Predicting Arm Nonuse in Individuals with Good Arm Motor Function after Stroke Rehabilitation: A Machine Learning Study

 NO, NO, NO! Don't predict failure to recover. Deliver interventions that lead to recovery. I can predict nonuse of the arm with one question; Is your hand functional? Y/N? If not there then deliver the protocols that allow full use of the hand. Do the research in the correct order; hand, then arm.

So solve the problems in the correct order! Duh! And don't give me crap about proximal to distal.

Predicting Arm Nonuse in Individuals with Good Arm Motor Function after Stroke Rehabilitation: A Machine Learning Study 

by

Yu-Wen Chen

^1,2,

Yi-Chun Li

^1,3,

Chien-Yu Huang

¹,

Chia-Jung Lin

¹,

Chia-Jui Tien

¹,

Wen-Shiang Chen

^4,5,6,

Chia-Ling Chen

^7,8 and

Keh-Chung Lin

^1,9,*

¹

School of Occupational Therapy, National Taiwan University College of Medicine, 17, F4, Xu-Zhou Road, Taipei 100, Taiwan

²

Department of Speech Language Pathology and Audiology, National Taipei University of Nursing and Health Sciences, 365, Mingde Road, Taipei 112, Taiwan

³

Department of Occupational Therapy, I-Shou University College of Medicine, 8, Yida Road, Jiaosu Village, Yanchao District, Kaohsiung 824, Taiwan

⁴

Department of Physical Medicine and Rehabilitation, College of Medicine, National Taiwan University, Taipei 10048, Taiwan

⁵

Department of Physical Medicine and Rehabilitation, National Taiwan University Hospital, Taipei 10048, Taiwan

⁶

Institute of Biomedical Engineering and Nanomedicine, National Health Research Institutes, 35, Keyan Road, Zhunan District, Miaoli 350, Taiwan

⁷

Department of Physical Medicine and Rehabilitation, Chang Gung Memorial Hospital at Linkou, 5 Fusing Street, Gueishan District, Taoyuan 333, Taiwan

⁸

Graduate Institute of Early Intervention, College of Medicine, Chang Gung University, 259 Wenhua 1st Road, Gueishan District, Taoyuan 333, Taiwan

⁹

Division of Occupational Therapy, Department of Physical Medicine and Rehabilitation, National Taiwan University Hospital, 7 Chung-Shan South Road, Taipei 100, Taiwan

^*

Author to whom correspondence should be addressed.

Int. J. Environ. Res. Public Health 2023, 20(5), 4123; https://doi.org/10.3390/ijerph20054123

Received: 16 January 2023 / Revised: 17 February 2023 / Accepted: 23 February 2023 / Published: 25 February 2023

(This article belongs to the Section Disabilities)

Download

Browse Figures

Versions Notes

Abstract

Many stroke survivors demonstrate arm nonuse despite good arm motor function. This retrospective secondary analysis aims to identify predictors of arm nonusers with good arm motor function after stroke rehabilitation. A total of 78 participants were categorized into 2 groups using the Fugl-Meyer Assessment Upper Extremity Scale (FMA-UE) and the Motor Activity Log Amount of Use (MAL-AOU). Group 1 comprised participants with good motor function (FMA-UE ≥ 31) and low daily upper limb use (MAL-AOU ≤ 2.5), and group 2 comprised all other participants. Feature selection analysis was performed on 20 potential predictors to identify the 5 most important predictors for group membership. Predictive models were built with the five most important predictors using four algorithms. The most important predictors were preintervention scores on the FMA-UE, MAL–Quality of Movement, Wolf Motor Function Test-Quality, MAL-AOU, and Stroke Self-Efficacy Questionnaire. Predictive models classified the participants with accuracies ranging from 0.75 to 0.94 and areas under the receiver operating characteristic curve ranging from 0.77 to 0.97. The result indicates that measures of arm motor function, arm use in activities of daily living, and self-efficacy could predict post intervention arm nonuse despite good arm motor function in stroke. These assessments should be prioritized in the evaluation process to facilitate the design of individualized stroke rehabilitation programs to reduce arm nonuse.

