Objective
This study compares school-age outcomes among preterm (PT) (32 0/7-<34 weeks), late PT (LP) (34 0/7-<37 weeks), and full-term (FT) infants to assess cognitive sequelae of LP births.
Study Design
We obtained linked birth and educational data for all nonanomalous singleton infants born 1994 through 1998 in New York City who had a third-grade standardized test score (n = 215,138).
Results
Children delivered LP and PT had 30% and 50% higher adjusted odds of needing special education than those delivered FT (adjusted odds ratio, 1.34; 95% confidence interval, 1.29–1.40; and adjusted odds ratio, 1.53; 95% confidence interval, 1.30–1.69). They also had lower adjusted math and English scores than those delivered FT (math: 7% and 10% of SD, respectively; English: 4% and 6% of SD). A linear association between gestational age and test scores was seen through 39 weeks’ gestation.
Conclusion
There is a significant risk of developmental differences in PT and LP infants compared with FT infants.
Despite multiple interventions to reduce the rate of preterm (PT) delivery, the PT birth rate has continued to increase over the past decade. Most of this increase is among infants born toward the end of the PT period, at 34-36 completed weeks of gestation, during the time known as late PT (LP). The rate of LP births increased 20% from 1990 through 2006 and accounted for >70% of all singleton PT deliveries in the United States during that time period.
For Editors’ Commentary, see Table of Contents
See related editorial, page 191
Mounting evidence has shown that infants born LP are less healthy than infants born at term. They are more likely than term infants to suffer from mortality and short-term morbidity such as increased temperature instability, hypothermia, hypoglycemia, hyperbilirubinemia, rehospitalization, and respiratory distress. Evidence has also suggested that infants born LP have similar risks of adverse neonatal morbidities as PT infants delivered at 32-33 weeks and 6 days’ gestation. Recent studies also suggest that risks decrease as gestation increases, even up to 39 completed weeks of gestation.
In 2005, the National Institute of Child Health and Human Development called for research regarding long-term morbidity for LP births as there is a paucity of data on long-term neurodevelopmental status. One study suggests that long-term health risks may include cerebral palsy, significantly higher rates of developmental delay and disability, and disorders of psychological development, behavior, and emotion. However, a significant knowledge gap remains about the educational outcomes of the large population of LP and PT infants once they reach school age.
We performed a retrospective cohort study to evaluate school-age educational outcomes among singleton LP infants compared with outcomes of PT and full-term (FT) infants. Differences in outcomes between PT and FT births and the impact of advancing gestational age in weeks on mean test score achievement were also examined. We hypothesized that school-age outcomes of LP infants would be significantly better than outcomes of PT infants, and significantly worse than those of FT infants.
Materials and Methods
The study population was drawn from the Longitudinal Study of Early Development data warehouse, which was constructed using probabilistic linkage technology to join child records from the New York City Department of Health and Mental Hygiene (DOHMH) Bureau of Vital Statistics birth and death certificates; the DOHMH Early Intervention Program and Lead Poisoning Prevention Program registries; and New York City Department of Education (DOE) enrollment, special education, and achievement data. The warehouse was constructed using probabilistic linkage software (Quality Stage; IBM, Armonk, NY) to link records based on values for identifiers including the child’s name, date of birth, sex, address, social security number, mother’s name and date of birth, and father’s name. Final samples were reviewed to verify a low false match rate (<1%).
Our study population included 215,138 children born 1994 through 1998 to mothers residing in New York City. We included all nonanomalous singleton infants who were delivered from 32 0/7-42 0/7 weeks’ gestation and were enrolled at DOE in grade 3. This cohort represents 47% of the eligible sample in New York City at birth as children who moved before beginning school, enrolled in private school, were homeschooled, or died were not part of the data set. Additionally, children who were born outside of New York City but who were enrolled in New York City public schools were excluded because their birth information was not available. Clinical estimate of gestational age came from the birth certificate, where it is reported in weeks and days. Exposure was defined as gestational age at delivery, which was classified into 3 groups: 32 0/7-33 6/7 weeks (PT), 34 0/7-36 6/7 weeks (LP), and ≥37 0/7 weeks (FT). Children recorded as having an Individualized Education Plan at any time <2007 (aged 9-13 years) were identified as having an educational disability necessitating special education services. Cognitive outcomes were assessed using standardized scores from DOE-administered third-grade English Language Arts (ELA) and math tests. These scores reflect mastery of the curriculum in math and ELA, and are used to measure performance within students over time as well as across students, schools, and school districts in all nonspecialized public schools in New York City. To permit comparisons across instruments and to facilitate interpretation of effect sizes, we transformed all test scores to z-scores by the child’s year of birth.
Maternal demographic and pregnancy characteristics, as well as birth outcomes of the child, all obtained from birth records, were examined as covariates and potential confounders. Maternal demographic characteristics included age (<20, 20-35, >35 years), level of education (≤high school, >high school), employment at the time of birth, race/ethnicity, insurance status, and parity (nulliparous or multiparous). Pregnancy characteristics included tobacco use during pregnancy, presence of any maternal medical risk factors (anemia, cardiac disease, lung disease, gestational diabetes, chronic diabetes, chronic hypertension, previous PT or small for gestational age infant, and/or renal disease), any complication of labor and delivery (preeclampsia, eclampsia, placenta previa, seizures during labor, cord prolapse, placental abruption, and/or fetal distress), and mode of delivery (vaginal or cesarean).
We controlled for whether the child was born at <10th percentile of birthweight for gestational age based on sex-specific US standard birthweights, reported need for neonatal intensive care, 5-minute Apgar score (<7, ≥7), and infant sex. We also included days absent in third grade (from the DOE records) as a covariate in our models to account for the direct relationship between attendance and test scores.
Statistical analysis
Criteria for significance were set at alpha <0.05 for all analyses. We assessed bivariate relationships between categorized gestational age and all other variables using the χ 2 statistic. We also examined bivariate relationships between each outcome variable and all potential predictor variables. Logistic regression was used to determine the odds that a child needed special education, and generalized linear models were constructed to assess the impact of gestational age category and of gestational age in weeks on test score achievement. We then assessed the effect of gestational age in continuous weeks on mean test scores adjusting for maternal characteristics, pregnancy characteristics, and birth outcomes, and determined the gestational age at which the relationships no longer hold. The multivariable models included all hypothesized covariates found to be significantly associated with test scores (chisq P < .05). All analyses were conducted using software (SAS System, version 9.2; SAS Institute, Cary NC). This study was approved by the New York City DOHMH Institutional Review Board.
Results
Of the 215,138 infants in the study, 2332 (1.1%) were PT, 13,207 (6.1%) were LP, and 199,599 (92.8%) were FT. Baseline characteristics of our study sample, stratified by gestational age category, are presented in Table 1 . Maternal race/ethnicity differed between all comparison groups as did tobacco use (both P < .001). There were a higher proportion of black, non-Hispanic women delivering PT than LP, and LT than FT. Similarly, tobacco use was more prevalent among mothers with PT than LT deliveries and those with LT than FT babies. The maternal demographic and pregnancy characteristics of the PT group were similar to the LP group for the majority of the other covariates, and the LP group differed significantly from the FT group. The proportion of women with complications of labor and delivery was significantly higher in the PT compared with the LP group and the LP group compared with women who delivered FT infants (24.9% vs 16.0% and 16.0% vs 7.8%).
