Joyce Molenaar

62 CHAPTER 3 The percentage of individuals with multidimensional vulnerability during pregnancy in the Netherlands was 8.1 in 2017 and decreased to 7.2 in 2021, as derived from the RFmodel (Figure 1). The percentages were slightly higher for XGBoost and lasso (respectively 8.0% and 9.1% in 2021), but showed a similar decreasing trend, as printed in Appendix 2. Appendix 2 additionally shows the complete case analysis. Figure 2 visualizes the geographical distribution of multidimensional vulnerability during pregnancy in the Netherlands over the years 2017 to 2021, based on predictions of the RF-model. There are differences between municipalities, with percentages ranging from 1.8 to 17.5%. Table 1. Results of the RF and the sensitivity analyses Metrics Mean from five-fold cross validation (SD) Confusion matrices for best fold Number in each category AUC F1-measure Precision Recall/ sensitivity Specificity Random Forest 0.98 (0.00) 0.70 (0.03) 0.74 (0.06) 0.66 (0.04) 0.98 (0.00) 30 (TP) 6 (FP) 14 (FN) 645 (TN) XGBoost 0.98 (0.00) 0.68 (0.04) 0.70 (0.02) 0.67 (0.08) 0.98 (0.00) 34 (TP) 10 (FP) 13 (FN) 638 (TN) Lasso regression 0.98 (0.01) 0.68 (0.04) 0.67 (0.07) 0.70 (0.07) 0.98 (0.01) 32 (TP) 12 (FP) 11 (FN) 640 (TN) AUC = Area Under the Curve, TP = True Positive, FN = False Negative, FP = False Positive, TN: True Negative Results based on analyses among study population of 4172 women Figure 1. Percentage of multidimensional vulnerability during pregnancy in the Netherlands during the years 2017 to 2021, based on the RF-model using routinely collected data prior to pregnancy. Results based on analyses among all unique pregnancies from 2017 – 2021 (n = 807.904)

RkJQdWJsaXNoZXIy MTk4NDMw