Data source and data collection
Women eligible for the study were found using CTcue®, an Artificial Intelligence searching program developed to collect data from Electronic Patient Files (EPF) of our clinic (CTcue®, version 2.1.1.10, Amsterdam, Netherlands). Data were collected in CTcue® and further processed in Rstudio.
Statistical methods
A directed acyclic graph (DAG) was constructed by using the online DAGitty software (http://www.dagitty.net/) to select covariates for inclusion in the statistical models.
The relationships between each of the variables were assigned based on knowledge of the publications regarding these associations. Concerning assumptions described in the DAG theory, a minimal sufficient set of adjustment variables was identified for estimating the effect of BDZs use on maternal and neonatal risks.