Volume 106, Issue 6, Pages 1453-1462
Sunni L. Mumford, Ph.D., Katherine J. Sapra, Ph.D., Rosalind B. King, Ph.D., Jean Fredo Louis, M.I.M., Germaine M. Buck Louis, Ph.D.
To estimate the prevalence of unintended pregnancies under relaxed assumptions regarding birth control use compared with a traditional constructed measure.
Nationally representative sample of U.S. women aged 15–44 years.
Main Outcome Measure(s)
Prevalence of intended and unintended pregnancies as estimated by  a traditional constructed measure from the National Survey of Family Growth (NSFG), and  a constructed measure relaxing assumptions regarding birth control use, reasons for nonuse, and pregnancy timing.
The prevalence of unintended pregnancies was 6% higher using the traditional constructed measure as compared with the approach with relaxed assumptions (NSFG: 44%, 95% confidence interval [CI] 41, 46; new construct 38%, 95% CI, 36, 41). Using the NSFG approach, only 92% of women who stopped birth control to become pregnant and 0 women who were not using contraceptives at the time of the pregnancy and reported that they did not mind getting pregnant were classified as having intended pregnancies, compared with 100% using the new construct.
Current measures of pregnancy intention may overestimate rates of unintended pregnancy, with over 340,000 pregnancies in the United States misclassified as unintended using the current approach, corresponding to an estimated savings of $678 million in public health-care expenditures. Current constructs make assumptions that may not reflect contemporary reproductive practices, so improved measures are needed.