Using Machine Learning to Identify Risk Factors Associated with Causes of Fetal Deaths in the United States, 2021

Fetal deaths are classified as the spontaneous intrauterine death of a fetus at any time during a pregnancy. Fetal death after 20 weeks or more of gestation are also referred to as still births. In the United States, state laws require reporting of fetal deaths, however the gestational restrictions around death reporting can vary from state to state–most states report deaths of 20 weeks or more or 350 grams birth weight. Death reports are published annually by the CDC’s National Center for Health Statistics (NVSS). Five selected causes account for 89.8% of all the reported deaths in 2018-2020: fetus affected by complications of placenta, cord and membranes; fetus affected by maternal complications of pregnancy; fetus affected by maternal conditions that may be unrelated to present pregnancy; and congenital malformations, deformations, and chromosomal abnormalities. The NVSS data provides information on the birth, the mother’s health history, the father’s health history, and potential risk factors during the pregnancy. The objective of this project is to look at causes of death in 2021 and create an accurate machine learning (ML) model to classify cases of fetal death after 20 weeks of gestation. XGBoost (Extreme Gradient Boosting) machine learning algorithm was used to classify deaths across 6 classes. The model was able to classify deaths amongst placental abnormalities, maternal complications, maternal conditions, unspecified cause, and other with 40 percent accuracy. In future work, hyperparameter tuning, SMOTE, and more complex training and testing techniques can be used to increase model fit.