Design
We used randomized controlled trials to assess the interventions. Participants were equally allocated into either the remodeled Chinese puppetry with the EV3® robot group or the standard puppet group. The research procedure for this trial followed the CONSORT 2010 guidelines and did not cause any harm to participants in the experimental group.
Participants
A total of 85 eligible participants aged 4-6 years were recruited from 16 preschools in Taiwan via school bulletins announcing recruitment information. The inclusion criteria were children diagnosed with developmental disabilities and a neuromuscular development index of ≤ 85 in the McCarron Assessment of Neuromuscular Development (MAND) test, indicating mild motor disability [15]. Children with severe cognitive impairment and visual disabilities, according to their disability cards in Taiwan, were excluded. All participants were screened by a research assistant; among them, 23 children with DD were excluded (19 children did not meet the inclusion criteria, one had severe medical problems, and four were rejected due to COVID-19 infection). A previous study showed that motor training can improve range of motion (ROM) (5-45°) from baseline to week 12, with a standard deviation (maximum degree) of 24°, as determined by sample size estimation in a clinical trial [6]. A clinically important difference of 0.5 and Cohen's d of 0.2 is considered acceptable. In this study, 31 children were recruited for each group according to the formula for calculating sample size [16]:


Sixty-two children with DD were randomly and equally assigned to either the experimental (n=31) or control group (n=31). The background information of the participants is presented in Table 1. The participants’ motor function was assessed using the McCarron Assessment of Neuromuscular Development (MAND) test [16]. The MAND test is used for children aged 3.5-16 years with motor problems, and comprises 10 items, namely, five for fine motor and five for gross motor.
Procedures
According to the experimental research design, pretests and posttests were conducted to determine the improvement in fine motor performance by operating the two kinds of glove puppetry over 12 weeks. We conducted the pretests on the 1st day of the 1st week of research, and the posttests on the last day of the 12th week of research. The experimental and control groups operated glove puppetry with and without remodeling by EV3® robots, separately. All subjects participated in 40-minute sessions of puppet play in the study hall twice a week for 12 weeks, totaling 24 sessions. The experimental process and experimental tools are shown in Fig. 1. A specified set of remodeled glove pupae was adopted in the experimental group, and traditional glove puppetry was used in the control group.
Remodeled glove puppets were prepared by a special education preschool teacher with 3 years of working experience. Two occupational therapists performed the tests and evaluations. A research assistant took care of the randomization and recruitment process to ensure that the evaluators were blinded to the process. The group number was chosen under one block, and block randomization was completed by MAND scores.
To increase the motivation of puppets playing in two groups, the well-known Chinese historical drama “Romance of the Three Kingdoms” was played. After children watched a period of drama, they operated puppets for role-playing. While the experimental group used switches to move the puppets, the control group put their hands in their puppets to move them. The study procedures were approved for human research by the Research Ethics Review Committee of the National Tsing Hua University (REC number: 107011HT075). All participants’ parents signed an informed consent form before testing.
TRAINING INSTRUMENTATION
We modeled the Chinese puppet using the robotic platform the Lego® Mindstorms® Education EV3 Core Set (45544). The EV3 Brick in the base set provided programming and data logging functions. It also includes interactive servomotors, sensors, and connection cables. The remodeled glove puppetry included three EV3® sensors attached to the robot puppets. The robot puppets’ arms and leg can be moved or rotated using one (touch sensor) and one (light sensor) servo motors, respectively, as shown in Fig. 1. The lion’s head and leg also can be controlled by the touch switch and light sensor. If the participants have problems with the EV3 switch, the modified switches also can be used in the experimental sessions, as shown in Fig. 1.
Outcome measures
We collected kinematic data, finger functional performance, and revised Knox Preschool Play Scale (PPS) scores. Kinematic analysis on the hand motion, including the angle and direction, was performed using Siliconcoach® Pro 7 software (Siliconcoach Ltd., Dunedin, New Zealand). A smartphone (OPPO-A55, BBK Electronics Ltd., Dongguan, China) was positioned 2 m away from the participants at a height of 100 cm for kinematic analysis. The piston finger device for restoring the motor function [17] was 172°– 140° at the metacarpophalangeal (MP) joint, and 155°– 86° at the proximal interphalangeal (PIP). The locations of kinematic markers for the MP, PIP, and DIP are shown in Fig. 1.
The Baseline® hydraulic pinch gauge (120226, LiTE™) is a reliable instrument to accurately measure tip, key and palmar pinch strength in both pounds and kilograms [18].
The revised Knox Preschool Play Scale (PPS) is based on developmental theories and assesses preschool children aged 0–6 years on their play performance [13]. The scale measures four aspects of play performance: play space management, material management, pretense symbolism, and participation (see Appendix A for details). The observer took the mean behavioral score as the dimension score. The inter-rater reliability of PPS for children with DD ranged from 0.984 to 0.986, and the test-retest reliability ranged from 0.861 to 0.961 [19]. Cameras were set up to record the first intervention for pretest, and the last intervention for posttest, and the play scores were generated from the recorded videos. Each behavior was scored at the upper age limit of the age group (for example, the 48- to 54-month level was scored at 54 months), according to the literature [20]. Assessment took approximately 40 minutes each on both the pretest and posttest, and two graduate assistants watched the videos to code behaviors.
Data analysis
Hand kinematics and preschool play scores were collected in both the pretest and posttest phases. Univariate analysis of variance (ANOVA) and analysis of covariance (ANCOVA) tests were used to measure the test results (Table 2). ANOVA can determine whether the intervention effect has a significant influence on these datasets (hand kinematics and PPS), and ANCOVA compares the adjusted means of two independent (experimental and control) groups, while using the pretest results as covariates (Table 3). These tests provide the benefit of statistically controlling a confounding variable (pretest score). The pretest score comparison between the two groups is also listed in Table 3.
Statistical analyses were performed using SPSS software (SPSS Inc., Chicago, USA) to analyze the results as follows: the effect size (partial eta squared, η2) as values for significant findings and the power analysis for the probability of rejecting the null hypothesis when incorrect. Using the effect size to support the null hypothesis could be due to the inadequate sample size, as listed in Table 3.