The Bead Maze Hand Function Test for Children

The skillful handling of objects requires an accurate application of finger forces (Johansson & Westling, 1988) that develops over time. This ability to control finger forces depends on prior experience with the object, as well as sensory information obtained during the ongoing interaction (Johansson & Cole, 1992; Johansson & Flanagan, 2009; Parikh et al., 2020). At about age 2 yr, children start demonstrating a finger force pattern that is characteristic of anticipatory control, a strategy necessary for the development of precision force control (Forssberg et al., 1991). By age 8 yr, this force pattern nearly matches that observed in adults (Forssberg et al., 1992). Recent work has shown that developmental improvements in finger force control continue into adolescence (Dayanidhi et al., 2013; Fuelscher et al., 2021; Rose & Parikh, 2022; Wolff, 2023). Accurately directing forces is also important, for instance, to rotate a doorknob or to avoid spillage while pouring liquids (Latash & Johnston, 2012; Parikh & Cole, 2012; Valero-Cuevas et al., 2003). Many children with neurological and developmental disorders have impaired finger force control (Dayanidhi et al., 2013; Gordon, 2016). Therefore, quantifying a child’s ability to control finger forces would provide a direct, specific measure of hand function tied to the quality of the movement. However, the examination of finger forces is not a part of common clinical practice.

Activity-based tests of hand function form an important component of a clinical evaluation of upper extremity motor function (Gogola et al., 2013). Most tests take a considerable amount of time and expertise to administer and may not accommodate the short attention span of children with a wide range of cognitive and physical abilities. Many tests assess movement quality through observation. Although trained observation can rate general activity performance, it does not quantify many important aspects of motor skill. Tests such as the Jebsen–Taylor Hand Function test, the Functional Dexterity Test, and the Purdue Pegboard Test^TM assess hand function but are fundamentally time-based measures (Vollmer et al., 2010; see Table 1 for common pediatric tests). Although it is an objective measure, a time-based motor score may not fully reflect functional differences in the ability to button and unbutton, use a pen or pencil, or hold a full cup of milk—daily activities that require manual skills. A time-based score is known to be insensitive to subtle but functionally relevant changes in hand function after an intervention (Olsen & Knudson, 2001; Sears & Chung, 2010; Wolff, 2023). It is important to note that time-based measures and forces are only modestly correlated (Cole et al., 2010; Parikh & Cole, 2012), which suggests that forces may provide crucial information about hand function beyond time. A comprehensive activity-based assessment of hand function must not only demonstrate whether or how quickly a task is accomplished but also quantify how well the person performs the activity.

We have developed a new activity-based pediatric hand function test, the Bead Maze Hand Function (BMHF) test that integrates measures of time and force control. The BMHF test is a common developmental toy enhanced with force sensors. During the BMHF test, children use their fingers to slide a bead over smooth wires (Figure 1A). The test includes three wires of increasing complexity (straight, single-curve, and double-curve) to assess motor skill at increasing levels of difficulty. The forces exerted onto the wire are recorded by a hidden sensor to provide information on precision in positioning and orienting the bead in relation to the wire. The sensorized toy is safe, because it has no loose ends or small parts and, therefore, can also be used to assess younger children. We argue that it can be administered with ease and minimal practice, which increases the usability of the BMHF test in busy clinical environments.

The objectives for this study were to examine

• ▪ how the BMHF test’s measures, such as total forces and task completion time, changed with age;

• ▪ the interrater reliability of the BMHF test scoring algorithm;

• ▪ the concurrent validity of the BMHF test measures; and

• ▪ the sensitivity of the force-based measure during a task requiring precision.

Because fine motor skills are known to improve with age, we expected that the total forces exerted onto the bead and the time to complete the task would decrease with age in typically developing children. We assessed reliability in the BMHF test scores when the sensor data were independently analyzed by two raters–researchers using separate computers. We also expected that the decrease in time to complete the BMHF test with age would correlate with improved scores on the Box and Block Test (BBT) with age (Cools et al., 2009). Last, we expected that the BMHF test’s force measure would be more sensitive to the precision requirement of the task (i.e., complexity of wire shapes) than the time measure.

Twenty-three typically developing children (9 girls, 14 boys; age range = 4.01–15.75 yr, M = 8.56, SD = 3.45) were recruited locally by flyer and word of mouth over a period of 1 yr. We included children up to the age of 16 yrs. because finger force control continues to develop in adolescence (Dayanidhi et al., 2013; Fuelscher et al., 2021; Wolff, 2023). Eligible participants had normal or corrected-to-normal vision; had no history of developmental disorders, neurological abnormality, or musculoskeletal disorders; had no history of hand surgery or significant hand or wrist injury; and attended a mainstream school, as reported by a parent. Informed written consent was obtained from a parent, and verbal assent was obtained from all participants. All participants had no prior experience with the apparatus. The study was approved by the University of Houston Institutional Review Board.

