A Partner-Related Risk Behavior Index to Identify People at Elevated Risk for Sexually Transmitted Infections
- First Online:
- Cite this article as:
- Crosby, R. & Shrier, L.A. J Primary Prevent (2013) 34: 81. doi:10.1007/s10935-013-0290-7
- 205 Views
The purpose of this study was to develop and test a sexual-partner-related risk behavior index to identify high-risk individuals most likely to have a sexually transmitted infection (STI). Patients from five STI and adolescent medical clinics in three US cities were recruited (N = 928; Mage = 29.2 years). Data were collected using audio—computer-assisted self-interviewing. Of seven sexual-partner-related variables, those that were significantly associated with the outcomes were combined into a partner-related risk behavior index. The dependent variables were laboratory-confirmed infection with Chlamydia trachomatis, Neisseria gonorrhoeae, and/or Trichomonas vaginalis. Nearly one-fifth of the sample (169/928; 18.4 %) tested positive for an STI. Three of the seven items were significantly associated with having one or more STIs: sex with a newly released prisoner, sex with a person known or suspected of having an STI, and sexual concurrency. In combined form, this three-item index was significantly associated with STI prevalence (p < .001). In the presence of three covariates (gender, race, and age), those classified as being at-risk by the index were 1.8 times more likely than those not classified as such to test positive for an STI (p < .001). Among individuals at risk for STIs, a three-item index predicted testing positive for one or more of three STIs. This index could be used to prioritize and guide intensified clinic-based counseling for high-risk patients of STI and other clinics.
KeywordsCondomsMenWomenSexually transmitted diseasesSexual Behavior
Recent surveillance data from the Centers for Disease Control and Prevention (CDC) indicate that age, gender, and racial disparities continue to characterize the epidemic of sexually transmitted infections (STIs) in the United States (CDC, 2010a, b). Of STIs, chlamydia and gonorrhea are most likely to affect those under the age of 25 years, those who are female, and those who are African American/Black (CDC, 2010a, b). Beyond these basic demographic variables, however, it is difficult to create a typology of a person most at risk of STI acquisition. Such a typology could be useful for prioritizing behavioral intervention efforts to prevent STIs—a task that has been described previously (Aral, Lipshutz, & Douglas, 2007). Although it is intuitively appealing to think of unprotected sex as a key predictor of STI acquisition, a considerably large body of empirical evidence suggests a lack of robust association between condom use and STI (Crosby, DiClemente, Wingood, Lang, & Harrington, 2003; Crosby et al., 2005; Warner et al., 2004, 2005; Warner, Stone, Macaluso, Buehler, & Austin, 2006; Zenilman et al., 1995).
A host of personal and relational characteristics may contribute to the prediction of STI acquisition (Crosby, Salazar, DiClemente, & Yarber, 2004; DiClemente et al. 2005; Salazar et al., 2007; Salazar et al., 2009). For example, cognitive states such as depression, oral and anal sexual behaviors, and sexual sensation seeking may predict genital STI acquisition (Crosby et al., 2004; DiClemente et al. 2005; Salazar et al., 2009). In addition to these past research findings, it may be useful to identify other, quickly posed triage questions, such as whether individuals have had recent sex with strangers, casual partners, or in exchange for drugs or money and whether their sex partners have other partners, may be more useful for determining high STI risk (Crosby et al., 2002; DiClemente, Salazar, Crosby, & Rosenthal, 2005; Gorbach, Stoner, Aral, Whittington, & Holmes, 2002; Koumans et al., 2001; Mertz et al., 2000; Rosenberg, Gurvey, Adler, Dulop, & Ellen, 1999). Accordingly, the purpose of this study was to use partner-related risk factors to develop a risk behavior index to predict STI among individuals at elevated risk for STI and to determine whether this index has differential effects on STI risk across age, gender, and racial/ethnic groups.
Data were collected from December 2007 through April 2011 as part of a larger study of the effectiveness of condoms against non-viral STIs. A convenience sample was recruited from patients at five sites: publicly funded sexually transmitted disease (STD) clinics in two cities in the Midwest and a general hospital STD clinic and two adolescent medical clinics affiliated with a pediatric hospital in a city in the Northeast. At the STD clinics, patients aged 18 and over were recruited. At the adolescent medical clinics, patients aged 15 and over were recruited. Additional eligibility criteria included being English-speaking; willing to be tested for chlamydia, gonorrhea, and trichomoniasis by providing a urine specimen; willing to provide contact information; reporting penile–vaginal intercourse in the preceding 3 months; and providing written informed consent (parental consent was waived by the institutional review board). At the Midwestern sites, clinic staff referred patients who met all eligibility criteria to a research assistant to confirm eligibility and to further describe the study. At the Northeastern sites, the study was listed on a research recruitment flag, which was attached to the appointment paperwork of age-eligible patients. The research flag included a brief description of the study and an area for providers to report participants’ recruitment status. This information was used to keep a running list of potential participants, which was used by the research assistant to contact eligible patients. Across the three remaining clinics, 1,424 patients agreed to be screened for eligibility. Of these, 1,297 were eligible and were thus invited to participate in the study, with 795 agreeing to do so, yielding a participation rate of 61.3 %. With the remaining patients from the adolescent medical clinics (n = 133), the baseline sample size for the current study was 928. The study protocol was approved by the institutional review boards at the investigators’ institutions.
