Mechanical Ventilation and Death During Pregnancy Complicated by COVID-19: A Prognostic Analysis from the Brazilian COVID-19 Registry Score

Assessing predictors of critical outcomes in COVID-19 may advise timely treatments and better prepare facilities to overcome extra adversities during pregnancy. However, many clinical parameters of existent scores are deeply modied by physiologic adaptations. Our aim was to assess the feasibility of a prognosis score developed for general hospitalized adults with COVID-19 in Brazil to predict clinical adverse outcomes in pregnant women upon hospital admission. retrospective substudy of a cohort analysis analysis the of on assess The and the or and were outcomes and Covid-19. The overall discrimination of the model was presented as the area under the receiver operating characteristic curve (AUROC).


Conclusions
The model ABC 2 -SPH developed in Brazilian general patients was not able to identify adverse clinical outcomes in pregnant women with COVID-19. We warn against the use of general inpatients COVID-19 prognosis in pregnant women. A more useful model for clinical prognosis is necessary concerning the speci cities of pregnancy affected by  Background Page 5/22 Coronavirus disease  has quickly spread worldwide with higher morbidity and lethality than other coronaviruses (1), threatening people's lives, mainly those more vulnerable or under adverse social contexts (2,3). Recent data raised concern about the impact of COVID-19 on pregnancy, since the pandemic severely hit more vulnerable countries with big birth rates (4)(5)(6). The impact of SARS-CoV-2 infection on pregnancy became more evident in controlled studies, revealing consistent increase in severe maternal morbidity and mortality and neonatal complications when comparing pregnant women with and without COVID-19 diagnosis (5). Pregnant women with comorbidities, such as diabetes, hypertensive diseases, heart disease and lung diseases deserve special attention, as they seem to be susceptible to the severe and critical forms of COVID-19, with higher risk of adverse outcomes (7).
Even though the majority of pregnant women are healthy and younger than most COVID-19 patients (8), during pregnancy there are signi cant anatomical and physiological changes that affect every organ system in the body (9). It is believed that such changes may interfere with the progression of COVID-19 (9,10).
The assessment of clinical characteristics and outcomes in pregnant women who are hospitalized with COVID-19, as well as the factors potentially associated with adverse maternal outcomes in those patients, is of utmost importance for public health. It may help health managers and stakeholders to better prepare facilities to overcome extra-adversities during the pregnancy-puerperal period (8). However, rapid scoring systems for prognosis applicable during pregnancy are challenging. There are speci cities in clinical parameters in pregnant women, so scores developed for non-pregnant cannot be applied in pregnant women without previous assessment. In this context, there is a lack of studies of risk or prognosis score for COVID-19 in pregnant women. Therefore, the primary aim of this pilot study is to assess the ability of a COVID-19 prognosis score, developed for general hospitalized adults with COVID-19 in Brazil, to predict mechanical ventilation and death in pregnant women upon hospital admission. Additionally, to assess the occurrence of pregnancy adverse outcomes, as well as severe and critical COVID-19.

Methods
This is a multicenter retrospective substudy of the Brazilian COVID-19 Registry, a multicenter cohort study of consecutive patients with laboratory-con rmed COVID-19 hospitalized between March and September 2020, in 37 Brazilian public and private hospitals, as previously described (11,12). It adheres to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline (13).
Study data were collected and managed by trained health professionals using Research Electronic Data Capture (REDCap) hosted at the Telehealth Center of the University Hospital, Universidade Federal de Minas Gerais (14,15). Over three hundred variables were collected from medical records, involving clinical, laboratory and imaging characteristics at admission, as well as in-hospital outcomes related to COVID-19. Obstetric data were gestational age, pregnancy complications at admission, whether there was delivery and, if so, mode of delivery, birth weight and vital state of the newborn. The study protocol and a coding manual guiding data collection with details and the de nition of each variable was agreed with the network of researchers (11). This Registry study previously established a prognostic scoring model for in-hospital mortality for COVID-19 patients, based on comorbidities, clinical characteristics, laboratory and imaging ndings at hospital presentation, the ABC 2 -SPH score (12). The score has shown high discriminatory value (AUROC 0.844, 95% CI 0.829 to 0.859), which was con rmed in the Brazilian (0.859 [95% CI 0.833 to 0.885]) and Spanish (0.894 [95% CI 0.870 to 0.919]) validation cohorts, and displayed better discrimination ability than other existing scores (12).
For the score development and validation, patients who developed the rst COVID-19 symptoms during admission due to other conditions and those who were admitted in another hospital rst (not part of the cohort) were excluded, as the score was intended to be applicable upon hospital presentation. For the present analysis, the external validation group gathered data of pregnant women who were admitted from March 1, 2020, to May 5, 2021. We kept the rst exclusion criteria, but opted to maintain women who were transferred between hospitals, as in Brazil only certain hospitals were selected to treat pregnant women with COVID-19 in the public health system, so if a pregnant women seek care in a non-reference center, she was transferred to a reference center and it would not be adequate to exclude those patients. After exclusion criteria, 85 pregnant women were identi ed in 19 of 37 multicentric network centers (Figure 1), in 12 different cities from 5 Brazilian states. Eight of them were public, 13 were teaching hospitals and 12 were reference centers for COVID-19 treatment, gathering average 322,8 beds (ranging from 60 to 784 beds).

