the Office of Management and Budget's (1995) MA classification of 50,000 or less used in some studies (e.g. Heckman et al., 2002), was too restrictive.

Demographic characteristics were first analyzed for differences across the two recruitment groups without regard to city size. Since the samples differed significantly by city size (e.g. <20,000; 20,000–49,999; and 50,000–75,000: χ² = 13.61, p < .001), all demographic variables were stratified by the three sizes of residence, using zip code, and reanalyzed. All risk variables were analyzed after stratifying by city size and recruitment group.

Results

Participants

Three hundred and twenty-seven MSM provided data for the study, with 44% of participants recruited through the Internet. Most men in the study were White (83%), had at least some college (81%), either worked full time or were students (78%), and lived in their own home (85%). There were no significant differences between the two samples by age, race/ethnicity, education, current work situation, or current living arrangements, regardless of city size. The mean age of the Internet sample was 32.39 years (SD = 12.50) and the conventional sample was 30.42 years (SD = 10.22). Seventy-seven percent of the Internet sample reported being single, compared to 64% of the conventional sample (χ² = 13.38, p < .01), but when city size was controlled, these differences disappeared. Rates of HIV testing were significantly different between the two recruitment groups (χ² = 11.07, p = .01). Examination of differences by city size indicated that the difference was among the men from cities under 20,000 (χ² = 8.41, p < .05). As shown in Table I, 38.0% of the men in the Internet sample reported that they had been never been tested for HIV, and only 25.4% of them had been tested in the last 6 months. In contrast, only 16.4% of the men in the conventional sample had never been tested and 41.0% had been testing in the last 6 month.

Table I.

HIV Testing^a

	Size of “town”
	0–19,999		20–49,999		50–75,000
	I.b (n) %	Conv.c (n) %	I.b (n) %	Conv.c (n) %	I.b (n) %	Conv.c (n) %
Sample size	(71)	(61)	(36)	(56)	(30)	(60)
Never tested	38.0d	16.4d	30.6	25.0	30.0	18.3
Neg. >6 month ago	32.4d	36.1d	47.2	42.9	43.3	36.7
Neg. <6 month ago	25.4d	41.0d	19.4	26.8	26.7	41.7
Positive	4.2d	6.6d	2.8	5.4	0	3.3

^aBold indicates significant differences.

^bInternet sample.

^cConventional sample.

^dp < .05 for the set of responses.

Internet Use (Table II)

Table II.

Internet Use^a

	Size of “town”
	0–19,999		20–49,999		50–75,000
	I.b (n) %	Conv.c (n) %	I.b (n) %	Conv.c (n) %	I.b (n) %	Conv.c (n) %
Location of computer	(76)	(60)	(36)	(60)	(31)	(63)
Don't surf for men	2.6d	1.7d	5.6d	0d	0	3.2
Home	93.4d	66.7d	86.1d	65.0d	83.9	63.5
Work/school/library	2.6d	6.7d	8.3d	8.3d	12.9	12.7
Other	1.3d	25.0d	0d	26.7d	3.2	20.6
Reasons for Internet use (check all that apply)	(76)	(61)	(36)	(59)	(31)	(63)
Information	35.5	24.6	25.0	20.3	29.0	30.2
Chat with men	77.6	68.9	72.2e	49.2e	80.6	65.1
Make friends	73.7	62.3	75.0e	50.8e	80.6	61.9
Find a long-term partner	36.8e	19.7e	25.0	11.9	41.9	27.0
Meet men for sex	60.5e	32.8e	63.9e	27.1e	64.5e	31.7e
Frequency of Internet use to meet MSM	(76)	(61)	(36)	(60)	(31)	(63)
Never	10.5	24.6	16.7d	36.7d	3.2	23.8
Less than once a month	22.4	16.4	11.1d	21.7d	29.0	27.0
2–3 Times a month	13.2	6.6	22.2d	18.3d	16.1	17.5
Once a week	18.4	13.1	13.9d	8.3d	12.9	14.3
More than once a week	35.5	39.3	36.1d	15.0d	38.7	17.5
Number of Internet sex partners	(56)	(41)	(31)	(35)	(21)	(49)
0	21.4d	58.5d	39.0	43.9	23.8d	55.1d
1	21.4d	14.6d	9.7	17.1	14.3d	10.2d
2–5	41.1d	22.0d	45.2	25.7	19.0d	24.5d
>5	16.1d	4.9d	16.1	14.3	42.9d	10.2d
Develop a long-term relationship with online partner (last 12 months)?	(52)	(41)	(30)	(35)	(20)	(49)
Yes	19.2	14.6	23.3	25.7	30.0d	8.2d

^aBold indicates significant differences.

