Materials and Methods

Birth certificate data from the National Center for Health Statistics (NCHS) were used to identify all women delivering in the United States from 1991 through 2006. Women who delivered a singleton between 37 ^0/7 and 42 ^6/7 weeks’ gestation were eligible to be included in the current study. Women who were not optimal candidates for labor induction were excluded from the analyses. Those with prior cesarean deliveries, fetal anomalies, with no prenatal care (as induction is generally a planned event), or with a nonvertex fetus were excluded. Also, those women whose records precluded an accurate determination of either race/ethnicity or the primary outcome measure were also excluded: those women with unknown gestational age at delivery, missing racial/ethnic designation, or unknown method of labor initiation were ineligible as well. Finally, women with pregnancy-induced hypertension (PIH) or with premature rupture of membranes (PROM) were not included in the analyses ( Figure 1 ). These last 2 groups of women were excluded given that these coded indications may develop as pregnancy progresses, and it cannot be ascertained from these data when the onset of these conditions occurred. Thus, the stability of the frequency of these conditions at 37-38 weeks over time is unknown, and the ability to adequately control for changes in these frequencies among or differences between populations, would not be possible.

Inclusion and exclusion of women delivering infants in the United States, 1991-2006

Exclusions are sequential. Congenital anomalies include anencephaly, spina bifida, gastrointestinal malformations, diaphragmatic hernia, and chromosomal anomalies.

Murthy. Early-term induction of labor. Am J Obstet Gynecol 2011.

The primary outcome measure was derived from the field named “induction of delivery” in the NCHS dataset. This field is defined as the “initiation of uterine contractions before the spontaneous onset of labor by medical and/or surgical means for the purpose of delivery.” Women who had labor augmentation after the spontaneous initiation of labor were not considered to have undergone labor induction.

Gestational age at delivery was based on the field designated as the “gestational age at birth” in the NCHS data set. Though some have contended that the variable coding for the “best clinical estimate of gestation” may be more valid, this variable was not collected by the state of California, which is also the state with the most annual births in each year. Thus, to optimize the sample size and the representativeness of the gravid population in the United States, the gestational age at birth variable was used. To ensure that use of the gestational age variable would not lead to biased estimates in ETI frequency, a sensitivity analysis was performed on the results using each of these 2 variables that coded for length of pregnancy and these analyses resulted in estimates that were quite similar (data not shown).

The annual ETI rate is a sum of 2 induction rates and was determined as follows: first, the number of inductions at 37 weeks’ gestation as a proportion of the number of eligible women delivering between 37-42 weeks’ gestation was calculated. Then, this rate was added to the proportion of women who were induced at 38 weeks’ as a proportion of those who remained pregnant at 38 weeks’ gestation. This sum is the annual ETI rate that is reported in these analyses (ie, Σ [# of inductions at 37 weeks/eligible women from 37-42 weeks] + [# inductions at 38 weeks/eligible women from 38-42 weeks]).

From the NCHS dataset, the study population was stratified by maternal race/ethnicity and was categorized into 4 groups: non-Hispanic white (NHW), Hispanic white (HW), Black, and Other. The “Black” and “Other” groups were not further subdivided by Hispanic ethnicity given the small proportion of women in each of these groups coded as Hispanic (3.0% and 3.3%, respectively).

Bivariable analyses were performed, using analysis of variance (ANOVA) and χ ² analysis, to determine the demographic and medical characteristics that were significantly associated with maternal race/ethnicity. Multiple comparisons were corrected using the Bonferroni method. Demographic factors, including advanced maternal age (≥35 years old), nulliparity, smoking status during pregnancy (yes/no), alcohol use during pregnancy (yes/no), and marital status. Medical factors (maternal diabetes [DM] and chronic hypertension [CHTN]) that may increase the chance of receiving a labor induction and that were consistently measured from 1991 through 2006 were further used for stratification in the analyses.

Unadjusted longitudinal trends in ETI were examined. These trends are reported stratified by maternal race/ethnicity and by the presence of DM or CHTN, and differences in trends according to these groups were evaluated using χ ² analyses with linear trends.

Multivariable logistic regression was used to estimate the independent association between maternal race/ethnicity and ETI (yes/no). Demographic and obstetric variables that were evaluated in the regression analysis were kept in the model if their inclusion changed the estimated odds ratio of the association between race/ethnicity and ETI by at least 10%. Maternal smoking and alcohol use during pregnancy were not evaluated in the models as potential confounding factors because of the large proportion of missing data. The models also included an interaction term between maternal race/ethnicity and the calendar year of birth to estimate whether the difference in the odds of ETI between racial/ethnic groups changed significantly over the 16-year study period.

All statistical tests were 2-tailed and α = .01 was used to define statistical significance because the large sample sizes. Analyses were performed with STATA v10.1 (STATA Inc, College Station, TX). Birth certificate data were accessed from the National Vital Statistics System website in April 2010. As these data are publicly available and deidentified, the Children’s Memorial Research Center Institutional Review Board exempted this study from review.