Adverse pregnancy outcomes, maternal complications, and severe illness among U.S. delivery hospitalizations with and without a COVID-19 diagnosis

Abstract Background Evidence on risk for adverse outcomes from COVID-19 among pregnant women is still emerging. We examined the association between COVID-19 at delivery and adverse pregnancy outcomes, maternal complications, and severe illness, whether these associations differ by race/ethnicity; and described discharge status by COVID-19 diagnosis and maternal complications. Methods Data from 703 hospitals in the Premier Healthcare Database during March–September 2020 were included. Adjusted risk ratios overall and stratified by race/ethnicity were estimated using Poisson regression with robust standard errors. Proportion not discharged home was calculated by maternal complications, stratified by COVID-19 diagnosis. Results Among 489,471 delivery hospitalizations, 6,550 (1.3%) had a COVID-19 diagnosis. In adjusted models, COVID-19 was associated with increased risk for: acute respiratory distress syndrome (adjusted risk ratio [aRR] = 34.4), death (aRR = 17.0), sepsis (aRR = 13.6), mechanical ventilation (aRR = 12.7), shock (aRR = 5.1), intensive care unit admission (aRR = 3.6), acute renal failure (aRR = 3.5), thromboembolic disease (aRR = 2.7), adverse cardiac event/outcome (aRR = 2.2) and preterm labor with preterm delivery (aRR = 1.2). Risk for any maternal complications or for any severe illness did not significantly differ by race/ethnicity. Discharge status did not differ by COVID-19; however, among women with concurrent maternal complications, a greater proportion of those with (versus without) COVID-19 were not discharged home. Conclusions These findings emphasize the importance of implementing recommended mitigation strategies to reduce risk for SARS-CoV-2 infection and further inform counseling and clinical care for pregnant women during the COVID-19 pandemic.


INTRODUCTION
Pregnant women are at increased risk for severe illness from coronavirus disease 2019  compared to nonpregnant women [1]. Pregnancy, labor, and delivery also increase a woman's risk for medical complications [2]. Further, chronic conditions, severe maternal morbidity, and COVID-19 have disproportionately affected racial and ethnic minority groups [3][4][5][6]. Three U.S.-based studies have investigated adverse outcomes among women with and without COVID-19 at delivery [7][8][9]. Among pregnant women who delivered at a hospital system in Dallas, Texas, those with COVID-19 were no more likely to have a preterm birth, severe preeclampsia, or a cesarean delivery for fetal complications compared to pregnant women without COVID-19 [7]. In New York City, pregnant women with COVID-19 were more likely to have postpartum fever, hypoxia, and to have a hospital readmission (12.9%) following delivery than pregnant women without COVID-19 (4.5%) [8]. Both studies were unable to evaluate if a COVID-19 diagnosis was independently associated with the adverse outcomes or adjust for underlying medical conditions. In a study of geographically diverse hospitals, rates of preterm birth, preeclampsia, thrombotic events and death were higher among women with COVID-19 than those without at delivery [9].
These previous studies were limited in sample size [7,8] or did not investigate differences by race/ethnicity [9].
Thus, to extend the evidence, we examined the association between documented COVID-19 diagnosis at delivery and risk of adverse outcomes adjusting for underlying medical conditions and demographic characteristics overall and stratified by race/ethnicity.
To better understand the impact of COVID-19, we also describe discharge status by COVID-

