Data include all US singleton live births and stillbirths from January 1, 2007, through December 31, 2015.
A, Relative change in the distribution of gestational ages at birth. B, Relative change in perinatal mortality rates. Data include all US singleton live births and stillbirths from January 1, 2007, through December 31, 2015, with gestational ages of 20 to 44 weeks.
eTable 1. Comparison of Maternal Characteristics: US Singleton Stillbirth and Live Births
eTable 2. Model-Fit Characteristics for Perinatal Mortality
eTable 3. Stillbirth and Neonatal Mortality Rates Based on Gestational Age: US Singleton Stillbirth and Live Births, 2006-2013
eTable 4. Temporal Changes in Early, Late, and Postneonatal Mortality Rates With Imputed Maternal Smoking Data: US Singleton Births, 2007-2015
Customize your JAMA Network experience by selecting one or more topics from the list below.
Ananth CV, Goldenberg RL, Friedman AM, Vintzileos AM. Association of Temporal Changes in Gestational Age With Perinatal Mortality in the United States, 2007-2015. JAMA Pediatr. 2018;172(7):627–634. doi:10.1001/jamapediatrics.2018.0249
Are changes in the gestational age distribution associated with changes in perinatal mortality?
In this cohort study of 34 236 577 US singleton births, births at all gestational ages except 39 to 40 weeks decreased from 2007 to 2015; an overall decrease in perinatal mortality rates was attributable to changes in gestational age distribution rather than gestational age–specific mortality. Although the proportion of births at gestational ages 34 to 36, 37 to 38, and 42 to 44 weeks decreased, perinatal mortality at these gestational ages increased.
Increased perinatal mortality in some gestational age groups may be associated with lower-risk pregnancies in which neonates are delivered preferentially at 39 to 40 weeks, leaving fetuses at higher risk for mortality at other gestational ages.
Whether the changing gestational age distribution in the United States since 2005 has affected perinatal mortality remains unknown.
To examine changes in gestational age distribution and gestational age–specific perinatal mortality.
Design, Setting, and Participants
This retrospective cohort study examined trends in US perinatal mortality by linking live birth and infant death data among more than 35 million singleton births from January 1, 2007, through December 31, 2015.
Year of birth and changes in gestational age distribution.
Main Outcomes and Measures
Changes in the proportion of births at gestational ages 20 to 27, 28 to 31, 32 to 33, 34 to 36, 37 to 38, 39 to 40, 41, and 42 to 44 weeks; changes in perinatal mortality (stillbirth at ≥20 weeks, and neonatal deaths at <28 days) rates; and contribution of gestational age changes to perinatal mortality. Trends were estimated from log-linear regression models adjusted for confounders.
Among the 34 236 577 singleton live births during the study period, the proportion of births at all gestational ages declined, except at 39 to 40 weeks, which increased (54.5% in 2007 to 60.2% in 2015). Overall perinatal mortality declined from 9.0 to 8.6 per 1000 births (P < .001). Stillbirths declined from 5.7 to 5.6 per 1000 births (P < .001), and neonatal mortality declined from 3.3 to 3.0 per 1000 births (P < .001). Although the proportion of births at gestational ages 34 to 36, 37 to 38, and 42 to 44 weeks declined, perinatal mortality rates at these gestational ages showed annual adjusted relative increases of 1.0% (95% CI, 0.6%-1.4%), 2.3% (95% CI, 1.9%-2.8%), and 4.2% (95% CI, 1.5%-7.0%), respectively. Neonatal mortality rates at gestational ages 34 to 36 and 37 to 38 weeks showed a relative adjusted annual increase of 0.9% (95% CI, 0.2%-1.6%) and 3.1% (95% CI, 2.1%-4.1%), respectively. Although the proportion of births at gestational age 39 to 40 weeks increased, perinatal mortality showed an annual relative adjusted decline of −1.3% (95% CI, −1.8% to −0.9%). The decline in neonatal mortality rate was largely attributable to changes in the gestational age distribution than to gestational age–specific mortality.
