Aral, Sevgi O. PhD, MSc, MA*; Patel, Divya A.†; Holmes, King K. MD, PhD‡; Foxman, Betsy†
IN INDUSTRIALIZED, DEVELOPED COUNTRIES, the family unit has been undergoing important changes.1 These changes, termed “the second demographic transition,”2 include declines in first marriages and remarriages, increases in divorce rates, increases in proportions of women gainfully employed outside the home, decreases in proportions of couples with children and increases in proportions of childless couples, delays in age at first marriage and age at first childbirth, as well as the average age of childbirth.3–7 In the United States, the percentage of adults ages 15 and older who were married declined from 69.3% and 65.9% in 1960 to 57.1% and 54% in 2003 for men and women, respectively.8 Conversely, over the same time period, percentage of adults ages 15 and older who were divorced increased from 1.8% and 2.6% in 1960 to 8.3% and 10.9% in 2003 among men and women, respectively.8 Data collected between 1995 and 1997 in Chicago showed that in the mid-1990s, Americans ages 18 to 59 spent 50% of their lives as singles, longer than ever before; were in the sex market for longer periods and at older ages; cohabited on the average 4 years; were married an average of 18 years; and dated or searched for a partner an average of 19 years.9 Sexual activity was no longer coupled with marriage; sex markets were embedded in social space and networks, and the structure of opportunities in sex markets varied over the life course.9
Investigators in Britain measured change in specific sexual behaviors among 16 to 44 years olds in the general population.10 They conducted 2 national surveys of sexual attitudes and lifestyles in 1990–1991 (Natsal 1990)11,12 and in 1999–2001 (Natsal 2000).10 Results showed an increase over the decade of 1990s in reporting of a wide range of behaviors associated with increased risk of HIV and sexually transmitted infection transmission, including numbers of heterosexual partners, homosexual partnerships, concurrent partnerships, heterosexual anal sex, and payment for sex. For many of these variables, the magnitude of the observed change was large. The number of heterosexual partners during their lifetime increased from 8.6 to 12.7 and 3.7 to 6.5 for men and women, respectively. The number of heterosexual partners, in the past 5 years, increased from 3.0 to 3.8 and 1.7 to 2.4 for men and women, respectively. The proportion reporting homosexual partnerships increased from 1.5% to 2.6% and from 0.8% to 2.6% among men and women, respectively. The proportion reporting concurrent partnerships increased from 11.4% to 14.6% among men and from 5.4% to 9.0% among women. Among men, the proportion that reported having paid for heterosexual or homosexual sex increased from 2.1% to 4.3%. During this period, the proportion reporting condom use on all occasions during the past 4 weeks increased from 18.3% to 24.4% among men and from 14.9% to 18.0% among women. In comparison with 1990, there was a significant increase in the proportion of men who have sex with men (MSM) in the population in 2000 and among these men in the proportion reporting receptive anal intercourse in the past year.13,14
Data similar to those collected in Britain do not exist for the United States. Results of the Youth Risk Behavior Survey (YRBS), an annual survey of a national probability sample of adolescents in public and private schools grades 9 to 12, conducted since 1991, show that the proportion of females 15 to 17 years attending school in the United States who reported ever having had sexual intercourse declined from 50.6% in 1991 to 42.7% in 2001.15 Conversely, the proportion of those who had initiated sexual activity who reported having had sexual intercourse during the past 3 months increased from 74.9% to 77.1% over the same period.15 In the absence of data, it is difficult to know whether sexual risk behaviors of older American adults have changed during this time period.
Earlier descriptions of sexual behavior among American adults have revealed interesting and important racial–ethnic differences in these behaviors.16,17 Such differences may have significant impact on sexually transmitted infection rates. In the absence of temporal trend data, it is also difficult to know whether racial–ethnic differences in sexual behaviors have increased or shrunk in the recent past.
