Sexual Behavior, Mental Health, Substance Use, and HIV Risk Among Agency-Based Male Escorts in a Small U.S. City

Abstract

Relatively little research has examined the personal sex lives of indoor male sex workers (MSWs) or possible connections in this group between sexual behavior and factors related to HIV risk. As part of a larger project, this study collected data from 30 agency-based indoor MSWs (mean = 22.4 years) about their sexual behavior, mental health, and substance use. Few HIV risk behaviors with clients occurred. Drug use and mental health problems were relatively frequent, but not related to increased risk behavior. Instead, MSWs appeared to employ rational decision-making and harm-reduction strategies. Conceptualization of MSW sexual behavior may be required where HIV risk is not attributed to sex work per se, but to other influences such as economic and relational factors.

During the past 20 years, much of the literature on the sexual behavior of male sex workers (MSWs) has focused on HIV risk behavior with clients. This line of study likely grew from concerns that MSWs may have been introducing HIV into the sexual networks of people who hired them (Bimbi, 2007). Thus, some researchers postulated that MSWs could be a transmission vector into less vulnerable populations (Morse, Simon, Osofsky, Balson, & Gaumer, 1991). Most of this initial research examined men working in a single type of venue, the street (Vanwesenbeeck, 2001). Street-based MSWs (i.e., hustlers) were found to exhibit relatively frequent sexual risk behaviors (Simon, Morse, Osofsky, & Balson, 1994), especially when they engaged in sex work to obtain basic needs or when they were using injection drugs (Estep, Waldorf, & Marotta, 1992). High rates of HIV risk also were found with non-clients due to unprotected sex and injection drugs (Rietmeijer, Wolitski, Fishbein, Corby, & Cohn, 1998). One study of street-based MSWs in Manhattan found that men reported the least sexual risk with paid clients, greater risk with male partners for pleasure, and the most risk with female partners for pleasure (Pleak & Meyer-Bahlburg, 1990). However, this study did not explore contextual factors accounting for the observed differences in sexual risk among these three types of partners.

Subsequent examination of MSWs who work in other settings has presented a more complex view. Independent and agency-based MSWs (i.e., escorts) exhibit fewer HIV risk behaviors with clients than men working on the street (Minichiello, Marino, Browne, Jamieson, Reuter, & Robinson, 2000; Simon, Morse, Osofsky, Balson, & Gaumer, 1992). Further, a growing body of research has suggested that indoor MSWs engage in less HIV risk behavior with clients than they do with non-client sexual partners (Joffe & Dockrell, 1995; Parsons, Koken, & Bimbi, 2004). Most studies find that when escorts do engage in sexual risk behavior with clients, it is neither at higher levels than in their personal lives, nor at a rate greater than that of other men having sex with men, MSM (Bimbi & Parsons, 2005; Parker, 2005); although some research suggests that rates unprotected anal intercourse with non-paying partners and substance use may be higher in some groups of MSWs (e.g., men who work as MSWs and who hire other men for sex) than in men not engaging in sex work (Koken, Parsons, Severino, & Bimbi, 2005). Moreover, HIV infection rates among MSWs are not significantly higher than among MSM who do not engage in commercial sex work (Vanwesenbeeck, 2001).

Taken as a whole, findings across work venues suggest that economic factors may be a stronger determinant of HIV risk behavior in a sex work context rather than whether a person has received pay for sexual services, per se (Bimbi, in press). Although the role of economic factors in determining MSWs' HIV sexual risk behavior with clients has not been extensively studied, the fact that street-based MSWs engage in higher levels of risk than indoor workers strongly suggests a leading role for socio-economic factors in this regard. Street-based MSWs have been shown to often experience economic stress (Earls & David, 1989; El-Bassel, Schilling, Gilbert, Faruque, Irwin, & Edlin, 2000; Williams, Timpson, Klovdal, Bowen, Ross, & Keel, 2003) at a higher level than indoor sex workers (DeGraff, Vanwesenbeeck, Zessen, Straver, & Visser, 1994). With this increased stress, they may perceive less ability to maintain safer sex boundaries with clients for fear of losing vital income.

