Maternal morbidity profile in hospitalizations in the Unified Health System in São Paulo, Brazil: Analysis using data mining, 2014 to 2019.

In Brazil, approximately 80% of births are publicly funded and are registered in the Hospital Information System of the Unified Health System (SIH/SUS). The SIH/SUS has morbidity data, but its use for maternal morbidity surveillance is still limited. The objective of this study was to apply data mining techniques to identify diagnoses registered in the SIH/SUS that alone or together are associated with maternal death. Obstetric hospitalizations at SIH/SUS of women aged 10-49 years from the largest state in Brazil (São Paulo) from 2014 to 2019 were analysed. Diagnoses were classified into 12 groups, adapting the WHO's classification proposal of maternal deaths. Groups of diagnoses that were related to maternal death were identified using association rules with the Apriori algorithm. Of the 2,742,467 hospitalizations, 831 (0.03%) resulted in death. The most frequent diagnoses associated with death, alone or in combination, were non-obstetric complications (62.0%). Pregnancy-specific hypertensive complications (20.1%), pregnancy-related infections (19.9%) and haemorrhages (13.6%) were also present in hospitalizations resulting in death. The risk of death was more significant for non-obstetric complications, unknown causes and external causes, with lifts above 10. The risk of death was at least three times greater for groups of diagnoses related to the most frequent causes of maternal death in the country (complications of hypertension, infections, and haemorrhages). Uncomplicated abortion, gestational diabetes and other causes of non-obstetric hospitalization were only associated with maternal death when together with other diagnoses. The study reinforces the data mining technique as an innovative approach in the study of maternal mortality in the country. The results also reaffirm the importance of hypertensive, haemorrhagic and infectious causes for maternal death and highlight the relevance of non-obstetric causes for the surveillance of maternal morbidity since women with these diagnoses, alone or associated with other complications, had a higher risk of death.