Symposium 'Meet the PhD Jury: Preterm birth prediction meets causal inference'

This thesis applies data science tasks to preterm birth, including description, prediction, and causal inference, and highlights the importance of international collaboration. It describes the causes of preterm birth, including spontaneous labor, premature rupture of membranes, and maternal/fetal pathology. Neonatal magnesium levels are discussed, with the highest levels observed in children exposed to maternal neuroprotection. A safe dosage regimen for fetal neuroprotection is proposed. A prediction model using time-to-event analysis and informative censoring is developed. The importance of analyzing the correct population for prediction is emphasized. Natural language processing and oversampling techniques are explored. Lastly, causal inference is used to study the interval between antenatal corticosteroid administration and birth, questioning the validity of the two to seven-day interval. A dynamic treatment regimen analysis is proposed as the appropriate approach.