Article
The Comparison and Interpretation of Machine-Learning
Models in Post-Stroke Functional Outcome Prediction
Shih-Chieh Chang
1
, Chan-Lin Chu
2,3
, Chih-Kuang Chen
2,4
, Hsiang-Ning Chang
1
, Alice M. K. Wong
4,5
,
Yueh-Peng Chen
6,
* and Yu-Cheng Pei
1,2,5,7,
*
Citation: Chang, S.-C.; Chu, C.-L.;
Chen, C.-K.; Chang, H.-N.; Wong,
A.M.K.; Chen, Y.-P.; Pei, Y.-C. The
Comparison and Interpretation of
Machine-Learning Models in
Post-Stroke Functional Outcome
Prediction. Diagnostics 2021, 11, 1784.
https://doi.org/10.3390/diagnostics
11101784
Academic Editor: Keun Ho Ryu
Received: 26 August 2021
Accepted: 23 September 2021
Published: 28 September 2021
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2021 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
1
Department of Physical Medicine and Rehabilitation, Chang Gung Memorial Hospital at Linkou,
Taoyuan 333, Taiwan; kenny92031@cgmh.org.tw (S.-C.C.); lilychang0412@gmail.com (H.-N.C.)
2
College of Medicine, Chang Gung University, Taoyuan 333, Taiwan; Dreamcheap2000@hotmail.com (C.-L.C.);
leonard@cgmh.org.tw (C.-K.C.)
3
Department of Neurology, New Taipei Municipal Tucheng Hospital, Chang Gung Memorial Hospital,
New Taipei City 236, Taiwan
4
Department of Physical Medicine and Rehabilitation, Chang Gung Memorial Hospital at Taoyuan,
Taoyuan 333, Taiwan; alicewong.mk@gmail.com
5
Healthy Aging Research Center, Chang Gung University, Taoyuan 333, Taiwan
6
Center for Artificial Intelligence in Medicine, Chang Gung Memorial Hospital at Linkou, Taoyuan 333, Taiwan
7
Center of Vascularized Tissue Allograft, Chang Gung Memorial Hospital at Linkou, Taoyuan 333, Taiwan
* Correspondence: yuepengc@gmail.com (Y.-P.C.); yspeii@gmail.com (Y.-C.P.);
Tel.: +886-33281200 (ext. 7711) (Y.-P.C.); +886-33281200 (ext. 3846) (Y.-C.P.)
Abstract:
Prediction of post-stroke functional outcomes is crucial for allocating medical resources.
In this study, a total of 577 patients were enrolled in the Post-Acute Care-Cerebrovascular Disease
(PAC-CVD) program, and 77 predictors were collected at admission. The outcome was whether
a patient could achieve a Barthel Index (BI) score of >60 upon discharge. Eight machine-learning
(ML) methods were applied, and their results were integrated by stacking method. The area under
the curve (AUC) of the eight ML models ranged from 0.83 to 0.887, with random forest, stacking,
logistic regression, and support vector machine demonstrating superior performance. The feature
importance analysis indicated that the initial Berg Balance Test (BBS-I), initial BI (BI-I), and initial
Concise Chinese Aphasia Test (CCAT-I) were the top three predictors of BI scores at discharge. The
partial dependence plot (PDP) and individual conditional expectation (ICE) plot indicated that the
predictors’ ability to predict outcomes was the most pronounced within a specific value range (e.g.,
BBS-I < 40 and BI-I < 60). BI at discharge could be predicted by information collected at admission
with the aid of various ML models, and the PDP and ICE plots indicated that the predictors could
predict outcomes at a certain value range.
Keywords:
machine learning; stroke; rehabilitation; post-acute care; functional recovery; activities of
daily living
1. Introduction
Stroke is a major cause of disability and thus imposes substantial social and eco-
nomic burdens [
1
,
2
]. Post-stroke rehabilitation is pivotal for managing disability and
improving quality of life [
3
]. Because of the high diversity of stroke-induced disabilities,
predicting their functional outcomes is difficult. Numerous factors may affect post-stroke
functional outcomes, including age [
4
], cognition [
5
,
6
], comorbidities [
7
], post-stroke in-
tervention [
8
,
9
], and stroke characteristics, such as severity [
10
], type [
11
], location [
12
,
13
],
and volume [
14
]. Therefore, medical resources must be allocated to patients with a more
favorable rehabilitation potential to help them achieve their rehabilitation goals.
Approximately 795,000 patients globally were newly diagnosed as having stroke in
2020 [
15
], and the cost of post-stroke care is expected to triple by 2035 [
16
]. Studies have
mainly used regression models to define predictors for post-stroke functional outcomes.
Diagnostics 2021, 11, 1784. https://doi.org/10.3390/diagnostics11101784 https://www.mdpi.com/journal/diagnostics
