Related Experiment Video
Updated: Jan 9, 2026
Author Spotlight: Enhancing Upper Limb Rehabilitation in Stroke Patients Through Advanced Robotic and Neuromodulation Technologies
Published on: October 11, 2024
Simplified assessment of upper limb dysfunction after stroke: decision tree analysis based on the International
Xiaobing Chen1,2, Kai L Catherine Chan3,4, Xinyue Wang3,4
1Department of Rehabilitation Medicine, Zhongda Hospital Southeast University, Nanjing, China.
Background:
The International Classification of Functioning, Disability and Health (ICF) provides a comprehensive framework for evaluating stroke patients across four key domains. However, due to the large number of ICF items and the complexity of its classification system, its practical application is often time-consuming, and requires substantial training for evaluators.
Aim:
To generate and validate a decision tree model based on ICF items for assessing upper limb dysfunction after stroke.
Design:
A cross-sectional study.
Setting:
Rehabilitation department of five hospitals.
Population:
This study included stroke patients (including ischemic or hemorrhagic; first-ever or recurrent), who were stratified according to disease phase: acute (1-7 days, 1.94%), subacute (8-180 days, 53.02%), and chronic (>180 days, 41.38%). All patients had CT/MRI-confirmed diagnosis, stable vital signs, and upper limb dysfunction.
Methods:
All participants completed the 56 items of the comprehensive ICF Core Set for stroke. The decision tree model of ICF items significantly associated with the Fugl-Meyer Upper Extremity Scale (FM-UE) was constructed by using the R package rpart.
Results:
A total of 464 participants after stroke were recruited. Ten ICF items that were strongly correlated with the FM-UE and P<0.05, and the items were "d4400," "d4401," "d4402," "d4403," "d4450," "d4451," "d4452," "d4453," "d4454," "d4455". Finally, the decision tree model included: "d4401: grasping," "d4553: turning or twisting the hands or arms" and "d4551: pushing". The statistical significant accuracy of the model confusion matrix in validation was 0.7381 (P=5.008e-13), and the AUC was 0.8406.
Conclusions:
This study identified ICF items highly correlated with the FMA-UE and developed a statistically significant decision tree model for the assessing of upper limb dysfunction after stroke.
Clinical Rehabilitation Impact:
The decision tree model based on key ICF items, substantially reduces evaluation time. It simplifies upper limb dysfunction assessment, enhances ICF application, and provides a simpler and more efficient assessment tool that represents a valuable addition to the clinical tool for stroke rehabilitation.
