Educational Performance of Children Born Prematurely

Key Points

Question

How well do children who were born prematurely perform in elementary and middle school?

Findings

In this population-based study of 1 301 497 children, those who were born prematurely frequently performed at levels comparable to the level of children born at full term on tests of kindergarten readiness and standardized mathematics and reading competences. A small proportion of children born near the limits of viability achieved gifted status.

Meaning

The effects of prematurity—including extreme prematurity—are not uniformly deleterious.

Abstract

Importance

Survivors of preterm birth often present with medical morbidities; however, variation in their long-term educational performance has not been well described.

Objective

To estimate the association between gestational age and 4 outcomes in school-aged children: readiness to enter kindergarten, scores on standardized tests in elementary and middle school, gifted status, and low performance.

Design, Setting, and Participants

In a retrospective cohort study, children born in Florida between 1992 and 2002 at 23 to 41 weeks’ gestation who entered Florida’s public schools between 1995 and 2012 were assessed for kindergarten readiness and tested in mathematics and reading in grades 3 through 8. Data analysis was performed from January 12, 2016, to March 1, 2017.

Exposures

Gestational age at birth.

Main Outcomes and Measures

Kindergarten readiness, scores on the Florida Comprehensive Achievement Test (FCAT), classified as gifted, and classified as low performance.

Results

A total of 1 527 113 singleton infants with gestational ages of 23 to 41 weeks born between 1992 and 2002 were matched to Florida public school records. Of these, 1 301 497 children were included in the analysis; 641 479 (49.3%) were girls. A total of 301 (65.0%) Florida children born at 23 to 24 weeks’ gestation were designated as ready to start kindergarten. When the FCAT test scores were adjusted for potentially confounding maternal and infant variables, children born at 23 to 24 weeks’ gestation performed 0.66 SD (95% CI, −0.73 to −0.59) lower compared with those born at full term. A total of 123 554 (9.5%) of all Florida-born public school students were considered gifted, including 17 (1.8%) of those born at 23 to 24 weeks’ gestation. In comparison, 75 458 (5.8%) of all Florida-born public school students were low performing; 310 (33.5%) of these children had been born at 23 to 24 weeks’ gestation. Kindergarten readiness, FCAT scores, and gifted status were positively related to gestational age, whereas low performance was inversely related to gestational age.

Conclusions and Relevance

Although gestational age has long been associated with poor educational performance, a sufficient proportion of children born near the limits of viability performed within expected school norms, warranting further investigation into how and why certain children are able to overcome the educational burdens that may follow preterm birth.

This population-based study examines the educational outcomes of children who were born prematurely.

Introduction

Preterm birth confers risks of mortality, medical morbidities, and adverse neurodevelopment in early childhood. Numerous studies have documented neurodevelopmental impairment among infants with shorter gestational ages (extremely premature, <28 weeks’ gestation), typically assessed close to 2 years’ corrected age. Both motor and cognitive disabilities have been related to antecedent maternal health characteristics and medical interventions. Neurodevelopmental impairment most frequently has been assessed within clinical or observational trials. The findings have been invaluable in generating thoughtful, evidence-based practice factors and parental counseling both before and after preterm birth. However, selected exclusion criteria, biases inherent to maternal or infant referrals to participating centers, and/or limited longitudinal follow-up may impair the external validity of their conclusions. Moreover, the associations of later markers of neurodevelopmental impairment or, conversely, academic achievement with early infant characteristics remain elusive, particularly as they pertain to outcomes achieved when children reach school age.

While clinicians tend to the medical needs of preterm infants, it is not surprising that their prognoses of school age outcomes can be imprecise. Clinicians may have difficulty describing these longer-run outcomes since they must rely on research reports that differ greatly because of variation in size of reporting institutions, receipt of perinatal care, number of countries with long-term data sets, stability of families, and methods of assessment. Parents in turn may have difficulty grasping how general outcomes for premature infants relate to their child and how the future might look for their infant. Outcomes with considerable variability, such as neurodevelopmental impairment, cerebral palsy, and cognitive impairments, may be hard concepts for parents to grasp as opposed to a more concrete outcome, such as school readiness. However, measures of school readiness or later assessment of school performance may not be standardized across states or nations.