^bInternet sample.

^cConventional sample.

^dp < .05 for the set of responses.

^ep < .05 for the pair.

Similar distribution patterns in Internet use variables were found across the three town sizes. Significant differences for the location of their computer were found only across the two groups for the smallest and middle-sized towns. For these two city sizes, significantly more men in the Internet sample reported home access (93.4% and 86.1%, respectively) than in the conventional sample (66.7%, χ² = 20.31, p < .001; 65.0%, χ² = 14.41, p < .001, respectively).

Significant differences in the reasons participants reported using the Internet varied by city size, although the general pattern of reasons was again similar. In cities with less than 20,000 population, a significantly higher percentage of the Internet sample used the Internet to find long-term partners (36.8 vs. 19.7, χ² = 4.83, p < .04) and to “meet men for sex” (60.5% vs. 32.8%, χ² = 10.43, p < .002). In the medium sized towns, a significantly higher percentage of the Internet sample used the Internet to “chat with men” (72.2% vs. 49.2%, χ² = 4.81, p < .04), to “make friends” (75.0% vs. 50.8%, χ² = 5.43, p < .03), and “to meet men for sex” (63.7% vs. 27.1%, χ² = 12.49, p < .001). Of the MSM in cities between 50,000 and 75,000, 64.5% of the men in the Internet sample were chatting to “meet men for sex” versus 31.7% of the conventional sample (χ² = 9.13, p < .004).

There was an overall trend for MSM in the conventional sample to use the Internet to meet men less frequently than those in the Internet sample. A significant difference was found for men from medium sized cities (χ² = 9.71, p < .01), such that 36.1% of the men in the Internet sample used the Internet “more than once a week” to meet men while only 15.0% of the conventional sample did so. Not surprisingly, 36.7% of the men in the conventional sample “never” used the Internet to meet men while only 16.7% of Internet sample reported “never.”

Internet sexual relationships were examined in two ways: number of internet sex partners in the last 12 months and whether the men had developed a long-term relationship with an Internet sex partner in the past 12 months. The number of partners located through the Internet showed similar trends for the two groups regardless of city size. The Internet sample tended to have more Internet partners than the conventional sample. Significant differences were found for the smallest (χ² = 14.61, p < .01) and largest (χ² = 11.39, p < .01) cities. For these two city sizes, 58.5% (smallest town) and 55.1% (largest town) of the conventional samples reported no Internet partners while less than only 21.4% and 23.8% of the Internet sample reported none. In contrast, 56.2% and 61.9% of the Internet sample reported more than two sex partners located through the Internet, while only 26.9% (smallest town) and 34.7% (largest town) of the conventional sample reported more than two Internet sex partners. Long-term relationships with Internet sex partners were reported more frequently by the Internet group (30.0%) than the conventional sample (8.2%) from the largest cities (χ² = 5.46, p < .05).

Sex Partners (Table III)

Table III.

Sex Partner(s)^a

	Size of “town”
	0–19,999		20–49,999		50–75,000
	I.b (n) %	Conv.c (n) %	I.b (n) %	Conv.c (n) %	I.b (n) %	Conv.c (n) %
Number of sex partners, last 12 months	(56)	(42)	(31)	(34)	(22)	(55)
0	12.5	7.1	6.5	17.6	4.5	6.0
1	10.7	26.2	22.6	20.6	4.5	30.0
2–5	42.9	42.9	41.9	35.3	36.4	35.0
>6	33.9	23.8	29.0	26.5	54.5	28.0
Describe current sex partner(s)	(31)	(41)	(19)	(34)	(15)	(48)
None	31.7	32.3	31.6	38.2	40.0	29.3
One steady	19.4	26.6	36.8	23.5	20.0	43.8
One casual	0	7.3	5.3	2.9	0	0
More than 1 steady	9.7	7.3	5.3	2.9	6.7	4.2
More than 1 casual	38.7	19.5	21.1	32.4	3.3	22.9
All situations had unprotected anal sex (check all that apply)	(64)	(42)	(34)	(35)	(21)	(49)
With a steady partner	23.4d	52.4d	32.4d	65.7d	38.1	61.2
When I'm in love	18.8d	40.5d	14.7d	37.1d	38.1	36.7
When turned on	26.6	31.0	29.4	22.9	23.8	24.5
After drinking alcohol	28.1	38.2	17.6	25.7	23.8	34.7
After using drugs	9.4	16.7	2.9	2.9	4.8	16.3

^aBold indicates significant differences.

