The risk of clinical complications and death among pregnant women with COVID-19 in the Cerner COVID-19 cohort: a retrospective analysis

The COVID-19 pandemic continues to wreak havoc in the United States and around the world. As of February,16, 2021, the cumulative case-counts in the US had surpassed twenty-seven million [1]. As researchers and experts alike look towards mitigation strategies, there is a critical need to investigate how the pandemic impacts specific subgroups. One critical population of interest is pregnant women. Pregnancy imposes physiological and immunological changes that have important implications on the severity and outcomes of viral conditions [2, 3]. Although data for COVID-19 pregnant women are still emerging, prior epidemiologic evidence suggests that many viral infections may result in a higher risk of fatalities and clinical complications among pregnant women [4, 5]. During the 2004 SARS pandemic, higher rates of complications and death were reported among pregnant women compared to non-pregnant women [6, 7]. Similar adverse outcomes were observed for pregnant women during the 2009 H1N1 influenza pandemic [8]. Data on MERS, albeit limited, also points to increased morbidity during pregnancy [9]. Accordingly, the Centers for Disease Control and Prevention (CDC) have emphasized the need for pregnant women to take rigorous precautionary measures against COVID-19 [10].

While some researchers suggest a reason for an increased risk of adverse outcomes for viral infections is the presence of a pregnancy-induced immune-suppressed state, others propose that the physiologic and immune responses during pregnancy in some instances may be protective rather than suppressive [11, 12]. Thus far, most studies on COVID-19 infections among pregnant women have been limited by small sample sizes, or have been focused on specific pregnancy outcomes such as miscarriage or stillbirths [13‐15]. Larger national studies of COVID-19 in pregnancy have been reported, but with limitations of substantial missing data on pre-existing chronic conditions and demographic variables [16]. Our study aims to fill this gap by assessing the impact of pregnancy on adverse outcomes of COVID-19 using national data from the Cerner COVID-19 De-Identified Data cohort. We aim to evaluate the adjusted risk of the following COVID-19 complications in relation to pregnancy: hospitalization, the maximum length of hospital stays, moderate ventilation, invasive ventilation, and death. As a secondary aim, we assessed the impact across a range of pre-existing co-morbidity.

Overall, there were 22,493 total female patients, 18–44 years old, from the COVID-19 “associated” cohort. The median (Q1-Q3) age was 31 [25‐38] and the majority of patients were Hispanic/Latino (41.0% (9232)) with 26.1% (5872) being NH White and 17.6% (3966) being NH Black. The majority of patients used private insurance (48.9% (11,010)) and 23.5% (5285) used Medicaid. Most patients came from the southeast United States (45.1% (10,134)). Pregnant patients (compared to non-pregnant patients) were younger (median age: 29 vs. 32), had a lower distribution of Hispanic/Latino ethnicity (34.8% vs. 41.5%), had a higher distribution of Medicaid use (36.0% vs. 22.5%) and lower distribution of self-pay (7.5% vs. 14.2%), and lower distribution from the southeast United States (39.3% vs. 45.5%). All P < 0.001 (Table 1). When looking at history of chronic diseases, non-pregnant women generally had higher percentages of all disease types. However, pregnant patients had a significantly higher percentage reporting obesity than non-pregnant patients (20.6% (331) vs. 17.1% (3567, P < 0.001). Pregnant patients had lower use reporting of all COVID-19 related medications than non-pregnant patients, except for anticoagulants (13.6% vs. 8.2%, P < 0.001). Pregnant patients also reported higher rates of gestational DM history than non-pregnant women (10.1% vs. 2.0%, P < 0.001) (Table 2).

Table 3 shows the comparisons in complications between pregnant and non-pregnant patients. Pregnant women, compared to non-pregnant women, had higher rates of hospitalization (60.5% vs. 17.0%, P < 0.001), higher maximum LOS among those who stayed less than one day in the hospital(0.15 days vs. 0.08 days, P < 0.001), and higher moderate ventilation use (1.7% vs. 0.7%, P < 0.001). Among those who stayed one day or greater in the hospital, pregnant women showed significantly lower max LOS than non-pregnant women (2.55 days vs. 3.32 days, P < 0.001). No significant differences were found in invasive ventilation or death.