The aim of this study was to estimate the association between gestational age at birth and school readiness and performance, including performance on school-based standardized assessments between third and eighth grade among surviving infants born in Florida between 1992 and 2002. We hypothesized that the degree of prematurity is associated with both the readiness to start kindergarten and the level of impairment throughout school in these cohorts.

Methods

Study Cohorts

All birth certificates from singleton infants born at 23 to 41 weeks’ gestation in Florida from 1992 to 2002 and surviving to 1 year were obtained from the Florida Bureau of Vital Statistics. Certificates were matched to Florida public school educational records maintained in the Education Data Warehouse of the Florida Department of Education according to 4 variables: first and last names, date of birth, and social security number. To maximize correct matches, transposition of letters or numbers was allowed in up to 2 instances as long as the transposition did not match more than 1 record. Children were included if they (1) were born in Florida, (2) remained in Florida until school age, and (3) attended Florida public schools. Complete data for children attending private schools were unavailable and excluded. Results were compared with findings from the American Community Survey, a 1% representative sample of the US population, to provide face validity in match rates to the proportion of Florida children who go on to attend public schools.

This study was approved by the institutional review boards of Northwestern University, the University of Florida, and Florida’s education and health agencies and involved a thorough review of the Family Educational Rights and Privacy Act as well as the Health Insurance Portability and Accountability Act regulations.

Exposure and Outcome Variables

Gestational age at birth as recorded in the birth certificate was the primary exposure and was stratified into 9 groups: 23 to 24, 25 to 26, 27 to 28, 29 to 30, 31 to 32, 33 to 34, 35 to 36, 37 to 38, and 39 to 41 weeks’ gestation. Four educational outcomes were assessed: kindergarten readiness, mean Florida Comprehensive Achievement Test (FCAT) scores, gifted status, and low performance.

Kindergarten readiness assessment was a dichotomous (ready/not ready) variable ascertained between ages 5 and 6 years. From the 1998-1999 to 2001-2002 school years when children entered school, readiness was determined by a checklist of academic and behavioral skills designed by the Florida Department of Education and administered by the teacher. From the 2006-2007 to 2008-2009 school years, readiness was determined by the Florida Dynamic Indicators of Basic Life Early Literacy Skills assessment. Readiness was not assessed during the 2002-2003 to 2005-2006 school years.

The FCAT score is an assessment administered in Florida schools that we have standardized at grade and school year as a mean (SD) score of 0 (1). Available standardized scores for children in grades 3 to 8 (aged 8-14 years) were averaged for each child across grades and the 2 tests (math and reading).

A child with gifted status is defined by the Florida Department of Education as “one who has superior intellectual development and is capable of high performance.” Gifted status is assessed throughout schooling, with a median age at assignment of 8 years among those classified as gifted. This status can generally be considered static as 82.8% of children classified as gifted were so identified in grade 4 or before, and only 1.5% of students classified as gifted have their status revoked. Thus, this was a dichotomous variable (yes/no) constructed as ever being classified gifted.

Low performance is defined as either being below the 5th percentile on the averaged FCAT score or having a missing test score and a disability such that participation in the statewide assessment program was deemed “not appropriate.” A total of 76.6% of children who are ever considered disabled had already received their disability classification prior to commencement of testing in grade 3. For this reason, as well as the fact that prior disability in this or any grade is highly path dependent on prior disability classification, this dichotomous variable (yes/no) was constructed as a single measure of ever observed.

Statistical Analysis

Descriptive statistics were used to quantify the maternal and infant characteristics after stratifying the eligible cohorts by gestational age. Differences in the characteristics between the gestational age strata were identified using regression analysis and F test. Two-tailed P values for all characteristics were <.001.

