Nelson, Sara J. MPH*; Manhart, Lisa E. PhD*; Gorbach, Pamina M. MHS, DrPH‡; Martin, David H. MD§; Stoner, Bradley P. MD, PhD∥; Aral, Sevgi O. PhD¶; Holmes, King K. MD, PhD*†**
SEX PARTNER CONCURRENCY, defined as having 2 or more sexual partnerships that overlap in time, amplifies the spread of sexually transmitted infections (STIs).1 Having concurrent partners removes the protective effect of partnership sequence,2 allowing pathogens to be acquired from 1 partner and transmitted to others during the overlapping period.3–5 Although having concurrent partners does not increase individual risk for acquiring STI beyond the risk associated with the number of partners, an individual’s concurrency may be a marker for participation in a higher risk sexual network and individuals with concurrent partners are often more likely to have STI.3,4,6,7
Concurrency has been assessed in a variety of ways. The most common methods include directly asking whether the individual had 2 or more sexual partnerships during a specified time period3,7,8; asking about start and end dates of sexual partnerships4,5,8–10; and having individuals keep coital diaries.11 Others have asked individuals if their partners have concurrent partners.5,12–14
Despite this range of concurrency measures, there have been few direct comparisons of the different methods. Overlapping dates of first and last sexual encounters between partners represents the most commonly used method and emerged from contact tracing data.4 Asking about individual partnerships, and implicitly constructing concurrency by calculating overlap, might facilitate more valid responses than asking explicitly about having concurrent partners15 and provides detail on the duration of overlap. However, this method requires accurate recall for dates of multiple partnerships, which may be challenging if partnerships are numerous or not particularly memorable. Furthermore, this approach may result in significant measurement error if the time between partnerships is short.16 In a previous random-digit dialing survey in Seattle, we found only fair agreement (κ = 0.34) when comparing concurrency assessed by a direct question versus overlapping dates.7 This suggests that the 2 measures may capture different phenomena, yet little is known about why they differ.
Given the increasing use of concurrency as a parameter in mathematical models of STI transmission, it is important to assess concurrency accurately. Understanding the differences between measures is an important step toward minimizing measurement error. Using data collected from young adult STD clinic attendees in 3 cities, we assessed agreement between the 2 most common measures of concurrency (direct question and overlapping dates), compared correlates, and explored reasons for differences between the measures.
Materials and Methods
Men and women attending public STD clinics in Seattle, New Orleans, and St. Louis, were recruited into the Young Adults, Partnerships, and STD Study between November 2001 and May 2004. Eligibility criteria included age 18 to 26 years and attending for a “new problem” (i.e., symptoms of, or suspected contact to someone with, STI). Potential participants in clinic waiting rooms were approached by a study interviewer, screened for eligibility, and provided a description of the study.
After informed consent, participants underwent an hour-long computer-assisted survey interview (CASI) consisting of questions from Wave III of the National Longitudinal Study of Adolescent Heath17 plus additional sexual behavior questions. Interviewer-administered questions elicited demographics, health characteristics, and detailed sociobehavioral information. A self-administered portion asked sensitive questions pertaining to sexual experiences, history of STI, delinquency, substance use, and detailed information about the 3 most recent romantic and sexual partners. The interview was followed by a routine clinical exam, unless patient flow patterns required that the exam occur first.
Clinical and Laboratory Methods
The clinical exam included elicitation of sexual history, symptoms, and inspection of the external genitalia. At each clinic, men provided a urine sample for Mycoplasma genitalium testing by PCR18 and a TMA assay.19 In Seattle, urine from men was also tested for Chlamydia trachomatis and Neisseria gonorrhoeae using the APTIMA CT/NG TMA assay (Gen-Probe, San Diego, CA). Among New Orleans and St. Louis men, a urethral swab was collected, and C. trachomatis and N. gonorrhoeae were detected using the Gen-Probe Pace2 and APTIMA CT/NG TMA assays, respectively. In both men and women genital herpes (HSV-2) infection was diagnosed by culture. All women received a speculum exam during which the clinician collected cervical swab specimens. C. trachomatis was detected by culture, and N. gonorrhoeae by Gram staining and culture in Seattle, by the Pace2 assay in New Orleans, and by the APTIMA CT/NG TMA assay in St. Louis. M. genitalium PCR18 and TMA assays19 were performed on a urine sample for all women and also on a cervical swab and a self-obtained vaginal swab for Seattle women. M. genitalium positivity was defined as a positive test result on any specimen by either test. Trichomonas vaginalis was diagnosed by the presence of motile trichomonads on wet-mount microscopy. All study procedures were approved by the institutional review boards at the University of Washington, Washington University, and Louisiana State University.