The degree of relationship and emotional intimacy between sexual partners also has been shown to be an important correlate of HIV sexual risk behaviors. Research with general population samples suggests that involvement in a steady, affectionate relationship is a strong predictor of HIV risk behavior among both MSM (Adams, Sears, & Schellenberg, 2000; Koblin, Chesney, Husnik, Bozeman, Celum, Buchbinder et al., 2003) and heterosexual men (Laumann, Gagnon, Michael, Michaels, 1994; Misovich, Fisher, & Fisher, 1997; Seal & Ehrhardt, 2004). The association between involvement in an affectionate relationship and increased sexual risk also has been documented among HIV-positive MSM (Semple, Patterson, & Grant, 2000; Wegner, Remien, & Carballo-Dieguez, 1998; Weinhardt, Kelly, Brondino, Rotheram-Borus, Kirshenbaum, Chesney et al., 2004). Men in affectionate relationships choose to have unprotected sex for reasons of pleasure, intimacy, and trust (Hays, Kegeles, & Coates, 1997; Seal, Kelly, Bloom, Stevenson, Coley, Broyles et al., 2000; Seal & Ehrhardt, 2003). MSWs report higher levels of HIV risk behavior with clients to whom they emotionally and/or physically attracted than with clients to whom they are less attracted (Joffe & Dockrell, 1995).

Therefore, as part of a larger research effort to understand the experiences of indoor MSWs, we examined the overall profile of MSWs' sexual experiences. In this article, we present data about the relative frequencies of various sexual activities and levels of associated risk with clients versus non-clients. Additionally, we expand previous research by examining sexual and sexual risk behavior within a broader context of men's social networks, current and lifetime drug and alcohol use, mental and physical health history, HIV/STI history and testing, experiences with child and/or adult victimization, and work as an escort. Questions regarding HIV risk among male escorts cannot be fully addressed, nor can comparisons between MSWs and other MSMs be adequately drawn without examining men's sexual risk behavior with both paid and non-paid partners. By placing sex for pay in the more general context of an escort's overall sexual behavior, we may begin to more fully understand the complex interplay between variables that have been associated with higher risk behavior and the escort's work setting.

Methods

Study Setting

A sample of 30 participants was recruited from a single male escort agency, which had been in operation for five years at the time of the study. Ninety-four percent of men escorting for the agency (30/32) took part in the research project. The agency was located within a small Mid-Atlantic city (population of about 49,000) that was located on the exurban fringe of a major metropolitan area, such that half of the neighboring counties were classified as rural areas by the most recent U.S. Census (2000). Escorts from the agency served approximately 180-200 clients per month with about 95% of calls taking place at a client's residence or hotel room. The remainder occurred at the agency.

Procedure

Due to issues surrounding confidentiality among a group of men ostensibly performing illegal activities, participants were recruited by the agency manager. He provided them with basic information about the purpose and methodology of the research project. MSWs interested in participation were scheduled by the manager into one of several pre-arranged appointments to meet with the first author. At this meeting, informed consent was reviewed with each participant and then data were collected. Data collection consisted of a semi-structured qualitative interview and a quantitative survey. The manager and his assistant also were interviewed to provide information about the agency and its operation.

Participants were asked to not provide any personally identifying information during their conversation with the investigator. Interviews were tape recorded and then transcribed, eliminating any personally identifiable information that had accidentally been included in the discussion. Audiotapes were then erased and destroyed. After their interviews, participants completed the two surveys. Time was provided afterward for MSWs to ask additional questions about, and to provide feedback on, the research project. These debriefing conversations were not recorded. MSWs were provided US$60.00 to compensate for their time and to acknowledge the sensitive nature of the information that they were asked to share.

Measures

Semi-structured, Face-To-Face Interview

The face-to-face interview took between 60 and 90 minutes to complete. Participants were first asked to provide basic demographic data, including detailed information about their educational and employment history. Next, in-depth information was elicited about men's (1) social networks (e.g., family, friends, dating partners); (2) current (past month) and lifetime drug and alcohol use; (3) mental and physical health history, including HIV/STI history and testing; (4) experiences with childhood trauma (e.g., sexual or physical abuse, domestic violence between caregivers) or adult victimization (e.g., sexual assault, crime victimization); and (5) work as an escort.

Quantitative Surveys

Upon finishing the qualitative interview, participants completed two self-administered quantitative surveys which together took about 45 minutes to complete. The first survey assessed sexual behavior in the past 30 days with both clients and non-paying partners. Information was collected about number of sexual partners, number of acts of intercourse, percentage of acts that included use of a condom, and the percentage of acts that involved semen exchange. Questions were delineated by partner type (client, non-client), type of intercourse (oral, anal, vaginal), and receptive or insertive partner (when applicable).

The second survey assessed mental health status using the Symptom Checklist-90-Revised (SCL-90-R). The standardized self-report SCL-90-R checklist asks respondents to indicate the severity to which they have been distressed by any of 90 different psychiatric symptoms in the past week. A summary score (the Global Severity Index, GSI) and scores for nine subscales (Somatization, Obsessive-Compulsive, Interpersonal Sensitivity, Depression, Anxiety, Hostility, Phobic Anxiety, Paranoid Ideation, and Psychoticism) are calculated as T-scores (mean = 50, SD = 10 using non-clinical as well as clinical norms). In two studies (Derogatis, 1994), Cronbach's alpha for the nine SCL-90-R subtests averaged .84 (range: .77-.90) and .86 (range: .80-.90), respectively. Men were rated as having a clinical level of psychiatric symptoms if they had a T-score of 70 or higher on the GSI, or (2) produced a T-score of 70 or higher on any two SCL-90R subscales (Derogatis, 1994).

