Preterm births have long been identified as a leading social and biological risk factor. They are the leading cause of U.S. infant deaths and a huge economic burden. Their effects span individual health consequences to national loss of workforce participation. It is important to fully understand the causes of preterm birth to properly address policy needs.
This dissertation’s purpose is threefold. First, I develop a more accurate Preterm Birth Rate (PtBR) by improving measurement of the population at risk of birth. Prior research depends on a measure with one outcome of pregnancy, a birth; and treats each birth as having the same time at risk no matter which week of gestation the birth occurs. To improve the conventional measure, I use an Event History Analysis approach that tracks person-pregnancy weeks, mapping an occurrence not by outcome but from the onset of a viable pregnancy, i.e., reaching week 20 of gestation. This more precisely measures time at risk of PtBR. Also, my measure includes pregnancies that ended with a fetal death, thus including pregnancies that were at risk of PtBR but do not experience a live birth. Finally, I include only pregnancies that reach a viable pregnancy in the calendar year of interest, disallowing births that reach a viable pregnancy in the prior calendar year and including pregnancies that reached a viable pregnancy in the focal year, but experienced the outcome in the following calendar year. My focal year is 2003, so I use 2003 and 2004 vital statistics Public Use Birth and Fetal Death files.
Second, I extend the understanding by testing a more complex model of relationships among predictors and PtBR via path analysis. This yields more precisely measured relationships and more accurate partitioning of variables’ effects on PtBRs. Prior research either looks at the simple relationships or uses a multivariate model that includes all the predictors in a single equation, yielding only direct effects of each variable on PtBR. I test a theory that aligns variables on the public use files in a causal chain and use a new path analysis method that assesses direct, indirect, and total effects with intuitive, comparable measures where combining coefficients from a variety of forms of multiple regression – e.g., OLS, multinomial logistic, and Poison regressions. The methodological complexity of this undertaking is extended further by applying path analysis in an event history analysis.
Third, I test a specific theory about a strong PTB differential, that between Blacks and Whites. I order 9 intervening mechanisms through which their statuses shape the likelihood of PtBR. Age (at this pregnancy), SES (measured by Education), and Marital Status are statuses shaped prior to gestation. They affect or are associated with two categories of mediators operating during gestation – Fetus Characteristics, i.e., Plurality and Sex; and Behavioral Risk Factors, i.e., Birth Order and the Triple Risk Factors of Tobacco Use, Efficacious Prenatal Care (measured by Month Began), and Mother’s Weight Gain. Findings clarify the state of understanding the relationship among these variables and PtBRs.