Variable selection for the classification of postoperative cardiac patients.

Area of Science:

• Critical Care Medicine
• Cardiovascular Surgery
• Respiratory Physiology

Background:

• Post-cardiac surgery patients in the ICU can be stratified into low- and high-risk groups.
• Identifying predictive variables is crucial for timely intervention and improved patient outcomes.

Purpose of the Study:

• To identify hemodynamic, ventilatory, and gas analysis variables that best discriminate between low- and high-risk patients post-cardiac surgery.
• To evaluate the performance of a Fisher linear classifier using these discriminating variables.

Main Methods:

• Measurement of 13 hemodynamic, ventilatory, and gas analysis variables in 200 ICU patients within 6 hours of cardiac surgery.
• Application of the divergence criterion to identify variables with maximal separation power at three observation times.
• Testing a Fisher linear classifier using cardiac index (CI), carbon dioxide production index (VCO2I), and arterio-venous oxygen difference (avO2D) via the rotation method.

Main Results:

• The cardiac index (CI), carbon dioxide production index (VCO2I), and arterio-venous oxygen difference (avO2D) consistently showed the greatest separation power.
• The Fisher linear classifier demonstrated a high probability of correct recognition, exceeding 87%.

Conclusions:

• CI, VCO2I, and avO2D are powerful predictors of risk stratification in post-cardiac surgery ICU patients.
• A simple linear classifier using these variables offers effective risk prediction.
• More advanced classifiers may yield further improvements in predictive accuracy.