The adjusted associations with the 5 clinical complications are reported in Table 4. Pregnant women, compared to non-pregnant women, saw higher odds in hospitalization (aOR: 12.26; 95% CI (10.69, 14.06)), moderate ventilation (aOR: 2.35; 95% CI (1.48, 3.74)), and higher max LOS among those staying less than one day in the hosptial (\( {e}^{\hat{\beta}} \): 1.78; 95% CI (1.67, 1.90)). Among those staying one day or more in the hospital, pregnant women had significantly lower max LOS than non-pregnant women (\( {e}^{\hat{\beta}} \): 0.90; 95% CI (0.85, 0.95)). No significant associations were found with invasive ventilation or death (Table 4). Figures 1A-F report the predicted outcomes by comorbidity scores (ECI) and grouped by pregnant and non-pregnant status. In all cases, as ECI increases so too does the predicted clinical complication. However, pregnant patients (compared to non-pregnant patients) have higher average predicted complications for hospitalization, max LOS (< 1 Day), and moderate ventilation.

The results of the stratified analyses reveal generally similar findings between all demographic/clinical groups for hospitalization, maximum LOS, invasive ventilation, and death (Table 5). Increased outcomes for pregnant patients (compared to non-pregnant) were seen across all groups for hospitalization and maximum LOS (< 1 Day). Among those staying a day or greater in the hospital, pregnant women showed lower max LOS than non-pregnant women. This was significant among patients 18–34 years old, whites, AIAN, Medicaid holders, and patients without a diabetes history. No significant associations were seen among invasive ventilation and death. For moderate ventilation, differences were seen among age groups and race/ethnicity groups. Among patients 35–44, pregnant women saw significantly higher odds of moderate ventilation compared to non pregnant women (aOR: 3.74; 95% CI (1.65, 8.52)). No such difference was detected among patients 18–34. NH Black and NH other race patients, that were pregnant, saw significantly higher odds of moderate ventilation than those that were not pregnant (aOR: 4.51; 95% CI (1.96, 10.39), and aOR: 6.40; 95% CI (2.68, 15.28) respectively). Again no such difference was seen among white and Hispanic/Latino patients. Additionally, when restricted to patients from the COVID-19 “positive” cohort (n = 8806; Pregnant n = 748, Non-pregnant n = 8058), the results are similar to that of the COVID-19 “associated” cohort (see Supplemental Tables A-C and Supplemental Figs. 1A-E).

The results of the sub-analysis (Table 6), looking at outcomes in postpartum women from the COVID-19 associated cohort, show higher complications for these women compared to pregnant women from the same cohort. When compared to pregnant women, postpartum women had higher percentages of hospitalization (90.5% vs. 60.5%, P < 0.001), higher max LOS less than one day (0.20 days vs. 0.15 days, P = 0.09), and higher invasive ventilation (5.3% vs. 1.6%, P = 0.003).

Our study investigated the extent to which pregnancy imposes the risk of severe COVID-19 complications among COVID-19 associated patients and COVID-19 positive patients, and examined the association across the range of prior comorbidity. We focused on five principal outcomes: maximum length of hospital stays, moderate ventilation, invasive ventilation, and death. In the COVID-19 associated group, we found significant differences in COVID-19 outcomes among COVID-19 associated pregnant women, and non-pregnant COVID-19 associated women. Notably, our findings showed that COVID-19 associated pregnant women were more than 12 times as likely to be hospitalized, more than twice as likely to require moderate ventilation and had a maximum LOS of less than 1 Day two times greater than that of COVID-19 associated non-pregnant. Perhaps most importantly, we did not observe an increased risk for invasive ventilation or increased risk of death among pregnant COVID-19 associated women. We observed similar results after stratifying by age and other demographic or clinical characteristics. We further restricted our analysis to confirmed COVID-19 positive patients and found comparable results. In the COVID-19 confirmed group, pregnant COVID-19 pregnant women were more than ten times as likely to be hospitalized, nearly three times as likely to require moderate ventilation and had a maximum LOS of less than 1 Day almost two times greater that of COVID-19 confirmed non-pregnant women.

