From 1998–2008, rates of preterm delivery (< 37 weeks of gestation) increased by 20% in the United States, a trend largely attributed to the rise in deliveries between 34–36 weeks (late-preterm birth (LPB)).^1,2 Although many secular trends accompany this pattern (e.g., multifetal gestations), nearly 75% of all singleton preterm deliveries can be accounted for by LPB.^1,3 Combined with mounting evidence suggesting that children born late-preterm are at increased risk for neonatal mortality and a range of neonatal morbidities,^4–6 the antecedents and sequelae associated with LPB have received much attention from the medical and public health communities.

LPB has also been associated with long-term cognitive and behavioral problems, including cerebral palsy, reading impairments, lower IQ scores, and higher levels of internalizing and attention-related problems at school-age.^7–12 Retrospective population-based investigations from Europe suggest that adults born late-preterm exhibit elevated risks for psychiatric diagnoses as well as socio-economic hardship,^13,14 providing further evidence of long-term impairment among LPBs.^13,14 Although little is known about factors contributing to such risks, there is increasing evidence that perinatal health may play a role. For example, children born late-preterm who were admitted to the neonatal intensive care unit (NICU) were more likely to exhibit deficits in non-verbal processing in relation to their full-term peers and late-preterm children not admitted to the NICU; no differences between the full-terms and non-NICU admitted LPBs were observed.¹⁵ Conversely, while LPB was unrelated to cognitive functioning, behavioral problems, and academic achievement in a longitudinal study spanning childhood and adolescence, analyses excluded pregnancies marked by perinatal health risks (e.g., maternal infection, substance use).¹⁶ Delivery circumstance (i.e., medically indicated or spontaneous) may also shed light on LPB associations with long-term child development outcomes,⁷ but this hypothesis has not been subjected to empirical investigation. Doing so may help identify etiological pathways that may be involved in any associations observed, given that delivery circumstance has been linked to specific indices of pre- and or postnatal complications.¹⁷

To address this issue, we used follow-up data from the Pregnancy Outcomes and Community Health (POUCH) Study, a prospective investigation of the etiological pathways leading to preterm delivery. Mothers provided reports of attention problems for the child resulting from the POUCH pregnancy using a norm-referenced instrument. To account for any long-term biological effects related to maternal mental health during pregnancy as well as reporting bias due to such characteristics at the time of the child survey, adjusted analyses incorporated measures of maternal psychosocial functioning obtained at each of these time periods. Our goals were to: 1) replicate previous findings linking LPB to higher levels of attention problems in childhood, and 2) examine whether delivery circumstances surrounding LPB (i.e., MI, spontaneous) were associated with these problems. In the event that significant associations were observed, we utilized additional medical record information and placental histology findings to investigate pathways that might be involved.

Relative to the subcohort, participation in the follow-up study and inclusion in the analytic sample was significantly associated with older maternal age, higher levels of maternal education, and married marital status at POUCH Study enrollment and older child age at POUCHChild Study participation (Table 1). Pregnancy outcomes and their association with maternal, child, and placental characteristics in the analytic sample are summarized in Table 2. Compared to FTB, MI LPB was associated with African-American race, diagnosis of hypertensive disorders during pregnancy, and SGA birth. No covariates were associated with spontaneous LPB.

In both the unadjusted and adjusted analyses, LPB collapsed across delivery circumstance was not associated with any aspect of parent-reported attention problems relative to FTB (unadjusted mean differences: oppositionality: 1.1 [95%CI −1.3,3.5]; inattention: 0.5 [95% CI −1.5,2.5]; hyperactivity: 0.7 [95%CI −1.5,2.9]; ADHD Index: −.10 [95%CI −0.3,0.1]) (Table 3).

