Risk and strength: determining the impact of area-level sentiment and protective factors on birth outcomes

PROJECT SUMMARY There are large and persistent differences in preterm birth and low birth weight. Individual-level risk factors do not fully explain the observed differences. There is increasing evidence for the role of area-level sentiment in explaining these disparities, but we currently lack both the measures, methods, and findings to empirically evaluate its influence. The proposed research will advance the research in all 3 areas. We will be using online and social media data and machine learning models to create two measures of area-level sentiment and implement a robust research design to determine whether area-level sentiment impacts birth outcomes. Our investigative team—comprised of experts in the field of epidemiology, health disparities, machine learning, social media data, biostatistics, and community engaged research—is uniquely suited to implement the study aims. Our Specific Aims are to 1) track and detect changes in area-level sentiment and identify local and national events during these time points, 2) determine the impact of changes in area-level sentiment on changes in adverse birth outcomes, and 3) identify protective factors for adverse birth outcomes. Because our data is collected repeatedly and finely across the United States, we can explicitly account for temporal trends and place effects. The proposed study uses new data to capture trends in sentiment with sophisticated machine learning models and represents a critical advancement in the investigation of differences in birth outcomes.