Predictive models of neonatal outcomes using continuous glucose monitoring and clinical records during pregnancy in women with previous bariatric surgery

This thesis aims to investigate the predictive modeling of neonatal outcomes, specifically focusing on birthweight, within two distinct populations of pregnant women: those with a history of bariatric surgery and a control group of overweight women. A gap in the literature reveals an incomplete understanding of why post-bariatric surgery mothers have an increased risk of giving birth to babies that are small for gestational age. A hypothesis is that the altered glucose metabolism post-surgery could be a contributing factor, which is why Continuous Glucose Monitoring (CGM) is employed in this research. Utilizing a comprehensive longitudinal dataset comprising clinical records during pregnancy and CGM features, the study seeks to identify significant predictors of birthweight and birthweight percentile and explore potential differences between the two cohorts. Statistical learning models constitute the core of the analysis, primarily aimed at investigating the association between CGM features and neonatal outcomes. Initial explorations involve descriptive statistics and visualizations to understand the foundational characteristics of both groups and to compare the distributions of variables within the dataset. Regression models are constructed to forecast birthweight, incorporating relevant predictors such as maternal characteristics and specific features derived from CGM. Random forest models are applied to investigate the presence of non-linear relationships between covariates and birthweight. These models are adept at unravelling complex patterns and dependencies that may not be adequately captured by linear modeling approaches. Ultimately, this research contributes insights into understanding and predicting neonatal outcomes, placing a particular spotlight on investigating the association between CGM features and birthweight in pregnant women with a history of bariatric surgery. The results suggest that indexes of glycemic control may play a predictive role in neonatal birth weight within the bariatric surgery population. While these initial insights point towards a potential association, further studies are warranted to delve deeper into the mechanisms of this interaction and to confirm the predictive significance of CGM-related features.