Decline in Unmet Need Is Not Main Reason for Rise In Contraceptive Prevalence

Abstract

Although contraceptive prevalence has increased substantially in Sub-Saharan Africa during the past two decades, for the most part the gains have not been driven by a decline in unmet need, according to an analysis of Demographic and Health Survey (DHS) data. (1) Examining trends in 22 countries that had multiple DHS surveys between 1990 and 2011, the investigators found that, on average, contraceptive prevalence rose by 13 percentage points--from 17% to 30%--between a country's first and most recent surveys. However, the proportion of women with unmet need fell by a mean of only four percentage points, a decline smaller than the simultaneous decrease in the proportion of women with no need (e.g., those wanting more children soon). Few studies have examined longitudinal changes in unmet need and lack of need among cohorts of women. However, Madsen and colleagues note that DHS data offer a useful substitute. Because the surveys in a particular country are typically administered about five years apart, one can use consecutive surveys to track changes in contraceptive use and nonuse among five-year cohorts of women; for example, a random sample of 25-29-year-old woman in a given survey will draw from the same pool of women as a sample of 20-24-year-old women in the previous survey, simulating a longitudinal cohort study. The current study used this approach to analyze DHS data collected between 1990 and 2011 from 22 countries in Sub-Saharan Africa. All countries had at least two surveys during that period, and most had four; in nearly all cases, the earliest survey for a particular country was conducted before 1998 and the most recent after 2005. For each survey, women aged 15-49 who were married or in a union were categorized, according to standard DHS definitions, as either being contraceptive users, having unmet need, having no need or being infecund. The investigators examined trends in these four categories in two ways. First, they performed period analyses to assess changes in the prevalence of the categories among women in each five-year age-group (e.g., whether the prevalence of contraceptive use, unmet need, no need and infecundity among 20-24-year-olds in a country's first survey differed from the prevalence among 20-24-year olds in subsequent surveys). These analyses included both descriptive and regression analyses. Second, as noted earlier, the investigators performed cohort analyses in which women in each five-year age-group were assumed to represent the same population as the next oldest five-year age-group in the following survey. Although the mean interval between surveys was, in fact, about five years, 21% of surveys were either less than four years apart or more than 6.5 years apart. Nonetheless, the authors retained these outliers in order to maintain comprehensive coverage and adequate sample size; supplementary analyses indicated that retention of the full sample did not meaningfully affect the results. Period analyses revealed that, on average, contraceptive prevalence in a given country rose by 13 percentage points (from 17% to 30%) between a country's first and last surveys (mean interval, 14 years). At the same time, countries had mean declines of four percentage points in the proportion of women with unmet need, five points in the proportion with no need and four points in the proportion who were infecund. Consistent with these findings, a multiple linear regression model more closely linked the increase in contraceptive prevalence to a decline in lack of need (coefficient, -1. …