Using game performance measure to assess upper limb motor function in children with neuromotor disorders: a systematic review

Abstract

Background

Upper limb impairments affect children with neuromotor disorders, limiting daily activities and participation. Videogames (ex: virtual reality, computer-based games) have emerged as promising tools for upper limb rehabilitation, offering better engagement and effects on upper limb movements. The capacity of video games to be used for assessing upper limb function with game-based metrics has not yet been explored, and their validation remains unclear. This systematic review aims to determine if game-based metrics can serve as relevant measures to evaluate upper limb impairments in children with neuromotor disorders.

Methods

A systematic review (PROSPERO: CRD42024550469) was conducted in PUBMED, MEDLINE, Web of Science, and Cochrane according to PRISMA guidelines. Articles published from inception to the 26th of May 2025 were screened according to inclusion/exclusion criteria and data were extracted focusing on game characteristics, outcomes measured, and their measurement properties.

Results

After screening, 26 studies on 1092 articles were included. In total, 443 children, mostly with cerebral palsy (n = 394), mean age 10.1 ± 2.4 years, underwent videogames with quantitative measurements. A total of 112 upper limb game-based outcomes were identified, measured either from the game itself or external instruments. These outcomes included kinematics (60%), game scores (24%), actimetry data (10%), electromyography (5%), and kinetics (>1%). Only 3 studies reported reliability data, with test–retest reliability ranging from poor to excellent across 7 outcomes. For discriminant validity, 8 studies included control participants, assessing 20 outcomes, of which 13 demonstrated the ability to differentiate between groups. Regarding responsiveness, pre-/post-therapy component was investigated through 12 studies across 52 outcomes. Only 29 showed significant improvements after intervention. Convergent validity, explored in 6 studies, reported moderate to high correlations with clinical assessments.

Conclusion

The present systematic review identified a wide range of game categories and upper limb game performance outcomes, involving different instrumented tools, and covering an interesting range of gestures. Despite the relevance of the game context and the use of instrumented outcomes, there is a lack of validation. To be used as research outcomes and to guide therapies in clinical practice, important work on the measurement properties has to be done.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12984-025-01699-5.

Introduction

Children with neuromotor disorders (ND), such as cerebral palsy (CP), often experience significant difficulties using their upper limb (UL) due to neurological deficits (e.g. motor deficits, spasticity, motor control impairment, and difficulties with motor planning and coordination) [1, 2]. These impairments significantly impact their function, reducing their ability to perform daily activities, and limiting their autonomy and overall quality of life [3]. Assessing these impairments is essential for therapeutic management to design targeted interventions, monitor progress, guide clinical decisions, and ensure meaningful outcomes. However, it remains difficult due to several factors such as the complexity of the UL function (many degrees of freedom, segment coordination, etc.), the diversity of tasks (bi- vs unimanual, gross vs fine motor tasks, ecological meaning), or even the heterogeneity of impairments in children with ND [4, 5].

Most of the time, UL function is assessed in a standardised/controlled environment which does not reflect what the children really do in the daily-life environment [4–6]. These concepts are defined as CAPACITY (i.e., an individual’s ability to execute a task or an action) and PERFORMANCE (i.e., what an individual does in his or her current environment), respectively according to the International Classification of Functioning, Disability, and Health [7]. The environment in which movements are performed plays a crucial role in shaping motor behaviour, influencing the quality and variety of gestures a child can achieve [8]. To reduce the gap between CAPACITY and PERFORMANCE, a gamification of the clinical UL assessment can be used. Some clinical assessments currently used in routine practice take the form of games (e.g. Assisting Hand Assessment – AHA – the bimanual performance-based score in children with CP) [9] and also instrumented assessments [10, 11]. These types of assessments allow more spontaneous movements and make evaluation more engaging and accessible for children. However, a major limitation of these assessments is their subjectivity. Many rely on observational scoring, which can vary depending on the evaluator and may not always reflect the child’s functional abilities outside the clinical setting [6].

In recent years, video games have been increasingly integrated into rehabilitation programs for children with ND [12]. They offer an engaging and interactive way to encourage movement, improve motor skills, and maintain motivation throughout therapy [12]. They allowed more spontaneous movements and the potential for more dynamic measures [13, 14]. Moreover, recent systematic reviews reported a better improvement in the UL function using video games than the conventional approaches [12, 15, 16]. In addition to serving as a clinical tool for rehabilitation, performance during video games (console games, virtual reality, computer-based games, etc.) (i.e. game performance) has also been used to assess UL characteristics in children with ND [17–20].

Given their interactive nature and ability to track movement, we hypothesized that video games could serve as valuable tools for objective motor assessment to monitor their condition and progress [18, 20]. They could provide quantitative data on movement patterns, range of motion, and coordination, offering a more comprehensive and ongoing evaluation of UL function [18]. However, the relevance and validation of these game-based metrics as outcome measures for UL impairments in children with ND remain unclear.

The purpose of this systematic review was to examine outcome measures and their measurement properties used to evaluate UL function during instrumented video games (game performance outcomes) in children with ND. Specifically, this review aimed to (1) describe the types of games used; (2) identify the methods employed for data collection and analysis (including outcomes, protocols, and instrumented tools); (3) summarize the available evidence of validation of these measures (reliability, validity, and responsiveness).

Methods

Reglementary issues

Before initiation, the protocol of this systematic review was registered on PROSPERO (CRD42024550469). The review was conducted and reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses 2020 (PRISMA) guidelines [21]. The PRISMA checklist are presented in Additional Data 1.

Search strategy

Searches were conducted in English in the electronic databases of Medline, Cochrane Library, Embase, and Web of Science from inception to the 26th of May 2025. The following combinations of keywords were used: (1) pathology: motor disorder OR nervous system diseases, (2) population: Child OR Teen OR Infant OR Adolescent, (3) body regions of interest: Upper Extremity OR Upper limb OR wrist OR hand OR finger OR forearm OR arm OR shoulder OR elbow and (4) game instrument: Game OR toy OR virtual reality OR instrumented OR Play OR exergaming OR software OR computer simulation OR data display. Search strings were formulated and tailored to the search syntax of each database to ensure a common search strategy (Additional File 2).

Study selection, inclusion, and exclusion criteria

The search results were imported into the Rayyan online application [22] which identified duplicates. Duplicates were checked individually to be removed. Two independent reviewers screened the title, abstract, and full text for inclusion of the studies in the analysis. Any disagreements were resolved through discussion until a consensus was reached. The inclusion/exclusion criteria were identified according to the PICOS framework (Table 1). Studies that included both typically developing (TD) children and children with ND were included, but the data from the TD children were not analysed.

Table 1.

Inclusion/exclusion criteria using the PICOS framework

PICOS	Inclusion	Exclusion
Population	• Children and adolescents (0–18 years old) • Neuromotor disorders (e.g. cerebral palsy, stroke, etc.)	• Neuromotor disorders without impact on UL function (ex: spastic paraplegia or lower limb impairments only)
Intervention	• Involving instrumented game, toy or tools (console game, virtual reality, computer-based interface, etc.)	• No specific UL evaluation • No oriented upper limb task (e.g. locomotion)
Comparison	/	/
Outcome	• All score or movement outcomes related to game performance • Obtained by instrumentation • Obtained during a game • Quantitative measures of UL characteristics (ex: spatiotemporal, kinematics, kinetics)	• No game-based metrics reported (e.g. game only use as intervention) • Outcome evaluated/calculated by clinician (ex: count of trial) • Not acquired during the game (ex: before or after the game)
Study design	• Full-text articles (psychometric studies, quasi-experimental studies, randomized controlled clinical trials, etc.)	• Articles written in a language other than French or English • Case report, book chapter, opinion letters, literature review, study protocol or conference abstracts • Case studies with participants n < 6, only descriptive

UL: Upper Limb

Quality assessment

To assess the methodological quality of the eligible studies and the quality of the game-based measure, a customized quality assessment scale of 20 items was developed using the modified McMaster Critical Review Form for Quantitative Studies [23] and existing systematic reviews in the field of biomechanics [4, 24]. It also included items of the COSMIN checklist (reliability, validity, responsiveness) for evaluating the methodological quality of studies on measurement properties [25]. Each item was rated on a scale of 0 (absent/insufficient), 1 (partial), or 2 (clear). The detailed item list is presented in Additional File 3. To ensure consistency and quality, the assessment was performed independently by two independent reviewers. As done previously, any disagreements were resolved by consensus.

Data extraction

The two reviewers extracted all the data independently and a custom-made Excel (Microsoft Office, Microsoft, Redmond, WA, USA) data extraction form was built.

In the first stage, the items were selected and reported into the following categories: article information, population characteristics, game characteristics, outcome characteristics, and results.

Article information included the first author's name, date of publication and the study design. The population characteristics section was divided into two subsections: the pathological and control group. The pathology, number of participants, age, gender, dominant side, and the Manual Ability Classification System (MACS) [26] in children with CP were collected.

In the game characteristic section, the following game data were extracted: name, manufacturer name, description, system, interface, command, movement category, motor requirements (uni-/bilateral), anatomical level involved, familiarisation phase, procedure description and context of use. The movement category classification was inspired by previous work on video game and rehabilitation studies [13, 27, 28] and described in Table 2.

Table 2.

Description of game movement category classification

Category	Description
Pointing and targeting	Games that require precise finger or hand pointing, such as tapping or selecting targets on a screen
Path following	Games where players must trace or follow a defined path, testing coordination and accuracy along a trajectory
Reaching and grasping	Tasks requiring players to reach for and interact with objects within the game, often testing range and control of motion
Object manipulation	Games that involve picking up, rotating, or moving objects, which assess grip strength and dexterity
Sport game	Continuous tracking of moving objects, which involves fine motor control and sustained attention
Reaction and timing	Games requiring quick responses, often testing speed and coordination of movements
Rhythm and sequencing	Games that rely on repetitive or sequenced hand movements in rhythm, assessing motor timing and precision

The outcome characteristics included the measurement system, the category (e.g. game score, kinematics, kinetic, actimetry and electromyography (EMG)), type (e.g. spatiotemporal, quality), a description, anatomical level involved, impaired/not impaired side, and measurement properties.

The measurement properties (reliability, discriminant validity, responsiveness, convergent validity) were collected for each study, according to the COSMIN guidelines [25]. Reliability (ability of a test to provide the same measurement twice or between several raters) was evaluated. Articles in which children with ND were compared to TD children were considered to assess discriminant validity (ability to discriminate children with ND from another population). Responsiveness consisted in the ability of an outcome to detect change over time in the construct to be measured (e.g. pre- vs post treatment). Convergent validity consisted in correlation with a gold standard measure or comparison with other outcome measurement instruments. Clinical assessments and research results were also reported.

