Background: Sexual partnership length affects sexual contact frequency between individuals and contact patterns in populations. Here we describe distributions of heterosexual partnership length, and how these have changed over time.
Methods: We analyzed data reported by respondents aged 16-39y on their most recent partner in three cycles of the US’ National Survey on Family Growth (NSFG), conducted 2002, 2006-2010, and 2011-2013; three cycles of Britain’s National Surveys of Sexual Attitudes and Lifestyles (Natsal), conducted 1990-1991, 1999-2001, and 2010-2012; and the Seattle Sex Survey (SSS), conducted 2003-2004. We calculated density and empirical cumulative density functions to examine partnership length distributions. We also estimated the proportion of respondents whose most recent partnership was less than or equal to one day, month, or year.
Results: In all surveys except SSS, most recent partnership length had a bimodal distribution, with separate short-and long-duration peaks. There was substantial heterogeneity in the frequency of short-duration partnerships between surveys. In Natsal 1 and 3, 14% (95%CI:13.9-16.2 and 12.5-14.2 respectively) of respondents reported their last partnership beginning and ending on the same day compared to 33% (95%CI:31.8-34.2) in Natsal 2. Short duration partnerships were less commonly reported in the US: 5.7% (95%CI:5.0-6.7), 6.4% (95%CI:5.8-6.9), and 7.2% (95%CI:6.3-8.1) in NSFG 6, 7, and 8 respectively, and only 2.5% (95%CI:1.4-3.9) in SSS. Long-duration partnerships in Natsal 1 were shorter than those in Natsal 2 and 3, with median durations of 609, 1981, and 1561 days respectively, indicating a shifted distribution.
Conclusions: Individuals’ partnership duration can be categorized into two types: casual short term and steady long term partnerships. However there is heterogeneity across time and place in the prevalence of shorter, casual partnerships. It is unclear how this heterogeneity affects infection transmission dynamics for pathogens with different characteristics, so future mathematical modeling studies should address this.