Advanced analytics for outcome prediction in intensive care units

Area of Science:

• Medical Informatics
• Artificial Intelligence in Medicine
• Critical Care Medicine

Background:

• Accurate prediction of intensive care unit (ICU) outcomes is crucial for patient management.
• Existing acuity scoring systems like SOFA and SAPS-III have limitations in predictive performance.
• There is a need for advanced clinical decision support systems (CDSS) leveraging comprehensive patient data.

Purpose of the Study:

• To develop and evaluate a novel expert knowledge-based CDSS for predicting ICU patient outcomes.
• To utilize physiological measurements from the first 48 hours of ICU admission for outcome prediction.
• To compare the performance of the developed CDSS against established scoring systems (SOFA, SAPS-III).

Main Methods:

• Data categorization by physiological organ system.
• Extraction and mutual information-based feature ranking.
• Development of an artificial neural network classifier and eight-fold cross-validation.

Main Results:

• The developed CDSS achieved an F-score of 42%.
• The SOFA system achieved an F-score of 26%.
• The SAPS-III system achieved an F-score of 29%.

Conclusions:

• The novel expert knowledge-based CDSS significantly outperforms SOFA and SAPS-III in predicting ICU outcomes.
• The system effectively identifies high-risk patients for critical deterioration.
• This CDSS offers a promising advancement in critical care decision support.