Figure 1 Flowchart of COVID-19 pregnant women included in the study
Clinical characteristics, laboratory data and obstetric characteristics at admission, as well as events that occurred during hospital stay were collected for the present analysis.

Outcomes
The primary outcomes were death, and the composite outcome of mechanical ventilation or death. The secondary outcomes included pregnancy outcomes and the occurrence of severe and critical COVID-19, according to World Health Organization criteria (16). Pregnancy outcomes included preterm birth, c-section, preeclampsia, maternal death, and immediate neonatal vital state. Occurrence of severe and critical COVID- Severe pneumonia demanding ventilatory support, peripheral oxygen saturation (SpO 2 ) <90% on room air, respiratory frequency >30ipm had classi cation as severe COVID-19.
Statistical analysis ABC 2 -SPH development and validation methods followed guidance from the Transparent Reporting of a

Multivariable Prediction Model for Individual Prediction or Diagnosis (TRIPOD) checklist and Prediction
Model Risk of Bias Assessment Tool (PROBAST), and are described elsewhere (12,17,18). In brief, generalized additive models (GAM) were used to examine the relationships between in-hospital mortality and potential predictors, selected based on clinical reasoning and literature review. Least absolute shrinkage and selection operator (LASSO) logistic regression was used to derive the mortality score, which was external validated (12).
Descriptive analysis of the pregnant women's clinical characteristics strati ed by COVID-19 severity and the cohort of (non-pregnant) patients was performed, concerning the frequency, variability, and central tendency measures. Continuous variables were summarized using medians and interquartile ranges (IQR), whereas counts and percentages were used for categorical variables. For comparisons, the Chi-squared test or Fisher test was used for the independence hypothesis, and the Mann-Whitney test compared the numerical variables between the groups. A p-value less than 0.05 was considered statistically signi cant. Calibration of the model applied to pregnant women was assessed graphically by plotting the predicted outcome of interest (death, composite outcome, or severe/critical disease) probabilities against the observed outcome, testing intercept equals zero and slope equals one. The area under the receiver operating characteristic curve (AUROC) described model's discrimination. Con dence intervals (95% CI) for AUROC were obtained through 2000 bootstrap samples.
Statistical analysis was performed with R software (version 4.0.2) with the tidyverse, pROC, rms packages.

