Related Experiment Video
Updated: Aug 6, 2026
The Impact of Motor Task Conditions on Goal-Directed Arm Reaching Kinematics and Trunk Compensation in Chronic Stroke Survivors
Published on: May 2, 2021
Prediction Models for Gait Independence in Patients with Post-Stroke Hemiplegia in a Convalescent Rehabilitation
Masakazu Murakami1, Ryo Sato2, Keita Ogasawara2
1Occupational Therapy Course, Department of Rehabilitation, Faculty of Health Sciences, Japan Healthcare University, Hokkaido, Japan.
Abstract:
Murakami M, Sato R, Ogasawara K, Kimura Y. Prediction Models for Gait Independence in Patients with Post-Stroke Hemiplegia in a Convalescent Rehabilitation Ward: Comparison of Predictive Performance Between Different Trunk Function Assessments. Jpn J Compr Rehabil Sci 2026; 17: 24-32.
Objective:
This study aimed to compare the predictive performance for walking independence at discharge using two trunk function assessments: the Functional Assessment for Control of Trunk (FACT) and the Trunk Control Test (TCT), alongside activities of daily living (ADL) ability and physical function at admission, in patients with post-stroke hemiplegia admitted to a convalescent rehabilitation ward.
Methods:
A cohort of 196 patients with post-stroke hemiplegia admitted to a convalescent rehabilitation ward between October 2020 and March 2024 was retrospectively analyzed. Patients achieving a walking score of 6 or higher on the Functional Independence Measure (FIM) at discharge were classified as the walking-independent group. Independent variables included motor FIM (m-FIM), cognitive FIM (c-FIM), paretic lower-extremity function on the Stroke Impairment Assessment Set (SIAS), Ueda's 12-Grade Hemiplegic Function Test, non-paretic-side SIAS function, age, sex, and either FACT or TCT. Decision-tree analysis served as the primary analytical method, with logistic regression and receiver operating characteristic (ROC) curve analyses conducted as supplementary approaches to examine discriminative performance.
Results:
The cohort included 102 men and 94 women, with a mean age of 77.7 ± 11.2 years. The walking-independent group comprised 47 patients (24.0%). Decision-tree analyses indicated that both the FACT and TCT models identified m-FIM and non-paretic-side SIAS function as predictive factors. Additionally, FACT and TCT were each selected as predictive factors in their respective models. ROC analysis demonstrated high discriminative ability for both models (FACT model: area under the curve [AUC] = 0.968; TCT model: AUC = 0.971).
Conclusion:
Both models utilizing FACT or TCT as trunk function assessments demonstrated high predictive accuracy, suggesting their potential utility for prognostic prediction in clinical practice.