Conclusions and Relevance
Although the proportion of births at gestational age 39 to 40 weeks increased, perinatal mortality at this gestational age declined. This finding may be owing to pregnancies delivered at 39 to 40 weeks that previously would have been unnecessarily delivered earlier, leaving fetuses at higher risk for mortality at other gestational ages.
Globally an estimated 15 million preterm deliveries (gestational age, <37 weeks) occur each year.1 In developed countries, more than 80% of stillbirths and one-third of neonatal deaths occur at preterm gestations.2-5 Among surviving infants, preterm birth is associated with substantially high risks of morbidity6-8 and neurodevelopmental and cognitive impairments in early childhood.9-12 This risk extends to infants born at gestational ages as late as 37 to 38 weeks who experience higher rates of adverse neonatal13,14 and pediatric15,16 outcomes compared with those born at 39 to 40 weeks.
Preterm delivery rates among live births have decreased in the United States from 2005 to 201414,17-19 and in several European countries.1,20,21 However, the rates began increasing in 2015 and 2016 in the United States.22 Because reductions in preterm and early term deliveries and perinatal mortality remain a global health priority,23 determining the association between gestational age distribution and perinatal mortality remains a challenge. Efforts expended to a more complete understanding of the effect of new interventions, policies, and practices on reducing the burden of early deliveries and, in turn, improvements in perinatal survival will be beneficial for clinical treatment of women and neonates during pregnancy and after giving birth.
Obstetrical interventions are intended for maternal and fetal reasons, particularly the burden of stillbirths24 at preterm gestations. However, an unintended (or intended) consequence of an increase in physician-initiated deliveries at preterm gestations may be a reduction of term or postterm deliveries. The objectives of this study were to (1) evaluate changes in the distribution of gestational age at birth among singletons from 2007 through 2015; (2) determine gestational age–specific changes in perinatal mortality rate and stillbirth and neonatal mortality rates; and (3) evaluate the association of changes in the gestational age distribution vs changes in gestational age–specific mortality with temporal trends in mortality during this period.
We undertook an analysis of US singleton births from January 1, 2007, through December 31, 2015. The data include fetal deaths and live births linked to corresponding infant deaths (period-linked files) for more than 99% of all births compiled by the National Center for Health Statistics of the US Centers for Disease Control and Prevention.22 The numerator for the period-linked file consists of all infant deaths occurring in a given data year linked to the corresponding birth certificates whether the birth occurred in that year or the previous year. The study population included women who delivered a singleton at a gestational age of 20 to 44 weeks. In this study we ascertained data on stillbirths from the combined 1989 and 2003 versions of the corresponding fetal death certificates; live births, from the live birth certificates; and infant deaths, from the infant death certificates. Thus, the study includes nearly all births in the United States from 2007 through 2015. These deidentified data are publicly available, and ethics approval under human subjects’ protection regulation was not required by the institutional review board of Columbia University.
Two sources of gestational age were used in this analysis: a clinical estimate based on the 1989 version of the birth certificate and one based on the best obstetric estimate (available in the 2003 revision of the birth certificate, which now forms the basis for gestational age reporting by the National Center for Health Statistics). A study from the National Center for Health Statistics25 has demonstrated that both estimates were comparable to each other. We examined births starting in 2007, given that the best obstetric estimate of gestational age was only available starting that year. Gestational age at birth was grouped according to the following categories26: extremely preterm (20-27 weeks), very preterm (28-31 weeks), moderately preterm (32-33 weeks), late preterm (34-36 weeks), early term (37-38 weeks), term (39-40 weeks), late term (41 weeks), and postterm (42-44 weeks).
Perinatal mortality included stillbirths and neonatal deaths. Stillbirths at 20 weeks or more were expressed per 1000 total births. Neonatal mortality was defined as the number of live-born infants who died within the first 28 days and was expressed per 1000 live births.
