Puerto Ricans

The life cycle of a Puerto Rican strain of Schistosoma mansoni (S. mansoni) was maintained in outbred NMR-I mice and Biomphalariaglabrata snails. Infective cercariae were obtained following exposure of snails with a patent infection to incandescent light for 2 h to induce the release of the parasites. Cercarial E/S products were produced as described previously [1 (link),3 (link),7 (link)]. Briefly, culture supernatants containing the 0–3 hour released preparation (0-3hRP) were collected (ensuring whole larvae and parasite tails were discarded), and stored at −20°C until required. Pooled supernatants were concentrated using filter spin columns with a molecular weight cut off of 3 kDa (GE Life Sciences) and the protein content measured using the BCA® protein assay (Thermo Scientific).
Recombinant Sm16 (rSm16), unlabelled or labelled with AlexaFluor® 546, was a gift from Dr Martin Gullberg, Umeå University, Sweden [9 (link),10 (link)].

Our analyses are based on restricted-use natality data for the years 2014–2019, collected by the National Center for Health Statistics (NCHS) as part of the National Vital Statistics System [38 –41 ]. We received permission to use the restricted natality files form the NCHS and the National Association for Public Health Statistics and Information Systems (NAPHSIS). Unlike the public-use data, these files contain information on maternal country of origin, allowing us to disaggregate births among immigrant Latina mothers and identify groups that have been disproportionately targeted by Trump, including populations from Mexico and Central America [1 –7 (link)]. Altogether, our analysis allows us to identify mothers born in the following countries/regions: Mexico, Central America, South America, Puerto Rico, and Cuba. Important to note is that, unlike the other groups mentioned here, people born in Puerto Rico are US citizens. Still, people born in Puerto Rico have different lived experiences from Puerto Ricans born in the US, and birth outcomes between these groups are also known to be different [42 (link),43 ]. To recognize these distinctions, we still refer to people born in Puerto Rico as “foreign-born” but acknowledge that this label is flawed.
In line with our study design, we limited our analytic dataset to include only births that occurred from July 1^st through December 31^st of each study year. Among these cases, we restricted our analysis to singleton live-births to Latina mothers who were residents of one of the 50 US states and who had complete information on maternal country/region of birth, age, formal schooling, parity, infant birthweight, and infant gestational age. Cases removed due to missing information made up less than 1% of all birth records in our study years.
The Institutional Review Board of the University of North Carolina at Chapel Hill approved the study protocol. All data was fully anonymized before we accessed them. All analyses and reporting of results were conducted in accordance with our data use agreement with NCHS/NAPHSIS.

We performed a multistep analysis. First, we conducted preliminary and descriptive analyses to evaluate the distributions and temporal variability of study variables. In these analyses, we estimated univariate frequencies of the control variables (maternal age, formal schooling, and parity) as well as the outcome variables (LBW, PTB, and inadequate prenatal care utilization) as proportions across each of the study periods. We assessed whether the proportions in the post-Trump periods differed from the pre-Trump period using two-tailed z-tests.
Second, we estimated multivariable logistic regression models to identify the temporal associations between exposure to the Trump treatment periods and each outcome of interest: LBW, PTB, and inadequate prenatal care utilization. Rather than comparing outcomes to a reference group with the lowest risk of the outcome (e.g., white mothers), we stratified the models for each population group of interest: US-born Latinas, foreign-born Latinas (aggregated), foreign-born Mexicans, foreign-born Central Americans, foreign-born South Americans, foreign-born Puerto Ricans, and foreign-born Cubans. By analyzing within-group variation over time, each ethnic population in essence serves as its own control group, providing a clearer assessment of how the exposure-outcome association differs across Latina mothers by nativity and maternal country/region of birth [48 (link)]. All models controlled for maternal age, formal schooling, and parity. All model equations and full model results can be viewed in the S1 File (see S1 File). The main results from the key independent variable measuring the Trump study periods are the adjusted odds ratios of the outcome (LBW, PTB, and inadequate prenatal care utilization) during each of the post-Trump periods compared with the pre-Trump period.
Third, we tested the sensitivity of our results by accounting for potential sources of bias. One potential source of bias in our statistical models in the exclusion of maternal marital status, which is a well-documented predictor of infant birth outcomes and use of prenatal care [49 (link)–51 (link)]. Despite its important relationship with this study’s outcomes of interest, we excluded maternal marital status from our analyses due to significant changes in the reporting of this information during our study period. In 2017, California stopped providing record-level data on maternal marital status for births occurring in the state. For all birth records in the years 2017–2019, approximately 6% are missing maternal marital status. In our analytic sample of Latina mothers, approximately 12% are missing on maternal marital status between 2017 and 2019. To avoid biasing our results based on this missing data, our analyses exclude maternal marital status as a control variable. The exclusion of marital status from our analyses, however, may still influence our results due to omitted variable bias. To address this potential important source of bias in our analyses, we used multiple imputation to replace missing data on maternal marital status, and re-estimated our multivariable logistic regression models for each population group using the imputed data.
For all analyses, statistical significance was assessed at 2-sided P < .05. All analyses were conducted using Stata 15 SE software.