THE PATTERN OF ACQUISITION of sexual partners and the number of partners have been shown by mathematical models to be important in the establishment and spread of sexually transmitted diseases (STD). 1,2 One HIV model has shown that a large number of overlapping partnerships in the population will accelerate the spread of disease throughout the sexual networks. 3 For example, a person with more than one sexual partner during a short period establishes a sexual network in which all members are at risk if one member acquires an STD. In contrast, someone with the same frequency of intercourse and the same number of partners will not form a similar network structure if the partners are serial because earlier partners are unaffected by subsequent introduction of an STD. Thus, it is important to examine sexual behavior beyond aggregate numbers.
No standardized classification of partnership patterns exists. Previous studies have classified partnership patterns based on combinations of steady and casual, 4 sequential or concurrent, 5,6 and the degree of monogamy, 7 and assessed their association with prevalent or incident sexually transmitted infection (STI). Each of these classification systems is based on different constructs and requires decisions about an interpersonal aspect of each relationship and the time frames in which these relationships occurred.
Sexual behavior is an important component in the study of STDs. Most previous studies used interview or survey instruments to describe and quantify each subject’s behaviors. The items on these instruments generally refer to a particular period (e.g., the previous 2 months) or try to assess “usual” behavior patterns for each subject, including some interpersonal aspect of the relationship. Typical areas of inquiry included the number of sexual partners, frequency of intercourse, “type” of partner, and the use of condoms. Coital diaries are an alternative data collection method that requires data to be recorded on a daily basis. For aggregated behaviors, diary and questionnaire data have excellent concordance. 8 However, the time-specific information available from coital diaries allows for a much more detailed description of sexual behavior, particularly for examining the temporal sequence of events.
The purpose of this report is to examine the coital diary data collected during the course of a 7-month longitudinal study of young women at high risk for STD 9 and to describe their sexual behaviors, with particular attention to issues of partner sequence and overlap.
Four hundred forty-four adolescents (108 males, 336 females) infected with or having a sexual contact infected with Neisseria gonorrhoeae, Chlamydia trachomatis, or Trichomonas vaginalis attending the STD clinic or one of four neighborhood adolescent health clinics in Indianapolis, Indiana were enrolled in a longitudinal study at the time of treatment. Written informed consent was obtained from all patients in accordance with US Department of Health and Human Services guidelines and a document approved by the Indiana University–Purdue University at Indianapolis Committee on Protection of Human Subjects. Repeat visits were scheduled at 1, 3, 5, and 7 months. At enrollment and each interim visit, subjects were given a daily diary in which they were to record the occurrence of sexual intercourse, partner initials, condom use, and use of alcohol or drugs before sex. Separate entries were made if more than one coital event occurred on a given day. The coital diary was a pocket-sized calendar, and data were to be recorded in the square corresponding to the day on which an event occurred. The analysis is limited to the 242 women (72.0%) who returned at least one diary in the 7-month period. Because not all women returned all of the diaries, analyses are presented either by time unit (e.g., per week) or stratified by total diary length.
One key component in the description of sexual behavior is the pattern of sexual partners beyond the total number. Because there are many possible patterns that are not easily categorized into homogeneous sets, and because terms like serial monogamy and concurrent partners do not have strict definitions, we chose to initially describe partner patterns based on runs of sexual encounters. Runs have a long history in statistics; the earliest reference, which is attributed to Karl Pearson 10 by Mood 11 in his report detailing the history of the runs statistic, develops its distribution. The total number of runs has been used as a statistic primarily to test for randomness in a series of data points. The basic runs test appears in many statistics textbooks, 12 and many modifications have been developed including a two-sample version, 13 use of the longest run, 14 number of runs up and down, 15 and a permutation test based on smoothed data. 16 We do not test the randomness of our data, but use the number of runs as a descriptive statistic for partner patterns.
A run is defined as a series of sexual encounters with the same partner (not necessarily on successive days) that is not interrupted by an encounter with a different partner. A new run begins each time a switch of partners occurs, whether to a new or existing partner. For a given subject the number of runs must be greater than or equal to their number of partners. Monogamy would be a single run of lifelong duration. When the number of runs exceeds the number of partners then some overlap in partnerships has occurred. Based on theoretical models that include the proportion of overlapping partnerships, 3 the more runs a person has, the more likely he or she should be to acquire and transmit one or more STDs.