Our findings parallel results from CDC’s initial report on SARS-CoV − 2 infection among pregnant women [16] yet contradicts certain findings from CDC’s updated report [25] on COVID-19 and pregnancy. For instance, similar to our results, CDC’s previous report showed that although pregnancy was associated with a heightened risk of hospitalization, intensive care admission, and mechanical ventilator receipt, pregnancy was not linked to increased risk of death among reproductive aged women with confirmed COVID-19 illness [16]. However, CDC’s newest study on COVID-19 and pregnancy reported that pregnant COVID-19 diagnosed women were nearly twice as likely to die compared to non-pregnant COVID-19 diagnosed women [25]. Similarly, researchers observed a case fatality rate 13.6 times higher among pregnant COVID-19 patients in Washington State [26]. In comparison to the above mentioned studies, our research is strengthened by a more complete assessment of demographic variables, clinical characteristics, and prior comorbidities. However, the overall low rate of deaths (only 4 deaths) among pregnant women in our sample may have hindered the ability to capture accurate mortality.

The current recommended standards of care for pregnant COVID-19 patients might explain our observed results. For instance, the CDC, the American College of Obstetricians and Gynecologists (ACOG), and the Society for Maternal-Fetal Medicine (SMFM) urge special considerations for the management of COVID-19 during pregnancy [27‐29]. Examples of these guidelines include hospitalization of pregnant women diagnosed with COVID-19 in facilities with maternal and fetal monitoring capabilities when warranted and the use of multispecialty team based approach during treatment [30]. We argue that because of these recommendations and the limited scientific knowledge on COVID-19 and pregnancy, providers are likely taking precautionary measures in caring for their pregnant COVID-19 patients. Therefore, the fact that pregnancy was linked to increased hospitalization, higher LOS of less than 1 Day, moderate ventilation but not invasive ventilation or death lends additional support to the notion that the observed results may in part reflect provider behavior or current clinical recommendations. The additional attention provided for pregnant women may reduce their risk of severe morbidity or death, but this cannot be definitively identified in these observational data. Further studies are needed to examine the contribution of provider behavior on COVID-19 health outcomes for pregnant women. Even so, the effects of specific physiologic changes during pregnancy such as decreased lung capacity, reduced cell-mediated immunity, and increased heart rate on the outcome of viral infections such as SARS-COV-2 infections cannot be overlooked [31, 32].

Other clinical factors associated with adverse COVID-19 outcomes were increasing comorbidity index score and being on Medicare or Medicaid. Several recent COVID-19 studies have reported a link between increasing comorbidities with increasing COVID-19 complications [33‐35]. Similarly, being on Medicare or Medicaid has also been associated with poor COVID-19 outcomes [36]. In particular, Medicaid can be used as a proxy for lower socioeconomic status as Medicaid eligibility is based on federal poverty level guidelines [37]. Medicare beneficiaries who are under the age of 65 such as those present in our study may have disabilities or severe medical conditions such as end-stage renal disease [38]. Taken together, lower socioeconomic status and underlying medical conditions have critical implications on worsening COVID-19 outcomes [33, 39‐41].