Data Analyses

Qualitative data analyses

Analyses of the semi-structured interviews were guided by an a priori coding scheme that was developed in correspondence with the open-ended questions used to elicit information relevant to the primary topics of interest (described above). Transcripts were initially coded by four raters. Inter-rater discrepancies were discussed until consensus about the appropriate code was obtained. Most of the discrepancies involved the omission or confusion of a specific code value by one of the coders rather than disagreement about a major category (i.e., whether or not a code should be applied or about its related theme). Decision trails were documented to assure that interpretations were supported by the data (Sandelowski, 1986; Hall & Stevens, 1991). This process was repeated until the four raters consistently achieved 80% or greater concordance on two successive transcripts (occurring at transcript five), after which the transcripts were rated by a single evaluator. Subsequent to the completion of coding, one of the remaining transcripts was randomly selected for evaluation by all of the raters in order to monitor continued concordance. This was confirmed at the 80% agreement level. Coding then proceeded until all transcripts had been completed. Subsequently, the codes present in each participant's interview were entered into an SPSS database for further analyses.

Descriptive data and analyses

We first present descriptive information about participants' demographic characteristics, mental health status and experiences, substance use behavior, HIV testing and infection history, and escort work behavior. We then report men's sexual behavior with clients and non-clients, and summarize comparison of behavior with these two types of partners. Finally, we report on univariate analyses of possible correlates of sexual behavior. Specifically, we assessed five participant characteristics: race (White versus non-White), sexual identity (gay, non-gay), education level (continuous), non-escort employment status (employed, not employees), and relationship status (steady partner, none). We also assessed five mental health variables: SCL-90R score (below clinical cut-off, above), childhood history of physical or sexual abuse (yes, no), history of victimization as an adult (yes, no), history of substance abuse (yes, no), and history of family violence or substance abuse (yes, no). Third, we examined eight substance use variables: any substance use (yes, no), any alcohol use (yes, no), any drug use (yes, no), and any drug use other than marijuana (yes, no). Each of these four variables were evaluated separately for current (past 30 days) and past (lifetime excluding the past 30 days) behavior.

All non-descriptive quantitative analyses were conducted using non-parametric statistical procedures in order to control for possible violations of the assumption of normality associated with the small sample size. These procedures included use of Wilcoxon Chi-Square (continuous outcomes-continuous predictors), Kruskal-Wallis Chi-Square (continuous outcomes-nonlinear categorical predictors), and Spearman Rank Correlation (continuous outcomes-linear categorical predictors; linear categorical outcomes-continuous predictors). For all other comparative analyses, standard Chi-Square statistical procedures were used. The statistical procedure used for each reported outcome is stated in the results section.

Results

Participant Characteristics

Demographics

As seen in Table 1, men ranged in age from 18-to-35 years-old (median = 22.4 years). The sample was predominantly Caucasian and the majority self-identified as homosexual. Most had completed high school and most had full-time jobs other than escorting. All men had current housing. Slightly over half of the MSWs were in a steady relationship. Of these men, 88% said their partner was aware of their escort work.

TABLE 1.

Characteristics of Agency-Based Male Sex Workers (n = 30)

Characteristic	Number (%)
Age
18–21	12 (40.0)
22–25	13 (43.3)
26–35	5 (16.7)
Education
Less than High School	4 (13.3)
High School	14 (46.7)
Current College Student	5 (16.7)
Two-Year College Degree	4 (13.3)
Four-Year College Degree	3 (10.0)
Ethnicity
African-American	2 (6.7)
Caucasian	21 (70.0)
Hispanic/Latino	3 (10.0)
Other/Mixed	4 (13.3)
Mental Health
SCL-90-R in Clinical Range	14(63.3)
Abused as Child	20 (66.7)
Physical Only	14 (63.3)
Sexual Only	3 (10.0)
Both	3 (10.O)
Violence/Drugs - Parents	12 (40.0)
Victimization - Adult	6 (20.0)
Non-MSW Employment
Unemployed	10 (33.3)
Part-Time	8 (26.7)
Full-Time	12 (40.0)
Relationship Status
In Relationship	16 (53.3)
Partner Knows of Escorting	14(87.5)*
Sexual Orientation
Bisexual	6 (20.0)
Heterosexual	5 (16.7)
Homosexual	19 (63.3)

^*Percentage of men in relationships (14/16)

Mental Health

Nearly half of the men scored above the clinical cut-off for psychiatric distress on the SCL-90R during the week prior to the study (shown in Table 1). Two-thirds of men reported a history of childhood sexual or physical abuse, while six men reported victimization experiences after the age of eighteen. A third of the participants reported a current or past problem with substance abuse. Forty-percent of men indicated a family history of domestic violence or substance abuse.