The baseline characteristics for the 242 young women included in this study are presented in Table 1. The study population was predominantly black, and more than 80% had an STI at enrollment. The mean age of first sexual intercourse was approximately 14 years, and the median number of lifetime sexual partners was five. Those who returned diaries did not differ significantly at baseline from those who did not (n = 94) in racial distribution, number of lifetime partners, or number of partners in the previous 2 months. However, those who returned the diary were half a year younger on average (17.5 versus 18.0 years;P = 0.024) and had an earlier age of sexual debut (14.1 versus 14.4 years;P = 0.044). Those who returned diaries were also more likely to have an STI at enrollment (82.5% versus 62.0%), and the prevalence of chlamydial infection was also higher among those returning diaries (54.6% versus 34.0%). The prevalence rates of other STIs at enrollment did not differ.
Not all of the 242 women who returned diaries did so for the entire 7-month duration of the study. The mean number of days covered was 121 (SD, 70 days). The distribution of days of diary coverage (data not shown) shows clear “peaks” at 1, 3, 5, and 7 months, which correspond to the study visit schedule and the most likely points for loss-to-follow-up examination. Examining the coverage data in 2-month intervals, 25.6% of the diaries covered 2 months or less, 27.7% covered 2 to 4 months, 17.8% covered 4 to 6 months, and 28.9% covered more than 6 months. Thus, as noted in the methods section, the diaries for individual subjects present varying observation periods. To maximize the available data, we present some analyses by week and some that are stratified by observation period (< 2 months, 2–4 months, 4–6 months, and > 6 months).
Distribution of Sexual Encounters
The average weekly number of coital events reported in the returned diaries is 0.94. Figure 1 presents the weekly average number of coital events over the 7-month follow-up period. The solid line represents a smoothed curve that was fitted using spline functions. Although this curve is not linear, there are no clear time trends and the weekly frequency of intercourse is essentially constant. In addition, there is no apparent difference between the early weeks, time after the diagnosis of the STI, and the rest of the period. We also examined the distribution of coital events over the days of the week to determine if sex occurred more frequently on certain days (e.g., weekends). The distribution is remarkably even, ranging from 11.1% of the Mondays to 12.5% of the Fridays. Using a chi-square test, the distribution per weekday does not differ significantly from a uniform distribution (P = 0.43).
The number of partners per subject was related to the length of the diary, but the relationship was not linear. Therefore, the data cannot be adequately summarized as the number of partners per month or other time unit. Table 2 presents the distribution of the number of partners stratified by the length of the diary. In all four strata the median number of partners was 1.0. In addition, the reporting of five or more partners was not a common occurrence even for diaries covering more than 6 months. The mean number of partners increases with diary length and is less than two for all of the strata, except for the diaries that cover more than 6 months.
Subjects without a partner would not have any runs, and subjects with one partner can have only one run. For those 132 women with one partner, the length of the partnership increased with diary length. The run lasted an average of 15 days in those diaries covering less than 2 months and 126 days for diaries covering more than 6 months (Table 3). The average percentage of the diary that the relationship covered (calculated as the time between the first and last coital event recorded) among these women with one partner was approximately 50% but varied in the four strata from 40% to 60%. Both the run length and percentage are underestimates of the true length of the relationship because the data may be censored at one or both ends (i.e., the relationship may have started before or continued beyond the diary).
The 72 subjects with two or more partners allow a more detailed examination of partnership overlap. There were 43 women who reported two partners in their coital diary. Of those women, 34 (79%) had only two runs, the minimum possible amount. Six women (14%) had three runs, indicating that they returned to their first partner after at least one coital event with a second partner. Three subjects had repeated switches between their two partners (one subject with five runs, two subjects with eight runs).
Of the 11 women reporting three sexual partners in their diary, 8 (72%) had only three runs. The remaining 3 women had four, five, and nine runs. Nine women reported four or five partners, four of whom (44%) had only four runs. There were also two subjects with six runs and one subject each with seven, 11, and 12 runs. For the 11 women reporting more than five partners, the number of runs ranged from seven to 37.
For each new run (i.e., partner switch) we examined the number of days between the first encounter of the new run and the last encounter of the preceding run. Table 4 presents the percentage of runs classified by the number of days since the previous run. More than 50% of the time a switch to a new partner took more than 1 week, compared with less than 30% of the time for a switch to an old partner. The mean time between runs for all runs was 14.6 days. The time between partners differed significantly (P < 0.001, Wilcoxon rank sum test) depending on whether the switch was to a new partner or a return to a previous partner. For a new partner, the mean time of 20.6 days was much greater than for an old partner (mean, 7.9 days).
