Detecting induced abortions from reports of pregnancy terminations in DHS calendar data.

ABSTRACT

PIP: The World Health Organization (WHO) classification scheme for pregnancy terminations was modified to evaluate the accuracy of pregnancy terminations as spontaneous or induced abortion in Demographic and Health Survey (DHS) calendar data. The decision algorithm considered length of gestation at time of termination, whether contraception was used during the period before termination and, if so, reason for discontinuation, whether the pregnancy was wanted or unwanted, and age, parity, and marital status at the time of the pregnancy termination. 1993 DHS calendar data on 1539 pregnancy terminations from Turkey were applied to the classification method. The algorithm classified 1413 (92%) of the terminations. It classified 71.8% (1015) as probably induced compared to 63% as reported by survey respondents. The algorithm classified 28.2% (398) as probably spontaneous compared to 37% as reported by survey respondents. The ratio of test positive to true positive (i.e., sensitivity of the method) was 92.5%, showing that the algorithm does a good job at identifying true cases of induced abortions. On the other hand, the ratio of test negatives to true negatives (i.e., specificity of the method) was only 65%, suggesting that the algorithm yields a relatively large number of false positives. When the investigators considered the effects of errors caused by misreporting of classification on the efficacy of the proposed scheme, both sensitivity and specificity increased (assuming 10% error, 92.7% and 67.3%; 25% error, 92.9% and 71.4%; and 50% error, 93.3% and 79.1%, respectively).