Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
January 9, 2026Open HeartOpen Access

The LightGBM model demonstrated superior predictive performance (AUROC of 0.9009) for cardiovascular risk scoring in patients undergoing non-cardiac surgery compared to traditional risk calculators.

View Full Paper
Ask AI
Bookmark
Share

Key result

The LightGBM model demonstrated superior predictive performance (AUROC of 0.9009) for cardiovascular risk scoring in patients undergoing non-cardiac surgery compared to traditional risk calculators.

Authors

NKNour Al KhatibACAli ChehabHTH Tamim

Discussion

Loading...

Member takes

Overview

Machine learning models assess cardiovascular risks in older patients undergoing surgery, suggesting improved patient safety.

Key Points

  • The study aims to develop a machine learning model to calculate cardiovascular risk scores for patients over 50 undergoing non-cardiac surgery.
  • Utilized the NSQIP 2022 dataset of 497,011 patients after data cleaning.
  • Primary endpoint included mortality, myocardial infarction, cardiac arrest, or stroke at 30 days post-surgery.
  • Implemented various machine learning algorithms including Logistic Regression, Naive Bayes, and Random Forest for analysis.
  • Models evaluated using area under the receiver operating characteristic curve (AUROC).
  • LightGBM model achieved the highest AUROC of 0.9009 with a 95% CI of 0.8889 to 0.9126.
  • Six data elements were significant: type of surgery, ASA classification, Blood Urea Nitrogen, sepsis, emergent surgery, and mechanical ventilation.

PICO

P
Population
Cardiovascular complications in patients undergoing non-cardiac surgery (n=497,011)
I
Intervention / Comparator
LightGBM machine learning model vs Conventional risk scores
O
Primary Outcome
Death, myocardial infarction, cardiac arrest, or stroke at 30 days postoperatively

Limitations

  • Heavy reliance on a single dataset for validation, potential issues with generalizability to other populations.

Cite This Study

Khatib et al. (2026) studied Cardiovascular complications in patients undergoing non-cardiac surgery (n=497,011). LightGBM machine learning model vs. Conventional risk scores was evaluated on Death, myocardial infarction, cardiac arrest, or stroke at 30 days postoperatively. The LightGBM model demonstrated superior predictive performance (AUROC of 0.9009) for cardiovascular risk scoring in patients undergoing non-cardiac surgery compared to traditional risk calculators.

synapsesocial.com/papers/696128f244c2cd6c68456c5bhttps://doi.org/10.1136/openhrt-2025-003565
View Full Paper
Ask AI
Bookmark
Share