The data extraction was double-checked. In case of discrepancy, the articles were rechecked by the two authors until a consensus was reached.

Data analysis

A descriptive analysis was conducted on the sample characteristics, game characteristics, types of instruments, outcome measures, and measurement properties. Quantitative data are expressed as mean values with standard deviations (SD) and categorical results as numbers (%). While categorical data are presented as frequencies and percentages.

Results

Study selection

The search across the selected databases initially identified 1113 records, of which 21 were duplicates, resulting in a total of 1092 unique papers. The screening excluded 957 records based on the title and abstract. The full-text screening excluded 108 records resulting in 26 articles included in the analysis. The PRISMA flowchart is presented in Fig. 1.

Fig. 1.

PRISMA flowchart of search and selection strategy

Quality assessment

The mean quality assessment score of the articles included was 28.3 ± 4.0 (18 to 36) with a maximal score of 40. The score frequency by item is provided in Fig. 2.

Fig. 2.

Frequency for quality assessment scores by item

Characteristics of included studies

Among the selected studies, 15 were interventional studies (including 4 non-randomized control trials and 4 randomized control trials) and 11 were observational studies (including 3 analytical studies, 7 descriptive studies and 1 metrological study) (Table 3). A global overview of the included studies and the link with the extracted data (including pathologies, system and type of game, regions of interest, type of measurement system, and outcome type) are presented in Fig. 3. The Fig. 3 shows the number of studies associated with each category.

Table 3.

Description of included studies and population

N°	Study			Pathological group						Control group
	Authors	Study design	Procedure	Diagnosis	n	Age (year)	Sex ratio (M/F)	Dominant or preferred side	MACS level	Diagnosis	n	Age (year)	Female (%)	MACS level	Dominant or preferred side
1	Howcroft et al., [42]	Interventional Study	Playing of 4 videogames	CP (15 uCP, 2 bilateral CP)	17	9.7 ± 1.7	0.41	/	/
2	Peper et al., [35]	Interventional Study	9 h of videogame therapy over a 6 weeks period	uCP (spastic)	6	8.4 ± 1.8	0.5	Right = 4 Left = 2	Level I = 1 Level II = 4 Level III = 1
3	Bertucco et al., [48]	Interventional Study (Non-Randomized Control Trial)	A single experimental videogame session	Dystonia (CP, primary dystonia)	16 (8 CP)	13.7 ± 4.2	0.31	Right = 6 Left = 10	/	TD	15	9.7 ± 2.5	/	Not applicable	/
4	Chiu et al., [32]	Interventional Study (Randomized Control Trial)	6 weeks of home-based videogame therapy + usual therapy	uCP	32	9.4 ± 1.9	0.53	Right = 16 Left = 16	Level I-III = 21 Level IV-V = 11	uCP	30	9.5 ± 1.9	0.57	Level I-III = 21 Level IV-V = 9	Right = 15 Left = 15
5	Sandlund et al., [38]	Interventional Study	4 weeks of home-based videogame therapy	CP (8 uCP, 7 bilateral CP) (13 spastic, 1 dyskinetic, 1 ataxic)	15	10.6 ± 2.9	0.47	/	Level I = 7 Level II = 6 Level III = 1 Level IV = 1
6	Weightman et al., [44]	Observational Study (Descriptive)	Videogame therapy	Spastic CP (5 uCP, 2 bilateral CP)	7	6.4 ± 0.8	0.14	Right = 2 Left = 5	/	TD	9	7.6 [7;9]	0.00	Not applicable	/
7	Shin et al., [51]	Observational Study (Descriptive)	Performance of a hand function test using touch screen technology	Children with neuromuscular diseases (CP & others)	45 (26 CP)	/	0.4	Right = 31 Left = 14	/	NHF	113	< 5 y (13) 5–6 y (14), 7–8 y (11), 9–11 y (33), 12–19 y (42)	0.50	Not applicable	Right = 100 Left = 13
										CHD	43	/	0.58	Not applicable	Right = 28 Left = 15
8	Zoccolillo et al., [46]	Observational Study (Descriptive)	Videogame therapy + usual therapy	uCP	8	6.5 ± 1.6	/	Right = 5 Left = 3	/
9	Preston et al., [50]	Interventional Study (Non-Randomized Control Trial)	Videogame therapy in single-user mode or dual-mode (with school friends)	CP (10 uCP 1 bilateral CP)	11	9.0 ± 1.9	0.27	/	/
10	Van Hedel et al., [39]	Observational Study (Descriptive)	Videogame session with a glove-based system	Brain lesions (CP, stroke, encephalitis, traumatic brain injury, others) (12 bilateral)	33 (15 CP)	12.6 ± 3.6	0.33	Right = 26 Left = 7	Level I = 14 Level II = 12 Level III = 5 Level IV = 2
11	Wilcox et al., [45]	Observational Study (Descriptive)	A structured in-clinic play session (2 toys, 2 videogames)	Upper extremity motor impairments (CP, DCD, chromosomal disorder, stroke, spina bifida, traumatic brain injury)	21 (9 CP)	8.5 ± 2.3	0.42	/	/
12	Keller et al., [19]	Interventional Study	Videogame Therapy	CP (4 uCP, 8 bilateral CP) (7 spastic, 2 ataxic,4 mixed)	13	12.9 ± 3.3	0.18	Right = 6 Left = 7	Level I = 2 Level II = 9 Level III = 2
13	Chen et al., [30]	Interventional Study (Non-Randomized Control Trial)	Home-based videogame therapy	CP (5 uCP, 2 bilateral CP; spastic)	7	9.9 ± 1.3	0.57	Right = 3 Left = 4	Level I = 1 Level II = 4 Level III = 2	TD	10	9.60 ± 1.26	0.7	Not applicable	Right = 10
14	Robert et al., [37]	Observational Study (Analytical)	3 gestures in virtual reality and matched physical environment	uCP	10	13.9 ± 2.1	0.10	Right = 5 Left = 5	Level I = 5 Level II = 5	TD	17	13.0 ± 2.1	0.53	Not applicable	Right = 14 Left = 3
15	Almeida Martins et al., [47]	Interventional Study (Non-Randomized Control Trial)	Pre-post-test virtual coincident timing task and real task	uCP (spastic)	10	6–19	/	/	/	TD	10	6–19	/	Not applicable	/
16	García-Hernández et al., [41]	Interventional Study	Reaching-transporting virtual objects with and without shoulder Kinesio Taping	CP	10	7.8 [7; 12.8]	/	/	/	CP	10	8.4 [6;13]	/	/	/
17	Leal et al., [49]	Observational Study (Analytical)	Videogame (concrete or abstract interface)	CP 2 uCP, 26 bilateral CP)	28	11.1 ± 2.3	0.25	/	Level I = 1 Level II = 23 Level III = 4	TD	28	11.1 ± 3.3	/	Not applicable	/
18	Kaya Cidi et Yilmaz, [43]	Observational Study (Descriptive)	2 exercise sessions with series of videogames	CP (9 uCP; 11 bilateral CP)	20	9.3 ± 3.7	/	/	/
19	Macintosh et al., [34]	Interventional Study	4-week home-based videogame therapy	uCP (11 spastic, 7 mixed tone, 1 dystonic)	19	11.7 ± 2.5	0.53	/	Level I = 12 Level II = 7
20	Adams et al., [40]	Observational Study (metrological)	Videogame therapy (affected and non-affected hand)	uCP	10	8 [4; 9]	0.40	Right = 5 Left = 5	/
21	Bautista et al., [29]	Interventional Study	10 weeks of videogame therapy + usual therapy	CP	13	13.8 ± 3.5	0.38	Right = 7 Left = 6	Level I = 3 Level II = 4 Level III = 2 Level IV = 3 Level V = 1
22	Kanitkar et al., [33]	Interventional Study (Randomized Control Trial)	A 16-week intervention program (videogame therapy or task-specific training (CIMT))	CP	33	7.3 ± 2.1	/	/	Level I = 4 Level II = 15 Level III = 8	CP	30	7.8 ± 1.9	/	Level I = 3 Level II = 16 Level III = 11	/
23	Tresser et al., [52]	Observational Study (Analytical)	A single 30-min videogame session	CP	20	9.4 ± 2.1	0.35	Right = 13 Left = 7	/	TD	60	7.8 ± 1.1	0.51	Not applicable	Right = 60
24	Baillet et al., [17]	Interventional Study (Randomized Control Trial)	Videogame therapy + usual therapy	CP (6 uCP, 4 bilateral CP)	10	14.3 ± 2.4	0.4	/	Level I = 0 Level II = 3 Level III = 3 Level IV = 4	CP	10	13.8 ± 2.0	0.3	Level I = 1 Level II = 2 Level III = 3 Level IV = 4	/
25	Cheng et al., [31]	Observational Study (Descriptive)	Visuomotor integration tasks in touchscreen, and two VR conditions	CP (10 uCP, 2 bilateral CP)	12	10.0 ± 2.4	0.67	Right = 6 Left = 6	Level I = 1 Level II = 4 Level III = 3 Level IV = 2 Level V = 2
26	Peramalaiah et al., [36]	Interventional Study (Randomized Control Trial)	8 week (3 session / week) of video game therapy	CP	17	7.2 ± 2.8	0.47	/	Level I = 6 Level II = 9 Level III = 6	CP	17	8.4 ± 2.5	0.47	Level I = 6 Level II = 5 Level III = 6	/

CIMT: Constraint Induced Movement Therapy; CP: Cerebral palsy; CHD: Congenital Hand Deformity; MACS: Manual Classification System; DCD: Developmental Coordination Disorder; NHF: Normal Hand Formation; y: years; uCP: unilateral Cerebral Palsy

/: not reported ; ± for mean ± standard deviation, [] for median [interquartile range]

Fig. 3.

Link between included studies and extracted data. Values correspond to the number of studies

Participants

A total of 443 children (39.8% of female, not reported in 4 studies) played video games with quantitative measurements. They were aged from 4 to 18 years (mean age 10.1 ± 2.4 years) and diagnosed mostly with CP (n = 394). Among all children, 190 had unilateral deficits and 71 had bilateral one (including 143 unilateral CP and 59 bilateral CP), while no information reported for 192 children. Regarding children with CP, 61 had spastic, 10 dyskinetic, 3 ataxic, and 11 mixed CP forms. Among them, 195 children were classified as MACS levels I to III and 26 children were classified as MACS levels IV to V [17, 19, 29–39] (Table 3).

