EP01.14: Prediction of preterm pre‐eclampsia via machine learning.

The second model was created by nested logistic regression adding anthropometric variables, serum, and ultrasound biomarkers to a previous model of maternal history using a stepwise method for variable selection. We performed an elastic net model that uses ridge and lasso regressions that automatically selects the best predictive variables for pPE, penalises non-statistically significant variables, and selects the best model using 10-fold cross-validation. To assess the performance of a machine learning algorithm compared to a logistic model for the prediction of preterm pre-eclampsia (< 37 weeks [pPE]) in Mexican population. [Extracted from the article]