The main contribution of the present analysis was to validate the ABC2-SPH score in 350 pregnant women from 24 Brazilian hospitals. The ABC2-SPH score has been shown to be a reliable tool in estimating in-hospital mortality risk in pregnant COVID-19 patients. In this population, the score had good overall performance for the primary outcomes and good discriminatory ability. Calibration was good for the prediction of in-hospital mortality, but it was poor for the composite outcome of in-hospital mortality and mechanical ventilation support. The score is simple, objective, uses variables easily available at hospital presentation and it may be easily calculated. Model performance comparison surpassed other existing scores, commonly used in the general population.
In fact, assessing predictors of critical outcomes in COVID-19 may advise timely treatments and better prepare facilities to overcome extra adversities during pregnancy. Our findings support the employment of the score as a tool in estimating in-hospital mortality at admission in pregnant patients. Therefore, it is of utmost importance to take into account that the score should be used in addition to the clinical judgment, to support clinical decision, for example, to help screening pregnant women who need more frequent reassessments, as well as to help to assess which one to refer to intensive care, in cases of limited resources. As a screening tool, it is of utmost importance to have a high sensitivity, to avoid missing as few cases as possible. In the present analysis, ABC2-SPH achieved 96.0% sensitivity, with a very precise confidence interval (91.8–98.4%), higher than any other score tested.
On the contrary, our findings evidence against the use of the score to predict the composite outcome of mechanical ventilation support and in-hospital mortality. In a recent analysis from our group (data not published yet), ABC2-SPH score did not have good overall performance to predict mechanical ventilation support in general (non-pregnant) patients. Therefore, the present results may reflect the fact that the score is not a good predictor for mechanical ventilation overall.
Several prediction scores have been proposed for use in the nonpregnant population with COVID-19 with varied success. The study conducted by Jones et al. (2021)  validated the 4C score for Canadian patients obtaining an AUC of 0.770 (95% CI 0.790–0.870). In addition, the accuracy of this prediction model (4C Mortality), beyond NEWS and CURB-65 was compared among the Romanian population with AUC of 0.818 (95% CI 0.718–0.919), 0.861 (95% CI 0784–0.939), and 0.801 (95% CI 0.681–0.922), respectively [39, 40]. In the present study, we tested these aforementioned scores, together with other scores commonly used for general COVID-19 patients, and ABC2-SPH outperformed all of them.
As aforementioned, many clinical parameters of existent scores developed for the general (non-pregnant) population are deeply modified by physiological adaptations of pregnancy. Notably, these adaptations are challenging for using the scores developed for the general population without further validation and can contribute to an understanding of the lack of prediction models for the prognosis of COVID-19 in this population, despite the fact that several prognostic scores have been developed for COVID-19 [41,42,43,44]. One multicenter retrospective cohort study including eight hospitals from four countries (n = 973) proposed two models to predict ICU admission and maternal death in pregnant women with symptomatic COVID-19 , however, both models are limited by methodological bias, with the absence of external (even geographic) validation.
Our study observed high in-hospital mortality in pregnant women (6%). A study based on secondary data from Brazil (975,109 cases) suggested that pregnant women with COVID-19 have approximately twice the mortality rates of men and non-pregnant women . Takemoto et al. (2020)  reported high mortality among Brazilian pregnant women with COVID-19, approximately 20 maternal deaths out of 125,218 overall cases and 8536 deaths (as of May 7, 2020), with lethality of 15.6% in 2021 . Similarly, another study found an association between COVID-19 and worse clinical outcomes for pregnant women in Brazil, with a 3.4 times higher death rate than any other acute respiratory distress syndromes (ARDS) etiologies . According to a systematic review with 2670 patients from seven countries, the differences in results for maternal characteristics reflect the profile of the patient of each country of origin . This study (n = 38 studies, 2670 patients, 52.6% from China), have shown a significant variation between maternal age among pregnant women with COVID-19, percentages of C-sections, maternal mortality rate and newborn outcomes . In fact, is it possible that pregnant Brazilian patients have different characteristics from those from other countries, placing them in the leadership of maternal deaths due to COVID-19 worldwide . One of them is the prevalence of underlying diseases, especially preeclampsia and obesity, conditions that are known inflammatory, risk factors for COVID-19 complications. Besides, an important contributor to greater mortality in the country were the barriers to access to prenatal care during the pandemic, inadequate monitoring of obstetric complications and barriers to access intensive care [4, 17, 47,48,49].