Games description

Among the 26 studies, 46 different games were played by children, including 27 research-designed video games (developed for computer, mobile application or medical devices) and 19 commercial video games, mainly from Nintendo^® Wii Sports^™ (Nintendo Entertainment Analysis & Development, Kyoto, Japan), and Microsoft Kinect (Microsoft Corporation, Redmond, United States of America). Only 2 studies used immersive virtual reality settings (e.g., head-mounted display), and are the most recent ones[17, 31]. Regarding the type of movements, “Reaching & Grasping” and “Sports” represent 30.2% each followed by “Pointing & Targeting” at 20.4%, “Path following” at 14.3%, “Reaction & timing” at 8.2% and “Object manipulation” at 4.1%. No game matched with the category “Rhythm and sequences”. Games were played in a laboratory setting in most studies (81.6%) but also at school (6.1%) and at home (12.2%). The games involved mainly unimanual tasks (63.3%) and the full UL (97.8%). The game characteristics are presented in Table 4.

Table 4.

Games description

Authors	Name (manufacturer)	Description	System	Interface	Command	Movement category	Motor requirements	Anatomical level involved	Familiarisation phase	Procedure	Context
Howcroft et al., [42]	Wii Sports: Bowling, (Nintendo)	Bowling	Commercial gaming console (Wii®)	Screen	Controller	Sports	Unimanual	Full UL	Yes	A single session in each game was played for 8 min at beginner level with 5 min of rest between games	Lab
	Wii Sports: Tennis, (Nintendo)	Tennis	Commercial gaming console (Wii®)	Screen	Controller	Sports	Unimanual	Full UL	Yes		Lab
	Wii Sports: Boxing, (Nintendo)	Boxing	Commercial gaming console (Wii®)	Screen	Controller	Sports	Unimanual	Full UL	Yes		Lab
	Wii Sports: Dance, (Nintendo)	Dance	Commercial gaming console (Wii®)	Screen	Controller	Sports	Unimanual	Full UL	Yes		Lab
Peper et al., [35]	Penguin	Fetch as many fish as possible. By moving the hands in antiphase (180°), a penguin could be steered along a bridge, to fetch a fish at the other side and take it to the igloo	Computer	Screen	Joystick	Reaction & timing	Bimanual	Full UL	Yes	An ABA-design including 4 sessions of assessment	Lab
Bertucco et al., [48]	/	To touch and ‘‘burst the bubbles."	Application (IOS)	Tablet	Touchscreen	Pointing & Targeting	Unilateral	Full UL	/	A 1 h single session composed of 4 blocks of 45 targets	Lab
Chiu et al., [32]	Wii Sports: Bowling, (Nintendo)	Bowling	Commercial gaming console (Wii®)	Screen	Controller	Sports	Bimanual	Full UL	/	Three sessions at baseline (before intervention), at six weeks (after intervention), and at 12 weeks (six weeks beyond the intervention). Each game was practiced for 10 min resulting in each session taking 40 min	Home
	Wii Sports: Air sports, (Nintendo)	Air Sports	Commercial gaming console (Wii®)	Screen	Controller	Sports	Bimanual	Full UL	/		Home
	Wii Sports: Frisbee, (Nintendo)	Frisbee	Commercial gaming console (Wii®)	Screen	Controller	Sports	Bimanual	Full UL	/		Home
	Wii Sports: Basketball, (Nintendo)	Basketball	Commercial gaming console (Wii®)	Screen	Controller	Sports	Bimanual	Full UL	/		Home
Sandlund et al., [38]	Eyetoy: Play3—Maestro game, (Sony Interactive Entertainment)	In this game, the child acts as the maestro of an orchestra. When playing the game, rhythm icons move up towards one of the five targets arranged in an arc at the top of the screen, when an icon passes through the target, the player should place his or her hand over that target	Commercial gaming console (Playstation 3®)	Screen	Gesture-based	Reaching & grasping	Unimanual	Full UL	Yes	A pre- and post- 4 weeks intervention assessment was performed. The children played at beginner’s level, and it took 120 secto complete one game trial. A total of three trials of 120 s each were recorded	Lab
Weightman et al., [44]	Spaceship	The child controlled a ‘‘spaceship’’ collecting ‘‘satellites’’	Computer	Screen	Joystick	Reaction & timing	Unimanual	Full UL	Yes	A single session composed of a total of 30 discrete movements away and towards the body	Lab
Shin et al., [51]	Minnesota Hand Function Test app	The first hand function, the ability to touch an item on the screen, was designed as “dots,” a 3*4 grid of dots that light up in random order. As the dots are lit, the subject must be able to accurately touch them. Once a dot has been touched, another will light up and the task will be repeated	Application (IOS)	Screen	Touchscreen	Pointing & Targeting	Unimanual	Full UL	/	We measured 4 tasks: touching dots on a 3*4 grid, dragging shapes, use of the touch screen camera, and typing a line of text. The test takes 60 to 120 s and includes a pre-test to familiarize the subject with the format. Each task is timed independently and the overall time is recorded	Lab
	Minnesota Hand Function Test app	The second hand function, the ability to touch and drag an item on the screen, is designed as “shapes,” which is 4 shapes displayed on the screen with an outline of a shape in the centre of the screen. The subject must be able to touch and drag the shape accurately to match the shape in the centre of the screen	Application (IOS)	Screen	Touchscreen	Pointing & Targeting	Unimanual	Full UL	/		Lab
Zoccolillo et al., [46]	Kinect Adventures Package: Space pops, (Microsoft Game Studios)	Pop bubbles appearing in the virtual environment by touching them. Subject should flap arms for flying around the virtual environment, and put arms back down for descending	Commercial gaming console (Xbox®)	Screen	Gesture-based	Reaching & grasping	Bimanual	Full UL	/	They performed 16–30 min sessions, twice a week for 8 weeks. Assessment of outcomes was performed at the beginning (T0) and at the end (T1) of the first phase of treatment, and at the beginning (T2) and at the end (T3) of the second phase	Lab
	Kinect Adventures Package: 20.000 Leaks, (Microsoft Game Studios)	Stop the water from filling the tank by placing a hand, or a foot, or any other body parts over a leak opened on a wall of an underwater glass tank	Commercial gaming console (Xbox®)	Screen	Gesture-based	Reaching & grasping	Bimanual	Full UL	/		Lab
	Kinect Adventures Package: Rally Ball, (Microsoft Game Studios)	Balls will shoot down a lane towards the avatar, and the child must hit the balls back with a part of body and destroy targets at the end of the lane	Commercial gaming console (Xbox®)	Screen	Gesture-based	Reaching & grasping	Bimanual	Full UL	/		Lab
	Kinect Sports: boxing, (Microsoft Game Studios)	Boxing	Commercial gaming console (Xbox®)	Screen	Gesture-based	Reaching & grasping	Bimanual	Full UL	/		Lab
	Kinect Sports: volley, (Microsoft Game Studios)	Volley	Commercial gaming console (Xbox®)	Screen	Gesture-based	Reaching & grasping	Bimanual	Full UL	/		Lab
	Kinect Sports: bowling, (Microsoft Game Studios)	Bowling	Commercial gaming console (Xbox®)	Screen	Gesture-based	Reaching & grasping	Unimanual	Full UL	/		Lab
Preston et al., [50]	Pentagram	A series of aiming movements around a Pentagram shape, guided by a target that moves with each successful aiming motion from point-to-point	Computer	Screen	Joystick	Path following	Unimanual	Full UL	Yes	A cross-over design (AB–BA), where Group A played with school friends (dual-user mode) and Group B played by themselves (single-user mode). Each group played the games for four weeks at a time, each separated by a minimum of three weeks ‘‘wash-out’’ periods. 30 min of games were played each school day	School
	Figure of 8	Four timed tracking tasks: the children track as closely as possible a target circle moving in a horizontally positioned Figure of 8	Computer	Screen	Joystick	Path following	Unimanual	Full UL	Yes		School
	Tracing	Four untimed tracing tasks: identical shape, rotated 90° each time	Computer	Screen	Joystick	Path following	Unimanual	Full UL	Yes		School
Van Hedel et al., [39]	Airplane, (YouRehab)	Keep the airplane in the middle of the cloud-free path by performing bilateral arm movements: the airplane should be steered upwards with elbow flexion movements and downwards with elbow extension movements	Medical device (Yougrabber® system)	Screen	Instrumented accessory	Path following	Bimanual	Elbow	Yes	A single session composed of 90 s*3 conditions of game play with 30 s of rest between conditions	Lab
Wilcox et al., [45]	Bouncing ball, (NanoGames Inc)	The controller is used to aim a cannon that shoots coloured balls at a conglomerate of balls moving down from the top of the screen	Computer	Screen	Controller	Pointing & Targeting	Unimanual	Wrist	/	All children participated in a structured in-clinic play session, during which they played with 2 toys and 2 computer games for approximately 5 min each	Lab
	Snowman, (NanoGames Inc)	A downhill skiing game where the controller is used to steer a snowman left and right in order to collect snowman body parts (sticks, coal, carrots), while avoiding trees	Computer	Screen	Controller	Sports	Unimanual	Wrist	/		Lab
Keller et al., [19]	Moorhuhn, (tronic Software & Services GmbH)	Hunting birds of various sizes (and thereby various points): move and position the arm and hand quickly and accurately in the virtual environment and timely grasp the joystick to shoot at a chicken	Computer	Screen	Joystick	Reaching & grasping	Unilateral	Wrist	/	The participants played 70 trials (each trial lasted 1 min) of Moorhuhn during these 3 days. The distribution of playtime across the 3 intervention days was as follows: day 1: 4 blocks × 5 trials (total: 20 trials), day 2: 6 blocks × 5 trials (total: 30 trials), and day 3: 4 blocks × 5 trials (total: 20 trials)	Lab
Chen et al., [30]	Super Pop VR	“Pop” as many bubbles as possible by moving her or his arms	Computer	Screen	Gesture-based	Reaching & grasping	Bimanual	Full UL	Yes	One single session composed of two games of Super Pop VR™ without Darwin’s feedback (baseline), three games of Super Pop VR™ with Darwin’s feedback (acquisition), and another two games of Super Pop VR™ without Darwin’s feedback (extinction). Each game lasted 75 s and included around 10 reaching movements	Home
Robert et al., [37]	/	Trace three different trajectory paths to reach final targets and then to return their hand to the initial position. In virtual environments, the hand avatar was a fish controlled directly by hand movements	Computer	VR	Gesture-based	Reaching & grasping	Unimanual	Full UL	Yes	A single session composed of 15 trials of three gestures in each of virtual reality and a matched physical environment	Lab
Almeida Martins et al., [47]	Movehero, (University of Sao Paulo)	React (using the upper limbs) and not let the balls pass from the fixed targets. The spheres should only be intercepted when they reach the targets allocated in parallel	Computer	Screen	Gesture-based	Reaction & timing	Unimanual	Full UL	Yes	The individuals initially performed a pre-test with five attempts at the coincident timing task using the environment with tactile feedback. After the pre-test, they all performed the virtual task practice using MoveHero (without tactile feedback) for 8 min (4 songs of 2 min each). Immediately after practice using MoveHero, the individuals performed 5 more trials in the post-test on the same initial task. Then, we applied a modification of the task to assess performance adaptation	Lab
García-Hernández et al., [41]	Transporting virtual task	Reach with the hand three virtual balls, then place them into a box in front of trunk	Computer	Screen	Gesture-based	Reaching & grasping	Unimanual	Full UL	Yes	A single session composed of 2 pre-test blocks and 2 post-test blocks. Each block included 3 movements (different levels) of 30 s maximum	Lab
Leal et al., [49]	Check Limit Game, (University of Sao Paulo)	A task in which 96 bubbles (distributed in rows and columns) are on the screen and must be touched using finger contact or sliding the finger to other bubbles (touchscreen group) to change colour from green to light purple	Computer	Screen	Gesture-based	Pointing & Targeting	Unilateral	Full UL	/	The protocol was divided into three phases—Acquisition phase: with 30 trials (15 s each), the acquisition phase lasted 7 min and 30 s; after acquisition phase participants stayed 15 min of no contact with the task and started the Retention phase: with 5 trials performed with the same interface used in acquisition and lasted 1 min and 15 s; and Transfer phase: after retention participants executed 5 trials with change of interface to verify transfer and lasted 1 min and 15 s	Lab
Kaya Cidi et al., [43]	Wii sports: Island cycling, (Nintendo)	Island Cycling	Commercial gaming console (Wii®)	Screen	Controller	Sports	Bimanual	Full UL	Yes	Two sessions on two consecutive days including 10 min of play by game with 2 min of rest between games	Lab
	Wii sports: Tilt Table Balance Board, (Nintendo)	Tilt Table Balance Board	Commercial gaming console (Wii®)	Screen	Controller	Sports	Bimanual	Full UL	Yes		Lab
	Wii sport: Soccer heading, (Nintendo)	Soccer Heading	Commercial gaming console (Wii®)	Screen	Controller	Sports	Bimanual	Full UL	Yes		Lab
	Wii Sports: Boxing, (Nintendo)	Boxing	Commercial gaming console (Wii®)	Screen	Controller	Sports	Bimanual	Full UL	Yes		Lab
Macintosh et al., [34]	Dashy Square, (KasSanity)	Move forward while avoiding obstacles	Computer	Screen	Instrumented accessory	Reaction & timing	Unimanual	Forearm	Yes	4-week intervention composed of around 16 sessions (3 to 5/week) of 30 min of game	Home
Adams et al., [40]	Dragon Feeding Time	Capture the food using a magic wand and then bring it to the dragon requires a point-to-point reaching movement	Computer	Screen	Instrumented accessory	Reaching & grasping	Unimanual	Full UL	Yes	After three visits of familiarisation with games and clinical assessments, the children completed the games during a fourth visit. Games were performed twice starting with the right or left UL and repeating with the opposite side	Lab
	Dragon Flight School	Sensed wrist rotation angles translate to the left–right movement of a baby dragon as it attempts to fly through a series of targets	Computer	Screen	Instrumented accessory	Path following	Unimanual	Full UL	Yes		Lab
Bautista et al., [29]	Maze game	Move the ball in the maze using the joystick	Computer	Screen	Joystick	Path following	Unimanual	Wrist	/	10 sessions of 30 min gameplay	Lab
Kanitkar et al., [33]	Computer game-based Upper Extremity	Rotate the test object (instrumented accessory) and move the game paddle to catch the moving target objects	Computer	Screen	Instrumented accessory	Object manipulation	Bimanual	Full UL	/	A pre- and post-intervention assessment composed of the computer game-based upper extremity assessment tool	Lab
Tresser et al., [52]	VG4Rehab	A ‘reach to touch’ (hit the balloon) task that is played using movement of the upper limbs while the user faces a large screen	Computer	Screen	Gesture-based	Reaching & grasping	Unimanual	Full UL	Yes	A single 40 min session composed of a series of six 120 s trials of the iVG4Rehab	Lab
Baillet et al., [17]	Tracking task (OpenMind Innovation Unity Technologies, San Francisco, US)	Move an effector with an external device to keep it as close as possible to a moving target	Computer	VR	Controller	Path following	Unimanual	Full UL	/	A 4-week intervention with weekly sessions was conducted. Children performed pointing and tracking tasks in VR. The number of trials increased weekly (from 5 × 6 to 8 × 6). Difficulty (speed, space) was adapted individually based on task accuracy	Lab
	Pointing task (OpenMind Innovation, Unity Technologies, San Francisco, US)	Reach stationary targets in the virtual environment with an effector	Computer	VR	Controller	Reaching & grasping	Unimanual	Full UL	/		Lab
Cheng et al., [31]	Touchscreen	Touch the bulls-eyes shape target on the touchscreen	Computer	Screen	Touchscreen	Pointing & Targeting	Unimanual	Full UL	Yes	A single session during participants completed the clinical test and performed visuomotor impairment tasks in touchscreen, visually simple VR and visually complex VR conditions	Lab
	Simple VR	Touch the bulls-eyes shape target in the VR environment	Computer	VR	Controller	Pointing & Targeting	Unimanual	Full UL	Yes		Lab
	Complex VR	Explode the balloons appearing in the virtual classroom to explode them	Computer	VR	Controller	Pointing & Targeting	Unimanual	Full UL	Yes		Lab
Peramalaiah et al., [36]	Computer game-based Upper Extremity	Rotate the test object (instrumented accessory) and move the game paddle to catch the moving target objects	Computer	Screen	Instrumented accessory	Object manipulation	Bimanual	Full UL	/	A 8-weeks computer game-based exercise program (3 sessions per week) with a pre- and post-intervention evaluation with the computer game–based upper extremity assessment of manual dexterity	Lab

