Statistical analyses
Prior to analyses, multiple imputations were performed to replace the missing values in the predictors (ranging from 1.1% to 36.2%) and timing of pregnancy recognition (25.5%) using multivariate imputation by chained equations.35 We created 50 imputed datasets with 100 iterations. Besides the study variables, we used body mass index at intake, gestational age at intake, and apgar score 5 minutes after birth as predictors for imputation.36
First, the inequalities, i.e. the associations between the different predictors and timing of antenatal care initiation, were estimated using linear regression models. The reference level was the level of the predictor with the earliest antenatal care initiation, e.g. ‘30-35 years’ in age, or ‘high’ in educational attainment, to ensure that the inequality was always positive because that simplified the interpretation. Second, the association between early pregnancy recognition and timing of antenatal care initiation was estimated using linear regression analyses adjusted for all predictors (age, migration background, relationship status, pregnancy intention, mental illness, Dutch language skills, parity, education, employment household income, housing, neighborhood deprivation, and cognitive functioning). This model was subsequently used to estimate the timing of antenatal care initiation if the intervention was set to pregnancy recognition within 6 weeks. By comparing the timing of antenatal care initiation with the hypothetical intervention (i.e. had all participants recognized the pregnancy within 6 weeks) and without the hypothetical intervention (i.e. timing of pregnancy recognition as it appears in the data), the reduction in the inequalities driven by the predictors was estimated.32 Bootstrapping with 1000 iterations was used to calculate the 95% confidence intervals. All analyses were conducted in IBM SPSS version 28 and R statistical software version 4.2.1.