We conducted a population-based retrospective cohort study that used Birth dataset, Fetal Death datasets made available by the Centers for Disease Control and Prevention, the National Vital Statistics System, and the National Center for Health Statistics; and the US population census data from the United Nations. [14] The Birth dataset had data on all births occurring within the US, while the Fetal Death dataset contained information on fetal death. The information available in these datasets was extracted from the birth and death certificates, respectively. The datasets included socio-demographics, health characteristics, and maternal risk factors associated with each live birth and stillbirth. For our study, we restricted our analyses to singleton birth within the gestational age from 20 to 42 weeks for the years 1982 through 2018 from the Birth dataset, and 1982 - 2017 from the Fetal Death data set.
Study Variables
We included the variables that were commonly available in the datasets for most of the study period. Our exposed group, childhood pregnancy, was defined as a pregnancy that resulted in singleton live birth or stillbirth among mothers ≤ 14 years of age. For our comparison group, we used mothers ages 15 to 19, whom are also referred to as adolescent mothers. [15] Our primary outcome, stillbirth, was defined as intrauterine death of a fetus at ≥20 weeks of gestation. [15] Covariates included in the study were 1) race, which was categorized as Whites, Blacks, and Others (which included all races except Whites and Blacks). We used this categorization as it was available for all years over the study period; 2) birthplace, classified as hospital, home, and others; 3) delivery method, categorized as vaginal and C-section; 4) birth attendant, classified as certified medical professional including medical doctors, nurses, midwives, etc. and others; 5) gestational age (which was measured based on the date of last menstrual period (LMP)), categorized as preterm (i.e., <37 weeks or term, i.e., ≥37 weeks); 6) maternal diabetes – the variable for diabetes was given without segregation between pre-pregnancy and gestational in the birth dataset till the year 2003, and in the fetal death dataset till 2013. So, we combined the variables to encapsulate any type of maternal diabetes; 7) chronic hypertension – the variable for chronic hypertension was changed to “pre-pregnancy hypertension” as from 2004 in the birth dataset, and as from 2014 in the fetal death dataset. Therefore, we combined the two variables and designated it as “chronic hypertension” for this study; 8) Pregnancy hypertension – the variable for pregnancy hypertension was changed to “gestational hypertension” as from 2004 in the birth dataset, and as from 2014 in the fetal death dataset. We decided to combine the two variables and called it “Pregnancy hypertension” for this study; 9) eclampsia - the variable for eclampsia was changed to “hypertension eclampsia” as from 2004 in the birth dataset, and as from 2014 in the fetal death dataset. We therefore, combined the two variables and called it “eclampsia” for this study.
Statistical Analysis
We calculated the rate of childhood viable pregnancy for each year by dividing the number of viable pregnancies (live birth and stillbirth) among girls aged ≤ 14 years by the total number of females aged ≤ 14 years in the US population. We compared these with similarly derived rates for adolescent mothers ages 15 – 19 years. This information was extracted from 1982 through 2017from the birth and fetal death datasets (numerator) as well as the US population census (denominator). Similarly, to derive the rate of childhood live birth for each year, we divided the number of live birth among girls aged ≤ 14 years by the total number of females aged ≤ 14 years in the US general population. Equivalent computation was performed for the comparison group (15-19 years old mothers). Live birth rates from 1982-2018 were computed to allow for comparison with previous studies that defined pregnancy rate using only live births. We also calculated stillbirth rates for both age groups by dividing stillbirth by the sum of live birth and stillbirth and then multiplying by 1000. The analyses involving stillbirth were restricted to the years 1982-2017 as the Fetal Death data was available only until 2017. We used joinpoint regression, which is a modeling technique, to detect the change in the rate of events over time. [16] We also calculated the average annual percentage change (AAPC) for trends in rates of viable pregnancies, live birth and stillbirth in both child and teen mothers over the study period.
We conducted bivariate analyses to uncover socio-demographic and maternal risk factors among child and adolescent mothers. We also examined the factors potentially contributing to live birth or stillbirth among children using Chi-square test that detects differences in proportion. To determine independent associations across the various socio-demographic and maternal comorbidities, including diabetes, chronic hypertension, pregnancy hypertension and eclampsia versus stillbirth, we generated adjusted hazard ratios (AHR) from Survey Cox Proportional Hazards Regression Models. In the adjusted models, we controlled for race, place of birth, delivery method, attendant at birth, gestational age, maternal diabetes, chronic hypertension, pregnancy hypertension, eclampsia; after removing missing values from all the covariates. All tests of hypothesis were two-tailed with a type 1-error rate set at 5%. Because the study was performed using publicly available de-identified data, it was approved as exempt by the Institutional Review Board of Baylor College of Medicine.