The mean FCAT scores of gestational age groups were compared using linear regression models; logistic regression models were used to estimate the odds for kindergarten readiness, gifted status, and low performance. In all models, each gestational age group presents contrasts relative to full-term (39-41 weeks’ gestation) infants. To account for family background characteristics, analyses were adjusted for available maternal and child variables contained in the birth certificate. The variables considered were those related to gestational age and known or hypothesized to be related to future educational attainment or performance. These maternal variables include indicators for African American race, Hispanic ethnicity, birth outside of the United States, age at the time of birth (classified as <20, 20-29, 30-35, and ≥36 years), 3 educational levels (classified as <12, 12-15, and >15 years of education), number of prior births, health problems (eg, hypertension, diabetes), language spoken at home, start of prenatal care in first trimester, and marital status. Infant characteristics were controlled for by including indicators for sex, congenital anomalies (eg, spina bifida, Down syndrome), abnormal conditions at birth (eg, meconium aspiration syndrome, seizures), and timing of birth (month and year). The FCAT score analysis further controlled for indicators of the child’s age in third grade to account for differences in “holding back” decisions that may have been related to gestational age.

Three sensitivity analyses were conducted with a subset of the data. First, we repeated the analysis, restricting it to 1994-2002 birth cohorts to focus on the children who were more likely to have had common access to antenatal corticosteroid and postnatal surfactant in routine perinatal care. Second, we incorporated the available zip code-level data and Medicaid eligibility indicators for this same 1994-2002 data set as additional covariates (both shown in the eFigure in the Supplement). Finally, we conducted a subanalysis of healthy pregnancies, that is, those that excluded maternal health problems, children with congenital anomalies, abnormal conditions at birth, advanced maternal age, and individuals who spoke neither English nor Spanish, and included only those starting their prenatal care in the first trimester (eTable 1 in the Supplement).

Analyses were conducted with Stata, version 13 (StataCorp), and SEs corrected for heteroscedasticity were used to calculate 95% CI. Data analysis was performed from January 12, 2016, to March 1, 2017.

Results

From 1992 through 2002, there were 1 896 674 singleton births with complete birth certificate information on infants born from 23 to 41 weeks’ gestation in Florida. Matches were made for 1 527 113 (80.5%) records on children born in Florida who attended public schools. Results from the American Community Survey indicated that 80.9% of Florida-born infants attended public schools, suggesting that the linkage rate was high and that nearly all potentially matchable children were matched in our data. In comparing the full birth sample with children who enrolled in Florida schools, we noted that children who were in the final study sample had less educated mothers who were more likely to be unmarried and African American compared with those who either left the state or enrolled in private schools (eTable 2 in the Supplement; column 1 vs column 4).

The matched data set contained FCAT scores for 1 276 943 (83.6%) Florida-born children who subsequently attended state public schools. Records on children who left the state prior to the start of testing in the third grade or had missing test score information accounted for 225 616 (14.8%) and 12 777 (0.8%) of the remaining sample, respectively. Disability that precluded FCAT scoring accounted for 11 777 (0.8%) of the records; however, these children were included when gifted status and low performance were assessed. Finally, the kindergarten readiness measure was present in 678 072 (44.4%) student records because the state did not test readiness every year.

Maternal and infant characteristics were tabulated by gestational age (Table). Groups with gestational ages classified as extremely preterm (<28 weeks’ gestation) had a higher proportion of African American mothers and a lower proportion of married mothers compared with the very preterm (28-31 weeks’ gestation), preterm (32-36 weeks’ gestation), and term (37-41 weeks’ gestation) groups. Although statistically significant differences were identified for all variables (P < .001), frequencies in male sex of the infant, high school education, Hispanic ethnicity, and nativity status appeared to be relatively static with respect to the duration of pregnancies.

Table. Maternal and Child Characteristics^a.