Keywords:

stroke rehabilitation; arm nonuse; predictors; chronic stroke; machine learning

1. Introduction

Stroke is a leading cause of disability [1] that can lead to chronic arm impairment [2,3]. Reduced arm function negatively affects stroke survivors’ quality of life [4]. A significant proportion of stroke survivors perceived nonuse in the affected arm as a major problem at 4 years after stroke [2]. Studies have reported constraint-induced movement therapy and its variants were effective in reducing arm nonuse and improving functional independence and quality of life in individuals with stroke [5,6]. Neurobiological changes corresponding to improvement in arm use after constraint-induced therapy were also observed with brain imaging [6]. Recently, researchers also advocated for a paradigm shift in stroke rehabilitation toward self-practice to reduce arm nonuse following motor rehabilitation [7]. However, early identification of individuals with stroke who may benefit the most from these programs is challenging. In the context of precision medicine, if individuals who are likely to develop arm nonuse can be identified early in the course of rehabilitation, the information would be able to guide and shape the individualized rehabilitation program.

Researchers have strived to investigate the prediction of arm nonuse in individuals with stroke. Many factors were found or proposed as being associated with the development of arm nonuse across the phases (i.e., acute, subacute, and chronic) after stroke. The factor with the most empirical support was upper-limb motor function. Substantial evidence supports the association between baseline or postintervention arm motor function and the amount of use or postintervention amount of use in the affected limb [8,9,10]. In the acute phase, age, stroke severity, sensory impairment, degree of disability at discharge, and hand grip strength were associated with the development of arm nonuse at 90 days after discharge [11,12]. In chronic stroke, upper-limb motor dysfunction, dependence in activities of daily living, and participants’ own perceived upper-limb function were associated with activities in the affected upper limb measured by wrist-worn accelerometers [13,14]. Self-efficacy was found to predict arm nonuse in a small group of participants [8] and to moderate the predicting relationship between upper-limb motor function and ratings of daily upper-limb use [15]. Emotional state, such as depression, was discussed as a potential factor in the context that it may affect compliance with constraint-induced movement therapy [16]. Other factors for daily upper-limb use after stroke rehabilitation, such as motivation, health behaviors, and environmental support, were also proposed but not explicitly tested [17].

Despite strong evidence suggesting the association between upper-limb motor function and daily use, good upper-limb motor function does not translate directly into more use of the affected limb. Studies have identified participants who demonstrated low daily use of the affected limb despite having good upper-limb motor function [18,19]. Some patients showed improvement in upper-limb function after rehabilitation, but continued to demonstrate low daily use of the affected limb [13,20]. This makes the prediction of nonuse a particularly challenging task. There are a few possible reasons behind the challenge. First, the relationship between arm function and arm use after stroke may be nonlinear [21,22] and may be subject to the moderating effects of other factors [15]. Second, the theoretical background that arm nonuse after stroke is learned [23] predicts that upper-limb motor function cannot be a sole predictor. Findings from earlier studies supported factors other than upper-limb motor function played a role in daily upper-limb activity [13,18]. Nevertheless, it may be a strong enough predictor to mask others, making it difficult to detect ancillary but important factors using traditional regression models.

The advances in artificial intelligence have provided us with another tool for data analysis, machine learning. Unlike traditional statistics, machine learning uses multidimensional linear and nonlinear methods to find patterns in the data [24], striving to achieve as high an accuracy as possible. It has the potential to identify factors that have nonlinear, complex relationships with upper-limb activity and achieve a high predicting capacity.

This study used machine learning methods to investigate predictors that can help identify individuals that were likely to develop arm nonuse but had good arm motor function after stroke rehabilitation. As discussed in the previous paragraphs, many potential predictors could be considered. We searched our database accumulated in the past few years to locate data with a set of likely predictors according to the literature. Furthermore, we grouped our participants in the database into two predefined groups according to their postintervention assessment results: those with low upper-limb use and good upper-limb motor function in the affected arm, and all others. We were interested in using preintervention measures to predict which patients would be likely to develop arm nonuse despite good upper-limb motor function after intervention. We hypothesize that measures other than motor function will emerge as important predictors for the classification. By doing so, we anticipate the results will inform clinical care planning when addressing personal and environmental factors in addition to motor function in the design of client-centered individualized rehabilitation programs.

More at link.