Imperial College London

This 18-page document outlines recommendations that emerged from a meeting of the Joint United Nations Programme on HIV/AIDS (UNAIDS) Reference Group on Estimates, Modelling and Projections. This meeting brought together 34 experts whose goals included: reaching consensus on a standard definition of "concurrent sexual partnerships", recommending methods for measuring concurrency in a population, and setting out a future research agenda around the study of concurrent sexual partnerships and its association with HIV transmission.

The recommendations are designed to provide UNAIDS and the World Health Organization (WHO) with guidance, and are (in summary form) as follows:

1) Definition of "concurrent sexual partnerships": Overlapping sexual partnerships where sexual intercourse with 1 partner occurs between 2 acts of intercourse with another partner.

Terminology: Participants discussed the abundance of terminology used to identify and describe concurrent sexual partnerships, pointing especially to confusion generated by the acronym "MCP". "Due to the ambiguity around the meaning of 'MCP', it is recommended that this acronym is not used to identify or describe concurrency, preferring the phrases 'concurrent sexual partnerships', 'concurrent partnerships', or simply 'concurrency'. If an acronym is required, 'CP' is recommended."

2) The Reference Group recommends that a measure of the amount of concurrent partnerships be included in the set of indicators for monitoring national HIV epidemics. The consensus is that the main indicator of concurrency should be: point prevalence of concurrency in a population, which is defined as: the percentage of women and men aged 15-49 with more than 1 ongoing sexual partnership at the point in time 6 months before the interview. This is calculated based on the dates of first and last intercourse with up to the last 3 partners in the past year. (There was debate about whether the entire population aged 15-49 or only the sexually active population is the more appropriate denominator in the proportion. The Reference Group opted to select the entire population). As explained here, the indicator should be presented as separate percentages for males and females and should be presented for age groups 15-24, and 25-49 (as sample size allows), in addition to the overall age group.

The Reference Group suggests that data be collected every 4 to 5 years, using national population-based surveys (e.g., demographic & health survey (DHS), AIDS indicator survey, multiple indicator clusters survey). Included amongst the appendices are specific recommendations for the DHS questionnaire.

The Reference Group recommends this indicator for aptly distinguishing between actual concurrency and simply having many (potentially monogamous) partners in the form of occasional one-off sexual encounters. "This indicator gives a picture of the proportion of population maintaining multiple ongoing sexual partnerships, which creates more connected sexual networks over which HIV may spread rapidly." That said, the proportion of concurrent partnerships "may not be directly related to risk of HIV transmission from concurrent partnerships, as this is also affected by the duration of overlap in partners, condom usage with concurrent partners, and patterns of coitus with each partner." Furthermore, "[w]hen interpreting the results it is important to note that if a person has concurrent partners it will affect their partners' risk of being HIV positive; while if a person has multiple partners it will affect their own risk of being HIV positive."

Limitations identified here include the potential for censoring bias with this indicator based on the collection of sexual histories for only the 3 most recent sex partners. Another potential bias is where sexual partnerships are ongoing, but the last sexual intercourse with the partner occurred more than 6 months before the interview. The indicator will provide a conservative (low) estimate of the amount of concurrency in the population. Finally, this indicator is only valid to the extent that the sexual partner history data collected in representative household samples are complete and accurate.

Aside from this main indicator, the Reference Group recommends 2 other indicators:

• The cumulative prevalence of concurrent partnerships, defined as the proportion of the adult population who have had any overlapping relationships in the past year. This is measured by identifying the individuals for whom any of the sexual partnerships reported in the past year have been overlapping based on the sexual partner histories.
• The proportion of multiple partnerships which are concurrent, calculated by dividing the number of adults with concurrent partnerships in the past year by the number of adults with multiple partnerships in the past year.

Other measures of concurrency which have been employed in the literature but that the Reference Group discussed and rejected (for reasons detailed in the document) are: the percentage of individuals who have had more than one sexual partner in the past 30 days, and the proportion of individuals who have ever had sexual intercourse with another person during their current or most recent partnership.

3) The Reference Group recommends that population-based household surveys include "sexual partner history" modules to collect information about the last 3 individuals with whom the respondent has had sexual intercourse within the previous 12 months. The interviewer is encouraged to ensure privacy before asking the following questions:

Q1: When was the last time you had sexual intercourse with this person? (Answer in days/weeks/months ago – [also years for the most recent partner])
Q2: When was the first time you had sexual intercourse with this person? (Answer in weeks/months/years ago)
Q3: Are you still having sex with this person?

The Reference Group indicates that questions about partners should be framed specifically around sexual partners and that questions about dates should specifically refer to acts of sex to distinguish between disease risk behaviour and culturally defined notions of relationships (for example, "When was the first time you had sexual intercourse with this person?" rather than "When did this relationship begin?"). The Reference Group stresses that interviewers should be well trained - e.g., to probe for all sexual partners in the past year, including those who are routinely under-reported in behavioural surveys.

It is recommended that surveys collect other information and risk behaviour about each partner, including: type of relationship (such as spouse, polygamous marriage, cohabiting partner, girlfriend/boyfriend not living with respondent, casual acquaintance, sex worker, etc.), the partner's age (for all partners), condom usage within the partnership, coital frequency within the partnership, and location of the partner. The Reference Group indicates that it may also be useful to collect information about the circumstances under which the respondent met the partner, alcohol and drug usage within the partnership, knowledge of the partner's HIV status, and the exchange of money or goods in the partnerships. Finally, additional routine information on lifetime and recent sexual behaviour, including attitudes towards and knowledge about HIV, should continue to be collected.

