Checklist-Driven Handover in Neurosurgical Care of Intracerebral Hemorrhage Patients: A Random Forest and SHAP-Based Outcome Analysis

Effective nursing handovers are critical in neurosurgical care for intracerebral hemorrhage (ICH) patients, who often present with rapid disease progression and high morbidity. However, traditional handover methods are frequently inconsistent and lack standardization, increasing the risk of nursing defects and communication errors. This study aims to evaluate the practical impact of implementing a checklist-based handover model among ICH patients in a neurosurgical department. A total of 100 inpatients were randomly assigned to either a control group receiving traditional handover or an observation group using a structured checklist. Handover duration, nursing defect incidence, and patient satisfaction were compared between groups. Additionally, a Random Forest Regressor was developed to model satisfaction outcomes, and SHapley Additive exPlanations (SHAP) analysis was applied to interpret variable contributions and feature interactions. Results showed that the checklist group had significantly shorter handover times, lower nursing defect rates, and higher patient satisfaction (all P < 0.05). Machine learning analysis further confirmed the checklist intervention as the most important predictor of satisfaction, with strong interaction effects enhancing model interpretability. These findings suggest that checklist-based handovers, supported by machine learning analysis, can improve clinical efficiency, ensure patient safety, and provide a scalable, evidence-based solution for quality enhancement in nursing care.