ABSTRACT
@#Objective To systematicly evaluate the constituent factors and accuracy of prediction models of outcome for patients with prolonged disorders of consciousness. Methods Articles about prediction models of outcome for patients with prolonged disorders of consciousness were retrieved from PubMed, Web of Science, CNKI and Wanfang Data until September 30th, 2021. The authors, publishing times, subjects, predictive indicators, outcomes and conclusions were extracted. Results A total of 4 313 articles were returned and 37 included, comprising randomized controlled trials and cohort studies, which published mainly from 2012 to 2021. The subjects were patients with prolonged disorders of consciousness, and their predictions consisted of clinical assessment scales, neuroimaging, neuroelectrophysiology and laboratory indicators. Conclusion Prediction models may be valuable for the long-term outcomes of patients with prolonged disorders of consciousness. Most of the current prediction models are composed of only a limited number of technical means, and the accuracy is uneven. Coma Recovery Scale-revised, default mode network and multiple evoked potentials-related prediction models are accurate, but lack a unified adaptation standard.