Measurement of Concurrency
In the self-administered portion of the questionnaire, participants listed up to 15 of their “romantic and sexual partners” in the previous 6 years, and ordered the 3 most recent sexual partners. To ensure that the partnership referenced in the direct question was theoretically the same partnership in the overlapping dates measure, analyses were restricted to participants listing 2 or more vaginal/anal sex partners. A small number who were not asked some partnership date questions because of a programming error were excluded.
Concurrency was defined as having another partner at any time during the relationship with the most recent partner and was measured by: (a) direct question, and (b) overlapping dates of the sexual relationships with the 2 most recent partners. In-depth questions asked about the dates of the first and most recent vaginal and/or anal sex with each partner (e.g., “In what month and year did you [first/most recently] have [vaginal/anal] intercourse with <Partner>?”). Concurrency measured by overlapping dates was computed by comparing the month and year of the first vaginal/anal sex encounter with the most recent partner to the last sexual encounter with the second partner. If the dates overlapped or were within the same month and year, the individual was classified as concurrent. No overlap was not concurrent. Following the date questions we asked: “Since you first had sex with <Partner>, how many other people did you have sex with during the sexual relationship?” Any response ≥1 indicated concurrent partners (i.e., concurrent), whereas a response of 0 indicated no concurrent partners (i.e., not concurrent).
Preprogrammed skip patterns required a response for each question, yet participants could opt out of a question by choosing “don’t know,” “decline to respond,” or “not applicable” as valid responses. Participants providing 1 of these responses for the direct question were classified as missing. Similarly, participants who opted out of reporting any year or month, which made it impossible to calculate overlap, were classified as missing for the overlapping dates measure. In addition, any out-of-sequence dates within a partnership (i.e., first sex happened before last sex) or between partnerships (i.e., last sex with the partner ordered “most recent” occurred before last sex with the partner ordered “second most recent”) were deemed uninterpretable and classified as missing for the overlapping dates measure.
Descriptive analyses of correlates of concurrency and missing data used Pearson χ2, Wilcoxon rank-sum, and Fisher exact tests. The κ statistic assessed the strength of agreement among concurrency measurements [0.00–0.20 (slight), 0.21–0.40 (fair), 0.41–0.60 (moderate), 0.61–0.80 (substantial), 0.81–1.00 (almost perfect)],20 and the McNemar χ2 tested whether the disagreement was random. Using multivariate logistic regression to identify independent correlates of concurrency, variables were selected for inclusion a priori if they were associated with concurrency in the literature or in univariate analyses (P < 0.10) (number of variables assessed = 38). Analyses involving missing data used only the Seattle and St. Louis data because the New Orleans questionnaire did not offer an opt-out option. STATA 9.0 was used for all statistical analyses (Stata Corp., College Station, TX).
Of the 1220 participants enrolled from Seattle (n = 605), St. Louis (n = 248), and New Orleans (n = 367), 158 (13.0%) did not answer any partner-specific questions, 203 (16.6%) answered questions about only 1 partner, 146 (12.0%) did not report vaginal/anal sex with both of their most recent partners, and 33 (2.7%) were omitted because of a programming error, leaving 680 (55.7%) in these analyses. Among these participants, 28.5% were diagnosed with an STI (M. genitalium, C. trachomatis, N. gonorrhoeae, T. vaginalis, or active HSV-2 lesions). M. genitalium was detected in 11.7%, C. trachomatis in 10.2%, N. gonorrhoeae in 6.2%, and T. vaginalis in 2.7%. The combined prevalence of STI among excluded participants (n = 540) was 38.7% compared with 28.5% among those who were included (P < 0.01).