The apparatus (Figure 1A) has three triaxial force transducers (Nano 25, ATI Industrial Automation; 1,000 Hz) affixed to a bottom plate and mounted on a wooden base. Straight and curved polished metal wires were attached by means of a top plate. This arrangement of the transducers allowed us to monitor forces applied to the wire along the vertical (z) and horizontal axes (x and y; Figure 1B). It is important to note that the sensors are installed in the base of the device, and no sensor is installed within the bead. Each wire is rigidly attached to the sensor to allow for continuous recording of forces as the bead is moved along the wire. A slight pushing, tilting, or rotating of the bead, as may happen while moving the bead over the wire, imparts forces onto the wire and is measured by the sensor attached to the wire. The apparatus uses three colored wires of increasing complexity according to their shapes—straight, single-curve, and double-curve—to assess motor skill at increasing levels of difficulty. The moving beads (MVBs; Figure 1A) that slide along the wire have an outside diameter of 18 mm and a lumen diameter of 6.5 mm. The outer surface is textured to minimize accidental slippage. Each wire also has a stationary bead (STB) affixed on it. A nut-and-rod apparatus with similar rod configuration, but with loose, small parts, has been used to study object manipulation in primates (Darling et al., 2006) and adults (Cole et al., 2010). The nut-and-rod apparatus was not designed for clinical use.

All testing took place over 1 yr in the same laboratory. Participants sat on a chair and rested their hands on their lap. The BMHF apparatus was positioned approximately 10 cm from the edge of the table. Participants were instructed to grasp the MVB, lift it, and draw it over the wire until it “bumped” the STB. Once the MVB contacted the STB, participants were instructed to release the grasp and return their hand to their lap. A demonstration was provided, followed by one practice trial per wire. Participants then performed the task five times per wire at a self-selected “comfortable” speed (blocked for each hand). The wire order was randomized and counterbalanced. Instructions on how to grip the bead, whether the bead should touch the wire, or how to perform the task were not provided. We reasoned that, for future clinical populations with cognitive deficit or for younger children, it would be impractical to standardize instructions regarding precision and speed.

Participants first performed a hand preference test (Gordon et al., 2006). They then performed the BMHF test with both their preferred and nonpreferred hands in a counterbalanced order. Participants also performed the BBT (Mathiowetz et al., 1985).

Forces were acquired through a custom program in LabVIEW software (National Instruments) and were analyzed using a custom written script in MATLAB (MathWorks; 30-Hz low-pass filter, fifth-order Butterworth; Parikh et al., 2020). The initial contact with the MVB was defined as the time point at which the force produced in the x, y, or z direction crossed and remained above a threshold—a mean ±2 SD of the baseline—for 10 ms. The end of the trial was defined by a stereotypical change in force caused by contact of the MVB with the STB during a phase just before forces returned to baseline (i.e., grasp release; Figure 1C). Trial end thresholds included +0.1 N z-force for Wire 1, +0.2 N y-force for Wire 2, and −0.2 N y-force for Wire 3. For each trial, the script identified the initial contact time and the end of the trial time and placed markers. For the reliability study, two independent researchers analyzed the outputs using the same custom written script on separate computers. The researchers visually checked these markers in relation to the force traces for accuracy. In case of errors, the markers were adjusted manually. The absolute value of the force signals from contact to end were summed to the total force for each trial.

Using trialwise total force and trial duration data, we determined whether there was an improvement in performance across five trials. We used maximum-likelihood linear mixed models with wire (straight, single-curve, and double-curve) and trial (1, 2, 3, 4, and 5) as fixed factors and subjects as a random factor. We examined changes by age in mean total force, mean horizontal and vertical forces, and mean trial duration using linear regressions and reported the fit to the line as r (calculated as the square root of R²). For interrater reliability, a third researcher calculated intraclass correlation coefficients, or ICC(2,1), using a mixed model (Tissue et al., 2017). For each wire, we used Pearson correlation coefficients to examine the relationships between the BMHF measures and the BBT score (concurrent validity). We examined how total force and trial duration comodulated with increasing difficulty of the wire shapes. Considering the wire length, it is reasonable for participants to take longer to complete a trial with longer curved wires in comparison with the straight wire. However, considering the complexity of curved wires, we expected that forces would not scale in proportion to trial duration. Both outputs that were obtained from the double-curve and single-curve wires were normalized to that of the straight wire for each subject, and we created a complexity index score as the difference between these normalized outputs. A positive complexity index value indicates that the participant increased total force more than trial duration on the more complex wire shape. An index value of 0 indicates that both outputs increased in proportion. For all tests, the significance level was set at .05, and post hoc comparisons applied Bonferroni corrections when appropriate.

All participants completed the BMHF test as instructed, with no adverse effects or incidents.

We found that the total force exerted on each wire did not change across five trials for the preferred hand: no effect for Trial, F(4, 307.03) = 1.057, p = .37, and no Wire × Trial interaction, F(8, 307.03) = 0.503, p = .85. However, we found that the total force differed between wires—effect for Wire, F(2, 307.03) = 306.14, p < .001—with force increasing as complexity increased (force difference, p < .001 for all wire post hoc comparisons). We also found that that the trial duration did not change over five trials for the preferred hand: no effect for Trial, F(4, 307.03) = 0.707, p = .58, and no Wire × Trial interaction, F(8, 307.03) = 0.373, p = .93. As expected, we found that the trial duration increased as the complexity of the wire increased: F(2, 307.03) = 377.738, p < .001; force difference, p < .001 for all wire post hoc comparisons. We found similar findings for the nonpreferred hand for both total force and trial duration. In summary, we found that total force and trial duration for both hands did not significantly change over repeated trials. Therefore, we considered these measures averaged across their respective five trials for subsequent analyses.