Characteristic	Gestational Age, No. (%), wk
	23-24	25-26	27-28	29-30	31-32	33-34	35-36	37-38	39-41b	Total No.
Sample size, No.	925	2108	3365	4692	8567	19 121	67 111	321 235	874 373	1 301 497
Maternal characteristics
African American	489 (52.9)	1059 (50.2)	1546 (45.9)	2011 (42.9)	3564 (41.6)	7017 (36.7)	21 431 (31.9)	85 399 (26.6)	205 717 (23.5)	328 233 (25.2)
Hispanic	257 (27.8)	493 (23.4)	741 (22.0)	1107 (23.6)	1933 (22.6)	4277 (22.4)	14 392 (21.4)	76 179 (23.7)	209 527 (24.0)	308 906 (23.7)
Foreign born	275 (29.7)	535 (25.4)	730 (21.7)	1074 (22.9)	1904 (22.2)	3978 (20.8)	13 776 (20.5)	71 978 (22.4)	204 757 (23.4)	299 007 (23.0)
Years of education
<12	211 (22.8)	583 (27.7)	985 (29.3)	1337 (28.5)	2448 (28.6)	5372 (28.1)	17 582 (26.2)	71 327 (22.2)	198 464 (22.7)	298 309 (22.9)
12-15	605 (65.4)	1316 (62.4)	2008 (59.7)	2801 (59.7)	5183 (60.5)	11 351 (59.4)	40 183 (59.9)	196 905 (61.3)	532 483 (60.9)	792 835 (60.9)
>15	109 (11.8)	209 (9.9)	372 (11.1)	554 (11.8)	936 (10.9)	2398 (12.5)	9346 (13.9)	53 003 (16.5)	143 426 (16.4)	210 353 (16.2)
Married at birth	430 (46.5)	946 (44.9)	1547 (46.0)	2272 (48.4)	4158 (48.5)	9884 (51.7)	37 485 (55.9)	200 399 (62.4)	547 294 (62.6)	804 415 (61.8)
Age at birth, mean (SD), y	27.1 (6.6)	26.7 (6.9)	26.3 (6.8)	26.5 (6.8)	26.3 (6.8)	26.5 (6.7)	26.5 (6.5)	27.0 (6.3)	26.6 (6.1)	26.7 (6.2)
No. of previous births, mean (SD)	1.0 (1.4)	1.0 (1.2)	1.0 (1.3)	1.0 (1.3)	1.0 (1.3)	1.1 (1.3)	1.1 (1.3)	1.1 (1.2)	1.0 (1.1)	1.0 (1.2)
Infant characteristics
Female	469 (50.7)	1053 (50.0)	1701 (50.5)	2306 (49.1)	4046 (47.2)	9026 (47.2)	32 145 (47.9)	156 446 (48.7)	434 287 (49.7)	641 479 (49.3)
Birthweight, mean (SD), g	845.7 (663.6)	958.0 (549.1)	1193.4 (545.4)	1621.6 (721.7)	1858.7 (528.9)	2270.6 (482.4)	2747.3 (472.5)	3191.5 (461.4)	3478.0 (447.7)	3322.7 (557.6)
Congenital anomalies	18 (1.9)	41 (1.9)	69 (2.1)	100 (2.1)	142 (1.7)	285 (1.5)	922 (1.4)	2734 (0.9)	5604 (0.6)	9915 (0.8)

^aData are based on final sample used in the analysis. The P value associated with the F statistic in regression of the maternal/child characteristic on indicators for 8 gestational groups (39-41 weeks omitted as a reference group) was <.001 for all 12 characteristics/regressions.

^bUsed as reference group.

Of Florida children born at 23 to 24 weeks’ gestation, 301 (65.0%) were designated as kindergarten ready. This proportion increased with gestational age, with 391 376 (85.3%) of children born at full term deemed kindergarten ready. The unadjusted and adjusted associations between gestational age and kindergarten readiness were similar (Figure, A). The adjusted odds ratios (ORs) for premature compared with full-term children were 0.35 (95% CI, 0.29-0.44) for children born at 23 to 24 weeks’ gestation, 0.77 (95% CI, 0.70-0.85) for those born at 29 to 30 weeks’ gestation, and 0.89 (95% CI, 0.86-0.91) for those born at 35 to 36 weeks’ gestation.