The median age was 22 years, and 7.1% identified as gay or bisexual. Over half of the sample (52.8%) self-identified as black, and 32.5% identified as white. Over 80% of participants had a high school diploma or equivalent, and 4.1% had been married. The majority (70.5%) had been with their most recent partner for at least 3 months, and 65.0% were in a current relationship. A similar proportion of participants reported condom use at last sex with their most recent partner (39.4%) as with their second partner (41.2%).
Agreement of Concurrency Measures
Overall, 55.5% (371 of 668) of participants reported another partner concurrently with their most recent partner when explicitly asked (direct question) (Table 1). Concurrency measured by overlapping dates between the 2 most recent partnerships (53.5%; 286 of 535) was not significantly different (P = 0.38). Despite the similar prevalences, the κ statistic was 0.395, suggesting only “fair” agreement between the 2 measures. Although 69.0% (365 of 529) of participants reported concurrent partners by at least 1 measure, 29.2% (85 of 291) of those reporting concurrent partners by the direct question did not do so by the overlapping dates measure, and 26.4% (74 of 280) of participants reporting concurrent partners by overlapping dates did not report this by direct questioning. The McNemar χ2 test was not significant (χ2 = 0.76, P > 0.05) suggesting that the 2 types of discrepant responses were equally likely. Participants who reported concurrency by both measures had longer duration of overlap (mean = 22 months) than those who only reported overlapping dates (mean = 11 months) (P < 0.01). In sensitivity analyses, we added or subtracted 1 month from the date of most recent sex with the second partner. (When the second partnership began and ended in the same month, we added or subtracted 1 month from both the start and end dates; persons with missing or illogical dates were excluded from sensitivity analyses.) We found that the prevalence of concurrency measured by overlapping dates increased after adding 1 month (62.1%) and decreased after subtracting 1 month (41.9%) (Table 2). However, the agreement between the 2 concurrency measures did not substantially change (κ = 0.354 and 0.413, respectively). The mean gap between nonoverlapping partnerships was 16 months and did not differ between respondents with and without discrepant concurrency reports.
There were no differences in agreement between the 2 measures by gender. Agreement was better among St. Louis participants (κ = 0.49), whereas weaker agreement was observed among younger (18–20 years) (κ = 0.25), gay (κ = 0.23), and Hispanic (κ = 0.20) participants, as well as those without a high school diploma (κ = 0.22), without a current partner (κ = 0.30), or with a history of injection drug use (κ = −0.13).
Correlates of Concurrency
In univariate analyses, a composite diagnosis of “any STI” (M. genitalium, C. trachomatis, N. gonorrhoeae, T. vaginalis, or active HSV-2 lesions) was not associated with either concurrency measure (Table 3), nor with the respondent’s perception of their most recent partner having concurrent partners (data not shown). Furthermore, no individual STI were associated with either measure.
Several factors were consistently more common among respondents reporting concurrent partners, regardless of the measure used. These included: less than a high school education, history of arrest, recent ecstasy use, history of injection drug use, higher lifetime number of sexual partners, younger sexual debut, ever having sex the same day as meeting a partner, receiving money for sex, and self-reported history of STI. Both measures of concurrency were also associated with pregnancy and longer duration (>3 months) of the most recent relationship.
In contrast, several characteristics in these univariate analyses were associated with concurrency measured by the direct question but not by the overlapping dates measure. Although concurrency measured by the direct question was more common among participants who were black (P ≤ 0.05) or who had black partners (P ≤ 0.05), were in a current relationship (P ≤ 0.01), had experienced childhood sexual abuse (P ≤ 0.05), or had ever had anal sex (P ≤ 0.01), none of these characteristics was significantly associated with concurrency measured by overlapping dates. Notably, gender, age, sexual orientation, and marital status did not differ by concurrency status by either definition.
