Dissertations and Theses

Date of Degree


Document Type


Degree Name

Doctor of Public Health (DPH)


Epidemiology and Biostatistics


Elizabeth A. Kelvin

Committee Members

Glen Johnson

Mary Huynh

Subject Categories

Public Health


Cesarean, public health, epidemiology, childbirth


Background: Childbirth is one of the most common reasons for hospitalization in the U.S., and Cesarean delivery (i.e., surgical childbirth) is costlier and has a higher likelihood of birth-related complications, maternal rehospitalization, and postpartum medical care utilization than vaginal delivery. The rate of Cesarean delivery in the United States (U.S.) has increased in recent years by over 60%, from 20.7% of all births in 1996 to 32.9% of all births in 2011. As Although this increasing trend of Cesarean delivery incidence has also been seen in other countries, the rate of Cesarean delivery has been rising more steadily within the U.S. than nearly anywhere else. While Cesarean delivery has been established as the safer delivery option for women with certain high-risk pregnancy complications that could put mother and/or baby in danger during vaginal delivery, the increasing rate of Cesarean delivery in the U.S. has not been accompanied with a concomitant decrease in maternal and neonatal morbidity and mortality. Therefore, it has been suggested that at least some Cesarean deliveries performed may be clinically unnecessary, and may put pregnant women at an avoidable higher risk of adverse health outcomes. Numerous factors across multiple levels of organization have been linked to influencing the likelihood of a pregnant woman receiving a Cesarean delivery. Firstly, a robust body of evidence has linked numerous clinical facets of pregnancy, either related to maternal health specifically (e.g., gestational diabetes) or fetal health presentations (e.g., fetal malpresentation), to increased Cesarean delivery likelihood. Additionally, certain sociodemographic characteristics, such as being of older age or being of black/African-American race, have been linked to a higher risk of having a Cesarean delivery. Numerous factors beyond the pregnant woman herself, however, have also been linked to the likelihood of a Cesarean delivery occurring, through a social-ecological framework. Practice-related (e.g., clinical experience, medical school location) and sociodemographic characteristics (e.g., age, gender) of the physician presiding over the birth have been shown to affect Cesarean delivery occurrence. Furthermore, aspects of the hospital where the birth occurs related to maternity health-related practices (e.g., vaginal birth after Cesarean occurrence) and ownership/affiliation (e.g., teaching status, private ownership) have been associated with influencing Cesarean delivery likelihood in numerous studies. Lastly, while there is a dearth of information as to how the health of communities where pregnant women live specifically affect Cesarean delivery likelihood, the sociodemographic profile of communities have been linked to other adverse pregnancy outcomes (e.g., preterm birth, low birthweight). As such, this dissertation research examined: 1) the role of sociodemographic and maternal health-related characteristics of communities related to overall and maternal health characterize in influencing Cesarean delivery incidence across ZIP codes in New York State (NYS); 2) how characteristics associated with pregnant women, physicians, hospitals and patient residential communities affect primary Cesarean delivery risk in pregnant women in NYS; and, 3) how the factors aforementioned in step 2 above affect repeat Cesarean delivery risk in pregnant women in NYS.