In this article, we describe the sexual behaviors of a representative sample of Seattle residents as reported in 2003–2004 and report on changes in such behaviors since 1995.
We compared the results of 2 random digit-dial (RDD) surveys conducted in the Seattle area among residents age 18 to 39 years of age with fluency in the English language, one in 1995 and the other in 2003–2004. Both surveys were conducted using similar methodologies, and identical inclusion and exclusion criteria. The RDD samples included listed and unlisted numbers randomly generated using a computer after determining all the working exchanges and working blocks within the area. Numbers were obtained from Survey Sampling, Inc. of Westport, Connecticut, for the 1995 survey and Genesys Sampling System Pennsylvania for the 2003–2004 survey. Up to 6 attempts were made to contact each number at different times of the day. Both surveys were conducted by the Social & Economic Sciences Research Center in Pullman, Washington.
In the 1995 survey (conducted between January 23, 1995, and February 23, 1995), we were able to contact 2049 (56.9%) of the 3600 telephone numbers in the initial sampling frame (1077 [29.9%] numbers were either nonresidential or disconnected; for 245 [6.8%], there were only answering machines; and 229 [6.4%] never answered). Of the 2049 contacted, 1241 (60.6%) did not meet eligibility requirements, leaving 808 eligible individuals. Five hundred forty-four of the 808 eligible individuals contacted (67.3%) agreed to participate and completed the interview. We limited our comparison to the 506 (92.9%) participants who reported ever engaging in vaginal, oral, or anal intercourse.
In the 2003–2004 survey (conducted between October 2, 2003, and January 26, 2004), we were able to contact 31,617 (84.1%) of the 37,000 in the initial sampling frame (21,294 [55.4%] numbers were either nonresidential or disconnected; 1527 [4.1%] were a data or fax line; on 1939, there were only answering machines [5.2%]; 2448 [6.7%] never answered; 482 [1.3%] were always busy; and 80 [0.2%] had telecommunication technical barriers). Of the 8683 households contacted, 6101 (70.2%) did not meet eligibility requirements, leaving 2582 eligible individuals. One thousand one hundred ninety-four of the 2582 eligible individuals contacted (46.2%) agreed to participate and completed the interview. We limited our comparison to the 1114 (93.3%) participants who reported ever engaging in vaginal, oral, or anal intercourse.
The survey instrument included questions on lifetime sexual history, partner and partnership characteristics, sexually transmitted disease history, and demographics. For almost all questions included here, the 1995 and 2003–2004 surveys had identical wording. In 2003–2004, we asked if the respondent was married to their sex partner and living together as married as separate questions. We also asked about hepatitis by type. Each survey was pretested on a sample of the study population and revised before initiating data collection. The telephone surveys required approximately 20 minutes to conduct and were administered using computer-assisted telephone interviewing software, which standardized the interview and minimized data entry errors. Respondents were given the option to select the gender of the available interviewers.
Descriptive analyses of characteristics were performed using contingency tables and chi-squared tests to assess significant differences between groups. We directly adjusted estimates for age, gender, income, marital status, and race using the combined survey populations as weights. To test if the year the survey was conducted explained the different estimates after adjusting for age, gender, income and number supported on each income, and marital status, we fit logistic regression models for binary variables. Because the distribution of lifetime number of partners, number of partners in the previous 12 months and age at first intercourse are not normally distributed, we adjusted for differences in age, income, marital status, and gender between the surveys using the negative binomial regression. The software package SAS was used to assist in summarizing the data and conduct the analyses.
The number of participants in the 2004 survey (1114) was larger than that in the 1995 survey (506) (Table 1). There were also some demographic and socioeconomic differences between the 2 samples. The 2004 sample was somewhat older; those 24 years and younger composed 25.6% of the 1995 sample but only 17.2% of the 2004 sample.
Men constituted a larger proportion (50.8%) of the 1995 sample compared with the 2004 sample (43.5%). The 2004 sample was more educated; the proportion that only completed up to and including a high school diploma was 19.9% in 1995 but only 7.4% in 2004. The 2004 sample had higher income; persons who made less than $25,000 per year constituted 41.9% of the 1995 sample compared with only 32% of the 2004 sample. There were no statistically significant (P <0.05) differences between the samples about race/ethnicity, marital status, number of children, current employment, or incarceration experience.
In 1995, after directly adjusting percentages for age, gender, income, marital status, and race, using the combined populations of the surveys as a standard, almost 40% of sexually active respondents reported having initiated sexual intercourse before age 17 and almost 22% reported having initiated intercourse after age 19; percentages initiating intercourse by specific ages were remarkably similar in 1995 and 2004 (Table 2). In 1995, respondents reported a median of 7 lifetime sex partners compared with a median of 8 lifetime sex partners reported in 2004. Between 1995 and 2004, the proportion of respondents reporting 3 to 14 lifetime sex partners increased from 51% to 57%; proportions reporting fewer than 3 and greater than 14 lifetime sex partners declined from 58% to 43%. The shift in the distribution of number of lifetime sex partners, away from the extremes and toward the middle, was not statistically significant. The reported median number of sex partners over the past 12 months was one in both surveys.
The proportion of respondents who reported sex with only opposite sex partners declined from 85.6% in 1995 to 82.8% in 2004, whereas the proportion who reported sex with only same-sex partners increased from 1% to 3.7%. The increase in same-sex partners was statistically significant (P <0.05). The proportion of respondents who reported sex with both opposite and same sex partners was 11.8% in 1995 and 12.7% in 2004 (Table 2).
The proportion of respondents who reported having concurrent sex partnerships was almost 24.5% in both 1995 and 2004. If either of the partners had other sex partners during the relationship, the partnership was considered concurrent.
Between 1995 and 2004, small declines were observed in reported history of human papillomavirus (HPV), chlamydial infection, genital herpes virus infection, syphilis, pelvic inflammatory disease, and nongonococcal urethritis. These declines and the small increase in HIV/AIDS were not statistically significant. However, the 3.4% decline in reported history of gonorrhea (from 5.8% to 2.4%) and the 8.1% decline in reported history of any sexually transmitted disease (from 28.6% to 20.5%) were statistically significant (Table 2).
The proportion of women who reported practicing vaginal douching declined from 15.5% in 1995 to 2.4% in 2004; this represents an important and statistically significant (P <0.05) change. We found no significant changes in reported vaginal intercourse during pregnancy. However, the trend suggested an increase in proportions engaging in this practice “sometimes,” whereas the proportion of those who reported “never” or “frequently” engaging in vaginal intercourse during pregnancy declined (Table 2).
Characteristics of the Most Recent Heterosexual Partnership
After directly adjusting percentages for age, gender, income, marital status, and race, using the combined populations of the surveys as a standard, we found no significant changes in sex partner recruitment patterns, as reflected in the most recent opposite sex partnership, between 1995 and 2004 (Table 3). In 2004, 3.1% of the respondents reported having found their most recent sex partner through the Internet. In 1995 and 2004, 73% and 77.6% of respondents were currently in a sex partnership, respectively. The proportion that were either married or living together as married did not change much between 1995 (53.8%) and 2004 (55.8%).
The proportion of respondents who reported that their partners knew their family and the reported time from meeting the sex partner to initiation of sexual activity increased somewhat between 1995 and 2004, but these changes were not statistically significant (Table 3).
In 1995, only 59.4% of respondents had reported condom use during first intercourse with their most recent sex partner. In 2004, 69.2% of respondents reported that they had used a condom the first time they engaged in sexual intercourse with their most recent sex partner. The increase in condom use between 1995 and 2004 was statistically significant (P <0.05) (Table 3).