^bInternet sample.

^cConventional sample.

^dp < .05 for the pair of responses.

Participants were asked to characterize their current sexual relationship(s). There were no significant differences between the two samples by city size for number or type of current sexual partners and the patterns were similar for the two groups across each city size. The majority of men reported more than two sex partners in the last 12 months and one-fourth reported more than six. The majority of men reported currently having a relatively low risk partner situation, with no current partner or one steady partner. Of the men with more than one sex partner, the majority reported more than one casual partner.

The situations in which the men reported engaging in unprotected anal sex were similar across groups for drug and alcohol use and when the men were “turned on.” On the other hand, for the men from the two smaller sized towns, the 52.4% and 65.7% of conventional sample reported that they were likely to engage in unprotected anal intercourse when they were with a steady partner while only 23.4% and 32.4% of the Internet sample were influenced by a steady partner (<20,000, χ² = 9.35, p = .003; 20–50,000, χ² = 7.68, p = .008). Again 40.5% and 37.1% of the men from the conventional sample reported unprotected sex when they were in love, while only 18.8% and 14.7% of the Internet group reported similar behavior (<20,000, χ² = 6.02, p = .02; 20–50,000, χ² = 4.50, p = .05).

Anal Sex With the Last Partner (Table IV)

Table IV.

Last Sex Partner With Whom Participant Had Anal Sex^a

	Size of “town”
	0–19,999		20–49,999		50–75,000
	I.b (n) %	Conv.c (n) %	I.b (n) %	Conv.c (n) %	I.b (n) %	Conv.c (n) %
Describe last partner had anal sex	(46)	(33)	(23)	(21)	(24)	(44)
New partner/one night stand	32.6	24.2	13.0	23.8	16.7	20.5
Casual/ex-partner	52.2	45.5	56.5	33.3	58.3	29.5
Monogamous/long-term	15.2	30.3	30.4	42.9	25.0	50.0
Where did you meet last anal sex partner	(46)	(34)	(23)	(22)	(23)	(42)
Internet	31.6d	13.1d	36.1d	10.0d	45.2d	14.3d
Bar	3.9d	8.2d	5.6d	3.3d	12.9d	17.5d
Face-to-face (party, work, etc.)	64.5d	78.7d	58.3d	86.7d	41.9d	68.3d
How long did you know him before having anal sex?	(48)	(37)	(23)	(32)	(24)	(46)
One day	45.8d	21.6d	21.7	21.9	29.2	26.1
Less than 1 month	31.3d	32.5d	34.8	40.6	37.5	43.5
1–6 Months	16.7d	18.9d	30.4	34.4	29.2	26.2
More than 6 months	6.3d	27.0d	13.0	3.1	4.2	4.3
Did you use a condom for last anal sex?	(48)	(37)	(23)	(33)	(25)	(46)
No	37.5	51.4	60.9	43.4	11.0	54.3

^aBold indicates significant differences.

^bInternet sample.

^cConventional sample.

^dp < .05 for the srt of responses.

There were no significant differences across the samples in the men's descriptions of their most recent anal sex partner. Among men in the Internet sample, the majority of men considered this a casual partner. Although this was true for the men in the conventional sample from the smallest towns (45.5%), the percentage decreased as city size increased (33.3% and 29.5%, respectively). There was a tendency for more of the men from the smallest towns to report the last partner as “new or one night stand” than the larger towns and the inverse was true for monogamous/long-term partners. The pattern of ways men reported meeting their last partner was similar for the three city sizes and significantly different across groups for all three city sizes (<20,000, χ² = 6.95, p < .05; 20–50,000, χ² = 10.39, p <.01; and >50,000, χ² = 10.78, p < .01). The Internet sample, regardless of city size reported that they found significantly more partners through the Internet (Table IV).

The length of time men knew their last sex partner before having anal sex was significantly different only for men residing in cities under 20,000 (χ² = 9.44, p < .02). Forty-six percent of the Internet sample knew their partner for 1 day or less while only 21.6% of the conventional sample was in this group. In contrast, 27.0% of conventional sample knew their sex partner for more than 6 months while only 6.3% of the Internet group knew their anal sex partner for more than 6 month. Although not significantly different, patterns of condom use varied across city size. More men in the conventional sample in the smallest (61.4%) and largest cities (54.3%) reported no condom use at last anal intercourse than the Internet sample (37.5% and 11.0%, respectively) and the opposite was true for the mid-sized town (conventional 43.4%, Internet 60.9%).

