Create an identification system that makes available the cases of maternal near misses in Brazil between 2002 and 2021, in addition to understanding the socioeconomic factors influencing the numbers, identifying regions with high vulnerability of cases, and forecasting the number of cases of maternal near miss in the coming years.
The Python programming language and the PySUS library were used to acquire the data from the Department of Informatics of the Unified Health System (DATASUS). The Pandas, Numpy, Scipy, Matplotlib, Statsmodels, and Scikit-learn libraries were used to manipulate the data and acquire the results. The data were filtered according to the research needs and analyzed for the creation of the prediction model.
Analysis of maternal near-miss data in Brazil from 2002 to 2021 reveals that regions with high population density, lower income levels, and higher rates of low education are more susceptible to maternal near misses. These factors are critical for evaluating the necessary interventions to reduce occurrences of maternal near misses.The maternal near-miss rates calculated using the Waterstone criteria showed that the Southeast (0.853 ± 0.415) and Northeast (0.817 ± 0.726) regions had the highest averages. Notably, the Northeast recorded the highest average in 2010 (M = 2.264). According to the Mantel criterion, the Southeast (0.059 ± 0.035) and South (0.058 ± 0.085) regions exhibited the highest averages, with the South experiencing its peak in 2013 (M = 0.315).Using WHO criteria, the North region had the highest mean rate (0.287 ± 0.207), followed closely by the Northeast (0.232 ± 0.243). The highest recorded rate in the Northeast occurred in 2010 (M = 0.946). Additionally, epidemiological data mapping cases of maternal near misses in Brazil from 2002 to 2021 has been organized by the 26 states and the Federal District, categorized according to their respective major geographic regions: North, Northeast, Central-West, Southeast, and South.
The project aimed to develop an innovative machine-learning system to classify near-miss cases using patient records from the hospital information system (SIH) and mortality data (SIM). While the mortality data excludes information about women who survived, it would still assist in the identification of patterns associated with maternal near-miss cases. Additionally, the models incorporated socio-economic data from the patients’ regions to deliver a more accurate and faster estimation of the number of maternal near-miss cases in specific areas.
Maternal near-miss is defined by the World Health Organization as cases of women who experienced life-threatening complications during pregnancy, childbirth, or the postpartum period but ultimately survived. This metric serves as a crucial indicator of the quality of primary healthcare within a region or country and is vital for assessing the impact of public health policies on access to healthcare. However, identifying near-miss cases within the Unified Health System (SUS) databases is complicated by issues of underreporting and incomplete records. The process of analyzing data to classify these cases is labor-intensive and conducted on an individual basis, often leading to delays of several months in accurately identifying women treated by the SUS.
The results of the project bring the picture of reproductive health care performed by the SUS. The potential of the project is to demonstrate what improvements are needed in public health care to reduce preventable deaths of pregnant women and, consequently, to achieve the goal of reducing maternal mortality established by the United Nations. Such results may drive new measures to reduce these maternal morbidity and mortality rates