Let's Talk About Sex: Improving Measurement of Contraceptive Use in Cross-Sectional Surveys by Accounting for Sexual Activity Recency.

Abstract

Standard measures of contraceptive prevalence have known biases given that they do not reflect sexual activity and true exposure to the risk of an unintended pregnancy. In this study, we aim to understand the extent to which women protect themselves against unintended pregnancy, taking into account exposure to sex, by examining contraceptive use patterns by marital status and recent sexual activity. Data come from population-based surveys of reproductive age women in 9 low-resource settings. We estimated contraceptive prevalence using (1) the standard current contraceptive use measure, (2) a new measure of contraceptive use at last sex, and (3) a comprehensive measure that combines current use and use at last sex. Analyses are stratified by site and examine patterns by marital status only, and by both marital status and sexual activity separately. We then examined method mix by each contraceptive measure. Study findings reveal distinct patterns in contraceptive use in relation to marital status and sexual recency across sites. Overall, married women tended to report higher levels of current contraceptive use compared to use at last sex, whereas unmarried women reported higher levels of contraceptive use at last sex. When examining these measures by sexual activity and marital status, results indicate lower levels of contraceptive use among women who had not had sex in the month prior to the survey, for both married and unmarried women. The comprehensive measure of contraceptive use yielded the highest estimates, by design. Method mix varied consistently by contraceptive measure, with current use tending to capture more permanent and long-acting methods and use at last sex more likely to capture short-acting and coital-dependent methods. These findings have important implications for how the family planning field evaluates unintended pregnancy risk and unmet need for contraception within low-resource settings, given different estimates yield discrepant estimates for who is "at risk."