1. Potential and limitations of machine meta-learning (ensemble) methods for predicting COVID-19 mortality in a large inhospital Brazilian dataset.
- Author
-
de Paiva BBM, Pereira PD, de Andrade CMV, Gomes VMR, Souza-Silva MVR, Martins KPMP, Sales TLS, de Carvalho RLR, Pires MC, Ramos LEF, Silva RT, de Freitas Martins Vieira A, Nunes AGS, de Oliveira Jorge A, de Oliveira Maurílio A, Scotton ALBA, da Silva CTCA, Cimini CCR, Ponce D, Pereira EC, Manenti ERF, Rodrigues FD, Anschau F, Botoni FA, Bartolazzi F, Grizende GMS, Noal HC, Duani H, Gomes IM, Costa JHSM, di Sabatino Santos Guimarães J, Tupinambás JT, Rugolo JM, Batista JDL, de Alvarenga JC, Chatkin JM, Ruschel KB, Zandoná LB, Pinheiro LS, Menezes LSM, de Oliveira LMC, Kopittke L, Assis LA, Marques LM, Raposo MC, Floriani MA, Bicalho MAC, Nogueira MCA, de Oliveira NR, Ziegelmann PK, Paraiso PG, de Lima Martelli PJ, Senger R, Menezes RM, Francisco SC, Araújo SF, Kurtz T, Fereguetti TO, de Oliveira TC, Ribeiro YCNMB, Ramires YC, Lima MCPB, Carneiro M, Bezerra AFB, Schwarzbold AV, de Moura Costa AS, Farace BL, Silveira DV, de Almeida Cenci EP, Lucas FB, Aranha FG, Bastos GAN, Vietta GG, Nascimento GF, Vianna HR, Guimarães HC, de Morais JDP, Moreira LB, de Oliveira LS, de Deus Sousa L, de Souza Viana L, de Souza Cabral MA, Ferreira MAP, de Godoy MF, de Figueiredo MP, Guimarães-Junior MH, de Paula de Sordi MA, da Cunha Severino Sampaio N, Assaf PL, Lutkmeier R, Valacio RA, Finger RG, de Freitas R, Guimarães SMM, Oliveira TF, Diniz THO, Gonçalves MA, and Marcolino MS
- Subjects
- Adult, Humans, Female, Middle Aged, Male, Brazil, Hospitals, Hospitalization, Machine Learning, COVID-19
- Abstract
The majority of early prediction scores and methods to predict COVID-19 mortality are bound by methodological flaws and technological limitations (e.g., the use of a single prediction model). Our aim is to provide a thorough comparative study that tackles those methodological issues, considering multiple techniques to build mortality prediction models, including modern machine learning (neural) algorithms and traditional statistical techniques, as well as meta-learning (ensemble) approaches. This study used a dataset from a multicenter cohort of 10,897 adult Brazilian COVID-19 patients, admitted from March/2020 to November/2021, including patients [median age 60 (interquartile range 48-71), 46% women]. We also proposed new original population-based meta-features that have not been devised in the literature. Stacking has shown to achieve the best results reported in the literature for the death prediction task, improving over previous state-of-the-art by more than 46% in Recall for predicting death, with AUROC 0.826 and MacroF1 of 65.4%. The newly proposed meta-features were highly discriminative of death, but fell short in producing large improvements in final prediction performance, demonstrating that we are possibly on the limits of the prediction capabilities that can be achieved with the current set of ML techniques and (meta-)features. Finally, we investigated how the trained models perform on different hospitals, showing that there are indeed large differences in classifier performance between different hospitals, further making the case that errors are produced by factors that cannot be modeled with the current predictors., (© 2023. The Author(s).)
- Published
- 2023
- Full Text
- View/download PDF