HIV and STI Testing and Infection History

Twenty-six participants (86.7%) had been tested for HIV. Among men who had obtained an HIV test, six (23.1%) had been tested only once, 15 (57.7%) obtained testing on an irregular basis, and five (19.2%) got tested on a regular schedule. Reasons for being tested or not being tested were not assessed. One person reported being HIV-positive. He had not escorted since learning of his serostatus two months prior. The remaining men said that they were HIV-negative. Six men (20%) said they had a lifetime history of STI infection. Of these men, two indicated they had pubic lice. The other four stated having either urethral gonorrhea or Chlamydia.

Escort work

About a fourth of participants had been working as an escort for over a year, whereas slightly more than half had been involved in escort work for three to twelve months. The average number of calls was relatively evenly distributed between those who reported one-two calls per month or less, one-two calls per week, and three or more calls per week.

Substance Use

About two-thirds of men reported using one or more substances in the past 30 days (see Table 2). Alcohol and marijuana were the most commonly used substances. About one-fourth of the men also reported use of substances other than alcohol or marijuana, none of which were injected. Combining both current and past substance use, nearly three-fourths of participants reported a lifetime history of substance use. More than half of the men reported lifetime use of substances other than alcohol or marijuana. One man reported lifetime injection heroin use. Slightly more than one fourth of men reported alcohol or drug use, during a call in the past 30 days.

TABLE 2.

Past and Current Substance Use Among Agency-Based Male Sex Workers (n = 30)

Alcohol and Drug Use
Substance	Number (%)
	Past 30 Days	Lifetime
No reported use	12 (36.7)	8 (26.7)
Alcohol	14 (46.7)	19 (63.3)
Cocaine/Crack	6 (16.7)	12 (36.7)
MDMA (ecstasy)	3 (10.0)	10 (33.3)
Marijuana	10 (33.3)	17 (56.7)
Other*	6 (20.0)	9 (30.0)
Use of drugs other than alcohol and/or marijuana	7 (23.3)	16 (53.3)
Self-reported substance use problem	8 (26.7)	10 (33.3)

^*non-alcohol depressants LSD/hallucinogens, and/or methamphelamine

Sexual Behavior-Prevalence and Frequency

Sexual behavior with clients

Table 3 presents participants' reported frequency of oral and anal intercourse with paying and non-paying sexual partners. Most men reported engaging in at least one of these activities with a paying sexual partner in the past month (mean = 6.1 partners; range: 0-27). Among men who reported having sex with one or more clients in the past month, approximately three-fourths of all appointments involved some type of oral or anal intercourse. All men who reported oral sex with a client had received oral sex, and about two-thirds had performed oral sex. Men were fairly evenly divided as to the role they took during anal sex with clients. The three men who did not report having sex with any clients during the assessment period all said they intended to do so in the near future, or had done so within the past two months.

TABLE 3.

Sexual Behavior in the Past 30 Days Among Agency-Based Male Sex Workers (n = 30)

Sexual Activity	Clients	Non-Clients
Median # of Partners*
men	3.0	2.0
women	0.0	2.2
Had Any Sexual Intercourse	90.0% (n=27)	86.7% (n=26)
male partners only**	100.0% (n=27)	66.7% (n=20)
female partners only**	0.0% (n= 0)	16.7% (n= 5)
both male and female**	0.0% (n= 0)	3.3% (n= 1)
Had Any Oral Intercourse	86.7% (n=26)	80.0% (n=24)
insertive only**	30.0% (n= 9)	0.0% (n= 0)
receptive only**	0.0%(n= 0)	0.0% (n= 0)
both**	56.7% (n=17)	80.0% (n=24)
Had Any Anal Intercourse	46.7% (n=14)	66.7% (n=20)
insertive only**	13.3% (n= 4)	13.3% (n= 4)
receptive only**	13.3% (n= 4)	20.0% (n= 6)
both**	20.0% (n= 6)	33.3% (n= 10)
Had Any Vaginal Intercourse	0.0% (n= 0)	16.7% (n= 5)
Had Sex Without A Condom
gave oral to male partner	50.0% (n=15)	63.3% (n=19)
receptive oral sex	80.0% (n=24)	63.3% (n=19)
anal receptive	3.3% (n= 1)	26.7% (n= 8)
anal insertive	3.3% (n= 1)	33.3% (n=10)
vaginal	0.0% (n= 0)	16.7% (n= 5)
Proportion of Sexual Acts With No Condom Use**
gave oral to male partner	.88	.95
receptive oral sex	.91	.95
anal receptive	.01	.56
anal insertive	.01	.41
vaginal	NA	.65
Had Sex With Semen Exchange
gave oral to male partner	3.3% (n= 1)	26.7% (n= 8)
receptive oral sex	23.3% (n= 7)	30.0% (n= 9)
anal receptive	0.0% (n= 0)	20.0% (n= 6)
anal insertive	0.0% (n= 0)	33.3% (n=10)
vaginal	0.0% (n= 0)	16.7% (n= 5)
Proportion of Sexual Acts With Semen Exchange**
gave oral to male partner	.01	.20
receptive oral sex	.11	.22
anal receptive	.00	.22
anal insertive	.00	.09
vaginal	NA	.35

