This cohort analysis examined a mega database to identify the association of PA in pregnant women with increased risk of IVH and severe IVH in their VLBW infants using the National Inpatient Sample (NIS) in the years 2016–2018.

The NIS dataset is produced by the Healthcare Cost and Utilization Project (HCUP) sponsored by the Agency for Healthcare Research and Quality (AHRQ). NIS is a de-identified, publicly available inpatient healthcare database. It contains data of more than 7 million hospital stays each year. NIS uses a stratified, single-stage cluster sampling design, with region, urban/rural location, teaching status, ownership, and bed size to identify strata. After stratification, a random sample of 20% of the hospitals from the target population is included. HCUP program requires the use of weighted samples to reflect national trends. NIS contains information on all patients including patient demographics, primary and secondary diagnoses, primary and secondary procedures, hospital characteristics, payment source, length of stay, and patients’ disposition. All types of admissions were included whether they were direct admissions, admissions from the emergency room or transfers from other hospitals. To avoid duplicate inclusion, infants who were transferred out of the delivery hospital were excluded.

Preterm VLBW infants were identified in the dataset using the code (NeoMat = 2) that is unique to neonatal hospitalization at birth in addition to the respective International Classification of Diseases codes - 10th version (ICD10) for VLBW infants with respective gestational age (GA) and BW categories. Infants born to mothers with specific medical or perinatal diagnosis were identified using respective ICD-10 codes. Similarly, infants developed postnatal condition or adverse effects were identified using respective ICD-10 codes. For different IVH grades, we used the ICD-10 codes: P520 (grade 1), P521 (grade 2), P522 (grades 3 and 4), P5221 (grade 3), and P5222 (grade 4). For PA, we used the code P021 (we extracted our samples from the neonatal records where all ICD-10 codes are designed for pediatrics population, but the obstetrical codes for PA, O45.9, O45.90, O45.91, O45.92, and O45.93, would be found only under maternal records and could not be used for this study). For all BW categories below 1500 grams, we used the codes: P0715, P0714, P0703, P0702, and P0701. Infants with central nervous system (CNS) anomalies, congenital heart disease (CHD), congenital diaphragmatic hernia (CDH), abdominal wall defects, hypoxic ischemic insult events at birth, and common genetic and chromosomal disorders were excluded from this analysis as any of these diagnoses may act as an independent risk factor associated with adverse neurological outcomes. This study involved publicly available de-identified data; therefore, it was exempted from review by the Institutional Review Board.

Infants included in the study were divided into two groups: infants born to mothers who suffered placental abruption and infants born to mothers who did not suffer placental abruption. Demographic, clinical, and perinatal characteristics were compared between the two groups. The odds ratios (OR) to develop IVH or severe IVH in VLBW infants born to pregnant women who suffered placental abruption were calculated using chi square testing. Such association was reexamined using logistic regression models to calculate adjusted OR while controlling for potential confounding variables including maternal conditions (maternal diabetes or hypertension), perinatal occurrences (breech or malpresentation, nuchal cord or cord prolapse, placental previa, or chorioamnionitis), infants’ demographics (sex, GA, BW, multiple gestation, small for gestational age [SGA] status), and postnatal conditions (respiratory distress syndrome [RDS], pneumothorax, pulmonary hemorrhage or hypertension, apnea or anemia of prematurity, necrotizing enterocolitis [NEC], or sepsis).

Binomial and categorical variables were described using frequencies and percentages. Chi-square and Fisher’s exact tests were used to compare groups. Statistical significance was set at p < 0.05. Regression analysis was performed to verify significant associations while controlling for confounders. Data analysis was performed using SAS version 9.4 (SAS Institute Inc., Cary, NC).