Concerning racial and ethnic differences, we found that the risk of hospitalization and higher LOS of < 1 Day among pregnant patients, compared to those who are not pregnant, to be significant and comparable across race/ethnicity groups (aOR ranges from 11.77–15.14 and \( {e}^{\hat{\beta}} \) ranges from 1.55–2.22 across races) (Table 5). However; NH Black and NH other race pregnant patients had significantly higher odds of moderate ventilation than those that were not pregnant; a finding that was not statistically significant among white and Hispanic/Latino patients. Current literature has uncovered the disproportionate overall burden of COVID-19 on minority populations, including Blacks and Native Americans [42, 43]. In a Brazillian study, researchers reported a nearly two-fold increase in maternal mortality due to COVID-19 for Black women [44]. Several reasons, including higher prevalence of underlying conditions, disproportionate representation in essential worker occupations, housing, and living conditions, and structural racism, have been identified as reasons for poor outcomes observed among minority populations [44‐47]. When focusing on pregnant women only, racial disparities in pregnancy complications persist among women of color when compared to whites with some exceptions for hispanic women which is consistent with our observation. For example, an investigation of racial disparity in pregnancy outcomes at a tertiary care medical center found that while black women were more likely, compared to whites, to remain in the hospital for > 4 days, have higher rates of preterm birth, small-for-gestational age infants, preeclampsia, and stillbirths, Hispanic women were found to have lower odds for preterm birth and when compared to black women, Hispanic women were less likely to experience any adverse pregnancy events, with the exception of gestational diabetes mellitus [48]. Here, we refer to the Hispanic paradox; the positive health outcomes observed among Hispanic populations despite risks associated with lower SES [49] as a likely explanation while acknowledging that other COVID-19 studies have reported adverse effects in Hispanic populations [50] in general but not among pregnant women in particular.

Finally, our stratified sensitivity analysis indicated a nearly fourfold increased risk for moderate ventilation among COVID-19 associated pregnant women aged 35 to 44 years. However, the risk of moderate ventilation among pregnant women aged 18 to 24 years did not yield significant results. In comparison, CDC’s newest study reported a heightened risk of invasive ventilation among COVID-19 pregnant women of all ages, with the highest risk observed among older pregnant women aged 35–44 years old [25]. Given that advanced maternal age has been associated with a host of adverse pregnancy outcomes [51] our findings highlight the need for providers to pay close attention to pregnant women of advanced ages who contract COVID-19.

Our analysis has several limitations. First, our data were limited to only participating health systems, and may not fully represent the general population. Second, our study only included individuals who sought medical care for COVID-19 and may under-represent medically underserved minority populations such as Hispanics/Latinos and NH Blacks who may not seek treatment due to lack of health insurance or other barriers. Thirdly, data on the indication for the intensive care unit (ICU) admission were not available and thus hospitalization may have occurred in other areas outside the ICU. Further, all cases and many of our outcomes and predictors were identified using ICD and LOINC EHR codes, thereby limiting the study’s specificity. Moreover, data from patient EHR are subject to ICD coding errors, may not include all relevant patient diagnoses, and may only capture primary patient complaints, resulting in lower estimates of vital patient health history components. Of note, our low estimates of maternal deaths in comparison to other studies might be partly attributable to possible data entry discrepancies between participating health systems and inconsistencies in data processing and restructuring. Lastly, our primary analysis did not include postpartum women, and given that the postpartum period may constitute a period of greater risks of maternal death [52], their exclusion in our study might underestimate maternal deaths in comparison to studies that include the postpartum period [26, 44]. To address this issue, we have added a subanalysis that compares complications in COVID-19 between postpartum women and pregnant women. Despite these limitations, our study has several strengths. We utilized a large sample size (8000–22,000), sufficient to capture important trends, and conduct informative stratified analyses for a wide variety of characteristics. Our large sample size also allowed for analyses of rare patient populations (i.e. pregnant women with COVID-19) that may be incredibly difficult to identify using alternative means. We also utilized a national database that allowed for an analysis of a diverse and more representative patient population. Finally, to reduce the effects of ICD coding limitations, we applied other methods such as text matching to capture every possible indication in the data.

Our study indicates that pregnancy confers additional risk with COVID-19 disease for medical complications, but may not ultimately increase mortality in the setting of sufficient medical care. Our results support current recommendations for particular attention and care to pregnant women who incur SARS-COV2 infection and highlight the importance of access to care for these women. Additionally, our findings provide further justification for the CDC and ACOG’s current recommendations [53, 54] regarding the vaccination of pregnant women who are part of recommended priority groups.