When grouped by delivery circumstance, MI LPB was associated with higher levels of hyperactivity and tended to be associated with higher ADHD Index scores relative to FTB in unadjusted analyses (mean differences: 3.9 [95%CI .5,7.3] and 3.0 [95%CI −0.3,6.3], respectively) (Table 3). These findings were replicated following adjustment for potential confounders, except that the association between MI LPB and the ADHD Index exceeded significance thresholds (mean differences: 3.8 [95%CI 0.6,7.2] and 3.1 [95%CI 0.0,6.1]); findings corresponded to effect sizes (Cohen’s d) between .3 and .4, a range often interpreted as moderate in magnitude.³⁴ We found no evidence that associations were modified by child sex or maternal symptoms of SMIs at the time of the child survey. Findings were similarly unaffected when birth weight z-scores, maternal pre-pregnancy body mass index, or multiparity between POUCH Study enrollment and POUCHChild Survey participation were included in the multivariate models (data not shown). With the exception of the inattention subscale, MI LPB was not associated with any CPRS dimension among the preschool-aged children. However, MI LPB was significantly associated with the inattention, hyperactivity, and the ADHD Index among the 6–9 year olds (Table 4); the size of these effects was considerable, with d’s ranging from .5–.6. No outliers were detected in these age-stratified analyses.

Because hypertensive conditions have been linked to MI LPB,^27,31 analyses were repeated excluding women with clinical diagnoses of hypertensive disorders (N=85) or placental findings linked to these conditions (N=134), and were performed on the entire analytic sample to maximize cell sizes (see Methods). In each of the analyses (Table 5), the unadjusted and adjusted associations of MI LPB with hyperactivity and ADHD Index scores were no longer significant; effect sizes also decreased (d=.06–.25). Removal of SGAs reduced associations between MI LPB and ADHD Index scores to non-significance and attenuated the associations between MI LPB and hyperactivity to trend levels (unadjusted mean difference: 3.6 [95%CI −0.3,7.3]). In contrast, associations of MI LPB with hyperactivity and ADHD Index scores were maintained when women exhibiting severe levels of histologic chorioamnionitis³⁵ (HCA) (N=63) or placental abruption³⁶ (N=16) were removed from analysis (data not shown).

On average, children born late-preterm did not exhibit higher levels of parent-reported attention problems compared to those born full-term. However, when LPB was grouped by delivery circumstance, MI LPB was associated with higher levels of hyperactivity and higher scores on a composite marking inattentive and hyperactivity symptoms. Following removal of women with hypertensive disorders or placental findings associated with hypertensive conditions, these findings were attenuated and no longer significant, suggesting that hypertensive disorders may play a role in the associations.

Delivery during the late-preterm period has received considerable epidemiologic attention. To date, conditions linked to LPB include infections (e.g., HCA)³³ as well as hypertensive disorders, fetal growth restriction, and placental abruption;^37–39 this latter group of conditions has been linked to deliveries prompted by medical intervention.^37,40 Despite the range of factors associated with LPB, we provide evidence suggesting that diagnoses of hypertensive conditions as well as placental findings implicated in these conditions contribute to the associations reported here. Indeed, removal of women with placental abruption or who exhibit severe HCA did not affect any of our findings. Although removal of SGA births did slightly attenuate links between MI LPB and hyperactivity symptoms, this group likely includes children who are constitutionally small and may not represent the most sensitive approach to identifying fetal growth restriction. Hypertensive conditions during pregnancy, as well as the placental vascular findings investigated here, may mark disturbances in placental perfusion which in turn impede gas, waste, and nutrient exchange between mother and fetus;²⁹ links between hypertensive disorders and alterations in placental hormone secretion have also been observed (e.g., corticotropin-releasing hormone).^41–43 The extent to which such biological alterations affect brain development, particularly circuits implicated in the development of attention problems (e.g., frontal-striatal and cerebellar networks), require further elucidation.^44–47

Interestingly, studies investigating links between hypertensive conditions and the development of attention problems have generally yielded null results.^48–50 There may be several factors contributing to this apparent contradiction to our findings, including a relative lack of socio-economic and racial/ethnic diversity in the study samples,⁴⁹ and more importantly, grouping preterms and terms together when evaluating these associations.⁵⁰ To the extent that the threshold motivating medical intervention changes across gestational age, such approaches may obscure associations that are due to symptom severity and may be revealed by testing for effect modification by gestational age at delivery. Many studies also employ composite indices of perinatal risk that represent a range of obstetric and neonatal complications,^49,51–54 thus obscuring associations between any given condition and the outcome of interest.

Nonetheless, our study findings may be relevant to outstanding issues in the LPB outcome literature. Although delivery prior to 34 weeks is more strongly associated with a range of developmental outcomes, including long-term behavioral problems,^55,56 we provide evidence suggesting that links between MI LPB and attention problems may be clinically-meaningful (0.3<d<0.4). Additionally, previous investigations suggest that only a subset of children born late-preterm appear to be at-risk for later attention problems.^7,15 Our study extends these findings by identifying maternal hypertensive disorders as one potential source/marker of such heterogeneity.

