Post-Doctoral Fellow Kim and Im Submit L'Oreal USA Women in Science Fellowship Application

Kudos to post-doctoral fellow Kate Kim and her advisor, Eun-Ok Im for the submission of her L’Oreal USA Women in Science Fellowship application entitled: "The Influences of Chronic Stress on Preterm Birth: Race/Ethnicity-specific Algorithms Using a Machine Learning Approach." This proposal requests funds for a one-year period with a start date of Nov. 1, 2018.  

Racial/ethnic disparities in preterm birth (PTB) are persistent in the U.S. with a higher prevalence rate of PTBs among racial/ethnic minority women than their non-Hispanic (N-H) White counterparts. However, the underlying mechanism of such racial/ethnic differences is not well understood. Even extensive biomedical, behavioral, and sociodemographic risk factors could explain only about a half of the incidences of PTBs. Instead, chronic stress has recently drawn a great attention as a robust predictor of PTBs particularly among racial/ethnic minority groups. Nevertheless, the existing chronic stress models produced inconsistent or even conflicting results on the associations of chronic stress to adverse birth outcomes mainly due to potential heterogeneity of the chronic stress pathways to PTBs in different racial/ethnic minority groups. Also, the traditional method, such as regression model, has several limitations in modeling the chronic stress pathways to PTBs due to strict model assumptions (e.g., normality, homogeneity of variances, linearity, and independence). Modeling chronic stress requires more rigorous and complicated techniques to improve the accuracy of its measurement because chronic stressors interact with one another, creating synergies to affect PTBs. Machine learning (ML) is promising to reveal variations in the chronic stress-PTB relationships among diverse racial/ethnic groups.

This study will analyze the large national database, Pregnancy Risk Assessment Monitoring System (2012-2015) collected by 37 states in the U.S. A total of 24 chronic stress items across stress domains (e.g., external stressors, buffers, and enhancers of stress) will be utilized. For each algorithm, models will be evaluated using a repeated 10-fold cross-validation. This study will help develop a screening method for chronic stress of racial/ethnic minority pregnant women at risk of PTBs. The long-term goals are to (a) screen pregnant women for chronic stress as a routine care during prenatal visits; (b) enhance the efficiency of allocating resource-intensive interventions to pregnant women at risk of PTBs based on the chronic stress  screening; and (c) ultimately reduce PTBs among racial/ethnic minority groups in the U.S.