We examined the following primary outcomes: (1) changes in the proportion of births by gestational age and (2) gestational age–specific changes in rates of perinatal mortality, stillbirth, and neonatal mortality. Temporal trends in perinatal mortality were evaluated during the entire 9-year period (2007-2015) and represent changes in the outcomes at the population level. This evaluation was accomplished by including an indicator for each year, with 2007 as the reference period. From this regression model, we extracted the effect for 2015, which is interpreted as the change in mortality rate from 2007 to 2015. We also examined the mean yearly change in mortality rates by including the period as a single continuous variable in the models. We examined the potential for nonlinear association between year and perinatal mortality nonparametrically with restricted cubic splines.27 Tests for nonlinearity were based on the likelihood ratio test comparing the model with the linear term only and the model with the linear and the cubic spline terms.28
Temporal changes in the distribution of gestational age during the study period were evaluated by fitting log-linear (Poisson models with robust variance) regression models. From this model, we estimated the rate ratio as the effect measure and converted the rate ratios for the period effect to percentage of change in the proportion of births for the period variable. We then examined gestational age–specific temporal changes in perinatal mortality during the same period. This process used 2 sequential regression models. In the first model, we estimated the crude changes in mortality rates from 2007 through 2015 within the strata of gestational age groups. In the second model, we adjusted these trends for confounding factors.
Potential confounders considered for adjustment in the analysis for perinatal mortality included maternal age (<20, 20-24, 25-29, 30-34, 35-39, and ≥40 years), single marital status, and self-reported maternal race/ethnicity (non-Hispanic white, non-Hispanic black, Hispanic, and other). Because race/ethnicity is associated with preterm delivery,29,30 we adjusted for this variable in the analyses. For analysis pertaining to neonatal deaths, we additionally adjusted for live-born parity (primiparous or multiparous) and smoking during pregnancy. However, the 2 latter confounders were unavailable in the stillbirth data files. Data on other confounding variables were unavailable (eg, cocaine use), were available in only a subset of the data (eg, body mass index), or were available in different formats across the years (eg, maternal educational level) that prevented a standardized approach to confounder adjustment.
We also undertook a decomposition analysis to evaluate changes in neonatal mortality.31 This method partitions the total temporal change in neonatal mortality rate attributable to the following 2 determinants: change in mortality rates attributable to changes in the underlying gestational age distribution and change in gestational age–specific mortality from 2007 to 2015 (SEs and corresponding CIs are not available for this approach). The sum of these components indicate the total difference in mortality rates between the 2 periods. Thus, a positive sum denotes the excess number of deaths in 2007 that could have been avoided had the distribution of gestational age and gestational age–specific mortality in 2007 been the same as in 2015. In contrast, a negative sum denotes the excess number of neonatal deaths in 2015 that could have been avoided had the distribution of gestational age and gestational age–specific mortality in 2015 been the same as in 2007. The assumptions implicit in this decomposition analysis are that no confounding bias occurs in the associations and that shifts in the gestational age distribution do not occur in specific subsets that may have different underlying risks of poor outcome.
Data on maternal smoking were missing in 4 179 044 births (11.8%); thus, we undertook a multiple imputation analysis. We assumed that the pattern of missing smoking data was random and generated 25 data sets with missing data imputed. This imputation was performed (after 25 burn-in iterations) under a logistic regression framework with year of birth, maternal age, primiparity, and single marital status variables in the model. We then pooled the results from the analysis of the 25 imputed data sets to obtain a single estimate of associations. We only present the results from the multiple imputation analysis.
We undertook 2 sensitivity analyses. First, we examined trends in stillbirth and early and late neonatal mortality. The early neonatal mortality rate was calculated as the number of deaths that occurred within the first week per 1000 live births; late neonatal mortality, as the number of deaths that occurred from 7 to 28 days per 1000 survivors of the early neonatal period. In the second analysis, we examined changes in postneonatal mortality rates (deaths from 28-365 days). The postneonatal mortality rate was calculated as the ratio of the number of deaths in the postneonatal period to the number of neonatal survivors.
We considered a 2-tailed P < .05 for the trend tests to denote statistical significance. Statistical analysis was performed using SAS software (version 9.3; SAS Institute).