"The design, wording, and ordering of questionnaires should be carefully considered to minimise non-response and elicit the most accurate answers as the order and way in which questions are asked can influence the findings of the survey."

4) Research Agenda for the Study of Concurrency and its Association with HIV Transmission

Methodological Research:

• Suggested research designs for validating date recall (accuracy of dates of first and last sex with previous partners) include: Conduct in-depth follow-up interviews using calendars on a subset of national household survey participants; re-visit a sample of survey participants several months to a year later and administer the same survey; and, in cohort studies, compare retrospective partnership histories with prospectively collected data.
• "If the respondent reports they are still planning on having sex with more than one of their partners again, then they are having concurrent partnerships. However, it is not known how well the reported intention to continue a partnership actually correlates to continuation of the sexual partnership. Cohort studies should investigate the validity of this question by investigating in subsequent rounds of data collection whether or not partnerships that were reported as ongoing at the previous round actually continued (and similarly whether partnerships that were reported as completed at the previous round in fact continued)."
• Proposed methods for validating the completeness of partnership histories:
  • Conduct in-depth interviews with prompting approach ("what about sex workers, truck drivers, etc"), comparing the results with response to the standard survey.
  • "Network census" surveys where both partners report partnerships. There may be opportunity to nest this sort of study within existing cohort studies.
  • "Coital diary" surveys where individuals prospectively record each of their acts of coitus. The Reference Group recommends improving reliable completion of such surveys by introducing short messaging service (SMS) or other new technologies. "The most common current methods for collecting information on coital frequency and condom usage tend to be fairly crude and ask respondents to report quantities that are difficult to reliably recall and estimate. Validation of, and perhaps innovation on, the existing instruments is necessary."
• "It should not be assumed that novel methods are necessarily better than standard face-to-face interviews administered in private settings with well-trained fieldworkers. Any benefits associated with novel methods should be weighed against potential drawbacks such as an increase in missing data for important but difficult to answer questions. Qualitative and methodological research is needed to understand why respondents refuse to answer or give inaccurate answers to certain questions....[S]ocial desirability bias is only one possible reason..."

Substantive Research:

Study designs that the Reference Group asserts may be able to demonstrate the empirical association between concurrency and HIV include:

• Incidence/Transmission Studies: Use existing HIV cohorts to identify sexual partners and link HIV transmission events with partner's sexual behaviour, including the incidence and prevalence of concurrency.
• Ecological Association Studies: Undertake analyses across multiple community-based cohorts to study the association between concurrency and HIV incidence, with community clusters as the unit for analysis.
• Contact tracing studies
• Evidence from intervention programmes aimed at reducing concurrent sexual partnerships that are currently being planned and rolled out. Ideally, the Reference Group stresses, such programmes will be tested in randomised and controlled trial (RCT) settings which "would provide firm evidence for the effectiveness of reducing concurrency for HIV prevention, and thereby give evidence that concurrency affects the spread of HIV....[However,] [a]s many education and prevention programs targeted at concurrency are also likely to include components aimed at reducing other risk factors, detailed monitoring of intervention of knowledge, behavioural, and disease outcomes of interventions are necessary in order to evaluate precisely which components of the interventions are most effective."

The Reference Group notes that not all "types" of concurrency may have the same risk of HIV associated with them; condom usage, patterns of coitus, and duration of overlap may vary. For example, partners in a faithful polygamous marriage would not be at risk of HIV as long as none were infected upon entering the marriage. Research into this area "requires more qualitative work to define the relevant categories of concurrency and quantitative work to estimate the relative frequency of different forms of concurrency. Secondly, research needs to understand the particular risk behaviours associated with types of concurrency. Finally information of the types of concurrency and the risk behaviour needs to be intersected with HIV pathogenesis..."

The Reference Group explains that understanding local social norms around concurrency is essential for creating locally relevant messaging aimed at reducing concurrency. Some areas that require research are: defining the reasons that people enter concurrent partnerships; understanding the social acceptability of concurrency; and identifying the social and structural drivers of concurrency, and how changing norms around concurrency will affect other social institutions.

"Limited research indicates that while education campaigns have been fairly successful at conveying the HIV risk associated with some risk behaviours, such as non-condom usage and very high numbers of multiple partners, understanding by the general population of concurrent partnerships and the potential HIV risk associated with them remains fairly low. As increasing knowledge and risk perception about concurrency are likely to be a key outcome of prevention programmes targeting concurrent sexual partnerships, collecting quantitative baseline data on these targets is important..."

Innovative Research Designs: "As our understanding of patterns of HIV spread becomes more detailed, the standard cross-sectional designs for epidemiological inquiry have become insufficient....The establishment of several HIV cohort studies have been an invaluable source of information about behavioural risk factors. More recently, partner studies including studies of sexual partnerships that span long distance labour migration and local network censuses have...provided unique data on sexual networks and HIV transmission....Established research programmes, such as cohort and surveillance sites, provide an organisational and scientific framework within which innovative studies such as local network surveys, partner tracing, or high frequency surveillance may be embedded." The Reference Group also notes that clinical trials may prove useful settings for investigating questions around sexual networks and HIV risk.

UNAIDS Reference Group on Estimates, Modelling and Projections website, accessed October 27 2009.