Discussion

The Internet is emerging as an important tool for reaching hidden populations, but little is known about rural MSM in regards to their Internet access, and use or their HIV risk behaviors. The Internet is a low cost and efficient method for HIV prevention with rural MSM if such men have ready access to the web, show similar demographic characteristics to men recruited by conventional methods, and demonstrate similar or higher HIV risk behaviors than men recruited by conventional methods. The purpose of this study was to compare the demographic characteristics, Internet use, and HIV risk behaviors of rural MSM living in the Rocky Mountain States recruited by using either the Internet or conventional methods.

In general, the patterns of Internet access and use and demographic characteristics were similar across recruiting samples, regardless of city size. It is encouraging, from a recruiting perspective, that almost all of the men across both samples had access to a computer in a secure location. Although fewer of the conventionally recruited sample had home computers, nearly all of the men had access to a computer either at home or in another location, such as a friend's home. Also noteworthy was the finding that a large proportion of the Internet sample resided in the smallest towns. This suggests that MSM in rural areas who are more geographically isolated and possibly more closeted not only have access to the Internet, but are utilizing it regularly.

The men were generally White, middle class, relatively well educated, single, and self-identified as homosexual. These findings suggest that the Internet is a good tool for contacting a subset of rural MSM, but these samples may not be generalizable to less well educated, or minority MSM residing in low-density areas.

Sexual risk behaviors, including numbers of partners, types of partners, and situations that promote unprotected anal intercourse, also showed similar patterns across the city sizes. Sex with current partners showed a mixed pattern of risk. First, the majority of men reported more than two partners in the last 12 months. On the other hand, the men described their current partner status as “no partner” or “one steady partner.” Condom use in higher risk situations was generally high for both samples, but more of the conventional sample reported that they would engage in unprotected sex with loved or steady partners. These findings suggest that rural men, regardless of how they are recruited, may prefer serial monogamy over maintaining concurrent sexual partners, which may lower their risk for contracting HIV (Hudson, 1993; Moris and Kretzschmar, 1997). On the other hand, if men in the conventional sample view their sexual relationships as “loved” or “steady,” regardless of how long they have known their sex partners, they may be more inclined to stop using condoms before they are entirely aware of their partners' sexual histories. These men's readiness to label relationships as “steady” or “loved” may increase their risk for HIV infection.

Anal intercourse constitutes the highest risk for contracting HIV for most MSM. In general, patterns of behavior between the two sampling groups were similar across city size, with the Internet sample engaging in higher risk behaviors. Previous research suggests that Internet sex partners present a higher risk than sex partners met in other ways (Bull et al., 2000). More of the men in the Internet sample reported surfing the web to meet men and reported doing so more frequently. For men in the Internet sample, the last anal sex partner was also more likely to have been found through the Internet and to be a casual partner or, for men in the smallest towns, to be a new partner. Additionally, men in the smallest towns in the Internet sample were more likely to have sex after knowing the partner for less than a day than MSM in the conventional sample. Of additional concern for this sample of Internet using men from extremely rural areas is their lack of HIV testing. HIV testing in rural areas may be especially problematic, since many small towns have not had anonymous testing until recently. Even if HIV testing is officially anonymous, it is a common belief among MSM in rural areas is that HIV testing is not anonymous and may indicate that one is gay or a drug user. Thus, they may be engaging in high risk sex, with relatively unknown Internet sex partners and not checking their HIV status.

Conclusion

Overall it would appear that the Internet may be an extremely useful and cost effective tool for recruiting rural MSM who are White and middle to upper middle class. Almost all of the men had Internet access at home or at a friend's home where they could contact other men on line. Sample demographics were similar for both groups supporting Internet sample generalizability, regardless of city size. On the other hand, the internet sample, especially men from the smallest towns, presented a somewhat higher sexual risk profile in terms of number of Internet sex partners, time to first anal intercourse, and frequency of condom use. Using the Internet as a primary recruiting resource for MSM is supported further by studies that show that Internet sex partners present a higher risk than partners found through conventional methods. The assumption of increased risk associated with Internet use for rural MSM requires additional study given the sampling limitations in this study (e.g. limited geography, lack of minority MSM, the use of retrospective self-report, and the small sample size). Finally, we used a p value of .05, with may be somewhat liberal, given the number of analyses, but given the exploratory nature of this study we were interested in identifying areas of future research rather than controlling for type 1 error rates. Future studies may need to be more conservative as hypotheses about risk differences across Internet and conventional samples begin to emerge.