Despite having similar symptoms, our analysis showed differences between pregnant women who died and those who were discharged. Those who died had higher age, prevalence of hypertension, obesity and, as expected, in-hospital complications than the ones who were discharged alive. These findings are consistent with a large study from the Centers for Disease Control and Prevention (CDC), comparing 386,028 positive nonpregnant women in their reproductive age (15–44 years), with 23,434 SARS-CoV-2 positive pregnant women, demonstrating that death occurred more frequently among women aged 35–44 years than among those aged 15–24 years. When stratified by age, all outcomes, such as hospitalization, ICU admission, receipt of mechanical ventilation, and death were more frequently in pregnant women aged 35–44 years than among those aged 15–24 years . Additionally, a brief communication conducted by Takemoto et al. (2020)  collected the effect of 978 pregnant women with COVID-19 in Brazil, indicating that women who died had higher maternal age (31.5 years). A living systematic review has shown that increasing age (odds ratio 1.83, 95% confidence interval 1.27 to 2.63; seven studies, 3561 women), high body mass index (2.37, 1.83 to 3.07; five studies, 3367 women), any pre-existing maternal comorbidity (1.81, 1.49 to 2.20; 3 studies; 2634 women), chronic hypertension (2.0, 1.14 to 3.48; two studies, 858 women), pre-eclampsia (4.21, 1.27 to 14.0; 4 studies; 274 women), and pre-existing diabetes (2.12, 1.62 to 2.78; 3 studies, 3333 women) are maternal risk factors associated with severe COVID-19 . Non-white ethnicity (1.61, 1.05 to 2.47; 3 studies, 31,469 women; 2.23, 1.25 to 3.97; 1 study, 669 women; respectively) and high body mass index (2.27, 1.20 to 4.31; 3 studies, 31,085 women; 6.61, 1.98 to 22.02; 2 studies, 485 women; respectively; Table 2) were associated with maternal death and the need for invasive ventilation .
In the present analysis, the most common laboratory findings among patients who died from COVID-19 were lower median SpO2/FiO2 ratio (333.0 vs 457.1, p = 0.002), higher median neutrophils-to-lymphocytes ratio (8.1 vs 5.3, p = 0.034), and lower partial pressure of oxygen (69.0 vs 86.0, p = 0.004). During pregnancy, vital signals had proper values, including a slight drop in SpO2 . It is important to mention that in this period the circulatory system undergoes physiological changes, 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 [52, 53]. Regarding inflammatory markers, the existing evidence is conflicting on whether the pregnancy is an immunological contributor to the severe progression of COVID-19 . 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 [54, 55]. The unit maternal and the fetoplacental immune system is responsive, protecting both the mother and the fetus against threats from the environment .
Nevertheless, we observed that childbirth had an impact of COVID-19. C-section was performed in 71.0% of childbirths, with lower birth weight in babies of pregnant women who died (low birth weight, 2000 vs 2813, p = 0.001), and 13 babies died. An aforementioned systematic review  analyzed cesarean delivery rates geographically and found rates to be considerably higher in China (83.9%), followed by the United Kingdom (71.9%), with Spain with the lowest rate of C-sections (35.9%). The reasons for these practices are unclear, but it may be attributable to the habitual medical practice of each country, in addition to the lack of guidelines and recommendations at the beginning of the pandemic. Regarding low birth weight, the placenta is a selective barrier able to protect the developing fetus against infections, including SARS-CoV-2 virus infection , and it acts as an immunity-modulating organ, regulating immune responses of cells present both at the implantation site and systemically . However, evidence of fetal vascular malperfusion or thrombosis has been observed in COVID-19, which may be related to an exacerbated maternal systemic inflammatory response and hypercoagulable state [59, 60]. In the meantime, our study observed lower median Apgar score in newborns of pregnant women who died from COVID-19, compared to those who were discharged alive (3.0 vs 8.0, p < 0.001). In cases of fetal distress, prematurity, and severe/critical maternal disease the Apgar scores are lower [61, 62].
Based on our results, we warn against the use of non-pregnant COVID-19 prognosis scores in pregnant patients to predict adverse outcomes without proper validation. While the control of COVID-19 pandemic is still challenging in many places, fast and efficient assessment of the prognosis of the COVID-19 is of utmost importance. We can expect the downstream effects of COVID-19 to be apparent for a number of years. Further studies with large sample sizes are required for the development and validation of a more accurate model to predict other adverse outcomes, such as mechanical ventilation and pregnancy outcomes, concerning the specificities of pregnant patients affected by COVID-19. Scores for pregnant women would be useful for early identification of cases at higher risk of worse outcomes in this highly vulnerable group of women. Further studies are also necessary to identify risks in pregnancy-related critical illness  due to COVID-19 or other causes. Evidence-based modeling could provide a proper prognosis score assessment tool that will help guide decision-making, develop patient care plans, and better allocate resources.
Even with its multiple strengths, the present study has some limitations. Recalibration of the ABC2-SPH score may improve its prediction of the effects of COVID-19 on pregnant women. However, our sample size is not large enough for this analysis (at least 100 events for recalibration) . This is a topic for future studies with larger sample sizes. Additionally, details about diagnosis of hypertensive syndromes of pregnancy, subtypes of diabetes during pregnancy may differ among the perinatal centers involved in this cohort. Besides, the study missed details to specify gestational complications. The newborn data was used to infer possible complications related to pregnancy and delivery, the analysis of newborn outcomes secondary to COVID-19 requires a different study design and was not the purpose of the present analysis. In fact, this is a topic for an ongoing study from our group. Additionally, we have not investigated the impact of each individual SARS-CoV-2 variant on pregnant women. Different variants had different rates of adverse obstetric outcomes and different prognosis . Lastly, due to the exclusion of pregnant/lactating women from the preliminary vaccine trials , the Brazilian vaccination campaign for pregnant women started in July 2021, and our sample size did not allow a stratified analysis. Further studies are required on both topics.