In contrast to other investigations, we did not observe an association between LPB and attention problems when collapsing across delivery circumstance. There may be several reasons for this. First, the majority of children in our study were younger than those in studies documenting associations between LPB and attention-related problems; to the extent that main effects of LPB are more readily observed at older ages, especially with respect to attention problems (next paragraph), this might have interfered with our ability to detect such differences across the LPB subgroups. Additionally, other studies documenting associations between LPB and attention problems might reflect the characteristics of the samples from which they were drawn. For example, links between LPB and attention problems may be more readily apparent in samples enriched with children who were growth restricted at birth or experienced neonatal complications.^7,15 Providing some support for this hypothesis, analyses utilizing NICHD Early Child Care data found no differences between LPB and FTB in terms of behavioral problems,¹⁶ but children from complicated pregnancies were excluded from study.

There are several caveats to consider when interpreting our findings. First, although we employed a norm-referenced measure of attention problems, the main outcome of this study was parent-reported. We therefore cannot rule out that maternal knowledge of children’s PTB status influenced their reports of attention problems. However, because our findings were limited to MI LPB, it is unlikely this issue explains the study findings. Additionally, longitudinal investigations of parenting quality suggest that while there may be alterations in maternal behavior during the first year of life as a function of PTB status, such differences are not observed in toddlerhood when measured using the Strange Situation, a lab-based observation considered to be the gold standard for characterizing parent-child relationship quality at that age.^57–59 Second, measures of socio-economic and demographic factors used in this report were obtained at POUCH Study enrollment. Changes in family structure and/or income over time may contribute to individual differences in behavioral problems in children,^60–63 and we were unable to investigate such effects or otherwise contextualize findings with contemporaneous socio-economic indicators. Third, although our analyses were drawn from a community sample, some were limited in power due to the relatively small number of MI LPBs. This is particularly true with respect to the age-stratified analyses as well as those involving the removal of subgroups based upon placental or medical record information. Given these concerns, we chose not to adjust for multiple comparisons. While we believe the consistency of findings across analyses helps assuage concerns regarding Type I error rate, replication of our findings is warranted and encouraged. Our analyses also assumed that non-missing data from a particular weighted sampling stratum were representative of that stratum. To address this assumption, we performed unweighted analyses and observed no alterations to the pattern of findings reported here. Caution should nonetheless be employed when considering the generalizability of the study findings, particularly given the follow-up rate for eligible LPBs (54%). Combined with evidence suggesting that POUCHChild Survey participation was associated with higher socio-economic status indicators (see Results), the current analysis may not reflect the full-spectrum of LPBs. Finally, it is important to note that MI LPB is a risk factor for subsequent neonatal morbidity,⁶⁴ and as such, this study is unable to identify whether prenatal, postnatal, or some combination of these factors contribute to the outcomes observed here.

Despite these caveats, there are notable strengths of this investigation, including the prospective design and demographic breadth from which the analytic sample was drawn. Additionally, the availability of detailed medical record information and placental findings allowed for an investigation of factors that may underlie the associations reported here. Information regarding maternal psychosocial functioning during pregnancy and at child survey completion helped address concerns that co-occurrences between maternal and child psychopathology contributed to the study findings.⁶⁵ Because there is no indication that women with attention problems are more likely to exhibit MI LPBs, we do not believe that other factors contributing to such co-occurrence fully account for the findings here (e.g., heritability).

This study provides many directions for future investigation. In addition to replicating the current findings, the extent to which associations between hypertension-associated LPB and attention problems persist across development requires elucidation. Additionally, LPB has been associated with other types of behavioral problems in childhood besides attention problems (e.g., anxious/depressive symptoms),^7,13 but at present, we have very little understanding of what factors contribute to the different types of outcomes observed among these children. This latter issue underscores the need to better understand the mechanisms involved in associations reported here and elsewhere.

In sum, we provide evidence suggesting that MI LPB is associated with higher levels of attention problems in childhood. Given that previous investigations point to marked heterogeneity in outcomes of children born late-preterm, studies such as ours may help identify children at highest risk for later difficulties while at the same time, uncover the mechanisms involved.