After all exclusions, 34 236 577 singleton live births and 199 839 singleton stillbirths delivered at 20 to 44 weeks’ completed gestational age from 2007 through 2015 remained (Figure 1). The proportion of births to women younger than 20 years decreased from 10.5% in 2007 to 5.9% in 2015, whereas births to women aged 35 to 39 years increased from 11.4% to 13.1%, and births to women 40 years or older, from 2.5% to 3.0% (eTable 1 in the Supplement). The proportion of women of white, African American, and Hispanic race/ethnicity remained stable from 2007 to 2015. Maternal rates of smoking decreased from 9.9% in 2007 to 7.8% in 2015.
The total number of births decreased from 4 183 633 in 2007 to 3 871 346 in 2015 (Table 1). With the exception of births at a gestational age of 39 to 40 weeks, the proportion of births at all other gestational age groups decreased from 2007 to 2015 (Figure 2A). The proportion of births at a gestational age of 34 to 36 weeks decreased from 6.4% in 2007 to 5.8% in 2015 (relative decrease of 9.7%); at a gestational age of 37 to 38 weeks, from 29.3% to 24.5%. The proportion of births at a gestational age of 39 to 40 weeks increased from 54.5% to 60.2%.
The overall perinatal mortality rate decreased from 9.0 per 1000 births in 2007 to 8.6 per 1000 births in 2015 (P < .001) (Table 1). Perinatal mortality rates increased by 7% at a gestational age of 32 to 33 weeks (55.4 to 59.4 per 1000), 15% at a gestational age of 34 to 36 weeks (15.1 to 17.4 per 1000 births; P < .001), 23% at a gestational age of 37 to 38 weeks (3.0 to 3.6 per 1000 births; P < .001), and 31% at a gestational age of 42 to 44 weeks (3.5 to 5.9 per 1000 births; P < .001). Perinatal mortality decreased at gestational ages of 20 to 27 and 39 to 40 weeks but showed annual adjusted relative increases of 1.0% (95% CI, 0.6%-1.4%) at 34 to 36 weeks, 2.3% (95% CI, 1.9%-2.8%) at 37 to 38 weeks, and 4.2% (95% CI, 1.5%-7.0%) at 42 to 44 weeks (Figure 2B). The assumption that changes in perinatal mortality rates were linear was met (eTable 2 in the Supplement).
Stillbirth rates increased at gestational ages of 20 to 27, 28 to 31, 32 to 33, 34 to 36, 37 to 38, and 42 to 44 weeks and remained unchanged at 41 weeks (eTable 3 in the Supplement). Neonatal mortality rates decreased at gestational ages of 20 to 27 and 28 to 31 weeks; increased at 34 to 36, 37 to 38, and 42 to 44 weeks; and remained unchanged at 41 weeks.
After adjustment for confounders, the relative annual decreases in perinatal mortality rates from 2007 to 2015 were −0.8% (95% CI, −0.9% to −0.7%) at a gestational age of 20 to 27 weeks and −0.4% (95% CI, −0.7% to −0.1%) at 28 to 31 weeks (Table 2). Although the proportion of births at a gestational age of 39 to 40 weeks increased, perinatal mortality rate at 39 to 40 weeks showed an adjusted annual decrease of −1.3% (95% CI, −1.8% to −0.9%). Perinatal mortality rates increased at gestational ages 34 to 36 (1.3%; 95% CI, 0.9%-1.7%), 37 to 38 (2.6%; 95% CI, 2.2%-3.0%), and 42 to 44 (4.2%; 95% CI, 1.5%-7.0%) weeks despite decreases in the proportion of births at these gestational ages.
The decomposition analysis showed that the overall decreases in neonatal mortality rates from 2007 to 2015 were largely associated with changes in the underlying gestational age distribution (by 63%), instead of gestational age–specific mortality (by 37%) (Table 3). Similar patterns were seen for the overall decrease in early and late neonatal mortality rates.
