METHODS

From 1992–4, heterosexual people 14–45 years of age with culture positive gonococcal or chlamydial infection in Seattle, Washington, were recruited for study participation. Patients were drawn from the local health department’s primary STD clinic, a health department adolescent health clinic serving minority and disadvantaged youth, and a random sample of culture positive people diagnosed at other clinical facilities. People with infection (index patients) were queried about all sex partners within the past 90 days, and partners residing in King County were contacted by a research disease intervention specialist, referred for evaluation and treatment, and offered participation in the study.

Participants were asked about their own sociodemographic characteristics and sexual and drug use behaviours, as well as the perceived characteristics and behaviours of each sex partner. Individual characteristics included age, race/ethnicity, number of sex partners in lifetime and in past 3 months, and history of injection drug use, commercial sex work (exchange of sex for money or other goods), sex with a commercial sex worker, and incarceration. Women were asked whether they had had sex with a bisexual man. Partnership characteristics included duration of presexual relationship (for analysis, dichotomised as less than 1 month v 1 month or more), duration of sexual partnership (less than 6 months v 6 months or more), presexual partner knowledge (how well people had known each other before having sex, measured as very well, fairly well, somewhat, or not at all), frequency of communication about sexual issues (often, sometimes, rarely, or never), and their judgment of the overall quality of communication between partners (very good, fairly good, fair, or poor). Condom use within partnerships was examined as a percentage of use during all reported acts of vaginal intercourse, and then dichotomised for analysis (always used v not always used). Perceived partner and partnership characteristics reported by index patients were then examined for agreement with the characteristics and behaviours reported by the partners themselves. The University of Washington human subjects committee approved all procedures.

Statistical analyses were conducted using spss (SPSS v 8.0, Chicago, IL, USA) and S-Plus statistical software (S-Plus v 3.1, Seattle, WA, USA). For age and number of sex partners (continuous variables), agreement was defined as a perceived value within 20% of the stated value by the partner, and chance corrected agreement was measured by the concordance correlation (ρ). The general formula for concordance correlation is:

where S₁ is the variance of index case responses, S₂ is the variance of partner responses, m₁ and m₂ are the respective means of index case and partners responses, and S₁₂ is the covariance between index cases and partners.¹³ In general, concordance correlation values of 0.4 or more are considered to indicate relatively high statistical agreement between observations. For discrete variables, agreement was defined as concordance between the index patient’s perception and the partner’s self reported description of that characteristic or behaviour. Kappa values were calculated to measure the chance corrected degree of concordance across all partnerships.¹⁴ The general formula for calculating kappa is:

where A_obs is observed agreement between partners, A_ch is total agreement expected by chance, and N is the total sample size. In general, kappa values of 0.4−0.6 indicate moderate agreement, and values more than 0.6 indicate considerable agreement between observations.^15,16 To compare kappa values or concordance correlations between subgroups, we used permutation tests.¹⁷

RESULTS

A total of 151 people with gonococcal or chlamydial infection (index patients) and 189 of their sex partners were enrolled (table 1). Index patients were predominantly female (68%), African-American (42%), or white (40%), with a mean age of 22.2 years; most index patients had chlamydial infection (62%). Enrolled partners were predominantly male (66%) and often African-American (48%), with a mean age of 23.4 years. Gonococcal or chlamydial infection was diagnosed in 78 (41%) of the 189 partners. Similar frequencies of risk behaviours were reported by the index case and partner groups, including history of incarceration (42–51%), injection drug use (9%), and having other sex partners in the past 3 months (44–46%). Approximately one third of index patients (34%) reported knowing their partners for less than 1 month before having sex, and a majority (53%) stated that the sexual partnerships were of less than 6 months’ duration (data not shown). Only 17% of index patients reported using condoms for all episodes of vaginal intercourse.