Figure. Adjusted and Unadjusted Associations for 4 Educational Outcomes.

A, Kindergarten readiness odds ratios (ORs); error bars indicate 95% CI. There is a mean (SD) value of 0.853 (0.354) for the reference group (39-41 weeks’ gestation). B, Linear regression of mean Florida Comprehensive Achievement Test (FCAT) scores. Error bars indicate 95% CI. There is a mean (SD) value of 0.028 (0.992) for the reference group. C, Gifted status ORs; error bars indicate 95% CI. There is a mean (SD) value of 0.099 (0.298) for the reference group. D, Low performance ORs; error bars indicate 95% CI. There is a mean (SD) value of 0.054 (0.255) for the reference group. All coefficients are relative to full-term gestation. Sample was based on 1992-2002 birth cohorts, but birth cohorts 1997-1999 were excluded in the kindergarten readiness analysis because readiness was not assessed in these 3 years. Standard errors corrected for heteroscedasticity were used to calculate 95% CI. The unadjusted analysis did not include controls. The adjusted analysis controlled for the following maternal characteristics: race, ethnicity, nativity status, marital status, educational level, age at the time of the child’s birth, number of previous births, prenatal care started in first trimester, maternal health problems, and language spoken at home. It also controlled for the following child characteristics: sex, month and year of birth indicators, congenital anomalies, abnormal conditions at birth, and age in third grade for test scores.

Among children who took the FCAT, both adjusted and unadjusted FCAT scores were higher for those born prematurely but with longer gestational age compared with full-term children (Figure, B). For children born at 23 to 24 weeks’ gestation, their adjusted FCAT score was 0.66 SD (95% CI, −0.73 to −0.59) lower than that of those born at full term. Correspondingly, the FCAT scores of children born at 29 to 30 weeks’ gestation (−0.18 SD; 95% CI, −0.20 to −0.15) and 35 to 36 weeks’ gestation (−0.03 SD; 95% CI, −0.04 to −0.02) were also lower than the scores of those born at full term.

A total of 123 554 (9.5%) Florida-born public school students were deemed gifted. The proportion of gifted students increased with gestational age, from 17 (1.8%) children among those born at 23 to 24 weeks’ gestation to 86 199 (9.9%) for those born at full term. In multivariable analysis, the adjusted ORs increased with rising gestational age (P < .001) (Figure, C) and, compared with children born at full term, were 0.23 (95% CI, 0.14-0.38) at 23 to 24 weeks’ gestation, 0.63 (95% CI, 0.55-0.72) at 29 to 30 weeks’ gestation, and 0.91 (95% CI, 0.88-0.94) at 35 to 36 weeks’ gestation.

Finally, 75 458 (5.8%) of all Florida-born public school students were considered low performers, and 310 (33.5%) children born at 23 to 24 weeks’ gestation had this poor cognitive outcome. The proportion of children defined as low performers decreased with gestational age to 46 743 (5.4%) of those born at full term. Compared with full-term children, the adjusted ORs for those born prematurely were 7.68 (95% CI, 6.57-8.98), 2.01 (95% CI, 1.84-2.20), and 1.26 (95% CI, 1.22-1.30) for children born at 23 to 24, 29 to 30, and 35 to 36 weeks’ gestation, respectively (Figure, D).

Discussion

In this longitudinal, secondary data analysis of prospectively collected data on nearly all matchable children born in Florida from 1992 to 2002, the degree of prematurity was associated with worse educational outcomes. Children born at a lower gestational age (1) were less likely to be kindergarten ready, (2) scored lower on standardized tests, (3) were less likely to be classified as gifted, and (4) were more likely to perform at a low level compared with children born at full term. This pattern aligns with a recent description of the relationship between gestational age and kindergarten academic achievement in a representative sample of the 2001 US birth cohort. However, the differences between the 9 gestational age groups in our Florida sample were less than expected, with mean adjusted test scores for children born near the limits of viability (23-24 weeks) of 0.66 SD (95% CI, −0.73 to −0.59) lower than those of children born at full term. Gifted students were also present throughout all gestational ages. These results suggest that some children and their families in Florida were able to manage the morbidities related to preterm birth and, when measured educationally in a time frame beyond the reported follow-up in most transition studies, perform well within expected school norms. To our knowledge, this is the first US-based study utilizing state registry data and delineating educational performance in school from kindergarten through 8th grade based on gestational age while accounting for confounding factors related to maternal and socioeconomic differences.