Characteristics of the Sample
The mean age of the sample was 29.2 years (standard deviation [SD] = 10.8). The majority self-identified as African American/Black (n = 617; 66.5 %) and were female (n = 510; 55.1 %). Most (61.7 %) of those aged 18 and older reported earning less than $1,000 per month in income or social assistance, and 45.0 % of those under 18 reported that they qualified for a free lunch at school. The mean number of lifetime sex partners was 29.7 (SD = 38.2), and the mean number of sex partners in the past 3 months was 2.9 (SD = 6.4). Slightly less than one-half of the sample (49.1 %) reported ever being diagnosed with an STI. Slightly more than three-quarters of the sample (77.6 %) reported having had any unprotected vaginal sex (UVS) in the past 3 months. Nearly one-fifth of the sample (169/928; 18.4 %) tested positive for Chlamydia trachomatis, Neisseria gonorrhoeae, and/or Trichomonas vaginalis.
Based on evidence suggesting the possibility of decreased reporting bias (Turner et al., 1998), data were collected using audio—computer-assisted self-interviewing (A-CASI). The A-CASI clearly and repeatedly defined sex as putting the penis into the vagina. Using a 3-month recall period, seven partner-related risk behaviors were assessed by A-CASI: (1) sex with someone recently released from prison; (2) sex with someone “you did not know at the time”; (3) sex with a “one-night stand”; (4) sex that involved an exchange of money or drugs; (5) sex with a partner known or suspected to have an STI; (6) sex with a partner known to be having a current relationship with another sex partner; and (7) having concurrent sex (defined in A-CASI as “sexual relationships with more than one person at time—meaning sex with partner A on one day, then with partner B on another day, and then with partner A again”).
To ascertain STI status at enrollment, first-catch urine specimens were collected for nucleic acid amplification testing using the BD ProbeTec ET C. trachomatis and N. gonorrhoeae Amplified DNA Assay (BD Diagnostics, Sparks MD) (Carroll et al., 1998). Aliquots were assayed for the presence of T. vaginalis using Taq-Man polymerase chain reaction. An Atlanta-based laboratory developed and validated this in-house polymerase chain reaction–enzyme-linked immunosorbent assay with established and acceptable estimates of sensitivity and specificity (Kaydos et al., 2002; Kaydos-Daniels et al., 2003).
First, bivariate associations between each of the seven partner-related variables and STI prevalence (testing positive for one or more STI versus none) were assessed using contingency table analysis, prevalence ratios, and their 95 % CI. Next, partner-related variables found to be significantly associated with STI prevalence on bivariate analysis were used to form a partner-related risk behavior index. Then, the index was tested for association with the prevalence of STIs and with participants’ reports of UVS in the past 3 months. Finally, logistic regression models were constructed to test the independent contribution of the index to the prediction of STI prevalence. The model was adjusted for age, gender, and racial/ethnic group. Age was dichotomized by a median split of 24 and younger versus 25 and older. Race was dichotomized as African American/Black versus White and other racial/ethnic backgrounds. Tests for interaction effects between the index and each of the three demographic variables were conducted with respect to STI prevalence. Additionally, self-reported history of STI and UVS were considered as covariates and retained in the model if significantly associated with STI positivity on bivariate analysis. All analyses were performed with SPSS (version 19.0), and statistical significance was defined as p < .05.
Bivariate associations between selected partner-related risk behaviors and laboratory-confirmed sexually transmitted infections
Partner-related risk behavior
% testing positive
95 % CI
Sex with a newly released prisoner (n)
Sex with a stranger (n)
Sex with a one-night stand (n)
Sex in exchange for money or drugs (n)
Sex with a partner known or suspected of having an STI (n)
Sex with a partner having sex with somebody else (n)
Sexual concurrency (n)
Sexual-partner-related Behavior Index
Based on the bivariate associations displayed in Table 1, we constructed a three-item index representing high-risk sexual-partner-related behaviors among this clinic-based sample of adolescents and adults. The index included: (1) recent sex with a newly released prisoner (2) recent sex with a person known or suspected to have an STI, and (3) sexual concurrency. Participants who reported one or more of these three behaviors (n = 379; 40.8 %) were thus compared to the remainder (n = 549; 59.1 %). Among those classified as at-risk by this index, STI prevalence was 23.6 %, compared with 14.7 % among those not classified as being at risk (p < .001). Among those classified as at-risk by this index, 81.3 % reported UVS, compared with 75.0 % among those not classified as being at risk (p = .025).