^*Median based on men who reported having sex with specific partner type (male, female)

^**Analyses limited to men who reported behavior

As further shown in Table 3, participants almost always used a condom during anal sex with a client regardless of their sexual role in the encounter. In contrast, most participants said they had never used a condom during oral sex with a client in the past month, regardless of their role. Nonetheless, semen exchange was not typical with clients. Men reported that they rarely ejaculated into their clients' mouths, and never permitted clients to do so with them. No semen exchange was reported during anal sex with clients.

Sexual behavior with non-clients

Table 3 also contains information about participants' sexual behavior with non-clients. Most men reported having some type of oral, anal, or vaginal intercourse with a non-client in the prior month (mean = 2.1 partners; range: 0-6). Most of these men reported only male partners. About a third of all sexual partners (32.3%) were someone with whom the participant was in a steady relationship. One-time sexual partners accounted for about half (48.4%) of total partners. Fifty percent (n = 15) of participants reported multiple anal or vaginal non-client partners.

Oral sex was the most common behavior with non-clients, followed by vaginal or anal intercourse. Among participants who had sex with men, half took both the insertive and receptive roles during anal sex with non-clients, with the remainder equally divided between one role or the other. About 60% of men reported having unprotected anal or vaginal sex with a non-client during the prior month. In all, about half of anal sex acts with male non-clients occurred without a condom, although ejaculation resulting in semen exchange was reported for less than 20% of these acts. Participants said they did not use condoms for the majority of vaginal sex acts, and reported ejaculation resulting in semen exchange during about a third of these acts. Condoms were rarely used during oral sex with a male partner and no participants reported using a condom during oral sex with a female partner. The percentage of oral sex acts in which ejaculation resulted in semen exchange was two times higher with female partners than with male partners.

Differences in sexual behavior with clients versus non-clients

Participants reported a higher number of client than non-client sexual partners [Wilcoxon Signed Ranks Test, z = −3.28, p = .001]. This difference was largely attributable to oral sex behavior with clients versus non-clients [Wilcoxon Signed Ranks Test, z = −2.22, p = .027], a finding that did not change if we limited analyses to male partners only [Wilcoxon Signed Ranks Test, z = −2.48, p = .013]. Men also reported getting oral sex from more client than non-client partners [Wilcoxon Signed Ranks Test, z = −3.04, p = .002].

However, if the frequencies of sex acts with clients and non-clients differed for anal and vaginal intercourse. Participants reported a higher number of anal and vaginal intercourse acts with non-clients than with clients [Wilcoxon Signed Ranks Test, z = 3.31, p = .001]. The same result also was obtained when we limited our analyses to the number of anal sex acts only [Wilcoxon Signed Ranks Test, z = 2.65, p = .008].

Examining safer sex behavior, having at least one act of unprotected anal sex was reported more often for non-clients than for clients [Wilcoxon Signed Ranks Test, z = H3.12, p = .002]. This finding remained statistically significant regardless of whether the participant was the insertive or receptive partner [both p's < .01]. Participants engaged in less semen exchange during anal sex with clients than non-clients [Wilcoxon Signed Ranks Test, z = −3.0, p = .003], a finding that remained significant for receptive [p < .05], but not insertive anal sex [p > .05].

Correlates of Sexual Behavior

Given that men did not report any female clients, and only six men reported female sexual partners who were non-clients, the analyses presented below are limited to either all partners or male partners only as described in the text. However, we note that the primary determinant of sexual behavior with women was sexual identity. All five heterosexually, identified and one bisexually, identified man reported having sex with a female non-client, whereas no one who identified as gay reported this behavior. Data regarding correlates are summarized in Table 4.

TABLE 4.

Association of Variables with Sexual Behavior in the Past 30 Days Among Agency-Based Male Sex Workers (n = 30)

Variable (Referent)	Clients	Non-Clients
Participant Characteristics
Race (White)	Less anal sex	More anal/vaginal intercourse More oral semen exchange
Sexual Identity (Gay)	More condom use for insertive anal intercourse	Fewer sexual partners More likely to use condoms for intercourse Less semen exchange during intercourse
Education (Higher)	None	None
Non-MSW Employment (Yes)	None	None
Relationship Status (Yes)	None	More receptive anal intercourse More anal intercourse with semen exchange Less condom use for anal intercourse More oral sex
Mental Health
SCL-90-R (Clinical Range)	None	None
Childhood Victimization (Yes)	More oral sex partners and acts	None
Adult Victimization (Yes)	More condoms during oral sex	More condoms during oral sex with men
Violence/Drugs In Childhood Family (Yes)	None	None
Substance Use
Any Use Ever (Yes)	More sex partners More oral sex partners and acts	Less sex with men and Fewer male partners Less anal intercourse with men Less anal intercourse with semen exchange
Alcohol–Past 30 Days (Yes)	Less oral sex	Less semen exchange during anal intercourse Less semen exchange during oral sex
Drugs–Past 30 Days (Yes)	None	None

Participant characteristics and sexual behavior with clients

Non-Whites were more likely than Whites to report having anal sex with a client [77.8% versus 33.3%; FET = .046]. Among men who had insertive anal sex with a client, gay-identified men used a condom during a higher proportion of acts than did non-gay-identified men [100.0% versus 94.5%; Kruskal-Wallis X²(l) = 4.0, p = .046].

Participant characteristics and sexual behavior with non-clients

Table 4 also shows that, compared to men who identified as gay, non-gay identified men reported a higher number of total non-client sexual partners [3.2 versus 1.4; Kruskal-Wallis X²(l) = 8.02, p = .005] and anal or vaginal non-client sex partners [2.4 versus 1.3; Kruskal-Wallis X²(l) = 4.31, p = .038]. A higher proportion of non-gay identified men also had engaged in at least one act of anal or vaginal sex without a condom [90.9% versus 36.8%; FET = .007] or which resulted in semen exchange [72.7% versus 26.3%; FET = .023]. Whites compared to non-Whites reported a higher total number of anal and vaginal sex acts combined [14.6 versus 2.2; Kruskal-Wallis X²(l) = 6.65, p = .01], a higher number of anal sex acts with a male non-client [10.0 versus 1.2; Kruskal-Wallis X²(l) = 4.45, p = .033], and were more likely to have engaged in at least one oral sex act that resulted in ejaculatory semen exchange [91.7% versus 8.3%; FET = .049], a finding that also was significant for male partners only [42.9% versus 0%; FET = .029].

Men in a steady relationship, compared to those who were not, were more likely to have engaged in receptive anal sex [76.9% versus 35.5%; FET = .033], to have engaged in at least one act of anal sex resulting in semen exchange [53.8% versus 11.8%; FET = .02], and to have had at least one act of receptive anal sex without a condom [61.5% versus 11.8%; FET = .007]. Among men who engaged in the specified behavior, men involved in a steady relationship used condoms for a lower proportion of insertive [84.3% versus 26.2%; Kruskal-Wallis X²(1) = 4.16, p = .041] and receptive [83.3% versus 21.0%; Kruskal-Wallis X²(l) = 6.57, p = .01] anal sex acts. Involved men also engaged in more acts of oral sex with any non-client [18.2 versus 9.3; Kruskal-Wallis X²(1) = 4.09, p = .043], more acts of oral sex with a male non-client [17.7 versus 7.9; Kruskal-Wallis X²(1) = 5.14, p = .023], and more oral sex acts in which the participant performed oral sex on a male partner [9.9 versus 3.7; Kruskal-Wallis X²(1) = 6.25, p = .012].

Mental health correlates of sexual behavior with clients

Victimization as a child was associated with a higher number of clients with whom the participant had engaged in oral sex [rho = .365, p = .048] and a higher number of total oral sex acts with a client [rho = .381, p = .038]. These findings remained significant for acts in which the client performed oral sex on the participant [both p's < .05], but not vice-versa [both p's > .10]. Among the 17 men reporting oral sex with a client, victimization as an adult was associated with a higher proportion of total oral sex acts during which a condom was used [rho = .393, p = .047], a finding that remained significant if we limited analyses to those occasions when the client performed oral sex on the participant [rho = .393, p = . 047].