Results
Clinical characteristics and laboratory ndings of the 85 pregnant women upon hospital presentation are shown in Table 1. Their comparison to the ones who were excluded is shown in Table S3.  With regards to patient outcomes (Table 2), in-hospital mortality was 20.4% and 3.5% for the modelderivation cohort compared to the pregnant women (p<0.001). Hospital and intensive care unit length of stay were also longer in the general group of patients than pregnant women (p=0.010 and p=0.048, respectively).  Our interpretation is that the set of prognosis markers of COVID-19 in pregnancy are not the same as the ones for the non-pregnant population admitted with the disease. This nding itself could contribute to an understanding of the poor outcomes as COVID-19 maternal mortality. According to Brazilian data, pregnancy complicated by COVID-19 is a serious burden for the hospital maternity services. Pregnant women accounted for 0.8/1000 of 20,350,142 con rmed cases in the country until Aug/2021 (19,20). However, the rate of mortality was 10.5%, 3.8 times higher than the rate of 2.8% of the national mortality (19).
Across the studies in non-pregnant populations, male sex, increasing age and underlying illness, such as cardiovascular diseases and diabetes, increased the risk of poor outcomes (21,22). The ABC 2 -SPH predictive model was developed upon a set of covariates upon hospital presentation to predict death: age, chronic diseases (hypertension, diabetes mellitus, obesity, coronary artery disease, heart failure, atrial brillation or utter, cirrhosis, cancer, chronic obstructive pulmonary disease and previous stroke), heart rate and SpO 2 /FiO 2 ratio, allied to low platelets, C-reactive protein and urea (12). Such predictors might score differently in pregnant women, since they are deeply modi ed by physiologic adaptations, such as the increase in heart rate, which itself may overestimate the risk estimated by the score in at least 5%, and maternal response to infections. Our analysis reveals how distinct these groups are in terms of age, laboratory analysis, and in-hospital complications. Previous diabetes, blood pressure, respiratory rate, heart rate, SpO 2 /FiO 2 ratio, neutrophils-to-lymphocytes ratio, and blood urea nitrogen at admission as associated with severe COVID-19. Preexisting comorbidities as diabetes and chronic hypertension have been shown to be associated with an increased risk for COVID-19 adverse outcomes in pregnant women (23,24).
Most pregnant women were young and healthy before the admission due to COVID-19, which partially explains why abnormal vital signals and in ammatory markers are associated with the in-hospital severe/critical progression, instead of pre-existent comorbidities. Neutrophil-to-lymphocyte ratio and Creactive protein were higher in the severe and critical COVID-19 group when compared to mild one. In fact, this ratio has been observed to be the most consistent abnormal hemocytometric nding in COVID-19 patients (25). In the multivariate modeling of ABC 2 -SPH score (12), C-reactive protein is the in ammatory marker which was signi cant in the nal model. We hypothesize that in ammatory markers could be covariables even more relevant for pregnant women, scoring proportionally higher than for general patients.
Besides, physiological adaptations to the pregnancy affect the organ system in the maternal body, modifying, as well the response to infections.
The existing evidence is con icting on whether pregnancy is an immunological contributor to severe progression of COVID-19 (26). A successful pregnancy depends on a responsive immune system, which explains reports of universal COVID-19 testing during pregnancy, that the vast majority is asymptomatic or has mild COVID- 19 (26, 27). The unit maternal and feto-placental immune system is responsive, protecting both the mother and the fetus against treats from the environment (28). The placenta is a selective barrier, able to protect the developing fetus against infections, including SARS-CoV-2 virus infection (29). It also acts as an immunity-modulating organ, regulating immune responses of cells present both at the implantation site and systemically (30). However, evidence of fetal vascular malperfusion or thrombosis has been observed in COVID-19, which may be related to an exacerbated maternal systemic in ammatory response and hypercoagulable state (31,32).
Notwithstanding, cardiopulmonary adaptive changes during pregnancy may increase the risk of hypoxemia and contribute to the increased severity of viral infections (33). The circulatory system is signi cantly adjusted during pregnancy, starting early in its course, driven by peripheral vasodilatation, increased heart rate and stroke volume, reduced pulmonary vascular resistance, and reduced pulmonary residual capacity.
These changes may affect the course of viral infections (9,33). For these reasons, although we believe that vital signals at admission might contribute to scoring in predictive models of COVID-19 outcomes during pregnancy, the expected cut offs are affected by physiological changes during pregnancy and might not coincide with non-pregnant women. Besides, with hemodilution and rising glomerular ltration rate, there are modi cations in the reference values for hemoglobin levels, proteins, creatinine, and urea (9), interfering in the performance of scores based on laboratory values. Therefore, it is comprehensible that scores used to predict mortality in general adults have limitations when used among pregnant women (34).
Another aspect of COVID-19 disease in pregnant women grounds on the overactivated renin-angiotensin system. This system plays a relevant role in maternal hemodynamic adaptations and in placentation and hypertensive disturbs during pregnancy (35). SARS-CoV-2 uses the protein angiotensin-converting enzyme the receptors 2 (ACE2) to invade cells, with potential implications for increased susceptibility to the virus during pregnancy (36). Obstetric outcomes were not the target in the current approach, even though they could be affected by COVID-19 (3,5,37).
Yet, this analysis has expected limitations that may affect the interpretation. Our sample size of pregnant women is limited. Maternal mortality was lower than Brazilian national rates (20), and the frequency of chronic hypertension was low. A speci c predictive model for COVID-19 prognosis for inpatient pregnant women was not proposed. The results also do not apply to antenatal care since the inclusion criteria was pregnant women admitted with COVID-19. As a retrospective analysis, the quality of data as incompletude might have occurred. Thus, we reduced the risk of inaccuracies by performing several quality checks and rechecking hospital medical records whenever necessary.
Based on our results, we warn against the use of non-pregnant COVID-19 prognosis scores in pregnant women to predict adverse outcomes without proper validation. While insu cient control of pandemic keeps worldwide, fast and e cient assessment of prognosis of the COVID-19 is of utmost importance for early identi cation of cases at higher risk of worse outcome in this highly vulnerable group of women. Although several studies developed and validated risk scores to estimate prognosis in COVID-19 patients, there is a lack of scores focused on pregnant women speci cities. Studies using data pools across national systems or healthcare data sharing frameworks are necessary to rapidly join and use clinical information relevant to

Declarations
Ethics approval and consent to participate The study protocol was approved by the Brazilian National Commission for Research Ethics (CAAE 30350820.5.1001.0008). Individual informed consent was waived due to the severity of the situation and the use of unidenti ed data, based on medical chart review only. Additionally, administrative permissions to access and use the medical records were obtained from each institution.

Consent for publication
Not applicable.

Availability of data and materials
All data generated or analyzed during this study are included in this published article and its supplementary information les.

Competing interests
The authors declare that they have no competing interests.