The sexual repertoire of the respondents apparently changed somewhat between 1995 and 2004. Those reporting only vaginal sex declined from 15.8% to 11.8%; those who reported only oral sex increased from 1.3% to 3.5%, and those reporting vaginal and oral sex declined from 76.1% to 74.3%. These changes were not statistically significant. Conversely, the proportion who reported vaginal, oral, and anal sex increased from 4.1% to 8%; those reporting any anal sex increased from 4.3% to 8.3%, and these increases were statistically significant (P <0.05) (Table 3).
Characteristics of the Most Recent Heterosexual Sex Partner
In 1995, majority of their most recent opposite sex partners were concordant with the respondents about race, age, years of education, and residence after direct adjustment (Table 4). There was a modest trend for increased concordance about these characteristics in 2004 compared with 1995; however, these changes were not statistically significant.
In 1995, 11.7% of respondents reported that their partners had concurrent partnerships, and 20.3% reported that they themselves had concurrent partnerships. In 2004, slightly lower proportions of respondents reported concurrent partnerships for their partners (12.9%) and for themselves (18.2%). These temporal changes were not statistically significant (Table 4).
Results of Multivariate Analyses
For dichotomous variables, we used logistic regression analysis to estimate the effect of the time of the survey on behaviors and health outcomes of interest. All logistic regression models adjusted for the effects of age, gender, income, numbers supported on the income, marital status, race, and education. Results of the multivariate analyses showed that after adjustment, the proportions of women who reported practicing vaginal douching (odds ratio [OR], 0.13; 95% confidence interval [CI], 0.07–0.22), the proportion of respondents who reported history of any sexually transmitted disease (OR, 0.70; 95% CI, 0.53–0.93), and history of gonorrhea (OR, 0.48; 95% CI, 0.25–0.89) declined between 1995 and 2004. Conversely, the proportion of respondents who reported only same-sex partners (OR, 3.27; 95% CI, 1.33–9.88), condom use at first sex with their most recent sex partner (OR, 1.38; 95% CI, 1.06–1.78), and reported practicing anal sex (OR, 2.01; 95% CI, 1.21–3.48) increased between 1995 and 2004 (Table 5). For continuous variables, we fit either a linear regression or negative binomial depending on the distribution of the independent variable. After adjustment, the only significant change in the continuous variables between the surveys was a decrease in the age at sexual debut between 1995 and 2004 (χ2 = 5.01; P = 0.025).
Sexual Behavior and Sexually Transmitted Disease History Among Asians and Blacks
Numbers of Asian and black respondents in the sample were small and data for racial–ethnic breakdowns are not presented here, and percentages are not adjusted. Overall, patterns of results in these 2 groups were similar to those among the general population. The median number of sex partners during their life lifetime was 4 in both surveys for Asian Americans; this figure was 5 in 1995 and 7 in 2004 for blacks. Interestingly, among blacks, the proportion of respondents who reported 15 or more sex partners during their lifetime increased from 20.6% in 1995 to 29.9% in 2004. Similarly, among blacks, the proportion of respondents who reported 2 or more sex partners over the past 12 months increased from 27.3% in 1995 to 35.8% in 2004. These differences were not statistically significant. Among blacks, the proportion of respondents who reported only same-sex partners increased from 2.9% in 1995 to 5.9% in 2004; the proportion reporting both opposite and same-sex partners increased from 5.9% in 1995 to 7.4% in 2004. In addition, the proportion of black respondents who reported that their partners were only 2 years older or younger than themselves declined from 64.3% to 25.9% between 1995 and 2004; this difference was statistically significant.
There was an important and statistically significant decline in the proportion of black women who reported practicing vaginal douching, from 70.6% in 1995 to 25.9% in 2004 (P = 0.004). There was a similar statistically significant decline in the proportion of Asian American women who reported practicing vaginal douching from 8.3% in 1995 to 5.9% in 2004 (P <0.05). The increase in the proportion of black respondents who reported condom use during first sex with their most recent sex partner was small, from 73.3% in 1995 to 78.2% in 2004. The increase in the proportion of Asian American respondents who reported condom use during first sex with their most recent sex partner was greater and statistically significant, from 55.6% in 1995 to 67.5% in 2004. The proportion of black respondents who reported meeting a sex partner through the Internet in 2004 (3.4%) was slightly higher than the proportion in the general population (3.1%); the proportion of Asian Americans who reported this behavior was lower (1.3%).
