BACKGROUND: Gastroschisis and omphalocele are common abdominal wall birth defects with prevalence (per 10,000 live births) in the United States of 4.3 and 2.6, respectively. Risk factors for these defects include gene variants and environmental (non-genetic) exposures. To date, there has been very limited use of machine learning to explore environmental risk factors associated with both gastroschisis and omphalocele. Using a retrospective population-based case-control study design, this study applied a contrast machine learning method to evaluate the relationships between several selected child and maternal risk factors and gastroschisis and omphalocele.
METHODS: Data for the current study were obtained from the Iowa Center of the National Birth Defects Prevention Study (NBDPS). Children diagnosed with gastroschisis or omphalocele (cases) and unaffected children (controls) were delivered from 1997-2011. Maternal NBDPS interview data for 133 children with gastroschisis and 31 children with omphalocele and 1,300 control children comprised the analytic sample. Gradient boosted regression was applied to evaluate selected child and maternal risk factors for predicting and classifying a delivery as being a case child (i.e., diagnosed with gastroschisis or omphalocele) or a control child. Variable importance and partial dependence plots were generated to examine effectiveness and classification probabilities of each risk factor.
RESULTS: For the 164 cases (gastroschisis and omphalocele) combined, important predictors with higher classification probabilities associated with being a case child were maternal age <20 years at delivery, high school or less education at delivery, multigravida, pre-pregnancy body mass index <18.5 kg/m2, periconceptional (one month prior through the third pregnancy month) active smoking, infection, and folic acid-containing supplementation. Child sex, maternal race/ethnicity, fever, chronic hypertension, chronic diabetes, periconceptional alcohol consumption and cannabis use were not identified as important predictors.
CONCLUSIONS: The current study is only the second to use a machine learning approach to examine risk factors for gastroschisis and first to examine those factors for omphalocele. Findings associated with increased risk of these defects observed in the current study tended to support some but not all factors identified using more traditional epidemiologic analyses. Future studies should aim to increase sample sizes to improve the sensitivity and accuracy of this analysis.
Table of Contents
ABSTRACT - iii
ACKNOWLEDGEMENTS - v-vi
KEYWORDS AND ABBREVIATIONS - vii
LIST OF TABLES - viii
TABLE OF CONTENTS - ix
CHAPTER I: LITERATURE REVIEW - 1
CHAPTER II: MANUSCRIPT - 18
CHAPTER III: DISCUSSION - 31
REFERENCES - 33
About this Master's Thesis
|Subfield / Discipline|
|Committee Chair / Thesis Advisor|
|A Machine Learning Approach to Investigate Epidemiologic Risk Factors of Gastroschisis and Omphalocele in Iowa, 1997-2011 ()||2022-04-21 12:10:16 -0400||