Clinical Scorecard: Predicting Adverse Outcomes in Neurosurgical Patients Using Machine Learning: Exploring the Impact of Liver Function Indicators

At a Glance

Key Highlights

• Machine learning model (CatBoost) achieved high accuracy (AUC=0.932) in predicting unfavorable functional outcomes at discharge
• Key predictors included admission GCS score, age, and liver function markers (AST, albumin, ALKP)
• Lower GCS, older age, higher AST and ALKP, and lower albumin levels were associated with worse prognosis

Guideline-Based Recommendations

Diagnosis

• Assess neurological status using Glasgow Coma Scale at admission
• Perform liver function tests including AST, albumin, and ALKP on admission for neurosurgical patients

Management

• Incorporate liver function markers into prognostic assessment to identify patients at risk of unfavorable outcomes
• Use machine learning models integrating neurological and hepatic parameters to guide early intervention strategies

Monitoring & Follow-up

• Monitor liver function markers dynamically during hospitalization to detect worsening hepatic dysfunction
• Regularly assess neurological status alongside liver function to evaluate prognosis

Risks

• Recognize that hepatic dysfunction may exacerbate neurological deterioration via the liver-brain axis
• Be aware that elevated AST and ALKP and hypoalbuminemia are risk factors for poor neurosurgical outcomes

Patient & Prescribing Data

Neurosurgical patients with TBI, ICH, or aSAH without severe pre-existing liver disease

Early identification of patients with abnormal liver function markers may inform tailored therapeutic approaches and improve prognosis

Clinical Best Practices

• Integrate liver function testing into routine evaluation of neurosurgical patients
• Utilize machine learning tools to enhance prognostic accuracy beyond traditional clinical scoring
• Consider the liver-brain axis in the pathophysiology and management of neurosurgical conditions
• Exclude patients with pre-existing severe liver disease when applying prognostic models based on liver function markers
• Validate predictive models externally in multicenter cohorts before widespread clinical implementation

References

• Tianjin Medical University General Hospital Ethics Approval
• Liver-Brain Axis and Neurological Outcomes
• Machine Learning in Neurosurgical Prognosis