Multivariate Analyses of Concurrency Measurements
In multivariate analyses adjusted for study site and lifetime number of sexual partners (Table 4), only duration of the relationship >3 months and pregnancy with the most recent partner were significantly associated with both measures of concurrency. Injection drug use was the strongest correlate of concurrency identified by direct questioning, but was not significantly associated with the overlapping dates measure. Also independently associated with concurrency measured by direct question were childhood sexual abuse, being black, ecstasy use, and having a current partner. In contrast, concurrency measured by overlapping dates was significantly associated with receiving money for sex, ever having sex the same day as meeting a partner, a black partner, and condom use at last sex with the second partner. Neither current STI nor history of STI was associated with concurrency, irrespective of the measure used.
Patterns of missing data differed substantially between the 2 measures. Only 12 (2.3%) of the 529 Seattle and St. Louis participants opted not to answer the direct question, whereas 121 (22.9%) did not provide adequate data to calculate concurrency by overlapping dates. Explicitly opting out of a date question was the most common reason for missing the overlapping dates measure (52 of 529; 9.8%), followed by out-of-sequence dates between partnerships (49 of 529; 9.3%) and out-of-sequence dates within a partnership (20 of 529; 3.8%).
There were no differences in prevalences of laboratory-diagnosed STI, gender, race/ethnicity, education, or marital status between those who did and did not provide adequate data, irrespective of concurrency measure. Those missing responses to the direct question, however, were more likely to have ever received money for sex (41.7%) compared to those answering the question (12.8%) (P = 0.01). In contrast, participants missing data for the overlapping dates measure were more likely to have injected drugs (6.6% vs. 2.2%, P ≤ 0.05), and less likely to have used a condom during last sex with the most recent partner (32.2% vs. 42.2%, P ≤ 0.05) or to be in a current relationship (50.4% vs. 67.4%, P ≤ 0.01) than those providing sufficient data. They also reported fewer sex partners in the past month (mean 1.2 vs. 1.4, P ≤ 0.05) and were less likely to have an age-discordant relationship (18.2% vs. 29.2%, P ≤ 0.05) than those with complete data.
More than half of these STD clinic attendees reported concurrent partners, whether measured by a direct question or by identifying overlapping start and end dates of sexual relationships. However, agreement between the 2 measures was only fair. Neither measure of concurrency was independently associated with laboratory-diagnosed STI, and multivariate models developed using traditional significance criteria (P < 0.05) contained fairly different risk factors for each measure of concurrency. Furthermore, the amount of missing and uninterpretable data for the overlapping dates measure was substantially more than for the direct question. This likely contributed to the observed differences in statistically significant risk factors for concurrency between definitions. In fact, many (but not all) risk factors associated with 1 measure but not the other had similar odds ratios, and their lack of statistical significance was possibly due to reduced statistical power. Nevertheless, in studies relying on traditional significance testing, each measure would be associated with different risk factors.
There are several possible explanations for the poor agreement between the 2 methods. If participants did not perceive and list short-term or casual sexual encounters as partnerships, they would not have been asked about start and end dates of sexual activity. Consequently, this concurrency would not be identified by the overlapping dates measure, but may have been captured by the direct question. If this were true, we would expect a higher prevalence of concurrency using the direct question, yet the difference we observed was small. Alternatively, if 1 partnership ended and another began within the same month without overlapping, it would have been misclassified as concurrent by our overlapping dates measure. The slightly higher proportion of discrepant reports among participants concurrent by overlapping dates only, compared to that among respondents classified as concurrent by direct questioning only (29% vs. 26%), suggests this may have occurred in some cases. Another possibility is that participants could not remember or provide accurate dates of first and most recent sexual activity. Our finding that participants classified as concurrent by both measures had significantly longer average partnership overlaps than those classified as concurrent by overlapping dates alone suggests that some shorter overlaps may reflect reporting error, rather than true overlap. In sensitivity analyses, however, adding and subtracting 1 month from the last sex date of the second partner did not substantially affect the agreement between the 2 measures, possibly because the average gap between nonoverlapping partnerships was large. Education and literacy may influence the concordance between measures, and the finding that participants without a high school education had poorer agreement between measures than those with more education (κ = 0.22 vs. 0.42) supports this. Neither measure may properly identify concurrency if relationships with 1 person started and ended multiple times, or if the respondent continued to have sex with this partner after the “relationship” ended. Finally, the overlapping dates measure only considered vaginal and/or anal sex partners whereas partners referenced in the direct question could have included oral sex partners.