IOS: Iphone Operating System; Lab: laboratory; UL: upper limb; VR: virtual reality

/: not reported

Measurements

Acquisition system

In the 26 studies, outcomes were measured with external instruments in 15 studies [17, 19, 30, 31, 34, 37–46] or directly by the game in 16 studies [19, 29–33, 35, 36, 38, 39, 47–52]. Among the external instruments, motion capture technology was the most frequently used across all different systems: markerless (3 studies) [30, 40, 41], optoelectronic (3 studies) [38, 42, 44], laser-based (2 studies) [17, 31], rotational potentiometers (2 studies) [19, 45], wireless electromagnetic tracking systems (1 study) [37], and an electronic goniometer (1 study) [39]. Accelerometer (3 studies) [34, 43, 46], and EMG (2 studies) [34, 42] were also used to measure UL-related outcomes.

Outcome measures

Among all studies, a total of 112 outcomes were identified and are presented in Table 5. Thirty-eight outcomes were calculated by the game, 58 with the motion capture system (optoelectronic: n = 26; markerless: n = 17; electromagnetic: n = 6; laser-based: n = 6; rotational potentiometer: n = 2; electronic goniometer: n = 1). The most reported outcome was kinematics, with 67 outcomes (60%), including spatiotemporal (n = 29), movement quality (n = 17), range of motion (n = 15), and error parameters (such as path accuracy, n = 6). The next most frequent outcome was a game score, with 27 outcomes (24%), including quantitative (n = 10), spatiotemporal (n = 9), error (n = 7), and game-level change (n = 1) parameters. The actimetry category contained 11 outcomes (10%), including spatiotemporal (n = 7), energy expenditure (n = 2), activity count (n = 1), and quality (n = 1) parameters. Six outcomes were classified in the EMG category (5%) encompassing intensity (n = 5) and quality (n = 1) parameters. Only one outcome – force production – was classified as kinetic parameters (0.9%).

Table 5.

