Statistical analysis
As we stratified neonatal outcomes according to timing of delivery, we
also compared how maternal characteristics, obstetric features, fetal
characteristics, and circumstances at parturition varied with each
completed week of gestation at the time of delivery. Categorical
variables were presented as n (%) and tested with Chi square test or
Fisher’s exact test as appropriate. Normally distributed continuous
variables were presented as mean ± SD and compared with One Way ANOVA.
Non-normally distributed continuous variables were presented as median
(IQR) and tested with One Way ANOVA on ranks. A level of statistical
significance of P ≤ 0.05 was considered
Multivariate logistic regression analysis was used to investigate if
gestational age and circumstances at delivery (spontaneous PTL, pPROM or
indicated delivery) independently affected the risk of adverse neonatal
outcomes. The following variables were tested as potential confounders:
maternal age, parity, history of spontaneous preterm birth, race, low
education, smoking, maternal BMI, excessive weight gain, utilization of
assisted reproductive technologies (ART), treatment with antenatal
corticosteroids, ASA, LMWH, or progesterone. Maternal medical disorders,
obstetric or fetal complications, type of labor other than spontaneous
(i.e induced or no labor), and route of delivery were not included in
the models as they were considered on the same causal pathway of
circumstances at delivery. Finally, a potential interaction between
gestational age at delivery and circumstances at delivery was
investigated.
Disorders of the mother or the fetus may affect neonatal outcomes, even
if they are not the primary indication for delivery. Therefore, we
tested the independent effect on adverse neonatal outcomes of maternal
medical conditions (diabetes mellitus, hypertensive disorders, liver
disorders), fetal characteristics (non-reassuring fetal status, IUGR,
prenatally diagnosed fetal anomalies or aneuploidies, and amniotic fluid
disorders), pregnancy complications (pPROM, spontaneous PTL,
chorioamnionitis, vaginal bleeding from placental abruption or abnormal
placentation), and gestational age at delivery using multivariate
logistic regression. Maternal age, parity, history of spontaneous
preterm birth, race, low education, smoking, maternal BMI, excessive
weight gain, utilization of ART, treatment with antenatal
corticosteroids, ASA, LMWH, or progesterone were tested as potential
confounders. Type of labor other than spontaneous (i.e induced or no
labor), and route of delivery were not included in the models as they
were considered on the same causal pathway of maternal, fetal, or
obstetrics complications.
The strength of the association between the covariates and the dependent
variable was estimated as area under the curve of a receiver operating
characteristic (ROC) curve plotted with the true-positive rate compared
with the false positive rate. Statistical analyses were performed using
Stata 15 (StataCorp, College Station, TX).