Análise da morbi-mortalidade materna e suas relações com a Covid-19 no Brasil através do machine learning

Abstract

Pregnancy is a natural process that involves physical and emotional changes, leaving a woman's body more vulnerable to illness. Maternal mortality is a serious public health problem, mainly affecting developing countries, where resources are scarce. In Brazil, between 2017 and 2018, the most frequent direct causes of maternal death were hypertension (pre-eclampsia and eclampsia), hemorrhages, and infections. Although there has been a reduction in the maternal mortality rate, the country is still above international targets. The COVID-19 pandemic significantly worsened the maternal health situation in Brazil, especially in 2020 and 2021, with an increase in maternal complications and deaths. The collapse of the Unified Health System (SUS) and institutional instability, reflected in disagreements between health management bodies, contributed to a weakened initial response to the pandemic. Pregnant women infected with SARS- CoV-2 face a higher risk of developing severe conditions, respiratory decompensation, and early delivery. Objective: To investigate the obstetric outcomes associated with the diagnosis of COVID- 19 in Brazil in 2020 and 2021. Method: This is a qualitative, quantitative, descriptive, and cross- sectional study, featuring the use of triangulation of methods. Triangulation of methods as a validation tool combines different methodological perspectives in the same study with the aim of adding rigor, richness, depth, and breadth to the investigation. Initially, the integrative systematic literature review method was used to understand the “state of the art” to respond to the first objective proposed in this study. The quantitative part of the study was developed using the database of the Influenza Epidemiological Surveillance Information System (SIVEP-Gripe). The data analyzed were for the years 2020 and 2021, corresponding to a dataset for each year. Results: COVID-19 directly affected maternal morbidity and mortality rates globally. SIM and SIVEP are high-quality systems that are essential for monitoring and decision-making in health, but there is a need to optimize the availability and qualification of their data. The present study revealed that among the COVID-19 data, girls, pregnant or in the postpartum period, under the age of 14, which repeatedly denounces the negligence inflicted by the Brazilian State, in denying these girls' right to live, play and study freely. Regarding hospitalizations, in 2020, out of a total of 8,364 notifications, 6,589 cases (95.1%) involved hospitalizations, 236 (3.4%) did not require hospitalization, and 104 cases had this information ignored (1.5%). Regarding hospitalizations, in 2021, out of a total of 14,506 notifications, 12,027 cases (97.3%) were hospitalized, 175 (1.4%) did not require hospitalization, and 156 cases had this information ignored (1.3%). Hospitalizations of pregnant and postpartum women in Intensive Care Unit (ICU) beds in 2020 had a total of 1,418 cases (20.5%) and in 2021 3,950 cases (32%). With an increase in the proportion of ICU admissions from 2020 to 2021, going from 20.5% in 2020 to 32% in 2021. Machine learning was used as a classifier, identifying the factors that most contributed to maternal death and they were: Invasive ventilatory support (importance of 15.8), ICU admission (importance of 6.62); Dyspnea (importance of 3.55), being a puerperal woman, variation in saturation, time between symptoms and evolution of the case. Conclusion: The identification of signs and symptoms that increase the risk of maternal death should be done even when classifying the risk of this pregnant body, and having an instrument to guide this action can anticipate adequate care and save lives. It is important to highlight the novelty of this study in using Machine Learning Modeling (ML) to identify and classify factors that define the outcome of death and cure. During the development of this study, some articles were found that worked on AM with the health interface, however, none of them with the interface, women's health and obstetric outcomes in the face of COVID-19.