The percentages of births, miscarriages, and abortions that resulted from unintended pregnancies were applied to the counts of each respective pregnancy outcome and then summed to determine the total number of unintended pregnancies. To calculate rates, we obtained population counts according to age and according to race and ethnic group from the U.S. Census Bureau.
24 All other distributions of population subgroups were derived from the Annual Social and Economic Supplements of the U.S. Census Bureau’s Current Population Survey,
25 except for religious affiliation, which was derived from the NSFG. Poor females were defined as those with incomes below 100% of the federal poverty level, and low-income females were those whose incomes were between 100% and 199% of the federal poverty level.
When calculating the percentage of unintended pregnancies that ended in abortion, we excluded miscarriages in order to assess only pregnancies in which the outcome was determined by the respondent. The rates of unintended pregnancy according to educational attainment were limited to women 20 years of age or older; this age cutoff excluded most females who had not yet completed schooling, yet still included young women, who have had historically high rates of unintended pregnancy. We also updated the rates of unintended pregnancy for 1981, 1987, 2001, and 2008 — years that the NSFG was fielded — to take into account updated population estimates and recent improvements in our analytic approach. Data on pregnancy intendedness were also collected in the 1995 survey of the NSFG but were excluded owing to concerns about the accuracy of the pregnancy intendedness data from that year.
26 We performed analyses at an aggregate level and separately for each population subgroup: we combined data on pregnancy intention, pregnancy outcomes, and populations from several different sources to calculate rates, which made it difficult to assess the reliability of our estimates and of the change over time. Because most of the uncertainty around the rate estimates was attributable to the percentage of pregnancies that were unintended (since the numbers of pregnancies and population denominators are based largely on generally complete census data), we performed a supplementary analysis to calculate 95% confidence intervals for the percentage of pregnancies that were unintended using a merged data set that combined the sample of births and miscarriages from the NSFG with the sample of abortions from the Abortion Patient Survey. We then used this range of percentages to calculate the 95% confidence intervals around the rate estimates. Although these percentages are expected to be less accurate than the ones calculated in the aggregate manner, the 95% confidence intervals around these percentages should represent the variance around the rate estimates.
The above approach uses two different data sources for pregnancy intention. We also used a single data set, the NSFG, to calculate a test statistic for the change between 2008 and 2011 in the percentage of pregnancies that were unintended. Using the NSFG alone for all pregnancy outcomes allows for a simple calculation of the test statistic. Abortions are underreported in the NSFG, and therefore the percentages calculated using this approach were expected to be lower than those in our main analysis. Nonetheless, we considered this analysis of trends to be reasonable, because the underreporting of abortions has not changed substantially over time.
27 (link),28 (link)
Finer L.B, & Zolna M.R. (2016). Declines in Unintended Pregnancy in the United States, 2008–2011. The New England journal of medicine, 374(9), 843-852.