Mental health correlates of sexual behavior with non-clients

One mental health variable was significantly correlated with any non-client sexual behavior: men with a history of familial violence or substance abuse used condoms for a greater proportion of oral sex acts with a male partner [rho = −.508, p = .022], a finding that also was statistically significant for acts in which the participant performed oral sex on his partner [rho = − .508, P = .022].

Substance use correlates of sexual behavior with clients

Two substance use variables were significantly associated with men's sexual behavior with clients. Past drug use was associated with a higher number of total sex partners [rho = .456, p = .011], oral sex partners [rho = .523, p = .003], and oral sex acts [rho = .575, p = .001]. The positive association between past drug use and a higher number of oral sex partners and oral sex acts remained significant regardless of whether the participant received or performed oral sex [both p's < .05]. Men who had used alcohol in the past 30 days were less likely than men who did not to have engaged in oral sex with a client resulting in semen exchange [7.1% versus 50.0%; FET = .013], a finding that also was statistically significant regardless of whether the participant received or performed oral sex [both p's < .05].

Substance use correlates of sexual behavior with non-clients

Any substance use in the past was associated with a decreased frequency of any sex with a male non-clients [57.1% versus 100%; FET = .029], anal sex with a male non-client [52.4% versus 100%; FET = .013], receptive anal sex with a male non-client [38.1% versus 88.9%; FET = .017], and at least one act of receptive anal sex with a male non-client that resulted in semen exchange [19.0% versus 66.7%; FET = .03]. A past history of substance use also was associated with a lower number of male anal sex partners [rho = −.367, p = .046]. Men who had a past history of alcohol use were less likely to have had anal sex with a male non-client [41.7% versus 83.3%; FET = .045].

Men who reported any substance use in the past 30 days compared to those who did not were less likely to have had at least one act of anal sex with a male non-client that resulted in semen exchange [15.8% versus 54.5%; FET = .042], a finding that also was observed for oral sex [15.8% versus 54.5%; FET = .042]. Men who reported alcohol use in the past 30 days compared to those who did not were less likely to have had at least one act of oral sex with a male non-client that resulted in semen exchange [7.1% versus 50.0%; FET = .017], a finding that also was observed for oral sex acts in which the participant performed oral sex on his partner [0% versus 54.5%; FET = .003]. More generally, current alcohol use was associated with a lower proportion of oral sex acts with male non-clients that resulted in ejaculatory semen exchange [rho = −.659, p = .002], a finding that remained statistically significant regardless of whether the participant received or performed oral sex [both p's < .005].

Discussion

We examined substance use, mental health, and sexual behavior among agency-based male escorts located in a small U.S. city. The men in our study engaged in relatively little sexual risk behavior with clients; participants reported no semen exchange during anal sex and very little during oral sex. Condoms were used during nearly all anal sex acts with clients, and escorts reported significantly fewer acts of intercourse per partner with clients than with non-clients, even though the total number of partners was not significantly different.

Although there was more risk behavior with non-clients MSWs frequency of sexual risk behavior with these partners appeared to be equivalent to comparable groups of young adult men (Koblin et al., 2003; Seal & Ehrhardt, 2004). Condoms were regularly used with non-client sexual partners by most men and the occurrence of semen exchange was limited. This pattern was particularly evident for anal sex with male partners, less so for vaginal sex. Gay-identified men reported higher rates of condom use in their sex lives at home and at work, whereas non-gay-identified men indicated a relatively higher number of non-client partners and somewhat lower rates of safer sex behavior. Comparing the rates of semen exchange to condom use with non-clients shows that insertive partners, regardless of orientation, often were withdrawing prior to ejaculation when they were not using condoms. Men also reported about one-third less, on average, personal than work-related sexual partners.

Fewer than half of non-paying sexual partners reflected a one-time encounter. Most non-paying partners were in a relationship of some type with the MSW. As in other studies, relationship status with a partner seemed to be an important mediating variable for sexual risk behavior. MSWs in relationships engaged in more sex overall and had riskier sex outside of their escort work than men who were not in a relationship. Given increased sexual risk behavior with romantic partners compared to non-paying one-time partners or to clients, it appears that desires for pleasure and intimacy and a sense of trust with relationship partners found to be associated with increased risk behavior in non-sex work samples also mattered for MSWs in our study.

In sum, despite the presence of many characteristics that have been associated with higher risk in previous research involving MSWs (e.g., mental distress, trauma, substance use), relatively low levels of sexual risk behavior were observed among men in this study. Although there were some differences in the frequencies of certain sexual behaviors among men who had experienced victimization (e.g., a higher frequency of oral sex with clients), there was no corresponding augmentation of risky sexual behavior.