Problems with missing and/or uninterpretable data likely contribute to the discrepancies. Previously, using telephone interviewers, we found that 10% of participants did not answer the direct question and 19% did not provide sufficient data to calculate overlapping dates.7 Using CASI and preprogrammed skip patterns, the current study found lower rates of missing data for the direct question, but a higher proportion of missing or uninterpretable data for the overlapping dates measure. Ten percent of our study population was overtly missing the necessary data and 13% reported out-of-sequence dates either within or between partnerships. A human interviewer may have been able to reconcile these date errors and provide other memory cues to help respondents recall forgotten partners.21 Alternatively, a more sophisticated CASI instrument could be programmed to reconcile out-of-sequence dates. The dates or the partnerships could be resequenced to correct this problem, but given the uncertainty about which was correct (partnership or date order), we excluded such people.
This study expands on earlier work investigating differences between 2 common measures of concurrency. The direct question was simple and yielded relatively complete data, and the self-administered CASI methodology probably increased willingness to answer the question.22 However, it may be less appropriate in face-to-face interviewing. Inquiring about the first and last dates of recent sexual partnerships allows respondents to indirectly report concurrency, provides information on the duration of overlap, and introduces only minimal bias by inaccurate recall of dates if the time between partnerships is large.16,23 However, this method requires adequate responses to multiple questions. Incomplete data for any of these renders the overlapping dates measure missing. Although coital diaries can elicit more complete, prospective data on sexual behavior and concurrency, they are not feasible for cross-sectional and/or case–control study designs. A partner's concurrency increases risk for STI, independent of one's own concurrency or number of partners,1 but knowledge of partner behavior is often poor.12,13,24
Restricting our analyses to participants who answered questions about their 2 most recent sexual partners ensured that each respondent had the opportunity to report a concurrent partner by either measure, but excluded those with only 1 partner (not concurrent by definition) from prevalence estimates. Furthermore, excluded participants who did not answer partner-specific questions were more likely to have STI, suggesting that some of the highest-risk individuals were not included in our analyses. However, after adjusting for study site the association between STI and this excluded subset was no longer significant; thus, any bias is likely small. We did not consider partnership type, and overlap between a main and casual partner may be a more accurate measure among adolescents.6 Finally, neither of these measures captures all aspects of concurrency germane to population-level STI transmission dynamics. From a transmission standpoint, only concurrency occurring during the infectious period of a given STI is relevant, and these periods vary for different pathogens.25,26 Thus, the frequency and duration of sexual activity with each concurrent partner is important, but only the overlapping dates measure captures duration and neither measure captures frequency. Complete network data with better information on the natural courses of infectivity for specific STI is needed to more accurately determine the impact of concurrency on transmission dynamics.
Concurrency is an important parameter in mathematical modeling and should be measured consistently. Our findings demonstrate that the most common measures of concurrency are not interchangeable and, among these young adult STD clinic attendees using CASI, the direct question provided more complete data. Incorporating a human interviewer, more sophisticated CASI programming, an event-history calendar, or other memory aids may result in more accurate reporting for the overlapping dates measure, but further studies are required to empirically demonstrate this.
1. Morris M, Kretzschmar M. Concurrent partnerships and the spread of HIV. AIDS 1997; 11:641–648.
2. Morris M, Goodreau S, Moody J. Sexual networks, concurrency, and STD/HIV. In: Holmes KK, Sparling PF, Mardh PA, et al., eds. Sexually Transmitted Diseases. 4th ed. New York: McGraw-Hill. In press.
3. Gorbach PM, Drumright LN, Holmes KK. Discord, discordance, and concurrency: Comparing individual and partnership-level analyses of new partnerships of young adults at risk of sexually transmitted infections. Sex Transm Dis 2005; 32:7–12.
4. Potterat JJ, Zimmerman-Rogers H, Muth SQ, et al. Chlamydia transmission: Concurrency, reproduction number, and the epidemic trajectory. Am J Epidemiol 1999; 150:1331–1339.
5. Koumans EH, Farley TA, Gibson JJ, et al. Characteristics of persons with syphilis in areas of persisting syphilis in the United States: Sustained transmission associated with concurrent partnerships. Sex Transm Dis 2001; 28:497–503.