By virtue of having an STI or being a contact to a person with an STD at baseline, the young women in this study are a high-risk group. During the follow-up period, the coital diaries revealed a study population in which approximately 15% had no partners, and the median number of partners among the rest of the population was one. Although there is not a universal pattern of risk for the acquisition of STD, it seems clear that many of these young women cannot be said to be at risk simply because of risky patterns of sexual behavior such as multiple partners, overlapping partners, or high frequency of intercourse. This is not a unique finding, and other studies of similar urban clinic populations also have shown more than 50% of young women with less than two partners. 5,6,17 A probability sample of young women in Detroit showed that 70% had one or no partners in a 1-year period. 7 In addition, these results do not differ markedly from a study of a lower-risk group of single white college seniors, 30% of whom had a single partner over the 3.5-year recall period during which they attended college. 4 Our reported mean frequency of intercourse (0.94 per week) is also not extremely high and is similar to a previous study. 17
Our results showed a uniform distribution of coital events over the days of the week. This is in contrast to our previous work where we found that sex occurred most frequently on Friday and Saturday. 18 The reasons for this difference are not clear.
Overall, 69% of the women reported zero or one partner, 19% had a pattern reflecting no partner overlap, and only 12% had some overlap of their sexual relationships. For those young women with one partner, the mean length of the sexual relationship was relatively long, at least half of the follow-up period covered by the diary. Although the classifications are not always defined in the same way, these pattern proportions are remarkably consistent with other studies that used methods of data collection other than coital diaries. For a probability sample in Detroit, Norris and Ford 7 report 70% relatively monogamous, 11% serially monogamous, and 19% nonmonogamous patterns of partnership. The partner patterns in a Planned Parenthood-based study population were 70% monogamous, 9% sequential, and 21% concurrent. 17 In a previous study in Indianapolis with a longer follow-up period (≤ 21 months), the percentages were 59, 28, and 13 for one partner, serial monogamous partners, and concurrent partners, respectively. 5 A higher-risk pattern distribution was reported among adolescents attending an STD clinic; however, the data are less comparable because they represent both men and women. 6 That study also found a significantly increased risk of infection with more concurrent partners.
Serial sexual relationships that do not overlap were common for those with two (79%) and three (72%) partners and accounted for a sizable proportion of those with four or five partners (44%). The pattern of partnerships among those reporting more than one partner was predominantly serial, and not the overlapping mixing pattern that models predict to be extremely high risk for the transmission of STDs.
Based on mathematical models, concurrent partnerships are clearly important in establishing infection in a population or community. 1,2 However, for individual risk, a person’s position in a smaller (local) network that is formed by one of their partners is more important than their position in the whole sexual network. Specifically, the number of partners that the person’s partner(s) has is a key predictor of that person’s risk. 1 Unfortunately, partners were not contacted or enrolled in this study, so data regarding their behavior are not available. Nevertheless, based on the relatively low-risk behavior patterns exhibited in our study population, it seems the choice of partners and the behavior of these partners may be the key factor in STI acquisition for adolescent women. The choice of an infected partner and subsequent infection could just be a case of bad luck among our study population; however, sexual partners in a similar population have been shown to live much closer to each other than would be expected in a random distribution. 19 Thus, women in a high prevalence area would have an increased likelihood of choosing an infected partner. In addition, Laumann and Youm 20 suggest that sexual network patterns in the black community may explain the higher infection rate for STIs. Blacks who have had only one partner in the past year are five times more likely to choose a partner who has had four or more partners in the last year (dissortative partner choice) than whites with one partner. In addition, STIs may stay within the black community because their partner choices appear to be more racially segregated (assortative mating) than other groups. A recent study in which STD prevalence and incidence rates were high even among adolescents with one lifetime partner concluded that population factors may be more important than individual risk behaviors. 21
The study population reflects a group of adolescent women at high risk for STD in a single inner city. Thus, care should be taken in generalizing these results. In addition, approximately one quarter of the originally enrolled subjects failed to return a diary, which may also effect the generalizability of our data. However, the subjects who returned a diary exhibited similar behaviors at baseline to those who did not and may have been at greater risk of infection because they were more likely to be infected. Given the similarity of our results to other studies and the generally low-risk behavior patterns that were exhibited, it is likely that other populations will exhibit similar or less risky behaviors. If, as our data and past studies suggest, partner choice and population factors are more important elements than personal high-risk behavior in the high prevalence of STI among inner city adolescent women, then interventions need to be developed that account for this finding. Clearly, personal protective measures such as condoms are still important; however, condom use is inversely related to relationship quality and is less frequent with existing as opposed to new partners, 9 and many of these existing relationships may not be mutually monogamous. Therefore, the current pattern of condom use among young women is not sufficient to prevent infection. Interventions must also be designed to increase testing of sexually active men and treatment of those men who are infected or who are contacts to persons with STDs. In addition, observational studies to determine the behaviors of both members of sexual dyads and not just women would be important in the design of future behavioral and educational interventions for these men. Successful interventions for these men will be needed because their behavior appears to be among the key factors in the establishment and continuation of STD in the community.
