# Multi-centered reassessment of CRS-R in disorders of consciousness: a dimensionality reduction study from cognition and motor function

## TL;DR

This study improves a medical scale for assessing consciousness by separating cognitive and motor functions into two dimensions, enabling better patient classification and treatment planning.

## Contribution

A novel two-dimensional model of the CRS-R scale that differentiates cognition and motor function for more accurate diagnosis of disorders of consciousness.

## Key findings

- The model achieved high diagnostic accuracy (0.94), precision (0.92), and recall (0.99) in classifying patients.
- Patients in minimally conscious states scored significantly higher than those in vegetative states (p < 0.05).
- A four-quadrant framework was developed to guide personalized interventions based on distinct clinical profiles.

## Abstract

This study aimed to enhance the Coma Recovery Scale-Revised (CRS-R) for disorders of consciousness (DoC) by developing a two-dimensional model differentiating cognition and motor function.

We analyzed 124 DoC patients retrospectively and validated findings using five multicenter datasets (n = 420). CRS-R subscores were decomposed into Consciousness_x (awareness) and Consciousness_y (arousal/motor function) using Projective Non-negative Matrix Factorization. Logistic regression established diagnostic thresholds, evaluated by accuracy, precision, recall, and F1-score.

The model achieved high accuracy (0.94), precision (0.92), and recall (0.99). Patients with minimally conscious state (MCS) or emerged MCS showed significantly higher scores than vegetative state (VS) patients (p < 0.05). The four-quadrant framework revealed distinct clinical profiles: Quadrant I (high awareness/arousal) identified patients for cognitive rehabilitation; Quadrant II (low awareness/high arousal) suggested arousal-enhancing therapies; Quadrant III (low awareness/arousal) indicated VS requiring basic support; Quadrant IV (high awareness/low arousal) highlighted needs for sensorimotor integration.

The two-dimensionally reduced representation of CRS-R scores maintains diagnostic accuracy while improving DoC classification. The four-quadrant model enables personalized interventions.

Our study has been verified by the Chinese Clinical Trial Registry with the registration number: ChiCTR2400085855, and the registration date is June 19, 2024.