Description of the game performance outcomes

Articles	Game	Game-based performance outcomes
	Name	System type	Outcome category—type	Description	Dominant/non-dominant UL	Anatomical level
Internal measure
Peper et al., [35]	Penguin	Game	Game score—Quantitative	Number of fish fetched	/	Full UL
Bertucco et al., [48]	/	Game	Game score—Quantitative	Success: the number of target bubble reached (Score (S) = (2.5 target Index of Difficulty)/Movement time)	Impaired	Full UL
		Game	Kinematic—Spatiotemporal	Time: time of movement between touches
Chiu et al., 32	Wii® Sports Resort	Game	Game score—Quantitative	Bowling point	Impaired	Full UL
		Game	Game score—Quantitative	Air sport performance
		Game	Game score—Quantitative	Frisbee performance
		Game	Game score—Quantitative	Basketball performance
Sandlund et al., [38]	Maestro game (EyeToy: Play3 collection)	Game	Game—Quantitative	Game score point	/	Hand
Shin et al., [51]	Minnesota Hand Function Test app	Game	Game score—Spatiotemporal	Total completion time	/	Hand
Preston et al., [50]	Pentagram (CKAT system)	Game	Game score—Spatiotemporal	Path length	/	Full UL
		Game	Game score—Spatiotemporal	Path length time
		Game	Kinematic—Quality	Normalized jerk index: measurement of the smoothness and time taken for a discrete movement
	Figure of 8 (CKAT system)	Game	Game score—Spatiotemporal	Path length
		Game	Game score—Spatiotemporal	Path length time
		Game	Kinematic—Quality	Path accuracy: The movement trajectory compared against a reference trajectory; RMS mean is a value of mean error
		Game	Kinematic—Quality	Normalized jerk index
	Tracing (CKAT system)	Game	Game score—Spatiotemporal	Path length
		Game	Game score—Spatiotemporal	Path Length time
		Game	Kinematic—Quality	Normalized jerk index
		Game	Kinematic—Quality	Time/Path Accuracy (TPA): a product of path accuracy and path length time; TPA allows comparison of children who sacrificed speed for accuracy and vice versa
Van Hedel et al., [39]	Airplane	Game	Game score—Error	The percentage of time on the cloud-free path. The derivative of the ideal path was used in the correlation analysis, as the arm angle reflects the vertical velocity of the airplane	Both	Elbow
Keller et al., [19]	The exergame Moorhuhn	Game	Game score—Spatiotemporal	Total time, measured in seconds, needed to catch all 12 targets	Impaired	Wrist
		Game	Game score—Quantitative	Moving targets are large, medium, and small birds and account for 5, 10, and 25 points, respectively. Additional points can be earned/lost by hitting “special targets
Chen et al., [30]	Super pop VR	Game	Game score—Error	Percentage of successful reached	Impaired	Full UL
Almeida Martins et al., [47]	Movehero	Game	Game score—Error	The timing error: the difference between the time the ball hit the target sphere and the time the individual managed to hit the target with the avatar’s hand (real coincident timing). It was used to analyse the constant error (directional tendency of the movement); the absolute error and the variable error (accuracy of the movement)	Both	Full UL
Leal et al., [49]	Check Limit Game	Game	Game score—Quantitative	Number of bubbles reached	Impaired	Full UL
Bautista et al., [29]	Maze game	Game	Game score—Spatiotemporal	Path: distance travelled by the ball	Impaired	Wrist
		Game	Game score—Spatiotemporal	Time: time to complete the level
		Game	Game score—Error	Collision: the number of times the ball contacts the rim
		Game	Game score—Quantitative	Performance: time * (colision + 1); The less represents a better performance
Kanitkar et al., [33]	Computer game-based Upper Extremity (CUE)	Game	Kinematic—Spatiotemporal	Average Movement Onset Time (MOT): The time from target appearance to the start of the game paddle movement. Values for MOT time are determined for each game movement response, and then the average is computed over the group of game movement responses for each direction	Both UL	Full UL
		Game	Game score—Error	Success rate (SR): the percentage of the total number of target objects that were caught in one game trial
		Game	Kinematic—Error	Movement Error (ME): the magnitude of the error when a target is missed (distance between paddle and target position). The average value for all misses is then computed as the ME. Units are a percentage of screen width
Tresser et al., [52]	VG4Rehab	Game	Game score—Changes in game levels	Increase in game level—the number and frequency of game level changes	Impaired	Full UL
		Game	Game score—Error	Percentage of success of reaching
Cheng et al., [31]		Game	Game score—spatiotemporal	Trial Completion Time: The period (in seconds) from the start of a trial (i.e., the appearance of target) to the end of the trial, defined as a target touch in the hand-only task, both eye gaze and hand touch on target in the eye-hand task, or the expiration of the 60-s trial period	Dominant	Full UL
Peramalaiah et al., [36]	Computer game-based Upper Extremity (CUE)	Game	Game score—Error	Success rate (SR): the percentage of the total number of target objects that were caught in one game trial	Impaired	Full UL
		Game	Kinematic—Error	Movement Error (ME): the magnitude of the error when a target is missed (distance between paddle and target position). The average value for all misses is then computed as the ME. Units are a percentage of screen width	Impaired	Full UL
External measure
Howcroft et al., [42]	Wii® sport	Optoelectro-nic	Kinematic—ROM	Wrist flexion–extension, forearm pronation-supination, elbow flexion–extension, shoulder flexion–extension, shoulder abduction–adduction	Both UL	Full UL
		Optoelectro-nic	Kinematic—Spatiotemporal	Angular velocity of Wrist flexion–extension, forearm pronation-supination, elbow flexion–extension, shoulder flexion–extension, shoulder abduction–adduction
		Optoelectro-nic	Kinematic—Spatiotemporal	Angular acceleration of Wrist flexion–extension, forearm pronation-supination, elbow flexion–extension, shoulder flexion–extension, shoulder abduction–adduction
		EMG	EMG—Intensity	Mean muscle activity of trapezius, triceps, biceps, flexor carpi radialis, wrist extensor bundle
Sandlund et al., [38]	Maestro game (EyeToy®: Play3 collection)	Optoelectro-nic	Kinematic—Spatiotemporal	Mean velocity	Both	Full UL
		Optoelectro-nic	Kinematic—Spatiotemporal	Peak velocity
		Optoelectro-nic	Kinematic—Spatiotemporal	Relative timing of peak velocity: the point in time at which the peak velocity occurred in relation to movement duration
		Optoelectro-nic	Kinematic—Quality	Straightness: the ratio of the actual length of the 3-d path of the hand divided by the straight line joining the start and end point of the movement
		Optoelectro-nic	Kinematic—Quality	Precision: the volume (cm3) defined by the end-positions of the hand in the six reaches towards each virtual target
		Optoelectro-nic	Kinematic—Quality	Smoothness: a zero-cross index, normalised to time and calculated as the number of times the hand acceleration curve crossed zero
		Optoelectro-nic	Kinematic—ROM	Maximal shoulder angle was calculated as the rotation angle of the UL relative to the upper body, i.e. the helical angle of the shoulder
Weightman et al., [44]	Spaceship	Optoelectro-nic	Kinematic—Spatiotemporal	The joystick end-point path length	Impaired	Full UL
		Optoelectro-nic	Kinematic—Spatiotemporal	Total completion time: the game start time to the end time
		Optoelectro-nic	Kinematic—Quality	Smoothness: normalized jerk of joystick/elbow/trunk
		Optoelectro-nic	Kinematic—ROM	shoulder movement (protraction/retraction)
Zoccolillo et al., [46]	Kinect Adventures Package & Kinect Sports Package	Accelero-meter	Actimetry—spatiotemporal	RMS of the acceleration	Both	Forearm
Van Hedel et al., [39]	Airplane	Electronical goniometer	Kinematic—Error	Selective voluntary motor control (SVMC): how well the desired elbow movements corresponded with the actually performed elbow movements	Both UL	Elbow
Wilcox et al., [45]	Bouncing ball and snowman	rotational potentio-meter	Actimetry—activity count	Play threshold frequency: the number of wrist movements for which the peak flexion and extension values were greater than the thresholds (thus eliciting a response from the toy or game), divided by the time of play. This measure represents how many goal movements are met per second of play	Impaired	Wrist
Keller et al., [19]	The exergame Moorhuhn	rotational potentio-meter	Kinematic—Spatiotemporal	Average path-ratio, as performed path divided by the shortest possible one, needed to catch all 12 targets	Impaired	Wrist
Chen et al., [30]	Super Pop VR	Markerless	Kinematic—Spatiotemporal	Path length: the distance the wrist point travelled	Impaired	Full UL
		Markerless	Kinematic—Spatiotemporal	Movement time: the duration from popping the second bubble to popping the third bubble
		Markerless	Kinematic—Spatiotemporal	Number of movement units: each movement unit was defined as one acceleration and one deceleration phase
		Markerless	Kinematic—Spatiotemporal	Mean speed of the hand: average speed of the hand
Robert et al., [37]	Not reported	Electroma-gnetic	Kinematic—Spatiotemporal	Time to peak velocity	Impaired	Full UL
		Electroma-gnetic	Kinematic—Spatiotemporal	Movement time
		Electroma-gnetic	Kinematic—ROM	Shoulder flexion and horizontal abduction		Shoulder
		Electroma-gnetic	Kinematic—ROM	elbow extension		Elbow
		Electroma-gnetic	Kinematic—Quality	Trajectory straightness was measured with the index of curvature as the ratio of actual endpoint path length to that of a straight line joining initial and final positions, where 1 indicates an ideal straight line		Full UL
García-Hernández et al., [41]	Transporting virtual task	Markerless	Kinematic—Spatiotemporal	Hand peak velocity as the maximum first derivative of hand position data	Impaired	Full UL
		Markerless	Kinematic—ROM	Shoulder Flexion, horizontal adduction, and elbow flexion
		Markerless	Kinematic—Quality	Smoothness: hand jerk as movement smoothness at end-effector
		Markerless	Kinematic—Quality	Smoothness: angle jerk as movement smoothness at the joint level
		Markerless	Kinematic—Quality	Straightness: Hand travel distance ratio as the ratio between the length of the real hand path and optimal trajectory
		Markerless	Kinetic—Force	Total force production: sum of joint torques
		Markerless	Actimetry—Energy expenditure	Absolute work achieved
Kaya Cidi et al., [43]	Wii® sport	Accelero-meter	Actimetry—Energy expenditure	Metabolic equivalents (MET) values derived from the accelerometer	Dominant	Wrist
		Accelero-meter	Actimetry—Spatiotemporal	Activity counts were assessed using vertical axis (VA) counts per minute
		Accelero-meter	Actimetry—Spatiotemporal	Activity counts were assessed using vector magnitude (VM) counts
		Accelero-meter	Actimetry—Spatiotemporal	Exercise intensity: sedentary time with time monitoring, moderate-to-vigorous physical activity, and light physical activity
		Accelero-meter	Actimetry—Spatiotemporal	Monitoring time (wear time)
MacIntosh et al., [34]	Dashy Square (KasSanity Inc., Toronto, Canada)	Accelero-meter	Kinematic -Spatiotemporal	Change in arm movement: the difference in resultant angular velocity variability of the forearm between the 5 s before and 5 s after the speed-change biofeedback event was calculated	Impaired	Forearm
		Accelero-meter	Kinematic—Quality	Three indicators of task performance: - “dodge point rate” = dodge points accumulated per minute of play, linked to correct timing of a gesture) - “style point rate” = style points accumulated per minute of play, linked to co-contraction quality - number of “practice panels shown per minute” (biofeedback)
		EMG	EMG—Quality	Q uality of co-contraction: style point rate per 60-min practice
Adams et al., [40]	1. Dragon Feeding Time 2. Dragon Flight School	Markerless	Kinematic—Spatiotemporal	MC-SCT is the average time to complete each of the individual point-to-point movements in a therapy videogame session	Impaired	Full UL
		Markerless	Kinematic—Quality	MC-NS is intended to represent movement smoothness. For each point-to-point movement, the estimated mean speed of the kinematic data segment is divided by peak speed to generate the reported MC-NS parameter
		Markerless	Kinematic—Quality	MC-LDJ is a measure of smoothness based on rate of change in acceleration. For each kinematic data segment, the formula provided by Balasubramanian et al. is used to calculate the inverse natural logarithm of a dimensionless jerk parameter
		Markerless	Kinematic—Quality	MC-SPARC is a measure of smoothness based on the calculated arc length of the Fourier spectrum of each captured speed profile. Using this approach, a point-to-point functional movement can be considered analogous to a sound, with a smooth movement corresponding to a pure tone
Cheng et al., [31]	/	Laser-based	Kinematic—Error	Hand endpoint accuracy is the distance (in cm) between finger position and the centre of the target at the last frame of each trial	Dominant	Hand
Baillet et al., [17]	/	Laser-Based	Kinematic—ROM	Elbow flexion ROM	Non-dominant UL	Elbow
			Kinematic—ROM	Elbow involvement: the implication of elbow in three dimensions performed between the position of the target in each dimension (x, y, z) and the angle of the elbow to determine with which dimension of the target motion the elbow movements were most coupled		Elbow
			Kinematic—Quality	Elbow fluency measured by jerk calculation		Elbow
			Kinematic—error	Absolute target/effector distance (in meter) for pursuit task		Full UL
			Kinematic—spatiotemporal	Trial's time (in second) for the pointing task		Full UL