Early and late neonatal mortality rates showed an adjusted annual relative decrease at a gestational age of 20 to 27 weeks of −0.7% (95% CI, −1.0% to −0.4%) and −2.1% (95% CI, −2.9% to −1.3%), respectively (eTable 4 in the Supplement). Early neonatal mortality rates showed an adjusted annual relative increase at gestational ages of 34 to 36 (1.5%; 95% CI, 0.6%-2.4%), 37 to 38 (3.1%; 95% CI, 2.1%-4.0%), and 42 to 44 (5.5%; 95% CI, 0.8%-10.4%) weeks. Postneonatal mortality rates showed an adjusted annual relative decrease at gestational ages of 20 to 27 (−5.6%; 95% CI, −6.4% to 4.8%), 28 to 31 (−5.0%; 95% CI, −6.2% to −3.7%), 32 to 33 (−2.3%; 95% CI, −3.8% to −0.8%), and 34 to 36 (−1.8%; 95% CI, −2.6% to −1.0%) weeks.
This population-based study of singleton stillbirths and live births in the United States from 2007 to 2015 revealed that the proportion of births at a gestational age of 39 to 40 weeks increased while proportions at all other gestational ages decreased. Although an overall decrease was observed in perinatal mortality during the study period, mortality rates decreased at a gestational age of 39 to 40 weeks but increased at 34 to 36, 37 to 38, and 42 to 44 weeks. The decrease in neonatal mortality rates was largely attributable to changes in the gestational age distribution rather than changes in gestational age–specific mortality.
Improvements in perinatal and neonatal care, particularly quality initiatives almost exclusively centered on the most high-risk infants at a gestational age of less than 32 weeks,32 are arguably important factors in outcomes for preterm neonates. Although the proportion of births at 20 to 27 weeks decreased by 8.1%, perinatal and neonatal mortality rates also decreased annually at this gestational age by 0.8% and 0.9%, respectively, after adjustments for confounders. Given the high mortality rates at extremely preterm gestational ages, even a small decrease in the number of deaths will have an influence on overall perinatal mortality. Furthermore, improvements in pregnancy dating at these early gestations may have likely shifted a small proportion of births at these low gestational ages to higher gestational ages.
The decreasing proportion of births at gestational ages of 34 to 36 and 37 to 38 weeks may be associated with changes in the timing of elective delivery, with hospital policies and quality initiatives effectively reducing unindicated deliveries before 39 completed weeks of gestation.33-35 Increased use of low-dose aspirin in women with ischemic placental disease36,37 may additionally have resulted in decreased need for indicated delivery before 39 weeks.38 The 1.3% and 0.9% adjusted annual relative increase in perinatal and neonatal mortality rates, respectively, at a gestational age of 34 to 36 weeks and 2.6% and 3.1%, respectively, at a gestational age of 37 to 38 weeks may have been an unintended consequence of the recommendation to postpone elective deliveries until 39 completed weeks.33 A possible reason for the increased mortality at a gestational age of 37 to 38 weeks could be that physicians may be more likely to defer to 39 weeks for delivery for women at moderately increased risk for adverse perinatal outcomes.39 With fewer obstetric interventions for early-term pregnancies, the risk may have increased for women approaching 39 weeks’ gestation.
The proportion of births at gestational age 39 to 40 weeks increased from 2007 to 2015, whereas perinatal mortality at this gestational age decreased secondary to reduced risk of stillbirth rather than neonatal mortality. Although the proportion of postterm births decreased during the study period, perinatal mortality rates increased substantially for this subgroup, supporting that improved clinical management may further obviate risk. For extremely preterm (20-27 weeks), very preterm (28-31 weeks), and moderately preterm (32-33 weeks) births, obstetrical intervention is based on maternal or fetal indications and is unlikely to be elective. A reason for the apparent increased perinatal mortality in some gestational age groups may be that neonates who previously would have unnecessarily been delivered earlier than 39 to 40 weeks were delivered at 39 to 40 weeks, leaving fetuses at higher risk for mortality at other gestational ages.
