Patterns of Prior Induced Abortions and the Likelihood of Subsequent Natural Pregnancy Loss: Exploratory Application of Pregnancy Outcome Sequencing
Background: Research concerning the long-term health consequences of induced abortion is constrained by both the limitations in the availability of data necessary to construct complete reproductive histories, as well as the limitations in the analytical methods necessary to interpret them. Objective: This study aims to determine the association of induced abortion and the likelihood of a subsequent natural loss by applying the pregnancy outcome sequence (POS), a research construct that defines the number and order of all pregnancy outcomes (births, induced abortions, natural losses) in each woman’s reproductive history. Methods: Using the Medicaid Analytic eXtract files from the Centers for Medicare and Medicaid Services Chronic Conditions Warehouse, we identified a study population of 508 unique POSs, representing 5455 women, each of whom had 1 to 16 pregnancy outcomes, for a total of 14,198 pregnancies. We applied an exploratory iterative sequential analytical approach, which included aggregate POS correlation analysis, logistic multiple regression, and simultaneous CIs (Agresti and Tukey-Kramer methods). We also established counting methods to populate the data tables for each analytical phase. Results: Overall, we found evidence to conclude that both prior abortions and natural losses are significantly associated with the risk of subsequent natural losses. For abortion, there is evidence of a dose-response relationship from 0 to 3 abortions and the likelihood of subsequent natural loss. For natural loss, the risk of a subsequent natural loss is significant after 2 natural losses and between the first and second. There is no association of prior births, or any combination of birth/abortion or birth/natural loss, with the risk of subsequent natural loss. Limitations imposed by the skewed distribution of the number of total pregnancies per reproductive history, resulting in small cell sizes, and exclusion of important covariates, restrain assurance in the results. Conclusions: The POS demonstrates that the order and combinations of pregnancy outcomes may result in varying conclusions that were previously undetectable. Therefore, further methodological development is indicated.