Between 1995 and 2004, there was a statistically significant (P <0.05) decline in history of nongonococcal urethritis among blacks (from 8.3% to 0%) and gonorrhea among Asian Americans (from 6.9% to 0%).
Results of logistic regression analyses, which adjusted for the effects of age, gender, income, numbers supported on the income, marital status, race, and education, revealed interesting patterns in risk behaviors among Asians and blacks. Among Asian Americans, the declines in the proportion of respondents who reported history of any sexually transmitted disease and being in a current relationship emerged as the statistically significant changes between 1995 and 2004. Among blacks, the declines in the proportion of respondents who reported practicing vaginal douching, having sex partners only 2 years younger or older than themselves, and the duration of time between meeting a sex partner and initiating sexual intercourse were the statistically significant changes over this period.
We compared sexual behaviors, sex partnership characteristics, and characteristics of sex partners among 18- to 39-year-old persons residing in Seattle, Washington, in 1995 and 2004. Overall, we found, after adjustment for sociodemographic variables, either no change in sexual behaviors, partners, and partnerships or a shift in the direction of decreased behavioral risk.
The changes that indicate a shift toward decreasing behavioral risk include the sizable decline in the practice of vaginal douching by women, in the history of gonorrhea and any sexually transmitted disease, and the sizable increase in the reported use of condoms during first sexual intercourse with the most recent heterosexual partner.
Conversely, we found a few changes that indicate a shift toward increasing behavioral risk in distinct parameters. These include the increase in the reported initiation of sexual intercourse before age 14, the sizable increase in the proportion of male respondents who reported having sexual intercourse only with same-sex partners, and the 2-fold increase in reported practice of anal sex.
Between 1995 and 2004, Seattle residents appeared to expand their sexual repertoires. Those who reported only vaginal sex declined, whereas reports of oral and anal sex increased and reports of combination of vaginal, oral, and anal sex almost doubled. In both 1995 and 2004, the sexual repertoire of the general population appeared to be more varied than that among Asians and blacks; however, sexual repertoire expanded in all 3 populations during this period.
It is noteworthy that despite being embedded in the sociodemographic context of the second demographic transition, our respondents' sexual behaviors, overall, had not changed much or had become less risky. The differences in the temporal trends in sexual behaviors in the United States and in Britain may reflect differences in the cultural and attitudinal mood in the 2 societies during this period. The basic demographic trends, which are common to both countries, may be coupled with different sets of normative and attitudinal tendencies in the 2 countries. Although in Britain behaviors reflect increasingly liberal and tolerant attitudinal shifts, in the United States, this period may be marked by increasingly cautious and perhaps more conservative attitudinal shifts.
Some of the changes in behavior observed among minority populations over the 9-year period were different than those observed among the general population. Among Asian Americans, a shift toward less risky sexual behaviors was reflected in the declines among women who reported practicing vaginal douching, reported history of any sexually transmitted disease, and in the proportion who reported being in a sexual relationship currently, and the increase in reported condom use during initial intercourse with their most recent partner. The decrease in elapsed time between meeting a sex partner and initiating intercourse with him or her indicates a shift toward greater risk in sexual behaviors. Among blacks, although the decline in the practice of douching indicates decreased behavioral risk, the increases in the proportions of men who report same-sex partners, more than 15 sex partners over their lifetime, and in age mixing between sex partners, and the decrease in time elapsed between meeting a sex partner and initiating intercourse all reflect shifts toward greater behavioral risk. However, these findings should be interpreted with caution as a result of the small numbers of Asians and blacks in our sample.