Measures of agreement between index patients’ perceptions and partner self reports are summarised in table 2. Patients and partners had high levels of agreement for partner race/ethnicity (kappa=0.83), partner age (ρ=0.88), and duration of the partnership (kappa =0.78). Agreement was moderate for injection drug use by the partner (kappa=0.59), partner history of incarceration (kappa=0.56), frequency of communication with the partner (kappa=0.46), and condom use within the partnership (kappa=0.44). By contrast, agreement was low for potentially important STD risk factors, such as whether the partner had other partners in the past 3 months (kappa=0.29), the partner’s total number of other partners in the past 3 months (ρ=0.23), history of sex work among female partners (kappa=0.27), and history of male partners having sex with sex workers (kappa=0.22). In general, partner agreement did not vary significantly by age, race/ethnicity, sex of the respondent, or infection status of the partner. Among 57 such comparisons, the only significant differences found were: (1) greater race/ethnicity agreement among respondents >20 years old (kappa=0.75 v kappa=0.91, p<0.05 for ⩽ 20 v >20, respectively); and (2) greater agreement for history of incarceration among male respondents (kappa=0.71 v kappa=0.42, p<0.05 for male v female, respectively).

Both underestimates and possible overestimates of reported partner risk were common (table 3). For example, six (4%) of 168 people who were perceived not to be injection drug users reported a history of injection drug use (that is, index case may have underestimated partner risk), while seven (39%) of 18 people perceived to be injection drug users denied such risk behaviour (that is, index case may have overestimated partner risk). Similarly, history of commercial sex work among female partners was underestimated for nine (15%) of 61 female partners, while history of having sex with a sex worker may have been overestimated for 20 (77%) of 26 male partners. Thirty six per cent of all people thought not to have other sex partners in the past 3 months in fact reported having multiple partners. Circumcision status of male partners was also commonly misperceived: by physical examination, 14 (15%) of 92 male partners thought to be circumcised were not, while six (40%) of 15 male partners thought not to be circumcised had been circumcised (data not shown).

DISCUSSION

Agreement between patients’ perceptions of partners’ risks and partners’ self reports of risk was highest for fixed personal characteristics such as partner age and race ethnicity, as well as for partnership duration. We found lower rates of agreement for partners’ numbers of other sex partners, and for measures of communication within the partnership. Patients’ perceptions often overtly disagreed with partners’ self reports for many important partner risk behaviours. Interestingly, even the accuracy of a fixed physical characteristic such as circumcision status of male partners was relatively low. This finding mirrors the results of previous research on circumcision, where low levels of agreement have been found between self reported status and confirmation by physical examination.¹⁸

This study differs from previous investigations of validity and reliability of partner self reports by examining perception of risk within STD affected partnerships, rather than mainly within established couples. Earlier studies have focused primarily on agreement within heterosexual couples in ongoing sexual relationships.^7–11 Agreement between partners’ responses in those studies was taken as a proxy measure of “reliability” of the sexual history. However, without independent validators of sexual behaviours (often not measurable, short of direct observation), individual responses defy validation. In this sense, our data do not address the reliability or validity of patient or partner responses, but rather the extent to which both individuals agree on various characteristics of partners and partnerships. In general, agreement about presence of risk may be more reliable than disagreement about risk, or even agreement about absence of risk. We find it noteworthy that, even among people recently diagnosed with an STD, partners’ risks were often not perceived.

Our results also corroborate and build upon the findings of Ellen et al¹¹, who found moderate to substantial agreement among STD clinic attendees and their main sex partners for risk behaviours such as “crack” cocaine use (kappa = 0.53−0.60), injection drug use (kappa = 0.51−0.80), and partners having other sex partners (kappa = 0.64−0.69). By contrast, our study documented substantially lower rates of agreement for partners having other sex partners (kappa = 0.29). This distinction is important, since the population studied by Ellen et al comprised couples at risk for STD, but not necessarily confirmed to have an STD. In fact, only 25% of the women and 1% of the men in that study were documented to have an STD at the time of examination. Such individuals arguably represent a lower risk pool than the individuals in the present study, who were recruited for participation only after the diagnosis of an STD, and who may or may not have been in an established couple relationship. It is possible that perception of partner risk is actually higher among people who seek screening in the absence of disease, when compared with those who actually acquired an STD.

