Recent Heterosexual Partnerships and Patterns of Condom Use: A Weighted Analysis

Copas, Andrew J.a,b; Mercer, Catherine H.a; Farewell, Vern T.c; Nanchahal, Kirand; Johnson, Anne M.a

doi: 10.1097/EDE.0b013e318187ac81
Methods: Original Article

Background: In epidemiologic studies of sexual partnerships, characteristics are often collected in part through detailed questions concerning recent partnerships. These data present challenges for analysis. First, although research interest generally lies in all partnerships in a certain time period, participants may be asked to provide detailed information only concerning their most recent, up to a fixed number. As more recent partnerships may differ from others, a simple analysis of these data may lead to bias. Second, the total number of partnerships for a study participant may be informative, so the analyst must choose between inference for the population of partnerships or for a typical partnership from the population of individuals. Third, data may be more fully recorded for study participants than their partners, and not all partners may be eligible to participate.

Methods: We propose weighting to deal with these challenges. Weighting provides a sensitivity analysis for the possible selection bias due to incomplete reporting. We analyze heterosexual condom use in Britain, using data from the National Survey of Sexual Attitudes and Lifestyles 2000.

Results: The sensitivity of estimates to possible selection bias is low. We find that the choice of population for inference is important for prevalence estimates, but has relatively little impact on measures of association. By defining within-participant partnership predictors we demonstrate how participants vary their condom use. We establish that, at least for male participants, shorter partnership duration is linked to a higher probability of condom use at last sex but lower probability at first sex.

Conclusion: We recommend a weighted analysis approach to recent partnership data, which can be simply implemented in standard survey analysis software. In other surveys the sensitivity of estimates to possible selection bias may be substantial and this will need to be assessed in each case.

Author Information

From the aCentre for Sexual Health and HIV Research, Research Department of Infection and Population Health, University College London, UK; bMRC Clinical Trials Unit, London, UK; cMRC Biostatistics Unit, Cambridge, UK; dPublic and Environmental Health Research Unit, Department of Public Health and Policy, London School of Hygiene and Tropical Medicine, UK.

Submitted 25 April 2007; accepted 22 July 2008; posted 23 September 2008.

Supported by grant G9811620 from the Medical Research Council, UK.

Supplemental material for this article is available with the online version of the journal at; click on “Article Plus.”

Correspondence: Andrew J. Copas, Centre for Sexual Health and HIV Research, University College London, The Mortimer Market Centre, Capper St, London WC1E 6JB, UK. E-mail:

© 2009 Lippincott Williams & Wilkins, Inc.