Articles	Clinical outcomes		Measurement properties				Research results
	Name	Results	Reliability	Discriminant validity	Convergent validity	Responsiveness	Research results
Internal measure
Peper et al., [35]	CROMS: AHA	The AHA results were mixed. Two children improved significantly, but at a group level, no significant effects were found. Baseline = 62.6 ± 023.5; end = 63.1 ± 23.5	/	/	/	✓	The corrected number of fish fetched per minute increased from 1.1 ± 0.8 in the first three sessions to 5.8 ± 2.3 in the last three sessions (t (5) = − 6.3; p < 0.005)
Bertucco et al., [48]	CROMS: Barry-Albright Dystonia scale	Barry-Albright Dystonia Scale (median [IQR]) = 2 [1: 3]	/	✓	/	/	CP had lower success than healthy children
			/	✓	✓	/	CP were slower than healthy children. Time showed a significant linear correlation with the Barry Albright dystonia scale score in children with dystonia: r = 0.329, f(1, 15) = 8.368, p < .05, 95% [0.024, 0.159]
Chiu et al., 32	CROMS: Nine-hole Peg Test and the JTTHF Coordination Clinical assessments: strength	No significant differences were found in grip strength or hand function between groups	/	/	/	✓	the participants’ Wii™ scores improved over the six weeks of training, this did not carry over to coordination or hand function. Wii™ training did not improve coordination, strength, or hand function
			/	/	/	✓
			/	/	/	✓
			/	/	/	✓
Sandlund et al., [38]	/	/	/	/	/	✓	At the post-training assessment, the children achieved, on average, 12% higher scores in the maestro game (p¼ 0.016 calculated with the Wilcoxon signed ranks test)
Shin et al., [51]	/	/	/	✓	/	/	When comparing children with normal hand development with those with NMD, in children aged 7 to 8 years with NMD took significantly longer total time; and those aged 12 years and older with NMD took significantly longer total time
Preston et al., [50]	CROMS: COPM PROMS: ABILHAND kids	On the ABILHAND-kids, 5 children improved, two deteriorated, and four showed no change. On the COPM, two improved, and nine showed no change. No significant differences were found (ABILHAND-kids, p = 0.424; COPM, p = 0.484)	/	/	/	✗	No difference between groups, p = 0. 445 (Friedman’s ANOVA) was found
			/	/	/	✗
			/	/	/	✗
			/	/	/	✗	There were kinematic improvements for Path Length (p = 0.022), Smoothness (p = 0.047) and Path Accuracy (p = 0.037) on the Figure of 8 task during the second deployment
			/	/	/	✗
			/	/	/	✗
			/	/	/	✗	No improvement was found from baseline to final assessment in any of the kinematic parameters except for TPA in the Tracing task (p = 0.007). On the Tracing task, Path Length for the second deployment (p = 0.028) and TPA for the first deployment (p = 0.022) were improved
			/	/	/	✗
			/	/	/	✗
			/	/	/	✗
			/	/	/	✓
Van Hedel et al., [39]	Clinical assessments: Active ROM, MAS, Manual Muscle Test (MMT), Trunk Control Measurement Scale (TCMS) and Test Of Non-verbal Intelligence (TONI-4)	Elbow flexion MMT scores were lower for the more affected side compared to the less affected side (median = 5.0; IQR = 3.8–5.0; p = 0.001)	/	/	✓	/	Friedman’s test showed significant differences between the game performance scores (i.e. the percentage of time on the cloud-free path) of different difficulty levels, i.e. different game conditions (p < .001) Total MMT scores of the more affected side (Youden Index 0.60) showed much smaller sensitivity and specificity than game performance (Youden Index was 0.76 for the % on the correct path as well as for the similarity between the ideal and actually flown trajectories)
Keller et al., [19]	CROMS: BBT, MA2	The BBT improved significantly (p = 0.008, d = 1.59). MA2 scores showed no significant difference but had a medium effect size (r = 0.40)	/	/	/	✗	The total time did not improve significantly (mean difference of week 1 =  − 0.91 s, SD = 3.69 s, versus mean difference of week 2 =  − 2.05 s, SD = 3.44 s; t(10) =  − 0.82, p = 0.433). The effect size was small (d = 0.32)
			/	/	/	✓	The average exergame score improved from 209.55 to 339.73 (p < 0.001, Cohen’s d = 1.80), The change in the Box and Block test improved from 0.45 (baseline week) to 3.95 (intervention week; p = 0. 008, d = 1.59)
Chen et al., [30]	/	/	/	✗	/	✓	No group differences All games during acquisition and extinction had larger “% successful reaches” than the 2 games during baseline
Almeida Martins et al., [47]	/	/	/	✓	/	✓	After practicing with MoveHero (no physical contact), only the CP group improved in both post-tests. However, they showed significantly lower precision and accuracy than the other group
Leal et al., [49]	/	/	/	✓	/	/	CP group presented worse performance than the control group in all phases of the study. CP group presented better performance in the abstract interface than in the concrete interface. Individuals from both groups were able to improve task performance and retain acquired information
Bautista et al., [29]	CROMS: BBT, MA2 Clinical assessments: somatosensory assessment (tactile and pressure pain thresholds)	The BBT showed a significant TIME·GROUP interaction, in 2 tasks (Reach Forward and Pronation/Supination) No significant effects on somatosensory measures	/	/	/	/	Decrease of the distance travelled by the ball (more efficiency of the path)
			/	/	/	/	A descriptive analysis of the task performance parameters (TIME, PATH, COLLISIONS, and PERFORMANCE) showed an increment of scores along the maze levels
			/	/	/	/
			/	/	/	/
Kanitkar et al., [33]	CROMS: Peabody Developmental Motor Scale-2 (PDMS-2): Grasp and Visual-Motor Integration (VMI) sub scores	Both groups improved in Grasp and VMI, with the experimental group showing greater PDMS-2 gains and larger effect sizes (0.69–0.8)	✓	/	/	✓	The CUE assessment has shown moderate to high test–retest reliability (0.17 < ICC < 0.83) There was a significant improvement observed in the majority of CUE object manipulation test scores for the experimental group (p < .01) with moderate to large effect sizes (0.50–1.2). No statistical difference in object manipulation test scores for control group (p > .01)
Tresser et al., [52]	/	/	/	✓	/	/	Higher increase level without weight on hand for both groups; Increase level was higher in TD
			/	✗	/	/	Higher %success without weight on hand but no difference between groups
Cheng et al., [31]	CROMS: Beery-Buktenica Visuomotor impairment,	/	/	/	✗	/	There were no differences in trial completion time between Simple and Complex VR No correlation between Beery-Buktenica scores and VR eye-hand coordination task performance was found
Peramalaiah et al., [36]	CROMS: Peabody Developmental Motor Scale-2 (PDMS-2): Grasp and Visual-Motor Integration (VMI) sub scores	The experimental group outperformed the control group, showing greater improvements in PDMS-2	✓	/	/	✓	Both groups showed significant improvements in the CUE assessment of manual dexterity, including success rates (tennis ball: p = .001; cone: p < .001; medicine ball: p = .001; and peanut ball: p < .001). In the CUE assessment, the experimental group demonstrated higher success rates (medicine ball: p = .001 and peanut ball: p = .02)
			✓	/	/	✓	Both groups showed significant improvements in the CUE assessment of manual dexterity, including movement errors (tennis ball: p = .01. In the CUE assessment, the experimental group demonstrated fewer movement errors (cone: p < .001)
External measure
Howcroft et al., [42]	/	/	/	/	/	/	Wii® bowling required the least wrist activity with significantly lower degrees of extension, flexion, and lateral deviation than in Wii® tennis (flexion and extension, p = 0.021). Elbow extension (of the dominant UL) was greatest during Wii® bowling, while DDR required significantly less elbow extension than the other games (p = 0.037)
			/	/	/	/	Angular velocities were significantly larger in the dominant UL than in the impaired UL during bilateral play
			/	/	/	/	Angular accelerations were significantly larger in the dominant UL than in the impaired UL during bilateral play
			/	/	/	/	Muscle activations did not exceed maximum voluntary exertions and were greatest for the boxing AVG and for the wrist extensor bundle
Sandlund et al., [38]	/	/	/	/	/	✗	The mean velocity decreased significantly in the physical task. The mean (F = 14.51, p = 0.000) were lower, on the non-dominant side in physical task
			/	/	/	✓	The peak velocity decreased significantly between pre- and post-training assessments in both the virtual and the physical task: the peak velocity (F = 7.35, p = 0.007) were lower, on the non-dominant side in the physical task
			/	/	/	✗	The relative timing of peak velocity showed no significant improvements post-training,
			/	/	/	✓	The movement straightness was significantly improved in the post-training assessment in the virtual task. The movements were less straight (F = 49.41, p = 0.000) on the non-dominant side in the physical task
			/	/	/	✓	The movement precision increased significantly after training in the virtual task (F = 15.40, p = 0.000). That is, the volume (cm3) defined by the end-positions of the hand in the six reaches towards each virtual target was significantly smaller at the post-test. In the physical task there were no corresponding significant improvements
			/	/	/	✓	Movement smoothness improved significantly after training in the physical task (F = 5.64, p = 0.020)., but did not increase significantly with practice in the virtual task
			/	/	/	✗	The maximal helical angle of the shoulder did not change between pre- and post-training assessments, in either the virtual or the physical task. The maximal helical shoulder angles were larger on the non-dominant side compared with the dominant side in both tasks (physical: F = 59.17, p = 0.000; virtual: F = 45.01, p = 0.000)
Weightman et al., [44]	CROMS: Movement Assessment Battery for Children (MABC)	MABC = 5 for all participants	/	✓	/	/	The joystick end-point path length was significantly greater in the CP group [F(2,22) = 4.527; p = 0.023]
			/	✗	/	/	The total time taken to complete the game was not statistically significant between the groups [F(2,22) = 2.905; p = 0.076]
			/	✓	/	/	There was a difference in the smoothness of the joystick control across the groups. A significant main effect of the group was found for a normalised jerk [F(2,22) = 7.111; p = 0.004]
			/	✓	/	/	Children with CP showed significant greater shoulder movement (protraction/retraction) than that shown by the adults and able-bodied children. [F(2,22) = 14.369; p < 0.001]
Zoccolillo et al., [46]	CROMS: QUEST and ABILHAND-kids, Visual-Motor Integration functioning scale (VMI)	QUEST scores improved during videogame therapy, mainly in grasping and dissociated movements, while ABILHAND-Kids scores improved after conventional therapy. Both improvements met minimal clinically important differences. No significant changes were observed in VMI scores during either videogame or conventional therapy	/	/	/	/	Hemiparetic side was moved less than healthy side (− 18 ± 8%, F = 39.303, P < 0.001). No significant effect of interactions between factors, except between therapy and side was found (F = 7.202, P = 0.028). Post-hoc analysis revealed that in VGT the paretic side was moved -20 ± 13% less than the other side (P = 0.001), while this difference was not significant in CT (− 10 ± 28%, P = 0.295)
Van Hedel et al., [39]	Clinical assessments: Active ROM, MAS, Manual Muscle Test (MMT), Trunk Control Measurement Scale (TCMS) and Test of Non-verbal Intelligence (TONI-4)	Elbow flexion MMT scores were lower for the more affected side compared to the less affected side (median = 5.0; IQR = 3.8–5.0; p = 0.001)	/	/	✓	/	Clinical motor and cognitive scores correlated moderately with SVMC (0.50–0.74). The TCMS correlated best with SVMC (r = 0.74). The SVMC measure varied widely among participants and did not differ between the conditions (more affected arm: p = 0.78; less affected arm: p = 0.44) and between the more and less affected side (0.36 ≤ p value ≤ 0.68). For each condition, the correlation between the ideal path and the actually flown path was significantly higher than the correlation coefficients reflecting our measure of SVMC (p < 0.001 for each condition)
Wilcox et al., [45]	AROM of wrist, Modified Ashworth Scale score	/	/	/	✓	/	A larger full ROM in wrist flexion and extension was associated with increased play threshold frequency, but the relationship was not significant (p = 0.09, R2 = 0.16). A significant correlation was found between play threshold frequency and maximum supination angle during AROM measurements (p.001, R2 = 0.585), where a greater ability to supinate was associated with a higher play threshold frequency. There was no correlation between MAS scores and play threshold frequency for all of the games and toys (p = 0.201 to 0.629)
Keller et al., [19]	CROMS: BBT, MA2	The BBT improved significantly (p = 0.008, d = 1.59). MA2 scores showed no significant difference but had a medium effect size (r = 0.40)	/	/	/	✗	For the path-ratio, a Wilcoxon signed rank test revealed no significant difference from week 1 (mean = 0.03, SD = 0.19) to week 2 (mean =  −  0.20, SD = 0.40; T = 20.50, p = 0.266) and the effect size was small (r = 0.24)
Chen et al., [30]	/	/	/	✗	/	✓	No group differences. All games during acquisition and extinction had faster speeds, and faster movement time than the 2 games during baseline
			/	✗	/	✓
			/	✗	/	✓
			/	✗	/	✓
Robert et al., [37]	Clinical assessments: Semmes-Wein- stein filaments, ROM, proprioception(Fugl-Meyer scale), spasticity (Tardieu)	See article results	/	✓	/	/	Only reliable kinematics for describing reaching in similarly aged children with CP were used (moderate to high test retest reliability) TD children made faster movements (97 ms, p = 0.042) for the sagittal gesture
			/	✓	/	/	Only reliable kinematics for describing reaching in similarly aged children with CP were used (moderate to high test retest reliability) TD children made faster movements (97 ms, p = 0.042) for the sagittal gesture
			✓	✓	~	/	Only reliable kinematics for describing reaching in similarly aged children with CP were used. (0.17 < ICC abduction < 0.51; 0.38 < ICC flexion < 0.93) The vertical gesture involved 24.4% more shoulder flexion (14.6°, p = 0.017) in TD children In virtual environments, greater sensory impairment (higher Semmes–Weinstein thresholds) was related to altered movement quality variables (frontal: less shoulder abduction, r = 0.70)
			✓	/	/	/	Only reliable kinematics for describing reaching in similarly aged children with CP were used (ICC > 0.87)
			✓	✓	/	/	Only reliable kinematics for describing reaching in similarly aged children with CP were used (0.59 < ICC < 0.81) Shoulder flexion was slightly more curved by 2.8% (p = 0.047) in children with CP
García-Hernández et al., [41]	CROMS: BBT	BBT scores showed that only intervention group obtained a significant increase from pre to post-test (t(9) = 0.9, p < 0.01)	/	/	/	/	No significant difference
			/	/	/	~	Improvements in meeting the required shoulder flexion angle. No improvement in other ROMs
			/	/	/	✓	Improvements in hand-movement smoothness
			/	/	/	✓	Improvements in arm joints smoothness
			/	/	/	✗	No significant difference
			/	/	/	✓	Improvements in joint force production
			/	/	/	✓	Improvement of energy expenditure
Kaya Cidi et al., [43]	/	/	/	/	/	/	The accelerometer counts accumulated per unit time obtained. from the wrist-worn data was 3.12 ± 0.86. The associated metabolic demand of performing video games exceeded the 3-MET moderate-to-vigorous physical activity threshold The cut-points determined and accelerometer counts per minute were significantly different between the wrist- and hip-worn outputs (p < 0.001). There was no difference in wrist-worn data between GMFCS levels
			/	/	/	/
			/	/	/	/
			/	/	/	/
			/	/	/	/
MacIntosh et al., [34]	/	/	/	/	/	✓	On average, participants reduced arm movement by 10.2 ± 4.0% in response to biofeedback (t18 = 7.68, p < 0.001, 95% CI =  − 11.9–− 8.5%). Participants continued to respond to the speed-change biofeedback across the intervention
			/	/	/	✓	Practice time was associated with: i. scoring points faster (p < .001, 95% CI 0.963–1.093, increase in dodge point rate per 60-min practice), ii. doing gestures with higher quality co-contraction (p < .001, 95% CI 22.555-30.042, increase in style point rate per 60 min practice), and iii. Seeing fewer practice panels (p < .001, 95% CI − 0.047–− 0.025, fewer practice panels shown per minute of play for every 60 min practice)
			/	/	/	✓	Practice time was associated with doing gestures with higher quality co-contraction (p < .001, 95% CI 22.555- 30.042, increase in style point rate per 60-min practice)
Adams et al., [40]	CROMS: JTHFT, WMFT, QUEST	/	/	/	✓	/	Moderate correlation with QUEST and WMFT (Time and FA) (p = 0.003–0.020). High correlation with JTHFT (p = 0.001)
			/	/	✓	/	Moderate correlation with JTHFT (p = 0.018)
			/	/	✓	/	Moderate correlation with JTHFT (p = 0.017) and WMFT (p = 0.026)
			/	/	✓	/	High correlation with JTHFT, QUEST and WMFT (Time and FA) (p < 0.001)
Cheng et al., [31]	CROMS: Beery-Buktenica Visuomotor impairment	/	/	/	✗	/	Hand endpoint accuracy was worse in VR compared to the touchscreen task for children with CP, potentially due to the lack of haptic feedback in VR No correlation between Beery-Buktenica scores and VR eye-hand coordination task performance was found
Baillet et al., [17]	CROMS: “manual dexterity” and the “ball skills” subtests of the MABC-2, BBT and CHEQ	A significant effect of Group for the “Total score of manual dexterity” in the MABC-2 test, showing a greater increasing of performances in post- and follow-up tests for the VR group, compared to the control group	/	/	/	✗	No significant changes were found for elbow range of motion in either group
			/	/	/	✓	A significant higher elbow involvement in the depth (z) dimension for the VR group in both post-test and follow-up evaluations was found
			/	/	/	✓	The VR group showed a significant 30% reduction in jerk values, compared to 1% reduction in the control group, with these improvements persisting after three months
			/	/	/	✓	The VR group showed a 56% reduction in target-effector distance after 12 sessions, maintained over time, versus only 9% in the control group
			/	/	/	✓	The VR group showed a 57% reduction in pointing task after 12 sessions, maintained over time, versus only 28% in the control group

