2.1 Databases and Linkage

Children born with EA/TEF, from January 1, 2000 to December 31, 2005, were identified from the Arkansas Reproductive Health Monitoring System (ARHMS), a statewide, population-based birth defects surveillance program whose trained health information systems' staff abstract clinical, demographic, and socioeconomic data from medical records at all delivering hospitals in Arkansas, the state's only pediatric specialty-care hospital and associated clinics, and the high-risk pregnancy and prenatal diagnosis centers. ARHMS staff code birth defects based on the six-digit British Pediatric Association extension of the International Classification of Diseases, Ninth Edition Clinical Modification (ICD-9-CM) coding system, as modified by the Division of Birth Defects and Developmental Disabilities of the Centers for Disease Control and Prevention and ARHMS. Eligibility for ARHMS requires Arkansas residency of the mother at the time of delivery and diagnosis of a structural congenital malformation during pregnancy, after stillbirth or termination of pregnancy, after live birth, and up to 2 years of age.

Using key identifiers such as mother's and child's first and last names, ARHMS data were linked to the ADH vital health statistics that track data on all births and deaths in Arkansas. For each child born with EA/TEF, up to two unaffected children (i.e., children who did not have a structural birth defect) were randomly selected from birth certificates and matched on hospital and month of birth, sex, and maternal race and ethnicity. The resultant file was linked to death records using birth certificate number. Demographic and socioeconomic characteristics including maternal education and limited clinical variables and paternal age, race, and ethnicity were obtained from birth certificate data on surviving EA/TEF children and unaffected matched children. We defined size for gestational age based on the U.S. sex-specific national standard growth curve. As such, small (SGA) and large for gestational age (LGA) infants had a birth weight < 10th or > 90th percentile, respectively (Olsen et al. 2010), while appropriate for gestational age (AGA) had a birth weight at or between the 10th and 90th percentile.

Using the birth certificate number, the ADH linked ARHMS surviving children with EA/TEF to the Arkansas Hospital Inpatient Discharge (AHID) database. AHID database claims track discharges from hospitals in Arkansas that had a hospital length of stay of ≥ 1 day and include their dates of service, length of stay, procedures, discharge diagnoses, and discharge status. We determined the number of rehospitalization and length of stay of hospital readmissions for surviving children with EA/TEF within 5 years of discharge from the neonatal intensive care unit (NICU) hospitalization where the initial surgery for EA/TEF repair was completed. The AHID database also documents payer status. Maternal education collected from the ADH vital health records and child's payer status abstracted from the AHID database were considered proxy measures of socioeconomic status.

Arkansas Children's Hospital (ACH) is the only pediatric specialty hospital in the state where > 90% of infants born with EA/TEF in Arkansas are transferred shortly after birth. If the medical record number was not available, the child's name and date of birth as well as mother's name were used to link children born with EA/TEF to their medical records at ACH. Data collected from ACH medical records included clinical and surgical characteristics of the first NICU admission of these children as well as length of stay and discharge diagnoses of any EA/TEF-related readmission within five years of NICU discharge.

The file including data from surviving children born with EA/TEF and surviving unaffected matched children was securely transmitted to the Arkansas Department of Education (ADE). The ADE database includes scores and proficiency designations from the Arkansas public school students who took the Augmented Benchmark Examination tests in literacy and mathematics (grades 3–8) that are administered annually. Scores designated as proficient or advanced (“proficient”) represented performance at or above grade level, while scores designated “non-proficient” represented performance below grade level (ElHassan et al. 2022; Walden et al. 2023). The threshold of proficiency is set by the ADE and is determined by the percentage of correct answers on the exams. The ADE database also documents the total days of school absence per child at 3rd through 8th grade and whether the child was referred to special education at any of these grades. The two groups of children were linked to their annual ADE records using their social security numbers. Students who were home schooled, attended private schools, or moved out of Arkansas, did not take the benchmark examination, could not be linked to their achievement test records, and were excluded. This final file was de-identified and securely transmitted to the UAMS researchers for analysis. The primary outcome measure was dichotomous variable of proficiency for each of the literacy and mathematics achievement tests at 3rd to 5th grade. The secondary outcomes measures were dichotomous variable of referral to special education at 3rd, 4th, or 5th grade and total days of school absence at 3rd through 5th grade.

2.2 Statistical Analysis

Continuous variables were summarized as median (minimum (min), maximum (max)); categorical variables were summarized as numbers and frequency (%). Mann–Whitney U test and Fisher's exact tests were used to compare statistics of numerical and categorical variables between EA/TEF children with links to education data versus those without links to education data. We compared characteristics of children born with EA/TEF and matched unaffected children using mixed-effect models for continuous variables and mixed-effect logistic and cumulative regression models for binary and categorical variables, respectively. A random effect was included in these models to account for correlations between matched pairs. In cases where small (≤ 3) number of children was observed for a binary variable among a comparison group, a Fisher's exact test was used. Characteristics with a p value < 0.2 were selected as covariates for the multivariate statistical models. At each grade, a multivariate logistic regression model was used to assess the association of a binary outcome of proficiency in literacy (or mathematics) with a diagnosis of EA/TEF (EA/TEF vs. unaffected), after adjusting for covariates and accounting for paired correlation using the generalized estimating equation (GEE) method. The same logistic regression model was also applied to the binary outcome of referral to special education. For the numerical outcome of total days of school absence, a Poisson regression model was used to assess its association with EA/TEF diagnosis after adjusting for the same covariates and accounting for the paired effect using GEE method. Statistical analyses were performed using SAS 9.4 (SAS Institute Inc., Cary, NC).

3.1 Study Cohort

A total of 46 infants were born, 2000–2005, with EA/TEF in Arkansas. Eight children with associated cardiovascular malformations, chromosomal abnormalities, or extreme prematurity did not survive to school age. Two children were excluded since they were possibly transferred out of state after birth and were not admitted to the NICU at ACH. Sixteen children were homeschooled, adopted, attended private school, or moved out of state and could not be linked to their 3rd grade education data. The final cohort included a total of 20 children with EA/TEF and 31 matched children without a structural birth defect (Figure 1).

3.2 Study Cohort Characteristics

There was no significant difference in parental or neonatal characteristics between those children who were linked versus not linked to their education data (Table S1). Most mothers of children born with EA/TEF were Caucasian (75%), received prenatal care (94%), and completed at least a high school education (81%). Infants had a median gestational age of 38 weeks (min = 30; max = 42) at birth and were typically singleton (83%) and AGA (75%). The majority of these infants had EA with distal TEF (83%) and more than half (56%) had associated congenital anomalies (12 infants had non-cyanotic heart disease including ventricular or atrial septal defect; 7 infants had a gastrointestinal anomaly including imperforate anus or duodenal atresia; 8 infants had a renal anomaly including absent kidney, horseshoe kidney, multi-dysplastic kidney, hydronephrosis, or ureteropelvic obstruction; 10 infants had a musculoskeletal anomaly including tethered cord, absent radii, rib anomaly; and 10 of these patients had at least anomalies in 2 systems). Eighty-nine percent had undergone an open surgical repair with almost half (47%) developing a stricture and 11% developing anastomosis leak, afterwards. These children received a median dose of 0.65 mg/kg of morphine and 0.3 mg/kg of lorazepam for pain and sedation control during their NICU hospitalization. The median length of their NICU stay was 20 days with almost half (46%) requiring EA/TEF related readmissions within the first 5 years of discharge from NICU hospitalization where the initial surgery for EA/TEF repair was completed.

The demographic, socioeconomic, and clinical characteristics of the matched unaffected children were obtained from the birth certificates and AHID database (Table 1). Among children who could be linked to their education data, unaffected children had a higher median birth weight (3260 (min = 879; max = 4649) grams) in comparison to children born with EA/TEF (median birth weight = 2864 (min = 1049; max = 4621) grams) (p = 0.02) though comparable gestational age and size for gestation.

3.3 Educational and School Outcomes

The number of children with EA/TEF and unaffected children who were linked to their education data decreased as they advanced from 3rd to 5th grade (Table 2). The ages of the children born with EA/TEF and unaffected matched children were comparable and were estimated at 9, 10, and 11 years-old at 3rd, 4th, 5th grade achievement-tests examination. No significant differences were observed in literacy (3rd: 70% vs. 71%; 4th: 83% vs. 81%; 5th: 94% vs. 87%, p > 0.05) or mathematics proficiency (3rd: 80% vs. 87%; 4th: 83% vs. 78%; 5th: 75% vs. 78%, p > 0.05), rate of referral to special education (20% vs. 19%, p = 0.65), or total (3rd through 5th grade) days of school absence (47 (min = 8; max = 488) vs. 26 (min = 1; max = 241) days, p = 0.07) between EA/TEF and unaffected children, while accounting for matching and after adjusting for differences in characteristics between the two groups (Table 2).