Our findings establish population-based baselines for the use of the Internet as a sex partner recruitment tool. In 2004, 3.1% of the general population, 1.3% of Asian Americans, and 3.4% of blacks reported using the Internet to recruit sex partners. Recent trends in incident syphilis among men who have sex with men revealed that Internet-based sex partner recruitment is a frequent and important risk behavior among men who have sex with men. Syphilis outbreaks in many urban areas in the United States have been associated with Internet-based sex partner recruitment in this population.18,19 Future research will reveal whether this practice diffuses further in the general population or remains contained in distinct high-risk populations.
Our study has limitations. The participation rate for RDD surveys has declined over the recent past.20 Although 80% of potential respondents who were contacted cooperated with our study, only slightly less than half of eligible respondents agreed to participate. This result is consistent with other reports of RDD surveys. In addition, the use of answering machines and mechanisms to screen calls and block calls from unidentified numbers increases persons removed from the sample before screening and may create difficulty in measuring selection biases. At the time of enrollment, participants were told that the purpose of the study was to learn about people's behaviors that might put them at risk of sexually transmitted infections. It is difficult to know if this statement selected for particularly high-risk (or particularly low-risk) participants. Moreover, like in other surveys of self-reported behaviors and sexually transmitted disease history, the only way we can determine the validity of responses is through checks for internal consistency and interviewer assessment. It is difficult to assess the impact of these limitations on our findings. However, if the differential selection was constant for both outcomes and exposures of interest, the estimated relationships between variables should be unbiased; and if the amount and pattern of selection remained constant between 1995 and 2004, our assessments of change in behaviors should be accurate. The finding of some change in the direction of higher and lower risk is supportive of the latter assumption. The population composition differed between 1995 and 2004; to adjust for the difference, we standardized to the pooled population and also compared the distribution of parameters using multivariate techniques and found similar results.
Our findings may not be generalizable beyond residents of Seattle. However, earlier observations of both behavior patterns and sexually transmitted infection trends indicate that observations in Seattle may be predictive of patterns in the rest of the country. Finally, it is difficult to assess the direction of change with information on only 2 points in time. The period under consideration may represent continuation of existing trends or an inflection in trends indicative of upcoming change. Only future assessment can clarify this issue.
Despite the limitations of the study, our findings have important implications. The stable and lower risk behavior patterns are consistent with the trends in gonorrhea incidence in King County, Washington, between 1995 and 2002 (Fig. 1), which was in general stable. These data support the assumption that the behavioral interventions put in place after the start of the HIV epidemic have, by and large, been successful. It is important to interpret this finding with caution. When sexually transmitted infection incidence and prevalence rates are really low, considerable increases in risk behavior may take place before such changes are reflected in sexually transmitted infection incidence and prevalence.
Over the study period, several trends were divergent between the general population and minority populations, particularly blacks. For example, although in the general population, the proportion reporting more than 15 sex partners over their lifetime declined, it increased among blacks; time between meeting a sex partner and initiation of intercourse increased in the general population but declined among both Asian and blacks; age mixing appeared stable in the general population but increased among blacks. These divergent trends are concerning because increases in risk behaviors and in sexually transmitted infection incidence and prevalence in minority populations may be reflected in similar increases in the general population only after a lag. Moreover, increasing divergence in risk behaviors and morbidity between minority populations and the general population may be an indicator of minority populations becoming marginalized in general and in interactions with the public health system in particular.
Our findings also reflect the multidimensional nature of sexual behavior and heterogeneity across populations. Because some sexual risk behaviors increased over the study period, others declined. Similarly, as some preventive behaviors such as condom use increased, others such as partnering with others of same or similar ages decreased in some subpopulations. These trends underscore the importance of multipronged approaches to sexually transmitted infection prevention, targeting interventions to specific subpopulations, and potential unintended consequences of behavioral interventions.