To develop a machine learning-based model to predict unfavorable outcomes, specifically mortality and functional decline, in neurosurgical patients, focusing on liver function markers.
Key Findings:
The CatBoost model achieved the best performance (AUC = 0.932, accuracy = 0.879).
Key predictive features included GCS score at admission, age, and liver function markers.
Lower GCS score and older age predicted unfavorable outcomes.
Higher AST and ALKP levels, along with lower albumin levels, were associated with poor outcomes.
Interpretation:
The CatBoost model effectively predicts adverse outcomes in neurosurgical patients by integrating neurological and liver function markers, highlighting the importance of the liver-brain axis in clinical decision-making.
Limitations:
Retrospective design may introduce bias.
Single-center study limits generalizability.
Need for external validation in multicenter studies.
Potential confounding factors affecting outcomes were not fully controlled.
Conclusion:
The study demonstrates the potential of machine learning models in predicting neurosurgical outcomes, emphasizing the role of liver function indicators.
