BACKGROUND: Recent studies comparing the number of reported abortions with external institutional estimates have suggested that Wave 4 (2008) of the United States’ National Longitudinal Study of Adolescent to Adult Health (Add Health) captured only 31% to 35% of expected abortions, compromising its ability to provide unbiased estimates of pregnancy-related experience.

MATERIAL AND METHODS: This study used multiple techniques to improve abortion capture, by examining external counts more closely than previous studies, using data from Alan Guttmacher Institute abortion incidence counts and patient surveys, Centers for Disease Control pregnancy and abortion surveillance, and U.S. Census population data; accounting for abortion patient under-reporting; and correcting the count of abortions unreported at one wave of Add Health with reports from other waves.

RESULTS: The abortion incidence rate increased to only 56% of external counts, confirming the caution regarding this metric. However, exposure (ever abortion), not incidence, is the predictor of choice for examining women’s pregnancy-related experiences on Add Health, and after improving capture, by Wave 5 (2017-2019) almost all (99%) of abortion exposure had been reported.

CONCLUSIONS: Substantial under-reporting remains a problem for analyses that depend on abortion incidence, but for analyses using abortion exposure, Add Health continues to be a robust and accurate resource to better understand women’s experiences of pregnancy, pregnancy predictors and outcomes, and the demography of U.S. fertility. In the absence of centralized national data on induced abortion, such data are of critical value for abortion policy and better understanding its place in American family life.

Keywords: Abortion, Induced, Data Accuracy, Longitudinal Studies, Pregnancy, Surveys and Questionnaires