Background: Estimates of sexual partnership durations, gaps between partnerships, and overlaps across partnerships are important for understanding sexual partnership patterns and developing interventions to prevent transmission of HIV/sexually transmitted infections (STIs). However, a validated, optimal approach for estimating these parameters, particularly when partnerships are ongoing, has not been established.
Methods: We assessed 4 approaches for estimating partnership parameters using cross-sectional reports on dates of first and most recent sex and partnership status (ongoing or not) from 654 adolescent girls in rural South Africa. The first, commonly used, approach assumes all partnerships have ended, resulting in underestimated durations for ongoing partnerships. The second approach treats reportedly ongoing partnerships as right-censored, resulting in bias if partnership status is reported with error. We propose 2 “hybrid” approaches, which assign partnership status to reportedly ongoing partnerships based on how recently girls last had sex with their partner. We estimate partnership duration, gap length, and overlap length under each approach using Kaplan-Meier methods with a robust variance estimator.
Results: Median partnership duration and overlap length varied considerably across approaches (from 368 to 1024 days and 168 to 409 days, respectively), but gap length was stable. Lifetime prevalence of concurrency ranged from 28% to 33%, and at least half of gap lengths were shorter than 6 months, suggesting considerable potential for HIV/STI transmission.
Conclusions: Estimates of partnership duration and overlap lengths are highly dependent on measurement approach. Understanding the effect of different approaches on estimates is critical for interpreting partnership data and using estimates to predict HIV/STI transmission rates.
Estimates of sexual partnership durations and gap length vary considerably by measurement approach; this variability may have implications for mathematical models using these estimates to predict HIV/sexually transmitted infection transmission.
From the *Department of Epidemiology, University of North Carolina at Chapel Hill, Chapel Hill, NC; †Department of Biostatistics, University of Washington, Seattle, WA; ‡University of New England, Armidale, New South Wales, Australia; §Department of Pathology, Johns Hopkins University, Baltimore, MD; ¶National Institutes of Health, Bethesda, MD; and ∥MRC/Wits Rural Public Health and Health Transitions Research Unit (Agincourt), Witwatersrand, South Africa
Financial support: This work was supported by the National Institutes of Health (T32 AI007001, KL2 TR001109), the National Institute of Mental Health (R01 MH087118), and the Division of Intramural Research, the National Institute of Allergy and Infectious Diseases. Additional support was provided, in part, by the National Institute of Allergy and Infectious Diseases, the National Institute on Drug Abuse, and the National Institute of Mental Health through Cooperative Agreements (UM1 AI068619, UM1 AI068617, and UM1 AI068613).
Conflicts of interest: We have no conflicts of interest to declare.
Correspondence: Nadia Nguyen, MSPH, 135 Dauer Drive, 2101 McGavran-Greenberg Hall, Chapel Hill, NC 27599-7435. E-mail: firstname.lastname@example.org.
Received for publication April 17, 2015, and accepted August 5, 2015.