Strengths of this study include the sociometric research design, which allowed comparison of index patients’ perceptions of partners with the self reported characteristics and behaviours of the partners themselves. The study collected data on a broad range of risk markers and behaviours, and the finding of moderately strong agreement for certain partnership characteristics and partner risk behaviours indicates that certain information is appropriately shared and internalised by members of these high risk partnerships. The study was restricted to people with gonorrhoea or chlamydial infection and their recent sex partners in order to focus upon those acquiring and transmitting STD. The research was designed to integrate easily into the ongoing STD partner services activities of the local public health department, as the research interview was included as an added component of routine partner notification activities.

The restriction of study eligibility to people with documented STD is also a weakness: results may not be generalisable to lower risk populations, people in established couple relationships, or general population samples not seeking STD care. Additionally, the study design required personal contact by a disease investigator to recruit and enrol partners for the study; people who were not successfully contacted or who declined enrolment may have more frequently had risky behaviours. Patients were asked whether they perceived their partners as ever having used drugs or practising commercial sex, rather than whether they perceived partners as having such risks in the recent past. A partner’s risk behaviours in the distant past may not have contributed to transmission of gonorrhoea or chlamydial infection to the patient. Also, certain risk behaviours were infrequently reported—for example, 9% of partners reported injection drug use and 1% of male partners reported same sex contact. This can lead to circumstances in which percentage of agreement is high but kappa values are low, and caution must be utilised in interpreting kappa values under these conditions.

The concordance correlation (ρ) evaluates agreement for continuous variables, while kappa evaluates agreement for discrete variables; comparisons between ρ and kappa warrant caution, however. Although ρ reduces to kappa when applied to binary variables, measurement of variables on different scales limits comparability. We dichotomised many continuous variables and evaluated agreement using kappa, but differences in measurement and coding of discrete and continuous variables may limit inferences from these dichotomisations.

Mathematical models suggest that choosing less risky sex partners may also influence the spread of infection at the population level.¹⁹ Our findings highlight the need for enhanced awareness among sexually active people of the current inaccuracies in assessment of partners’ risks, particularly with regard to whether a partner has multiple other partners. Recent research on the typology of sexual partnerships demonstrates that partnership concurrency is common and perhaps normative during certain life phases (for example, when a long term relationship is ending, or soon after sexual debut).²⁰ Public health efforts are needed to develop and evaluate interventions to promote enhanced awareness of the normative nature of partnership concurrency among sexually active individuals, and the need for enhanced risk assessment and risk disclosure through skills building and self efficacy training. In this sense, “knowing one’s partner” represents a harm reduction approach to STD prevention.²¹

Perception of a partner’s risks depends largely on that partner’s truthful and accurate disclosure of risk within the partnership. STD/HIV prevention efforts must therefore strive to improve early and complete risk disclosure to reduce spread of infection within partnerships, and within networks of partnerships. Recent research on disclosure of HIV serostatus highlights the importance of risk perception and risk disclosure between partners, prior to sexual contact.^22–26 The process of disclosing STD/HIV infection status and personal risk behaviours may jeopardise a relationship before sexual intercourse has occurred, and could further jeopardise one’s reputation and other relationships. Public health efforts to promote routine questioning of partners about risk behaviours and infection status, to promote risk disclosure, and to overcome the associated stigma, could create greater awareness of disparities between perceived and reported risk; enhance individual skills and self efficacy in such discussions of mutual risk; and attempt to make such discussions normative. Additional research should examine whether enhanced disclosure of risk within, rather than between, risk groups (for example, among men having sex with men, among drug users) can lead to avoidance of risky partners and more consistent use of condoms and other protective behaviours within these risk populations.

In summary, interventions to promote sexual health communication within partnerships may generate greater awareness of partner risk behaviours and promote adherence to protective health strategies to reduce risk of STD transmission. Further research is required to develop, test, and evaluate interventions that promote sexual health communication within partnerships and within partnership networks.