This study investigates bias and fairness in a MIMIC-III trained benchmarking model for in-hospital mortality prediction. Using a fairness and generalizability assessment framework, the researchers found a strong class imbalance and fairness concerns for Black and publicly insured ICU patients. The study highlights the need for comprehensive fairness and performance assessment frameworks when using open data resources to build benchmark models.
Publisher
Scientific Data
Published On
Jan 24, 2022
Authors
Eliane Röösli, Selen Bozkurt, Tina Hernandez-Boussard
Tags
bias
fairness
MIMIC-III
in-hospital mortality
ICU patients
benchmark models
healthcare
Related Publications
Explore these studies to deepen your understanding of the subject.