6. Rosenberg MD, Gurvey JE, Adler N, et al. Concurrent sex partners and risk for sexually transmitted diseases among adolescents. Sex Transm Dis 1999; 26:208–212.
7. Manhart LE, Aral SO, Holmes KK, et al. Sex partner concurrency: Measurement, prevalence, and correlates among urban 18–39-year-olds. Sex Transm Dis 2002; 29:133–143.
8. Le Pont F, Pech N, Boelle PY, et al. A new scale for measuring dynamic patterns of sexual partnership and concurrency: Application to three French Caribbean regions. Sex Transm Dis 2003; 30:6–9.
9. Flom PL, Friedman SR, Kottiri BJ, et al. Stigmatized drug use, sexual partner concurrency, and other sex risk network and behavior characteristics of 18- to 24-year-old youth in a high-risk neighborhood. Sex Transm Dis 2001; 28:598–607.
10. Adimora AA, Schoenbach VJ, Bonas DM, et al. Concurrent sexual partnerships among women in the United States. Epidemiology 2002; 13:320–327.
11. Howard MM, Fortenberry JD, Blythe MJ, et al. Patterns of sexual partnerships among adolescent females. J Adolesc Health 1999; 24:300–303.
12. Drumright LN, Gorbach PM, Holmes KK. Do people really know their sex partners? Concurrency, knowledge of partner behavior, and sexually transmitted infections within partnerships. Sex Transm Dis 2004; 31:437–442.
13. Lenoir CD, Adler NE, Borzekowski DL, et al. What you don't know can hurt you: Perceptions of sex-partner concurrency and partner-reported behavior. J Adolesc Health 2006; 38:179–185.
14. Gregson S, Nyamukapa CA, Garnett GP, et al. Sexual mixing patterns and sex-differentials in teenage exposure to HIV infection in rural Zimbabwe. Lancet 2002; 359:1896–1903.
15. Morris M, ed. Network Epidemiology: A Handbook for Survey Design and Data Collection. Oxford, UK: Oxford University Press, 2004.
16. Morris M, O'Gorman J. The impact of measurement error on survey estimates of concurrency. Math Popul Stud 2000; 8:231–249.
17. Udry JR. The National Longitudinal Study of Adolescent Health (Add Health), Waves I & II, 1994–1996; Wave III, 2001–2002 [machine-readable data file and documentation]. Chapel Hill, NC: Carolina Population Center, University of North Carolina at Chapel Hill, 2003.
18. Dutro SM, Hebb JK, Garin CA, et al. Development and performance of a microwell-plate-based polymerase chain reaction assay for Mycoplasma genitalium.
Sex Transm Dis 2003; 30:756–763.
19. Wroblewski JK, Manhart LE, Dickey KA, et al. Comparison of transcription-mediated amplification and PCR assay results for various genital specimen types for detection of Mycoplasma genitalium
. J Clin Microbiol 2006; 44:3306–3312.
20. Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977; 33:159–174.
21. Brewer DD, Potterat JJ, Muth SQ, et al. Randomized trial of supplementary interviewing techniques to enhance recall of sexual partners in contact interviews. Sex Transm Dis 2005; 32:189–193.
22. Turner CF, Ku L, Rogers SM, et al. Adolescent sexual behavior, drug use, and violence: Increased reporting with computer survey technology. Science 1998; 280:867–873.
23. Brewer DD, Rothenberg RB, Muth SQ, et al. Agreement in reported sexual partnership dates and implications for measuring concurrency. Sex Transm Dis 2006; 33:277–283.
24. Stoner BP, Whittington WL, Aral SO, et al. Avoiding risky sex partners: Perception of partners' risks v partners' self reported risks. Sex Transm Infect 2003; 79:197–201.
25. Kraut-Becher JR, Aral SO. Gap length: An important factor in sexually transmitted disease transmission. Sex Transm Dis 2003; 30:221–225.
26. Brunham RC, Plummer FA. A general model of sexually transmitted disease epidemiology and its implications for control. Med Clin North Am 1990; 74:1339–1352.