Methods: There were two separate analysis plans for this dissertation research, specific to the approach that occurred for the first aim, and the approach that occurred for the second and third aims. For the first aim, an ecological approach was taken, aggregating Cesarean delivery incidence taken from NYS hospital discharge data from 2011-2014 across ZIP codes in NYS with pertinent perinatal health data and where enough births occurred so that they were not excluded via the NYSDOH small cell policy (N=1,316). Predictors utilized that were aggregated and weighted by ZIP code included those related to hospital maternity procedure-related data aggregated and weighted by ZIP code (i.e., Proportion of births attended by a midwife; with augmented labor; with induced labor; and, that were VBAC); hospital characteristic data (i.e., Proportion of births occurring in a teaching hospital; in a private hospital; and in a religiously-affiliated hospital); patient sociodemographic data (i.e., mean age of women giving birth in a ZIP code); ecological perinatal health data (i.e., Proportion of births that were low birthweight; premature births; Medicaid or self-pay; and had late or no prenatal care); and, sociodemographic community data (i.e., Proportion of ZIP code residents that were black; Hispanic; lived below the Federal Poverty Level; spoke a language other than English at home; had at least some college as their highest level of education; had public and private health insurance; median household income; and, urbanization of ZIP code). For the statistical analysis for this AIM, the first step included conducting a naïve complete model negative binomial regression to capture model residuals. Second, a univariate Moran’s I analysis of complete naïve model residuals was conducted to examine spatial autocorrelation of residuals across NYS ZIP codes. Thirdly, based on the results of the univariate Moran’s I analysis, a negative binomial regression was conducted with generalized estimating equations methodology that utilized a working correlation matrix that accounted for spatial autocorrelation (i.e., first-order autoregressive, using the county as the nesting unit). For the second and third aim, a fixed effects regression multilevel analysis was conducted to assess the role of factors associated with pregnant women (patients), physicians, hospitals, and patient residential communities affected primary and repeat Cesarean delivery, respectively. Specifically, a blockwise modified Poisson regression analysis was conducted with generalized estimating equations methodology with the first-order autoregressive working correlation matrix. Predictors used in the analysis include patient-level clinical characteristics of pregnancy from hospital discharge data records (i.e., diagnosis of gestational hypertension; gestational diabetes; fetal malpresentation; obstructed labor; and multiple gestation) and patient sociodemographic characteristics from birth certificate records (i.e., maternal nativity, black/African-American race, and Hispanic ethnicity, for births that occurred in NYC only) as well as physician characteristics from American Medical Association Physician Masterfile and NYS Department of Health (DOH) doctor profile data (i.e., physician age, gender, nativity, medical school location, time since graduated medical school, and malpractice history), hospital procedural data from NYSDOH data (i.e., percentage of VBACs in a hospital, percentage of births with induced labor, with augmented labor, and midwife-attended births), hospital characteristic data from NYSDOH and Medicare data (i.e., religious affiliation, private ownership, and teaching status), and community-level health indicators from NYSDOH, Centers for Disease Control and Prevention, and American Community Survey Census data (i.e., premature birth rate, low birthweight rate, Medicaid birth rate, late or no prenatal care rate, black/African-American population rate, Hispanic population rate, rate of those in the population living below the Federal Poverty Level, and urbanization of ZIP codes). Additionally, multiplicative interaction analysis was conducted for physician gender and private hospital ownership for both primary and repeat Cesarean delivery.

Results: There was significant spatial autocorrelation (p=0.001) of model residuals across NYS ZIP code, as per the univariate Moran’s I analysis. As such, the negative binomial regression for the first study aim was conducted using the first-order autoregressive working correlation matrix within generalized estimating equations methodology. Within this analysis, numerous predictors were found to increase Cesarean delivery incidence. Specifically, across NYS ZIP codes, the following were all associated with increased Cesarean delivery incidence: a higher percent of births in a teaching hospital: IRR: 1.13 (p=0.009); a higher percent of births that were low birthweight: IRR: 1.07 (p=0.001); a higher percent of residents that were black/African-American: IRR: 1.29 (p

Discussion: This study is one of the first to consider predictors of Cesarean delivery from a multifaceted, social-ecological lens that considered a multitude of factors that likely affect the lived experiences of pregnant women prior to their labor and delivery, as well as during the labor and delivery process. While numerous limitation related to potential misclassification of secondary data sources used, residual confounding, and lacking variables that likely contribute to birth-related decisions (e.g., patient preference, physician fear of malpractice), it provides a perspective that justifies the consideration of factors beyond the patient-physician dyad that have been a main focus of method of delivery literature for decades. Additionally, it adds credence to consider Cesarean delivery as a health-related outcome where “place matters”, as is so often discussed within many other areas of study in public health. Ultimately, this study helps to characterize the role that hospitals, health systems, and communities play in shaping perspectives and expectations that likely impact the method of delivery that a pregnant woman ultimately undergoes. Future studies should attempt to ascertain the specific causal mechanism by which these distal factors actually shape the behaviors, perspectives, and implicit biases of pregnant women, their family members, their physicians and other involved clinicians (e.g., nurses, certified nurse midwives), and hospital and health administrators that implement institutional policies that may shape clinical practice.

Included in

Public Health Commons



To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.