AHA: Assisting Hand Assessment; COPM: Canadian Occupational Performance Measure; CP: Cerebral palsy; CROM: CHD: Congenital hand deformity; Consumer Reported Outcome Measures; CHEQ: Children’s Hand-use Experience Questionnaire; CUE: Computer game-based Upper Extremity; GMFCS: Gross Motor Function Classification System; JTHFT: Jebsen-Taylor Hand Function Test; MA2: Melbourne Assessment Test; MABC-2: Movement Assessment Battery for Children–Second Edition; METs: metabolic equivalents; MMT: Manual Muscle Test; NHF: Normal hand formation; NMD: neuromotor disorders; PDMS-2: Peabody Developmental Motor Scale-2; QUEST: Quality of Upper Extremity Skills Test; ROM: Range of Motion; SVMC: Selective voluntary motor control; TCMS: Trunk Control Measurement Scale; TD: typically developing children; TONI-4: Test Of Non-verbal Intelligence; TPA; UL: upper limb; VMI: Visual-Motor Integration; VR: virtual reality; WMFT: Wolf Motor Function Test

/: not reported

✓: positive result reported

✗: negative result reported

 ~ : mixed result reported

Measurement properties

In total, 23 studies reported one or more measurement properties of their game-based measures (Table 5).

Reliability

Of the 26 studies, only 3 studies [33, 36, 37], with 7 different outcomes, reported reliability results using the Intra-class Correlation Coefficient (ICC). Kanitkar et al. [33] and Peramalaiah et al. [36] reported a poor to high test–retest reliability (0.17 < ICC < 0.83 [53] of their outcomes (movement onset time, success rate and movement error), and Robert et Levin [37] reported moderate to excellent test–retest reliability [54] of 4 of their outcomes (movement straightness, elbow extension, shoulder abduction, and flexion). All these outcomes are described in Table 5.

Discriminant validity

Eight studies [30, 37, 44, 47–49, 51, 52] included healthy control participants, involving 20 outcomes. Among these outcomes, 13 demonstrated a significant discriminant validity (p < 0.05). All these outcomes are described in Table 5.

