Clinical Sciences/Health Conditions
Ye Zhang, MD
Dr.
Department of Rehabilitation Medicine，Xuan Wu Hospital, Capital Medical University
Beijing, Beijing, China (People's Republic)
Zilong Zhu, MM
Dr.
Department of Rehabilitation Medicine，Xuan Wu Hospital, Capital Medical University
Beijing, Beijing, China (People's Republic)
Weiqun Song, MD
Dr.
Department of Rehabilitation Medicine，Xuan Wu Hospital, Capital Medical University
Beijing, Beijing, China (People's Republic)
Accurate prognosis for patients with prolonged disorders of consciousness (pDoC) is critical for clinical management but remains challenging. Current machine learning models are often hindered by small sample sizes, inconsistent data standards, and limited interpretability. This study aims to overcome these issues and develop a machine learning-based prognostic model which could support clinical decision-making and enhance the quality of life for pDoC patients.
Design:
A retrospective study of pDoC patients hospitalized at Xuanwu Hospital from January 2013 to August 2024 was conducted. Data on clinical information, behavioral scales, and event-related potentials (ERPs) were collected, using 6-month consciousness recovery as the outcome. The least absolute shrinkage and selection operator (LASSO) regression identified key variables, and model performance was assessed using metrics such as macro-averaged receiver operating characteristic area under the curve (ROC-AUC), macro-averaged precision–recall area under the curve (PR-AUC) and macro-averaged F1 score. Clinical utility was evaluated via Decision Curve Analysis (DCA), and interpretability was supported by SHapley Additive exPlanations (SHAP) method.
Results:
A total of 110 pDoC patients, comprising 72 with minimally conscious state (MCS) and 38 with vegetative state/unresponsive wakefulness syndrome (VS/UWS), were enrolled in the study. Follow-up assessments identified 53 patients with improved prognosis and 57 without improvement. LASSO regression analysis selected eight variables: sex, skull state, etiology, duration, Coma Recovery Scale-Revised (CRS-R) score, diagnosis, mismatch negativity (MMN) pattern, and epilepsy status. The efficient logistic regression model demonstrates robust performance with a macro-averaged ROC-AUC of 0.7504 (95% CI: 0.6590–0.8404). In the decision curve analysis, the threshold probabilities for net benefit ranged from 13% to 89%. The CRS-R score ranked highest in SHAP importance.
Conclusion:
This study developed a reliable prognostic model for consciousness recovery in pDoC patients by integrating multimodal data using efficient logistic regression. The model can predict 6-month recovery outcomes and supports optimized treatment decisions for pDoC patients.
