AUTHOR=Zhou Liang , Chen Yuanyi , Liu Ziyuan , You Jia , Chen Siming , Liu Ganzhi , Yu Yang , Wang Jian , Chen Xin TITLE=A predictive model for consciousness recovery of comatose patients after acute brain injury JOURNAL=Frontiers in Neuroscience VOLUME=17 YEAR=2023 URL=https://www.frontiersin.org/journals/neuroscience/articles/10.3389/fnins.2023.1088666 DOI=10.3389/fnins.2023.1088666 ISSN=1662-453X ABSTRACT=Background

Predicting the consciousness recovery for comatose patients with acute brain injury is an important issue. Although some efforts have been made in the study of prognostic assessment methods, it is still unclear which factors can be used to establish model to directly predict the probability of consciousness recovery.

Objectives

We aimed to establish a model using clinical and neuroelectrophysiological indicators to predict consciousness recovery of comatose patients after acute brain injury.

Methods

The clinical data of patients with acute brain injury admitted to the neurosurgical intensive care unit of Xiangya Hospital of Central South University from May 2019 to May 2022, who underwent electroencephalogram (EEG) and auditory mismatch negativity (MMN) examinations within 28 days after coma onset, were collected. The prognosis was assessed by Glasgow Outcome Scale (GOS) at 3 months after coma onset. The least absolute shrinkage and selection operator (LASSO) regression analysis was applied to select the most relevant predictors. We combined Glasgow coma scale (GCS), EEG, and absolute amplitude of MMN at Fz to develop a predictive model using binary logistic regression and then presented by a nomogram. The predictive efficiency of the model was evaluated with AUC and verified by calibration curve. The decision curve analysis (DCA) was used to evaluate the clinical utility of the prediction model.

Results

A total of 116 patients were enrolled for analysis, of which 60 had favorable prognosis (GOS ≥ 3). Five predictors, including GCS (OR = 13.400, P < 0.001), absolute amplitude of MMN at Fz site (FzMMNA, OR = 1.855, P = 0.038), EEG background activity (OR = 4.309, P = 0.023), EEG reactivity (OR = 4.154, P = 0.030), and sleep spindles (OR = 4.316, P = 0.031), were selected in the model by LASSO and binary logistic regression analysis. This model showed favorable predictive power, with an AUC of 0.939 (95% CI: 0.899–0.979), and calibration. The threshold probability of net benefit was between 5% and 92% in the DCA.

Conclusion

This predictive model for consciousness recovery in patients with acute brain injury is based on a nomogram incorporating GCS, EEG background activity, EEG reactivity, sleep spindles, and FzMMNA, which can be conveniently obtained during hospitalization. It provides a basis for care givers to make subsequent medical decisions.