1. Ghani AC, Swintin J, Garnett G. The role of sexual partnership networks in the epidemiology of gonorrhea. Sex Transm Dis 1997; 24: 45–56.
2. Morris M, Kretchmer M. Concurrent partnerships and the spread of HIV. AIDS 1997; 11: 641–648.
3. Watts CH, May RM. The influence of concurrent partnerships on the dynamics of HIV/AIDS. Math Biosci 1992; 109: 89–104.
4. Joffe GP, Foxman B, Schmidt AJ, et al. Multiple partners and partner choice as risk factors for sexually transmitted diseases among female college students. Sex Transm Dis 1992; 19: 272–278.
5. Howard MM, Fortenberry JD, Blythe MJ, Zimet GD, Orr DP. Patterns of sexual partnerships among adolescent females. J Adolesc Health 1999; 24: 300–303.
6. Rosenberg MD, Gurvey JE, Adler N, Dunlop MBV, Ellen JM. Concurrent sexual partners and risk for sexually transmitted diseases among adolescents. Sex Transm Dis 1999; 26: 208–212.
7. Norris AE, Ford K. Sexual experiences and condom use of heterosexual, low-income African American and Hispanic youth practicing relative monogamy, serial monogamy, and nonmonogamy. Sex Transm Dis 1999; 26: 17–25.
8. Fortenberry JD, Cecil H, Zimet GD, Orr DP. Concordance between self-report questionnaires and coital diaries for sexual behaviors of adolescent women with sexually transmitted infections. In: Bancroft J, ed. Researching Sexual Behavior. Bloomington, IN: Indiana University Press, 1997: 237–249.
9. Katz BP, Fortenberry JD, Zimet GD, Blythe MJ, Orr BP. Partner-specific relationship characteristics and condom use among young people with sexually transmitted diseases. J Sex Res 2000; 37: 69–75.
10. Pearson K. The Chances of Death and Other Studies in Evolution. London: Cambridge University Press, 1897.
11. Mood AM. The distribution theory of runs. Ann Math Stat 1940; 12: 228–232.
12. Dixon WJ, Massey FJ Jr. Introduction to Statistical Analysis. New York: McGraw Hill, 1969.
13. Wald A, Wolfowitz J. On a test on whether two samples are from the same population. Ann Math Stat 1940; 11: 147–162.
14. Mosteller F. Note on applications of runs to quality control charts. Ann Math Stat 1941; 12: 228–232.
15. Wallis WA, Moore GH. A significance test for time series. J Am Stat Assoc 1941; 36: 401–409.
16. Katz BP. A computer enhanced runs test for smoothed data. Stat Med 1988; 7: 795–804.
17. Weisman CS, Plichta S, Nathanson CA, Ensminger M, Robinson JC. Consistency of condom use for disease prevention among adolescent users of oral contraceptives. Fam Plann Perspect 1991; 23: 71–74.
18. Fortenberry JD, Orr DP, Zimet GD, Blythe MJ. Weekly and seasonal variation in sexual behaviors among adolescent women with sexually transmitted diseases. J Adolesc Health 1997; 20: 420–425.
19. Zenilman JM, Ellish N, Fresia A, Glass G. The geography of sexual partnerships in Baltimore: applications of core theory dynamics using a geographic information system. Sex Transm Dis 1999; 26: 75–81.
20. Laumann EO, Youm Y. Racial/ethnic group differences in the prevalence of sexually transmitted diseases in the United States: a network explanation. Sex Transm Dis 1999; 26: 250–261.
21. Bunnell RE, Dahlberg L, Rolfs R, et al. High prevalence and incidence of sexually transmitted diseases in urban adolescent females despite moderate risk behaviors. J Infect Dis 1999; 180: 1624–1631.