PulseJournal ClubResearchersJournalsExplore
Instagram
HomeJournal ClubExplore
Synapse
⌘+K
Synapse
June 11, 2026BMC Medical Informatics and Decision MakingOpen Access

Data balancing improves mortality prediction for emergency department patients

View Full Paper
Ask AI
Bookmark
Share

Authors

CTChinyang Henry TsengYLYu-Sheng LoYLYu-Juin Lin

Discussion

Loading...

Member takes

Overview

Randomized trial improves mortality prediction accuracy in emergency department patients, suggesting better clinical outcomes.

Key Points

  • The study aims to improve patient mortality prediction accuracy in the emergency department using data balancing methods.
  • Analyzed 2,437,341 non-traumatic adult ED visit records from 2008 to 2016.
  • Evaluated data balancing methods including Random Under Sampling, SMOTE, and Random Over Sampling.
  • Employed machine learning models: Random Forest, AdaBoost, XG Boost, with Logistic Regression as meta learner.
  • XGB model achieved AUROC of 91.41% in 168-hour mortality timeframe, surpassing previous study results.
  • True Positive Rate and True Negative Rate improved significantly to 79.88% and 86.73%, respectively.
  • Random Over Sampling outperformed other methods, achieving best AUROC in 24-hour mortality timeframe at 93.72%.

Cite This Study

Tseng et al. (2026) studied this question.

synapsesocial.com/papers/6a2a4fa380c8f91e7f39c909https://doi.org/10.1186/s12911-026-03608-9
View Full Paper
Ask AI
Bookmark
Share