Multivariate associations of demographic characteristics and sexual risk behaviors with laboratory-confirmed sexually transmitted infections
95 % CI
African American/Black race
Age of less than 25 years
Classified as high-risk by index
Unprotected vaginal sex in the past 3 months
Interaction Effects with STI Prevalence
First, an interaction effect with gender was tested in the regression model. The effect was not significant, yielding an adjusted odds ratio of 1.83 (95 % CI [.92, 3.67]; p = .086). Next, an interaction effect with race was tested. The effect was not significant, yielding an adjusted odds ratio of 1.34 (95 % CI [.59, 3.05]; p = .48). Finally, an interaction effect with age was tested. The effect was also not significant, yielding an adjusted odds ratio of .95 (95 % CI [.48, 1.89]; p = .89).
In this study of more than 900 adolescents and adults attending publicly funded STD clinics and other clinics that diagnose and treat STDs, a three-item partner-related risk behavior index was strongly predictive of STI prevalence. Contrary to previous studies (De, Singh, Wong, & Kaida, 2007; DiClemente et al., 2002; Mehta, Erbelding, Zenilman, & Rompalo, 2003; Thomas, Weiner, Schoenbach, & Earp, 2000), this study did not find that past STI acquisition was predictive of STI prevalence. As might be expected, however, race and age were highly predictive of STI prevalence, with those self-identifying as African American/Black being more likely and those being less than 25 years of age being less likely to test positive for an STI. Gender was not significantly associated with STI prevalence. It must, however, be kept in mind that all three of these demographic variables—age, gender, and race—may function quite differently within a clinical sample. Nonetheless, after controlling for these demographic variables, a three-item risk behavior index made a significant independent contribution to the prediction of STI prevalence. The observed 1.8-fold increase in odds of testing positive occurred even after adjusting for recent UVS. Those classified as being at high risk by this index were about 90 % more likely than those not classified as such to test positive for an STI. Thus, it appears that this three-item index may be quite useful in the identification of individuals at high risk of STI acquisition.
A reasonable alternative to the use of a brief index is to simply rely on gender, race, and age as predictors of STI. In a post hoc analysis, however, the index identified 322 people who were not female, African American/Black, or younger than 25 years of age and accounted for 68 additional cases of testing positive for an STI above 42 cases accounted for merely by this demographic identification. Thus, use of the index in this sample increased the predictive accuracy of STI prevalence by more than 50 % above that achieved by demographic information alone. Consequently, it may be that asking three simple questions (i.e., Have you recently had sex with a person newly released from prison? Or with a person you know or suspect may have an STI? Or with more than one partner during the same period of time, e.g., the past 6 months) could serve as an effective method of prioritizing screening efforts for STIs.
Finding that sex with a person newly released from prison was a key component of the index is important, as this has potential implications for research and policy regarding prisons and STIs. Although entry screening of new inmates for STIs is a fairly common practice, exit screening is not common, perhaps because it would imply that sex does indeed occur in the confines of prison. Further research may also add to STI prevention efforts by investigating potential post-release safer sex programs for males.
The composition of the index is quite interesting given that four of the seven partner-related variables tested proved to be unimportant in predicting STI prevalence. Having sex with a stranger, having a “one-night stand,” exchanging sex for money or drugs, and having sex with a partner who was not being monogamous were each—surprisingly—not significantly associated with STI prevalence.
A novel component of the index is the item assessing whether people “knew or suspected” that a recent sex partner was infected with an STI. The importance of this item in the index is interesting because it clearly suggests that people may indeed have a sense of STI risk either before or after having sex with a partner. Peterman et al. (2000) initially suggested that these perceptions of risk may be accurate and, thus, may prompt condom use more often among those perceiving their partner to be risky. Our findings, however, suggest that this may not have been the case in this population of clinic attendees.
As is true for most sexuality research, findings are limited by the validity of self-report. The use of a convenience sample limits the generalizability of the findings to other at-risk populations. Also, our decision to compare African American/Black participants to those of all other racial/ethnic backgrounds, rather than to analyze each race separately, is a necessary limitation given the lack of racial/ethnic diversity other than African Americans/Blacks and Whites in our sample. Another limitation is our use of only seven partner-related risk behaviors for testing and creating the index; certainly other partner-related risk behaviors could have been assessed.
A subset of partner-related risk behaviors may be used as an index for predicting STI acquisition. Although gender, race, and age are useful indicators of likely STI acquisition, the three-item index appears to provide substantial added precision in predicting who will have one of three STIs (chlamydia, gonorrhea, or trichomoniasis). Clearly, STI intervention and screening programs cannot feasibly be directed at all sexually active persons. Instead, it is important to focus these efforts on persons most at-risk of subsequent STI acquisition. Because STIs are a product of risky sociosexual environments, an index capturing this added risk is a potentially valuable tool for prioritizing primary prevention as well as secondary prevention efforts in otherwise apparently homogenous populations.
Support for this project was provided by a grant to the first author from the National Institute of Allergies and Infectious Diseases, grant #5 R01 AI068119.