Statistical analyses
To investigate possible associations of maternal country and region of birth, reason for immigration and length of residence with placental abruption, we estimated odds ratios (ORs) with 95% confidence intervals (CIs) using binary logistics regression analysis. Maternal country of birth, maternal region of birth, reason for immigration and length of residence were included in the regression models as categorical variables using non-immigrant women as the reference group.
Adjustments were made for year of birth, maternal age, parity, multiple pregnancies, chronic hypertension and level of education. Year of birth and maternal age at the birth were included as polynomial quadratic terms in the regression. To account for dependency among pregnancies to the same woman, we used robust standard errors that allowed for within-mother clustering (18).
Missing values were imputed with the mi suite of commands in Stata, using the multivariate normal model with five imputations (19). The imputations were performed for each exposure-outcome association and included the same variables as in the analytic regression models. To obtain ORs with 95% CI across the five imputed datasets, we used Rubin’s combination rules, adjusted for the variability between imputation sets.
To investigate the possible impact of smoking on study results, we performed analyses for the sub-period 1999-2016 for which smoking data were available. Adjustment for smoking had little impact on the reported results (data not shown). Similarly, adjustments for consanguinity between mother and father or Norwegian health region for the birth did not change the results and were therefore not included in the models.
All analyses were performed using Stata IC version 16 (Stata Statistical Software, College Station, TX, USA).