Responsiveness

Twelve studies [17, 19, 30, 32–36, 38, 41, 47, 50] were designed to evaluate a pre-post therapy effect, involving 52 outcomes. No other responsiveness components were reported. Only 29 outcomes showed significant differences after intervention (videogame therapies, etc.). All these outcomes are described in Table 5.

Convergent validity

Six studies [31, 37, 39, 40, 45, 48] reported correlations between investigated outcomes and existing clinical assessments, involving 12 outcomes. Significant moderate to high correlations for 5 outcomes and significant low correlations for one outcome were found with validated scores (e.g. AHA, MACS, Jebsen Taylor Hand Function Test, etc.). A significant moderate correlation was found for one outcome (play threshold frequency) with clinical examination measure (wrist supination range of motion) but not with clinical classification (MACS) [45]. The sensibility, specificity, and area under the curve from receiver operating analyses showed similar values of validity for the outcome “percentage of time on the cloud-free path”, and clinical assessments (manual muscle testing) [39]. All these outcomes are described in Table 5.

Discussion

This systematic review investigated the relevance of game performance outcomes as metrics in the analysis of UL impairments in children with ND. Results highlighted the use of a large variety of games mainly played by children, mainly with CP, aged from 4 to 18 years. The games were either commercial or specifically designed for research purposes or used as UL therapies. Various outcomes, either directly recorded from the game and/or measured with external instruments were used (mostly with motion capture system) to objectively quantify UL movements during games. Spatiotemporal variables were most often calculated, independently of the measurement system (game, markerless, optoelectronic, etc.). Game performance measurements showed moderate to good discriminant validity. Only a few studies investigated convergent validity but mainly showed moderate to good correlations with clinical assessments. However, reliability assessments were largely missing. Game scores and externally instrumented outcomes were used to evaluate UL therapy efficacy, yielding relevant results and highlighting their potential for both clinical and research applications.

Games

A wide variety of games were used through the selected studies, with approximately half consisting of commercial games and the other half being serious games. It has been reported that the two types of games show similar results in physical rehabilitation [12, 55], allowing the results to be interpreted without distinction despite more frequently demonstrating significant clinical outcomes linked to patient improvement with serious games [12, 28]. The types of games played primarily focused on the most commonly studied UL movements in motion analysis protocols: reaching and grasping [4, 5, 56]. However, game selection was mostly driven by expected rehabilitation improvements, often adopting a capacity-based approach (e.g., improving wrist motion). As a result, most UL movements explored were basic discrete actions (i.e. measuring the child's CAPACITY in a specific activity),[57] rather than being representative of daily activities—PERFORMANCE–, which often require bimanual coordination [58]. Despite some studies that have implemented their protocol in daily-life context (at school or at home) [30, 32, 34, 50], there is a lack of evaluation of PERFORMANCE using video games. Measuring PERFORMANCE is crucial for understanding the true impact of interventions. Future studies should prioritize assessments in a daily environment to capture meaningful improvements in everyday motor skills.

The game environment has the potential to facilitate more comprehensive motor engagement and spontaneity compared to other motion analysis protocols [4]. It can support the integration of the impaired UL by incorporating coordination, fine and gross motor skills, and dynamic control across a broad spectrum of gestures. Moreover, game engagement and user satisfaction are key factors that could influence game performance, i.e. speed, precision and regularity [34, 45, 52]. As previously noted, the environment can impact movement execution [59, 60]. Additionally, learning effects must also be considered when interpreting repeated UL assessments, particularly concerning studies using video games both as intervention and evaluation. An improvement in the game performance outcome could be related to a better comprehension of the game and not only an improvement of the UL function. Indeed, it was reported that exergames affect both function and cognitive performance [20].

Notably, games involving “Rhythm and Sequences” movements were absent from the selected studies. This may be due to the challenges of fine motor detection (e.g., hand movements) and the mixed cognitive and motor demands of these tasks, making it difficult to isolate pure motor gestures. A concurrent cognitive task can affect motor performance under dual-task conditions, as shown in gait studies [61]. However, assessing UL motor function in such conditions—particularly through game-based tasks—more closely reflects the real-life context in which gestures are performed. Another important consideration is the significant cognitive component involved in-game performance. Understanding game objectives, motor planning and organization, and information processing all contribute to gameplay, making it challenging to separate motor involvement from cognitive demands. Several studies have demonstrated a link between cognitive function and hand performance [62]. Nevertheless, the diversity of game categories presents both a challenge and an opportunity. While it limits the generalizability of the review’s results, it also offers significant potential for providing a more comprehensive assessment of UL function.

Acquisition system and outcomes

Among the 112 outcomes reported, motion capture emerged as the predominant system category, accounting for 60% of the measurements, and with varied measurement systems (motion capture systems, accelerometers, etc.). Twenty-five of the outcomes were measured with an optoelectronic system, highlighting the importance of movement in the assessment of UL function. However, it is important to note that only 3 studies used an optoelectronic system (Fig. 3). The proportion of outcomes measured by this system is high because one study [42] investigated the range of motion, angular velocity, and acceleration across multiple joints and planes, thereby increasing the number of outcomes in this category. Spatiotemporal parameters were predominantly analyzed regardless of the measurement system used, whether the game itself or external instruments. This is consistent with findings from systematic reviews of UL motion analysis protocols in children with ND, which identified spatiotemporal parameters as the most commonly used quantitative measures for evaluating and describing UL movements [4, 5]. Spatiotemporal parameters—such as movement duration, velocity, and smoothness—are directly related to UL performance [63], as they offer objective and quantifiable insights into motor control and coordination. These measures are particularly valuable in rehabilitation, where goal-oriented therapy focuses on improving the quality of movement to enhance children’s autonomy in daily activities [64]. On the other hand, a unique aspect of game-based protocols is the use of game scores, which are specifically designed for the games themselves. These scores introduced new quantitative measures not available in standard motion analysis protocols, offering promising potential for real-time tracking of movement progression. However, it is essential to demonstrate the clinical relevance of these scores, especially as numerous clinical assessments have been developed to evaluate UL motor function. It remains crucial to understand what these scores reflect and how they can be used to personalize and guide therapeutic interventions.

In terms of study frequency, “game” was the most frequently used, appearing in 14 of the selected studies (Fig. 3) and accounting for 34% of the measured outcomes. This proportion can be attributed to the accessibility of these systems, as they are low-cost, user-friendly, and directly integrated into the rehabilitation task (functioning as both an assessment and therapy tool). The integration of game-based acquisition methods contributed to the ongoing expansion of markerless tracking systems, contrasting with motion capture systems such as optoelectronic which require synchronization with the game, expertise, and expansive material, making them more complex to implement, especially in a daily environment (ex: at home).

Measurement properties validation

The results of this review show that most UL outcomes have not been fully validated according to the COSMIN guidelines [25], highlighting a lack of measurement properties validation. Several systematic reviews on UL motion analysis, with or without game, also reported the same lack of measurement properties validation [4, 5]. Discriminant validity and responsiveness (only one of its components: pre-post intervention) were the most frequently evaluated properties; most outcomes successfully differentiated between children with ND and control subjects across all types of games and UL movements. However, reliability has not been sufficiently assessed whereas the accuracy and interpretation of the measurements could be impacted especially considering the complexity of UL movements and the spontaneity of movements induced by the game setting. In this review, most studies reported changes in outcome values following various interventions (e.g., rehabilitation game therapy), suggesting that instrumented assessments may be effective indicators of therapeutic efficacy. Regarding convergent validity, the highest correlations were found for 5 outcomes with performance-based assessments used in clinical practice (ex: Jebsen-Taylor Hand Function Test [40]). Only the shoulder range of motion in the sagittal and frontal plane during a reaching task was validated for 3 measurement properties: a good moderate to high test–retest reliability, a significant difference between children with CP and control children (discriminant validity), and a good correlation with sensory impairments (Semmes–Weinstein thresholds) (convergent validity) [37]. Spatiotemporal parameters were the most frequently explored, but only a few have been validated: time of movement between touches demonstrated strong discriminant and convergent validity [48], while average movement onset time showed good reliability and responsiveness [33, 36]. This suggests that these measures may serve as more effective indicators of improvement than others, such as peak velocity or end-point path length also explored.

Limitations

This systematic review has some limitations. As previously mentioned, the selected studies involved mainly children with CP, the leading cause of motor disorders [1], so the results could not be generalized to all children with ND (i.e. stroke, traumatic brain injury, etc.). Secondly, the lack of standardization of games and systems made direct comparisons of results between studies challenging. However, the diversity of game categories represents both a limitation in generalizing the review’s results and a significant potential for providing a comprehensive assessment of UL function. Thirdly, the quality of the articles included ranged widely (from 18 to 35/40) reflecting a variable quality between studies. In addition to the lack of measurement properties discussed previously, one-third of the description of study methods was scored as “partial”, making it challenging to reproduce each gameplay protocol and analysis. Furthermore, no studies reported a priori sample size calculations, making it unclear whether they had sufficient statistical power to detect meaningful differences. Then, the age range of the children who were compared in this review was very wide but younger children could not be included because they are too young to use screens and videogames. In addition, despite its rapid expansion, the accessibility to new technologies is also a limiting factor in recommending this type of assessment in clinical practice [65]. Finally, we did not include the “grey literature” in the study search, which may limit the comprehensiveness of the systematic review results. However, focusing on peer-reviewed literature ensures a higher level of methodological rigour and quality.

Conclusion

The present systematic review identified a wide range of game categories and game performance outcomes, covering an interesting range of gestures. Despite their use, there is a lack of measurement properties validation. Nevertheless, developing video games that integrate advanced motion-tracking technology could reduce the gap between CAPACITY and PERFORMANCE, allowing clinicians to obtain reliable, real-time data while keeping children actively engaged in their rehabilitation therapy in daily environments. To be used as research outcomes and to guide therapies in clinical practice, important work on the measurement properties has to be done.

Abbreviation

ND

Neuromotor disorders

CP

Cerebral palsy

UL

Upper limb

ICF

International classification of functioning, disability, and health

TD

Typically developing

MACS

Manual ability classification system

AHA

Assisting hand assessment

ICC

Intra-class correlation coefficient

ROM

Range of motion

Author contributions

KRD wrote the main manuscript with substantial contributions from MC. KRD and MC performed data extraction and analysis. All authors contributed to the study design, data interpretation, and critical revision of the manuscript. All authors have read and approved the final version.

Funding

Open access funding provided by University of Geneva. This study was part of a project funded by the Prim’Enfance Foundation and the Research Fund of Geneva University Hospital.

Availability of data and materials

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.