diagnosis and select outcomes.
A c c e p t e d M a n u s c r i p t 5

METHODS
We performed a retrospective cohort analysis of all delivery hospitalizations using the Premier Healthcare Database Special COVID-19 Release (https://www.premierinc.com/; PHD-SR, release date 01/08/21). PHD-SR is a U.S. hospital-based, service-level, all-payer database that includes data on inpatient discharges from >849 geographically diverse nonprofit, non-governmental, community, and teaching hospitals and health systems from rural and urban areas [10], representing approximately 20% of all U.S. inpatient discharges. We included data from the 703 hospitals with delivery hospitalizations during March-September 2020.
Delivery hospitalizations among females aged 12-55 years were identified using ICD-10-CM diagnostic and procedure codes pertaining to obstetric delivery and diagnosis-related group delivery codes, excluding pregnancy losses (Supplemental Table)  [left against medical advice, discharged to court/law enforcement, still a patient, and information not available]; and death). Composite measures of any maternal complication and any indicator of severe illness were also created. Readmission was defined as any subsequent hospitalization at the same hospital within 30 days of delivery hospitalization discharge.
Discharge status was identified from the patient discharge status code. ICU admission was identified using the hospital chargemaster records (i.e., the comprehensive list of all items billable to a hospital patient or to a patient's insurance provider). Mechanical ventilation was identified through a combination of the hospital chargemaster records and ICD-10-CM procedure codes. All other outcomes were identified from ICD-10-CM diagnosis and procedure codes (Supplemental Tables). Pearson chi-squared tests were used to ascertain differences in demographic and hospital characteristics by COVID-19 diagnosis. Median maternal age was calculated overall and stratified by COVID-19 diagnosis; a t-test was used to determine if median maternal age differed by COVID-19 diagnosis. Poisson regression models with robust standard errors were used to calculate crude and adjusted relative risks for individual outcomes, accounting for within-hospital correlation [14]. Covariates included maternal age, race/ethnicity (Hispanic, non-Hispanic Black, non-Hispanic White, and non-Hispanic other/unknown), primary payor (Medicaid, private insurance, and other), obesity, any diabetes, any hypertension, and asthma.
Additionally, crude and adjusted relative risks were calculated for composite measures of any maternal complication and any severity of illness stratified by race/ethnicity. Individual outcomes stratified by race/ethnicity could not be assessed due to a lack of statistical power to detect significant differences. The proportion of women not discharged home (women who died, were discharged to other care [e.g., skilled nursing facility, hospice, other long-term care], and with other/missing discharge status) was calculated by maternal complications of interest, stratified by COVID-19 diagnosis. All analyses were performed in SAS Version 9.4. This activity was reviewed by 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).
Overall, 4.2% of all delivery hospitalizations had any adverse pregnancy outcomes of interest (Table 2) Most (98.9%) patients were discharged home from the delivery hospitalization (Table   2); however, 1.1% were discharged to other care, had other or missing outcomes, or died.
Overall, 1.2% were readmitted within 30 days; among these, there were no differences by  Table 3). The risk for any maternal complication by race/ethnicity was highest for non-Hispanic other/unknown (aRR= 8.1), followed by Hispanic (aRR= 7.5), non-Hispanic Black (aRR= 5.1), and non-Hispanic White (aRR=5.1). Similarly, risk for any severe disease was highest among non-Hispanic other/unknown (aRR=4.2), followed by Hispanic (aRR=3.5), non-Hispanic Black (aRR=3.3) and non-Hispanic White (aRR=2.9). However, the race/ethnicity specific 95% confidence intervals for any maternal complication and any severe disease overlap, indicating that the risks do not vary significantly by race/ethnicity in this population.

DISCUSSION
While the absolute risk of maternal complications and severe illness was low, individuals with a documented COVID-19 diagnosis at delivery hospitalization had 17 times the risk for death, almost 14 times the risk for sepsis, 13 times the risk for mechanical ventilation, five times the risk for shock, four times the risk for acute renal failure, and more than twice the risk for adverse cardiac event or thromboembolic disease compared with those A c c e p t e d M a n u s c r i p t 10 who did not have a COVID-19 diagnosis. Risks for any maternal complications or severe illness did not differ statistically by race/ethnicity. Our estimates of the risks for adverse pregnancy outcomes, maternal complications, and severe illness are low compared to some other studies of COVID-19 among pregnant women [15,16]. For example, in a cohort of hospitalized symptomatic pregnant women with laboratory-confirmed SARS-CoV-2, 16.2% were admitted to an ICU, 8.5% required invasive mechanical ventilation, and 0.7% died [15]. Differences in estimates are likely due to varied population denominators of pregnant women with COVID-19-associated hospitalizations [15,16] versus our population of all delivery hospitalizations, which provides a more appropriate denominator to establish risk. Our estimates of maternal complications and maternal severe disease were similar to a recent analysis using the same dataset [9]. However, our analysis adds to this previous work by stratifying by maternal race/ethnicity and by examining the proportion of women discharged home by maternal complications and COVID-19 diagnosis.
We also directly estimated relative risks instead of using odds ratios to approximate risk; our findings are consistent with the previous study's results.
In this study, individuals with a documented COVID-19 diagnosis were 34 times more likely to have a diagnosis of acute respiratory distress syndrome, a well-documented complication of COVID-19 [17,18], than those without a COVID-19 diagnosis. While COVID-19 appears to be a systemic disease [19], pulmonary disease is the most welldocumented manifestation.  [7,15]. Race and ethnicity minority populations have been A c c e p t e d M a n u s c r i p t 11 disproportionately affected by COVID- 19 [21]. Race and ethnicity may be markers for other factors that affect health, including neighborhood and physical environment, socioeconomic status, access to health care, and occupational exposure (e.g., frontline, essential, and critical infrastructure work) [3]. While the risk for any maternal complication or severe illness was highest for non-Hispanic other/unknown, followed by Hispanic and non-Hispanic Black individuals, each of the 95% confidence intervals associated with race/ethnicity specific adjusted risk ratios overlapped with each other and also with non-Hispanic White. Thus, in this study, we did not find statistically significant differences in risk of any maternal complications or severe illness by race/ethnicity. Although these data provide a large sample size, the individual outcomes of interest were rare, precluding further exploration of individual outcomes by race/ethnicity at this time.
We found women with COVID-19 were at slightly increased risk of cesarean delivery and preterm labor with preterm delivery than women without COVID-19 at delivery hospitalization. However, statistical significance does not imply clinical significance, especially in studies with very large sample sizes. Previous studies of pregnant women with COVID-19 admitted for delivery have conflicting findings related to cesarean delivery. One study found that women with COVID-19 were less likely to have a cesarean delivery (unadjusted risk ratio: 0.80; 95% CI: 0.64-0.99) than women without COVID-19 [7]. In contrast, cesarean delivery rates were higher among women with symptomatic (46.7%) or asymptomatic (45.5%) COVID-19 compared to women without COVID-19 (30.9%) [8]. In other studies of pregnant women admitted for delivery, those with COVID-19 were no more likely to deliver a preterm infant [7,8], but in a recent meta-analysis of primarily small sample sizes and case studies, preterm birth was higher among pregnant women with COVID-19 compared to those without COVID-19 [22]. Inconsistent findings may be due differences in obstetric intervention practices across populations and geography, our outcome A c c e p t e d M a n u s c r i p t 12 of preterm labor with preterm delivery (not preterm birth), and our ability to adjust for underlying medical conditions. Most women in our study population were discharged home after delivery hospitalization (98.9%). This was true even among women with an ICU admission, with 85.7% of women with and 97.6% of women without a COVID-19 diagnosis discharged home. While we did not have indication for ICU admission at the time of delivery hospitalization, reasons for admission may include precautionary infection control in the case of COVID-19 and obstetric-related issues such as follow-up of hypertensive disorders in pregnancy, control of sepsis, and hemorrhage. Hypertensive disease, hemorrhage, and cardiomyopathy or other cardiac disease have been documented as common diagnoses associated with delivery-ICU admissions [23]. However, the proportion of individuals not discharged home who had a diagnosis of adverse cardiac event/outcome, thromboembolic disease, and shock or who required mechanical ventilation was at least 20 percentage points higher among individuals with a COVID-19 diagnosis compared to individuals without a documented COVID-19 diagnosis. These findings suggest that some individuals have had a more complicated disease course and a longer recovery. Further study is needed to understand whether these findings are reflective of complications of acute infection, postacute hyperinflammatory disease, or late sequelae [24].
In our overall sample, 1.2% of women were readmitted within 30 days and this estimate was not significantly different by COVID-19 status at delivery hospitalization. Our readmission estimate was similar to a study of 30-day readmission after discharge from delivery hospitalization in 21 states [25]. In a study of 675 pregnant women admitted for delivery, postpartum readmission differed among women with symptomatic COVID-19 (6.7%), asymptomatic COVID-19 (3.6%) and without COVID-19 (1.5%) [8]; however, this study did not have sufficient sample size to adjust for underlying medical conditions. A c c e p t e d M a n u s c r i p t 13 Further investigation of risk factors for severe clinical outcomes to understand longterm recovery and to inform optimal treatment and management of laboratory-confirmed SARS-CoV-2 infection in pregnant women is needed. Further, whether maternal SARS-CoV-2 infection affects fetal/infant health is not well understood.