Rates of drug use also were high for the men in our study. Participants in our study reported that they used non-alcohol drugs (especially cocaine and ecstasy) on a relatively frequent basis. Moreover, nearly a fourth of our sample admitted that their current alcohol and substance use caused significant problems in their lives. What was surprising to find was that substance use was related to decreased sexual risk behavior in terms of higher use of condoms, less semen exchange, fewer non-paying sexual partners, and less anal sex (particularly receptive anal sex). Further, drug or alcohol use on client calls was not typical, occurring only about 25% of the time. Perhaps men with a history of using substances were more skilled at navigating potentially risky situations, and used this experience during sex with clients to further reduce sexual risk.

Overall, our data suggest that escorts employed a rational decision-making process during sex with both client and non-client partners. From a rational decision-making perspective, people are expected to select the highest reward behavior associated with the fewest negative outcomes (Calhoun & Weaver, 1996; Pinkerton & Abramson, 1992). This view would suggest that the MSWs in our study saw less incentive and/or more risk in sexual situations with clients relative to non-clients. Thus, men engaged in fewer acts of anal sex (especially in the receptive role), permitted less semen exchange, and used condoms significantly more often than with clients versus non-clients.

Although research indicates that escorts can earn more income from performing higher risk activities (Bimbi & Parsons, 2005), this potential financial incentive did not motivate men in the current study to engage in HIV risk behavior with clients. In contrast, somewhat greater risk was observed with non-clients, particularly within the context of a steady relationship. Previous research with MSWs has found decreased safer sex behavior among partners who were physically and emotionally attracted to one another (DeGraff, Vanwesenbeeck, van Zessen, Straver, & Visser, 1994; Joffe & Dockrell, 1995), conditions which one might reasonably assume occur more frequently with non-clients. Without as much emotional or erotic incentive with clients, escorts may have had less motivation to engage in higher risk sexual activities when on the job. Moreover, MSWs in our sample were not lacking in basic survival needs. Thus, there was less economic incentive to engage in unsafe sexual practices for increased income.

Even though there was more risk behavior with non-clients, escorts still appeared to employ in harm-reduction strategies when having sex outside of work. For instance, insertive partners frequently withdrew prior to ejaculation, and condoms were used for many acts of intercourse. Harm-reduction strategies with non-client partners permitted some involvement in what might be higher incentive, albeit riskier sexual behaviors (e.g., anal or vaginal intercourse without a condom), while limiting the degree of risk by constraining the highest risk aspects of the interaction (e.g., preventing semen exchange).

The lack of an association between sexual risk behavior, mental health, and substance use among men in our study is somewhat contradictory to previous findings. In part, this discrepancy may reflect a confound between mental health and substance problems with work setting and economic situation. Most studies reporting an association between mental health and substance abuse problems and increased risk behavior have been conducted with MSWs working on the street. Not only are mental health and substance difficulties higher among street- versus agency-based MSWs, the latter group is more likely to possess basic needs, such as food or shelter. Without such basic needs, cost-benefit analyses related to safer sex behaviors would be altered such that the potential loss of income resulting from enforcement of safer sex boundaries could be unacceptable if that income were needed to provide for one's survival needs. As most of the men in our study were employed outside of escorting and used their sex work income to supplement other sources of revenue, there may have been less economic incentive for MSWs to cross safer sex boundaries with clients. This may not be the case with other groups of MSWs, especially those experiencing homelessness or other socio-economic difficulties and may, in fact, be one of the primary determinants of their safer sex behavior with clients. Further research will be needed to specifically address this issue.

Limitations of our study include the relatively low number of participants, all of whom worked for the same agency. Thus, the extent to which our findings would generalize to other MSWs is uncertain. Ethnic and cultural diversity in the sample also were limited, perhaps by the agency's location in an exurban area adjacent to several rural counties. Thus, our data are probably best viewed as a single snapshot in time of a particular group of young men working for a particular escort agency.

Despite these shortcomings, our findings concur with current research showing that indoor MSWs pose relatively little HIV risk to their clients. Such results argue for a holistic focus on the male sex worker as more than a potential HIV risk to the community, looking instead at how MSWs' needs and varying work situations affect their overall well-being. Moreover, our research suggests that economic factors (such as the availability of non-sex work income, socio-economic status, and the degree to which sex work income is used to obtain survival needs) play a critical role in determining the type of sexual behavior and associated level of risk with paying sexual partners for MSWs. As with other people employed in an underground sector of the economy, the basic rights and welfare of MSWs have not often been a public concern. Risk behavior observed in earlier studies or with men working on the street may reflect economic and social disadvantage instead of problems arising directly from sex work itself. Thus, we believe that attention to the male sex worker as a whole individual engaged in a particular occupation within his individual socio-economic and relational context is long overdue.