Back

Assessing the Performance of Machine Learning Models to Predict Neonatal Mortality Risk in Brazil, 2000-2016

Alves, L. C.; Beluzo, C. E.; Arruda, N. M.; Bressan, R.; Carvalho, T.

2020-05-25 epidemiology
10.1101/2020.05.22.20109165 medRxiv
Show abstract

Neonatal mortality figures are an important healths problem, as the first month of life is the most vulnerable time for survival. Factors associated with neonatal mortality are complexly and influenced by the maternal and newborn biological characteristics, social conditions and the care provided by the health services. The aim of this study was investigated the association between features related and neonatal mortality risk in Brazil. Data came from two surveys: The Mortality Information System and Information System on Live Births. The final sample was composed of 302,943 children between 2006 and 2016. We highlight the proposition of a new approach based on machine learning to address the problem of neonatal mortality death risk classification. The results using three different machine learning classifiers points toward expressiveness of features, being newborn weight, Apgar at the first and fifth minute, congenital malformations, gestational weeks and number of prenatal appointments the six more expressive.

Matching journals

The top 4 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.