We performed a retrospective cohort analysis of birth hospitalizations using the Premier Healthcare Database Special COVID-19 Release (PHD) (release date: July 19, 2022), an all-payer, administrative database containing patient-level discharge records from nongovernmental, community, and teaching hospitals representing 20% of hospital admissions throughout the U.S [13]. We included data from 523 hospitals reporting birth hospitalizations for newborns during May 01, 2020–February 28, 2022.

We identified birth hospitalizations using International Classification of Diseases, 10th Revision, Clinical Modification (ICD-10-CM) diagnostic codes (Supplementary Table). From May 2020–February 2022, 1,518,554 inpatient birth hospitalizations were reported in the PHD. We excluded 1094 duplicate birth records from 544 newborns, 118,329 newborns from 127 hospitals that did not report newborns in all comparison periods of this study, and 10,419 newborns who were born ≥35 gestational weeks and had charges for ventilation support and were discharged home with no evidence of being admitted to a higher-level nursery or ICU. The resulting study sample included 1,388,712 newborns (Supplementary Fig. 1).

COVID-19 status during the birth hospitalization was based on ICD-10-CM code U07.1 (COVID-19, virus identified) during May 2020–February 2022 [14] or positive RNA laboratory test results for SARS-CoV-2 (available for 29.7% of hospitals). Time periods were defined using admission month and to reflect the times at which the Delta and Omicron variants became the predominant circulating variants of SARS-CoV-2: [15] Pre-delta (May 2020–June 2021), Delta (July–November 2021), and Omicron (December 2021–February 2022).

Ventilation support was defined through hospital chargemaster records, ICD-10-CM procedure codes, and Current Procedural Terminology codes and categorized to represent the highest level of ventilatory support received (invasive vs noninvasive) (Appendix 1). Nursery level of care (Level I, Level II, Level III, or higher) was defined using hospital chargemaster records (Appendix 1). Newborns with billing codes for admission to an ICU, step-down unit, or a Level III or Level IV nursery were considered to have admission to a Level III or higher nursery; newborns with chargemaster codes for admission to a Level II nursery were categorized accordingly; and newborns with codes for basic room and board or without codes for higher-level nurseries were categorized as admitted to a Level I nursery. Level of care was categorized to represent the highest level of care received (Level III + vs. Level II) (Appendix 1). There were 7656 newborns who had ventilation in the absence of specific charges for Level II or higher care. Of these, 1491 were discharged to another facility, discharged to hospice, or deceased within a day of birth; these newborns were included for analysis without imputing a higher-level nursery. For the remaining 6165 newborns, we backfilled Level III or higher care when [1] they were born ≤35 weeks’ gestation (1876/6165) or [2] had a billing record that indicated likely ICU-level care (e.g., caffeine, surfactant, total parenteral nutrition, furosemide, lorazepam, nitric oxide, umbilical catheter) (4289/6165).

Newborn demographic and health characteristics (sex, race/ethnicity, payor, gestational age, small for gestational age, and discharge status) were described along with hospital characteristics (urbanicity and U.S. census region). Race and ethnicity were combined for this analysis: non-Hispanic Black, Hispanic, non-Hispanic Other/Unknown, and non-Hispanic White. Newborns with unknown Hispanic origin were assumed to be non-Hispanic if race was available. The non-Hispanic Other/Unknown category includes persons reported as a race other than Black or White for which sample size was too small to individually report and are combined into a single category to comply with the Health Insurance Portability and Accountability Act and other regulatory requirements: Asian, American Indian/Alaska Native, Native Hawaiian/Pacific Islander, persons of more than one race, and persons with missing race. Gestational age and small for gestational age were defined using ICD-10-CM codes in the newborn health record. Gestational weeks at the time of delivery are reported through ICD-10-CM codes for preterm newborns; newborns without an ICD-10-CM code for preterm birth were assumed to be born full-term (Appendix 1). Gestational age strata were chosen to reflect the critical threshold at which the majority of infants would not routinely receive respiratory support (32 gestational weeks) [16] nor require admission to an ICU after birth for prematurity (35 gestational weeks) [17].

Disposition was identified from the patient discharge status codes (home; other care [includes discharge/transfer to other facilities]; other/unknown [left against medical advice, information not available]; and death/discharge to hospice). Because data are not reported into the PHD until discharge, analyses of length of hospital stay are limited to newborns discharged within 90 days of birth.

Pearson chi-squared tests were used to assess differences in infant and hospital characteristics by COVID-19 status. Median length of stay in days was calculated for newborns born ≥35 gestational weeks, the gestational age cutoff at which ICU admission is not routine, stratified by ICU admission. Levene’s test for homogeneity of variance was used to verify the assumption of equal variance. Wilcoxon rank sum tests were used to compare length of stay by COVID-19 status. Poisson regression models with robust standard errors were used to calculate relative risks for critical care, accounting for within-hospital correlation. Relative risks were estimated controlling for race/ethnicity and provider U.S. census region and stratified by gestational age (<35/≥35 weeks for higher level nursery care and <32/≥32 weeks for ventilation support). A priori significance level was set to p < 0.05. Analyses of large datasets can result in statistically significant findings that may not be meaningful; consequently, we a priori chose to report significant results with a risk ratio (RR) indicating at least 5% difference for newborns with and without COVID-19 or for newborns with COVID-19 between time periods (Pre-delta: May 2020–June 2021; Delta: July–November 2021; Omicron: December 2021–February 2022). All analyses were performed in SAS® 9.4 (SAS Institute, Cary NC). The data that support the findings of this study are available from Premier Inc. and were used under license for the current study. Restrictions apply to the availability of these data, and so line-level data are not publicly available. Analytic code is available upon request to the first author.

We conducted five sensitivity analyses to assess the impact of exclusion criteria and definitions on the outcomes. First, we excluded all 7 656 newborns who received ventilation support with no indication of being admitted to a Level II nursery or higher level of care who were not discharged home. Second, to assess potential conflation of newborn infection with maternal infection, we excluded newborns with an ICD-10-CM COVID-19 code, no positive laboratory result when laboratory results were available, and an ICD-10-CM code for Contact with and (suspected) exposure to COVID-19 (Z20.822) to account for the possibility that the COVID-19 diagnosis code on the newborn record was only indicative of a maternal infection. Because the Z20.822 ICD-10-CM code went into effect January 1, 2021, this sensitivity analysis was limited to newborns born after that date. Third, because newborns could be transferred to another facility to receive higher level care at the end of their birth hospitalization, we excluded newborns who were transferred to other facilities or had an unknown discharge status. Fourth, to address potential misclassification of newborns in time periods of variant predominance, we excluded newborns from June 2021 and December 2021 from the analysis because the Delta and Omicron variants rose to be the dominant circulating variants the weeks ending June 26, 2021 and December 25, 2021, respectively [15]. Fifth, in order to test a more generous threshold of when most newborns would be expected to require mechanical ventilation support due to their gestational age, we recalculated the risk for ventilation support, stratifying at 34 instead of 32 gestational weeks.

This activity was reviewed by the Centers for Disease Control and Prevention (CDC) and was conducted consistent with applicable federal law and CDC policy; the activity was determined to meet the requirements of public health surveillance as defined in 45 CFR 46.102(l) [2].