Callaghan and colleagues40 estimated that 69% of the decrease in the US infant mortality rate from 2007 to 2013 was associated with gestational age–specific increases in infant survival, and 31% was associated with changes in the gestational age distribution. We found that the decrease in neonatal mortality rates from 2007 to 2015 in the United States was largely associated with changes in the underlying gestational age distribution and less associated with changes in gestational age–specific mortality. Despite the contrasts between the study by Callaghan and colleagues40 (infant mortality, births at a gestational age of 20-21 weeks, and births with malformations were excluded) and the present study, the cumulative findings support that reduction in neonatal mortality may be attributable, at least in part, to gestational age–specific improvements.
This study has several strengths. Gestational age, based on the best obstetric estimate, has been found to be highly sensitive and specific, with predictive values exceeding 98% for preterm delivery.41 Although previous research20,42 has evaluated population-level changes in gestational age at delivery in the United States and other industrialized countries, none, to the best of our knowledge, has evaluated the extent to which such changes are associated with recent trends in US perinatal and neonatal mortality. An impetus to examine perinatal deaths as the primary end point overcomes several shortcomings. First, registration of live births at the borderline of viability and who die shortly after birth may encourage birth attendants to complete a fetal death registration instead of a live birth and an infant death registration, contributing to registration artifacts.43-45 Second, physician-initiated obstetric interventions are designed to end the pregnancy in the context of serious maternal or fetal compromise.39 Although preventing fetal demise can often be achieved by such interventions, these interventions in turn may increase the risk of neonatal deaths. A combined examination of stillbirths and neonatal deaths may provide additional insights to understand temporal trends.
This study also has limitations. First, the potential for misclassification of gestational age on the birth record remains. A study of vital records for 2 states found obstetric estimates to be within 2 weeks of data recorded in medical records in 98% to 99% of cases; these rates were 67% to 92% for dating the same week.46 Second, gestational age is likely to be overestimated among antepartum stillbirths. However, given that the proportion of stillbirths relative to live births is small, errors in dating stillbirths are unlikely to have shifted the temporal patterns of perinatal death. Nevertheless, we caution against a causal interpretation of this analysis. Third, the extent to which early pregnancy terminations (at <20 weeks) with signs of life may have affected the trends remain unknown. Fourth, the associations for neonatal mortality may be affected by confounding due to indication bias, particularly among neonatal deaths occurring after a major obstetric complication.47 However, the potential for bias of confounding by indication is addressed by the decomposition analysis that determines the contributions of temporal changes in mortality rates relative to underlying changes in the gestational age distribution vs gestational age–specific mortality.48 Fifth, we were unable to adjust the perinatal mortality trends by maternal smoking. Furthermore, whether the trends were influenced by other unmeasured sociodemographic characteristics remains unknown. Sixth, women may have had more than 1 pregnancy during the study period; however, owing to data limitations, we were unable to adjust for this clustering.
In this study, from 2007 to 2015 in the United States, an increase in singleton births at a gestational age of 39 to 40 weeks occurred, and perinatal mortality rates decreased at this gestation window. Changes in the distribution of gestational ages were largely associated with the overall decrease in mortality rates. A study of changes in spontaneous and iatrogenic deliveries may provide additional important insights to understand the prevailing trends in perinatal mortality.
Accepted for Publication: January 24, 2018.
Corresponding Author: Cande V. Ananth, PhD, MPH, Department of Health Policy and Management, Joseph L. Mailman School of Public Health, Columbia University, 722 W 168th St, New York, NY 10032 (firstname.lastname@example.org).
Published Online: May 14, 2018. doi:10.1001/jamapediatrics.2018.0249
Author Contributions: Dr Ananth had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: All authors.
Acquisition, analysis, or interpretation of data: Ananth, Goldenberg, Friedman, Vintzileos.
Drafting of the manuscript: All authors.
Critical revision of the manuscript for important intellectual content: Ananth, Goldenberg, Vintzileos.
Statistical analysis: Ananth.
Administrative, technical, or material support: Ananth.
Study supervision: Ananth, Friedman, Vintzileos.
Conflict of Interest Disclosures: None reported.
Additional Contributions: Jennifer Zeitlin, DSc, Institut National de la Santé et de la Recherche Medicale, Paris, France, provided thoughtful and constructive comments on this manuscript. She was not compensated for this work.
Create a personal account or sign in to: