Study Cohort
Our cohort included children with CHD born in NC between 1998–2003 identified from electronic medical records (EMR) between January 1, 2008, and December 31, 2013, across five tertiary care centers (Carolinas Medical Center, Duke University, East Carolina University, University of North Carolina, Wake Forest School of Medicine) in NC, which provide comprehensive care for patients with CHD. Cases were identified using International Classification of Diseases (ICD), Ninth Revisions codes, and were classified by hierarchical native anatomic complexity groupings similar to previously published algorithms based on the individuals’ hemodynamic severity and basic anatomy: severe disease, shunt disease, valve disease, and other.15 Cases with an isolated code of 745.5 were excluded from this analysis since secundum atrial septal defect, a CHD, cannot be distinguished by ICD codes from patent foramen ovale (PFO), a normal variant and, therefore, inclusion of 745.5 may overestimate cases with CHD.15 Severe CHD included endocardial cushion defects, interrupted aortic arch, tetralogy of Fallot, total anomalous pulmonary venous return, transposition complexes, truncus arteriosus and univentricular hearts. We then linked these data with the North Carolina Department of Public Instruction educational records. The NC Department of Public Instruction collects information on end of grade (EOG) testing, academic promotion, and receipt of exceptional services for all children in public schools in NC. The study was approved by all of the participating institutional review boards. Only individuals authorized by the NC Department of Public Health had access to confidential information.
Data Linkage
We identified children born with CHD between 1998–2003 in the North Carolina Birth Defects Monitoring Program (NCBDMP) using ICD-9 codes. This cohort was linked to the NC Department of Education records by NC Department of Public Instruction. Records from the NCBDMP were matched to educational files using the child’s first name, last name, date of birth, and the last four digits of the child’s social security number.11 A child’s file was considered to be linked with the educational file if it had at least one record from third grade or later (Fig. 1).
Episodes of Care
We ascertained episodes of cardiac care from the EMR across the five tertiary care centers. Any inpatient or outpatient cardiac encounter on a date of service was considered an encounter in the health care system. We calculated the number of cardiac episodes of care by adding the number of inpatient and outpatient hospital days across the five tertiary care centers. To establish temporality between receipt of care and timing of when a child completed the EOG assessment, we included only episodes of care prior to the estimated date of the EOG assessment over the period between 1/1/2008 and 4/30/2012. Given the variability in timing of assessment across schools, we used April 30 of the year the student took the EOG as the latest date of the assessment (Fig. 1). We modeled episodes of care by stratifying children into those with number of episodes of care above the 50th percentile and those below the 50th percentile of the cohort distribution.
Outcome Measures
North Carolina End-of-Grade Assessments
All students attending NC schools are required to take end of grade (EOG) tests from third through eighth grade. These assessments evaluate student achievement in common core grade level standards including reading and math. All students in NC participating in the standard course of study are required to take the EOG assessment. Students with disabilities who have an individualized education plan, may qualify to take an alternative assessment and were not included. Previous studies showed this to be approximately 11% of the CHD population.11 The competencies assessed represent core knowledge that students should master at the end of each grade level. The educational records from the NC Department of Public Instruction included testing scores for both third-grade reading and math as well as scores for alternative assessments. Our analysis was restricted to those students who completed the EOG assessment in reading and math.
End of grade test scores were reported as percentile scores, developmental scale scores, and achievement levels. Developmental scale scores were calculated based on the number of items that were correctly answered and are based on the mean and standard deviation for each grade level. Developmental scale scores can be used to assess each student’s individual progress over time. Achievement levels represent students’ command of the common core state standard and are differentiated based on cut points from the developmental scale scores.
Achievement levels were reported over the following five categories: Level 1: limited command of the Common Core State Standards; Level 2: partial command of the Common Core State Standards; Level 3: solid command of the Common Core Standards; Level 4: thorough command of the Common Core Standards; Level 5: comprehensive command of the Common Core Standards and prepared for advanced content. We considered children meeting grade level performance if they scored from level 3 through level 5 on the EOG assessment. A score of level 1 or level 2 was considered a failing EOG score. If a child was retained, we used the child’s initial grade level performance to represent their achievement level.
Covariates
We included confounders that were associated with the presence of CHD as well as academic performance among children without a structural birth defect. In our model, we included CHD type, maternal education, sex, birth weight, gestational age, as well as race/ethnicity. We ascertained maternal education, race/ethnicity, birth weight, and gestational age from the birth record. Maternal education was modeled as less than high school or high school or greater. Race/ethnicity was modeled as a categorical variable, classified as: White Non-Hispanic, Black Non-Hispanic, Hispanic, and Other.
Statistical Analysis
We compared the distribution of means and proportions of baseline characteristics among children with CHD among those with episodes of care above and below the 50th percentile by t- tests and chi squared tests respectively. The odds and 95% confidence interval of not meeting grade level standards among children with number of visits above the 50th percentile compared to below the 50th percentile was estimated using logistic regression models controlling for CHD type, maternal education, sex, birth weight, gestational age, and race/ethnicity. In addition, we modeled the mean change in the developmental scale scores comparing those children with visits above the 50th percentile compared to below the 50th percentile using linear regression adjusting for these covariates. We modeled these associations for both reading and math scores in third grade.