(B-206) Methodological Considerations about Policy Impact on Studies of Pregnancy Registries and Maternal/Infant Outcomes

Background: The Dobbs ruling, a United States (US) Supreme Court decision that the Constitution does not protect the right to induced abortion, may introduce bias to the analysis and interpretation of pregnancy registry data. For example, the ruling may increase the proportion of subjects who are lost to follow-up and adversely affect registry enrollment and data completeness. Statistical analyses of US pregnancy registry data in the post-Dobbs era should identify and adjust for sources of bias resulting from this decision.

Objectives: There were two objectives for this targeted literature review.

1. Identify potential types of bias resulting from the Dobbs decision in the collection and/or analysis of pregnancy data in the US.

2. Determine which methods have been used to adjust for those sources of bias and assess their appropriateness.

Methods: We performed a targeted literature review (TLR) in PubMed using the following terms related to the Dobbs decision: induced abortion, bias, methods, and pregnancy registries.

Results: The initial TLR identified 192 potential articles. Filtering the search to articles published after the Dobbs decision was made reduced the number of results to 37. The results were further narrowed to select articles that focused on identifying and/or adjusting for sources of bias in US pregnancy data.

After the results were narrowed, four articles were confirmed to be truly relevant. The articles identified several potential sources of bias, including loss to follow up, selection bias, reporting bias, and misclassification of abortion interventions and/or outcomes.

Only one of the four articles conducted a statistical analysis to assess the impact of bias on reporting treatment effect. A covariate associated with the bias (loss to follow-up) and the outcome (treatment failure) was identified and then was used to conduct multiple imputation for missed treatment outcomes. This simulation-based method allows an evaluation of how the exclusion of those with unknown outcomes may have biased estimated treatment effects.

Conclusions: The Dobbs ruling can introduce potential bias to pregnancy registry studies. Appropriate statistical methods should be considered to adjust for those biases in the analysis to correctly interpret the study outcomes.