Application of Advanced Analytical Tools to Predict Preterm Birth and Travel for Prenatal Care

Loading...
Thumbnail Image
Date
2020
Journal Title
Journal ISSN
Volume Title
Publisher
University of Alabama Libraries
Abstract

Preterm birth is one of the many widely studied pregnancy related issues. It is often associated with individual health status, but it has also been studied with respect to neighborhood characteristics. It has been shown that preterm birth rates are particularly higher in groups with a lower socioeconomic status such as Medicaid enrolled populations. High preterm birth rate is one of the contributing factors of infant mortality rate, which in turn brings tragedy to the family along with potential socioeconomic impact to the society. In this dissertation, individual and neighborhood level information of a sample population of Medicaid enrolled pregnant women from the state of Alabama were studied. The neighborhood level characteristics were used in addition to individual level health information to compare the predictive potential of traditional and machine learning analytical methods on classifying preterm births. In the first chapter, advanced analytical methods such as gradient boosting, random forest and logistic regression were used to predict the pregnancy outcome. In the second chapter, network analysis was used to construct a network based on travel information of Medicaid enrolled pregnant women from Alabama. The network features were studied to shed light upon the most central nodes in the travel network using various centrality measures such as betweenness and closeness centralities. The travels to receive healthcare by pregnant women were compared to the benchmark physical degree of separation from the origin to the destination locations. In the third chapter, weighted networks constructed using the travel frequencies among counties were used to compare link weight prediction models. The frequency of travel between any two counties in the state were predicted using two supervised learning models and a deep learning model. Overall, this dissertation showed use cases of various advanced analytical tools on the healthcare sector. The chapters in the dissertation cover descriptive analysis (Chapter II), predictive analysis (Chapter I & Chapter III) and prescriptive analysis (Chapter I).

Description
Electronic Thesis or Dissertation
Keywords
Network Analytics, Predictive Analytics
Citation