Study population and outcome ascertainment
This analysis includes all birth records issued to non-Hispanic black women (black women, hereafter) in Louisiana in 2014 (N = 23,954), data accessed through the National Center for Health Statistics. We identified preterm births (those occurring < 37 weeks’ gestation) and low birth weight births (those < 2500 g) as our primary outcomes of interest. Birth records included a geographic identifier for maternal parish of residence for linkage to data on incarceration and other parish-level contextual factors. As all data were deidentified, this study was deemed exempt by Tulane University Institutional Review Board.
Jail incarceration prevalence
Our primary exposure of interest was the parish-level jail incarceration prevalence among black individuals (count of black individuals age 16 to 64 in jail per 1000 black non-incarcerated residents). Jail data were isolated from prison data for this analysis in order to specifically estimate parish-level effects. Data were provided by the Vera Institute of Justice based on analysis of incarceration trend data from the Bureau of Justice Statistics, the Census of Jails, and the Annual Survey of Jails [17]. Since some counties are too small to have their own jail facilities, they may rely on nearby jurisdictions for detention purposes. Incarceration rate estimates for these counties are derived based on data from multi-jurisdictional jails and the sending county’s share of the combined resident population [18]. The incarceration prevalence was scaled by its interquartile range (IQR) for the purposes of estimating risk of preterm birth or low birth weight associated with an IQR increase in incarceration.
Additional covariates
Other measures included in the analysis were individual-level covariates available on the birth record and known to be associated with adverse birth outcomes, including insurance type (Medicaid, private, and other, including self-pay), maternal age (< 20, 20–24, 25–29, > 30), parity (nulliparous, multiparous), and the Kotelchuck Index of Prenatal Care Adequacy based on the ratio of observed prenatal visits to the number that would be expected given dates of care initiation and delivery (< 50% = inadequate, 50–79% = intermediate, 80–109% = adequate, and 110% or more = adequate plus) [19, 20]. In order to isolate the association between the parish-level black incarceration prevalence and adverse birth outcomes within the black community independent of differences in socioeconomic contextual factors between parishes, additional parish-level covariates were derived from the 2014 U.S. Census American Community Survey, including the Gini Index, which measures population income inequality (0–1 scale), racial income inequality (absolute difference in white to black median household income), the 2010 U.S. Census classification of parishes by urbanicity (mostly urban, mostly rural, completely rural), and the percent of blacks with a bachelor’s degree or higher.
Statistical analysis
We identified the distribution of maternal and parish-level characteristics within the total study population and stratified by outcome status. We mapped parish-level incidence of preterm birth, low birth weight, and jail incarceration across Louisiana in ArcGIS in order to visualize patterns and variation. Next, we fit modified Poisson regression models with generalized estimating equations to account for women clustered within parishes. The first model (Model A), estimated the relative risk (RR) and 95% confidence interval (CI) for preterm birth associated with an IQR increase in the prevalence of incarcerated black persons, adjusted for maternal and parish-level characteristics. Model B included the same covariates with low birth weight as the outcome.
Lastly, we conducted sensitivity analyses to test the robustness of our findings. Given the growing body of literature highlighting the association between area violence and adverse reproductive health outcomes [21,22,23], we sought to explore the degree to which high incarceration prevalence simply reflected greater exposure to crime (and its negative health consequences) as opposed to representing a broader marker of social inequality and structural determinants of health inequities for black women. Therefore, we fit the same models described above and additionally included the parish-level index crime prevalence to estimate associations with preterm birth (Model C) and low birth weight (Model D). The index crime prevalence is the count of crimes (criminal homicide, rape, robbery, aggravated assault, burglary, motor vehicle theft, larceny-theft, and arson) per 1000 residents, as described in the 2014 Uniform Crime Reporting System Report from the Federal Bureau of Investigation [24]. All data analyses were conducted in SAS 9.4. All maps used for this study were created using ArcGIS software by Esri. (ArcGIS® and ArcMap™ are the intellectual property of Esri and are used herein under license. Copyright© Esri. All rights reserved).