STRENGTHS AND LIMITATIONS
Using a large database of electronic health records and administrative data from geographically diverse hospitals across the U.S., this study investigated the risk of a number of adverse outcomes and their association with a documented COVID-19 diagnosis, adjusting for sociodemographic characteristics and underlying chronic diseases. The large sample size also enabled stratification by race/ethnicity; however, we were limited to examining non-Hispanic White, non-Hispanic Black, Hispanic, and other non-Hispanic women due to small cell sizes. Nevertheless, interpretation of these data is subject to several limitations. First, the analysis relied on ICD-10-CM codes in electronic health records and administrative data.
Thus, the exposure of interest (documented COVID-19), identified from the use of two ICD-10-CM codes during the study, and underlying medical conditions and outcomes could be misclassified or miscoded. However, a recent study found that the U07.1 code, used to identify COVID-19 during March-September, had a high sensitivity and specificity [26].
Second, information on timing of SARS-CoV-2 infection and course of illness was not available. The inclusion of asymptomatic cases of COVID-19 may bias our results towards the null. Third, many hospitals implemented universal SARS-CoV-2 testing among pregnant women admitted to labor units in spring 2020 [7,27,28], but this information was unavailable and might result in a misclassification bias if hospitals had different practices during the study period. Fourth, because outpatient records were not universally available and linkage across different hospital systems was not possible, the analysis was restricted to delivery hospitalizations and did not examine COVID-19 diagnoses, underlying medical A c c e p t e d M a n u s c r i p t 14 conditions recorded before the delivery hospitalization (e.g., during a prenatal visit), prenatal care utilization, or maternal delivery history (e.g., parity, previous cesarean delivery, or history of preterm delivery) which may impact the risk for maternal complications. Fifth, it is unknown how maternal race and ethnicity data were collected. Practices may vary by hospital and maternal race and ethnicity may have been misclassified. Lastly, although the Premier Healthcare Database included a large population across U.S. census regions, data are not nationally representative.

CONCLUSION
A documented COVID-19 diagnosis, identified through ICD-10-CM codes in delivery hospitalizations, was associated with multiple concurrently documented adverse pregnancy outcomes, maternal complications, and indicators of severe illness. However, the absolute risks were low. Risk for any maternal complication or for severe illness did not differ by race/ethnicity in this study. These findings emphasize the importance of implementing