UC Berkeley

This dissertation is comprised of three article-length essays, all of which concern important issues in early-life health and well-being. All three essays focus on the decision to formula feed or breastfeed--one of the first decisions a mother makes in her child's life. Two of the papers--one quantitative and one qualitative--study the environment of the Neonatal Intensive Care Unit and the potential effect of hospital policy and procedures on breast milk feeding. The remaining paper investigates a large federal policy and the consequences of altering the costs of infant formula relative to breast milk. All papers are tied together by an eye towards the plasticity of these early life experiences, as well as the troubling persistence of health disparities by race, class and maternal education.

Essay One: Breast milk feeding in the Neonatal Intensive Care Unit (NICU) is associated with a host of improved health outcomes. However, breast milk feeding rates differ by socioeconomic status, race, ethnicity and maternal education indicating that these results are vulnerable to selection bias. Qualitative work by this author and others suggests that women giving birth in the late-night hours are less likely to begin a successful milk expression regimen due to the lack of experienced clinicians working during these shifts. Using the hour of birth as an instrument for breast milk feeding, this study attempts to isolate the effects of breast milk feeding on incidence of deadly conditions in the NICU, as well as the infant's growth patterns and length of stay. This study also uses innovative measures of the indications for delivery type in order to construct a sub-sample whose distribution of delivery times is the most random, thereby increasing the validity of the analysis. The first-stage of the analysis revealed no significant relationship between late-night births and breast milk feeding at discharge, contrary to the claims of clinicians and mothers interviewed in a separate study. C-Section delivery and shorter maternal lengths of stay significantly predictive of decreased breast milk feeding at discharge, even after controlling for potential confounders. The reduced-form analysis suggests that infants born in the evening (5pm-Midnight) are roughly 2-4% more likely to contract Necrotizing Enterocolitis at some point during their stay in the NICU. The majority of associations between hour of birth and other health outcomes were insignificant. Evidence of heterogeneity in hour of birth effect size by birth weight, gestational age, race/ethnicity and maternal age were also explored.

Essay Two: It is impossible in most countries to randomize assignment into child health programs that may offer benefits. In the absence of this gold standard of program evaluation, researchers face the threat of selection bias--the possibility that there are unmeasured differences, relevant to outcomes, between those who are treated and those to whom they are compared. A common concern is that people who are eligible for a program but choose not to enroll may differ from those who do enroll. Because policies geared towards a country's most vulnerable people are determinants of health inequities, it is imperative that sources of selection bias be identified and that evaluation methods minimize the impact of selection bias on our estimations of treatment effects. Using a case study of a large Federal nutrition program in the United States, this study reviews how researchers have attempted to minimize selection bias and presents an analysis illustrating how the decision to take up the program can highlight sources of this bias. Relying on data from a longitudinal study of mothers and infants, I show that prenatal attitudes and beliefs may determine postnatal program enrollment, and that the direction of the bias differs by demographic variables. Further, I show that magnitude of supposed program effects vary significantly as a function of these prenatal beliefs. In sum, this paper makes the case for more careful study of the factors that determine take-up of a program, and inclusion of those factors in an evaluation of the program

Essay Three: The third paper in this series diverges from the methodology of the first two essays. This paper is the culmination of a year-long survey data collection effort; the work is a collaboration between UC Berkeley, UC San Francisco and Alta Bates Summit Medical Center (Berkeley, CA). The objective of this study was to investigate determinants of breast milk feeding in the NICU, and to try and account for the pervasive racial, ethnic, socioeconomic and language disparities in breast milk outcomes. The survey was developed by the authors of this essay based on established theories of decision making about infant feeding. Over the course of the study period, mothers giving birth at less than 32 weeks gestational age were invited to participate in the study, either through filling out a survey in the hospital, participating in a one-on-one interview, or both. This essay focuses on the results from the survey which were later linked to medical outcome data of the infant upon discharge home. An innovation of this study is the collection of breast milk exclusivity--that is, if a dose-response relationship between breast milk and outcomes did exist, our data collection method would be able to capture it. Results indicate that mothers who participated in the study were less likely to breast milk feed if they were: of black race, non-Hispanic (any race), low-income, or living a long distance from the NICU. Measures of social support, peer effects, and attitudes towards breast milk feeding also predicted the proportion of an infant's feeding that was breast milk.. Implications of these findings are discussed, as are the lessons learned from pursuing this type of study.