Risks for adverse outcomes in prolonged, low-risk pregnancies - Machine learning approach using Swedish population-based Health Registers

• Applicant: Dr. Stefanie Schmauder
• Subject Area: Gynaecology and Obstetrics ; Epidemiology and Medical Biometry/Statistics
• Term: since 2022
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 504494065

Risks for adverse maternal and fetal outcomes are increasing after the estimated delivery date (> 40 completed gestational weeks), even in low-risk pregnancies. Adverse outcomes, which must be considered, are often life-threatening or impair future health. Examples are post-partum hemorrhage (PPH) or severe birth injuries for the mother and asphyxia or stillbirth for the child. Therefore, the induction of labor (IOL) after 41 full gestational weeks is mainly recommended worldwide. This led to an increase in artificial IOL, even in middle- and low-income countries. However, despite several randomized controlled trials over the past decades, it is still under discussion at which gestational age an IOL is superior to an expectant management. The aim of the present project is to investigate at which exact gestational age after the estimated delivery date (40 full weeks of gestation) an induction of labor (IOL) can reduce the risks for adverse maternal and fetal outcomes compared to an expectant management until 42 full weeks of gestation. As a novel approach, statistical means of machine learning algorithms will be applied. The study population will include healthy pregnant women with an uncomplicated, singleton pregnancy and a gestational age of at least 40 completed weeks. Eligible pregnancies will be identified from the Swedish Medical Birth Register and the Swedish Pregnancy Register between January 1, 1997 and October 31, 2021 resulting in a study population of almost two million pregnancies. Utilizing the unique Swedish personal identity number (PIN) a further linkage to other Swedish National Health Care Registers is possible. This provide the possibility to consider a wide variety of variables (“features”) into the statistical models, including granular data on medical history, medication and socioeconomic status. Three infant (neonatal death, stillbirth, Apgar-Score < 7 at 5 minutes post-delivery) and 2 maternal (operative vaginal delivery, cesarean section) outcomes will be investigated. With unsupervised methods correlations between not only known (e.g. maternal age, Body-Mass-Index), but also still uninvestigated factors (e.g. birth history of the mother) and the outcomes will be identified and quantified by applying supervised methods. Therefore, the individual clinical characteristics of pregnant women can be considered to identify the gestational age at which the benefit of an IOL outweighs its potential harms. This could provide an individualized and evidenced-based decision-making tool for the expectant mothers and maternity care providers in the future.

• DFG Programme: WBP Fellowship
• International Connection: Sweden
• Host: Professor Olof Stephansson