Our analysis used a novel population-based cohort linking statewide births in Florida and educational assessments measured in nearly all live-born children over an 11-year period. This analysis also took advantage of a high rate of follow-up, with 67.3% in the full birth cohort population and 99.3% among potentially observable records. This high rate of follow-up occurred outside the context of a specific randomized trial or observational study and incorporated population-based outcome assessments in children’s educational settings. In doing so, we demonstrate the utility of a potentially reproducible method of assessing these cognitive outcomes and, specifically, prematurity in ways that may be recapitulated in other states or countries.

These results contribute to the growing body of knowledge that, although preterm birth—even near the traditional limits of viability—confers risks of long-term developmental impairments, the outcomes may not be uniformly deleterious. Considerable variability exists in the type and severity of impairment in extremely preterm infants. Our results suggest that, by 5 to 6 years, the negative effects of preterm birth on kindergarten readiness or later achieving competence in mathematics and reading tests might be less than expected when compared with full-term infants. These findings may assist clinicians in reassuring parents during and after their infant’s discharge from the neonatal intensive care unit. Moreover, these findings are supported by the large, longitudinal design covering 11 cohorts that corresponded to the time frame during which many advances in perinatal medicine (eg, antenatal corticosteroids and surfactant) were achieved.

Limitations

Limitations exist for this study. As an analysis of administrative data, our findings are associations and do not imply causality. Although our follow-up rate was high, it does not exclude the possibility that children who were not tested were systematically different from those who were. For example, our public schools sample comprised relatively fewer affluent families because affluent families were more likely to move away from Florida or educate their children in private schools where state-standardized testing was not required. In addition, our initial sample was positively selected because we required an infant to survive at least to 1 year. The Centers for Disease Control and Prevention–linked birth-death cohort data for Florida for 1995-2002 births suggest that the 1-year mortality rate is as high as 49% among infants born at 23 to 24 weeks’ gestation.

As in any analysis of secondary data, unknown and unmeasured factors (eg, incidence of bronchopulmonary dysplasia, intraventricular hemorrhage, caffeine or corticosteroid use, or even family and environmental factors) may have modified the observed associations. Moreover, coding errors within administrative data are always a concern; for example, gestational age may have been misclassified, which is a factor documented in other geographic regions. In our sample, the median birthweight for infants born near the limits of viability (23-24 weeks) was approximately 9% larger than the published values found in Centers for Disease Control and Prevention natality files. The median birthweights for all other gestational age groups were equal to or lower than official statistics that include 1-year mortality. However, we found the pattern of results to be identical in the subcohort of low-risk pregnancies.

Finally, 2 of our outcomes—kindergarten readiness and FCAT score—are objective measures based on results using standardized assessment instruments. Our other 2 outcomes—gifted status and low performance—have some level of subjectivity because these classifications are the result of parent and teacher conferences that culminate in drafting an individualized learning plan.

Conclusions

The limitations do not diminish the findings in this linked birth certificate and school-based data set that a portion of prematurely born children—even those of extremely preterm birth—will enter kindergarten ready to learn, do well on mathematics and reading tests, and be considered gifted. Understanding why certain children are able to overcome the educational burdens of preterm birth and thrive is an area for future investigation.

Supplement.

eFigure. Robustness: Restricted Sample and Additional Controls

eTable 1. Robustness: Estimates for “Healthier” Pregnancies

eTable 2. Demographic Characteristics of Mothers and Children