Determining the Optimal Virtual Reality Exergame Approach for Balance Therapy in Persons With Neurological Disorders Using a Rasch Analysis: Longitudinal Observational Study

Abstract

Background: Virtual reality (VR) exergames have gained popularity in the rehabilitation of persons with neurological disorders as an add-on therapy to increase intensity of training. Intensity is strongly dependent on the motivation of the patient. Motivation can be increased by delivering variation within training and challenging exercises. However, patients are often underchallenged, as exergame difficulty often does not match the patient's ability. A Rasch analysis can establish hierarchy of exergame items in order to assist the delivery of patient-centered therapy.

Objective: The aim of this study was to apply the Rasch model to create a hierarchical order of existing VR balance exergames and to relate these exergames to the abilities of persons with neurological disorders, in order to deliver challenge and variation.

Methods: A total of 30 persons with stroke and 51 persons with multiple sclerosis (MS) were included in the study. All participants performed a training program, lasting 3 weeks for persons with MS and 4 weeks for persons with stroke, in which they performed VR balance exergames with a movement recognition-based system (MindMotion GO; MindMaze SA). VR exercise scores, Berg Balance Scale scores, and clinical descriptive data were collected. Berg Balance Scale and device scores were analyzed with the Rasch model using a repeated-measures approach to examine whether the distribution of exercise scores fitted the Rasch model. Secondly, a person-item map was created to show the hierarchy of exercise difficulty and person ability.

Results: Participants completed a selection of 56 balance exercises (ie, items), which consisted of a combination of various balance tasks and levels (ie, exercises). Using repeated measures, this resulted in a count of 785 observations. Analysis showed strong evidence for unidimensionality of the data. A total of 47 exercises (ie, items) had a sufficiently good fit to the Rasch model. Six items showed underfit, with outfit mean square values above 1.5. One item showed underfit but was kept in the analysis. Three items had negative point-biserial correlations. The final model consisted of 47 exercises, which were provided for persons with low to moderate balance ability.

Conclusions: The VR exercises sufficiently fitted the Rasch model and resulted in a hierarchical order of VR balance exercises for persons with stroke and MS with low to moderate balance ability. In combination with the Berg Balance Scale, the results can guide clinical decision-making in the selection of patient-focused VR balance exercises.

Trial registration: ClinicalTrials.gov NCT03993275; https://ichgcp.net/clinical-trials-registry/NCT03993275.

Keywords: Rasch analysis; balance; digital therapeutics; exergaming; multiple sclerosis; neurorehabilitation; stroke; virtual reality.

Conflict of interest statement

Conflicts of Interest: JK declares that part of this work was funded by MindMaze SA, Switzerland. The study was carried out independently and there was no influence from the company on the design, analysis, interpretation, and presentation of the results. EW is employed at Hocoma Medical GmbH, Switzerland, a manufacturer of rehabilitation robotics.

©Evelyne Wiskerke, Jan Kool, Roger Hilfiker, Karl Martin Sattelmayer, Geert Verheyden. Originally published in JMIR Serious Games (https://games.jmir.org), 22.03.2022.

Figures

Figure 1

A patient performing exercises with the device under investigation. Exercises in the standing position were performed without aid or physical support, and a chair was placed at arm’s length of the participant for security.

Figure 2

Details of the various exergames. Games are divided into static sitting and standing balance exercises, weight shifting in standing exercises, and dynamic standing balance exercises. Games are described using the following parameters, where a green checkmark indicates the game contains it, while a red X indicates it does not: high precision—high precision of movement is needed in order to steer the avatar well; speed predefined—the speed is constant within the game and cannot be influenced by the player; speed increase over levels—the predefined speed increases with higher level; obstacles in levels 2 & 4 or levels 6, 8, & 10—whether obstacles occur in these levels or not; moving obstacles—obstacles move from left to right or up to down, interfering with the players trajectory; high cognitive demand—the game contains elements such as go-no-go reactions or choices between collectables with different point counts. FBW: forward-backward step.

Figure 3

Bubble plot with the outfit mean square statistics. Game descriptions can be found in Figure 2. Outfit mean square values should range from 0.5 to 1.5; items below 0.5 show overfit and items above 1.5 show underfit. The size of the bubble shows the model SE. The number beside each exergame represents the exergame level, and the ● denotes the games that are performed in a seated position. The number beside BBS represents the scale item number. BBS: Berg Balance Scale; Cross FBW: Cross the Road (forward-backward step); Cross Free: Cross the Road (free steps); Garden: Veggie Guard; LineR Bi: Line Roller (bilateral); LineR Uni: Line Roller (unilateral); Ski: Skiline.

Figure 4

Person-item map. This map shows the person ability and item difficulty in one scale expressed in logit values. Person ability is shown on the left side, with the lowest person ability at the bottom and highest at the top. Item difficulty is shown on the right side, whereby items are organized from least difficult at the bottom to most difficult at the top. The number beside each exergame represents the exergame level, and the ● denotes the games that are performed in a seated position. The number beside BBS represents the scale item number. BBS: Berg Balance Scale; Cross FBW: Cross the Road (forward-backward step); Cross Free: Cross the Road (free steps); Garden: Veggie Guard; LineR Bi: Line Roller (bilateral); LineR Uni: Line Roller (unilateral); M: mean; S: one SD; Ski: Skiline; T: two SD.