We found that both mean total force and mean trial duration decreased with age (Figure 2). The trends appeared stronger when the preferred hand was used and were statistically significant. For the straight wire, both trial duration (r = .655, p = .001) and vertical (z) forces significantly decreased with age (r = .479, p = .02). For the single-curve wire, both trial duration (r = .622, p < .01) and horizontal forces significantly decreased with age (r = .466; p = .03). For the double-curve wire, both trial duration (r = .679, p < .001) and horizontal forces significantly decreased with age (r = .460, p = .02). Similar trends were observed for the nonpreferred hand.

For the trial duration measure, we found that the ICC(2,1) between the two raters (absolute agreement) was .95, 95% confidence interval (CI) [.884, .979] for the low-difficulty condition; .984, 95% CI [.962, .993] for the medium-difficulty condition; and .98, 95% CI [.96, .99] for the high-difficulty condition. For the total force measure, we found that the ICC(2,1) between the two raters was .967, 95% CI [.921, .986] for the low-difficulty condition; .929 95% CI [.833, .97] for the medium-difficulty condition; and 0.979, 95% CI [.952, .991] for the high-difficulty condition.

A longer trial duration was modestly associated with a lower score (a smaller number of boxes transferred) on the BBT for each wire shape (all rs were >.599 for the preferred hand and >.475 for the nonpreferred hand; all ps were less than an adjusted α of .025). However, the amount of force exerted on each wire had no significant correlation with the number of boxes transferred on the BBT.

The BMHF test apparatus was familiar, and the administration protocol was engaging for children. We kept the protocol and instructions simple to accommodate children with a wide range of ages and cognitive abilities in busy clinical environments. We found that participants did not change their performance with repeated attempts, which suggests that the activity was more like playing with a toy rather than taking a motor assessment. The use of a sensorized toy that is familiar to a child possibly reduced the impact of testing on behavior (Duff & Gordon, 2003).

We found excellent interrater reliability in our sample of 23 children, ages 4 to 15 yr. This finding is likely due to the use of an objective method to compute the BMHF scores with a minimal need for subjective interference (<10% of trials). The BBT score was significantly correlated with the time to complete the BMHF test for trials performed on all three wire shapes with both preferred and nonpreferred hands, thus establishing concurrent validity. However, as expected, the total force failed to correlate with the BBT score. This finding is consistent with earlier reports of modest or no correlation between finger forces and time-based measures during manual tasks (Cole et al., 2010; Olsen & Knudson, 2001). The BBT score informs about the number of blocks transferred from one compartment to the other in a 1-min trial but fails to explain how the blocks are grasped and transferred. It is likely that two participants may perform the activity in a distinct manner (Wong & Whishaw, 2004) yet produce similar BBT scores. In our study, two children, ages 7 and 9 yr, had nearly identical BBT scores (32 and 33, respectively; Figure 2B, circled) and similar BMHF test completion times (∼2 s for the double-curve wire). However, they produced significantly different total forces on the double-curve wire (2,415.6 N and 981.6 N, respectively; Figure 2A, circled). Thus, a force-based measure may provide distinct but complementary information about hand function.

We found that total force decreased with age for the double-curve wire. This was driven by a decrease in the horizontal forces imparted onto the wire (Figure 2E), possibly because of the improved steadiness of contractions of hand, forearm, or arm muscles with age (Deutsch & Newell, 2004). Our finding that the vertical forces decreased with age for the straight wire (Figure 2C) are similar to findings in laboratory-based grip and lift studies (Forssberg et al., 1991, 1992) and might be due to an improved ability to orient the bead (i.e., less tilting) relative to the wire with age (Cole et al., 2010). Similarly, horizontal and vertical forces imparted onto the single-curve wire decreased with age (Figure 2D). Better BMHF test scores in older children provide evidence that the BMHF test was motivating enough for participants at the higher end of the age range. We found that using greater force versus greater time was the preferred strategy on the double-curve wire, as indicated by higher scores on the complexity index. The complexity of the double-curve wire demands greater precision in orienting the bead along the wire. This finding indicates that force is the more relevant measure of skill on tasks that require more precise control. In contrast, for the single-curve wire, force and time scaled proportionally, which suggests that force and time are equally indicative of skill on the single-curve wire. The single-curve wire may not be complex enough to challenge manual precision in typically developing children. However, in clinical populations, an easy- to intermediate-level wire shape may prove useful in delineating changes in function after treatment.

Despite a smaller sample size, the findings from this work support larger psychometric studies in clinical populations. Future studies may include kinematic assessment of the bead and hand, wrist, or arm to understand the mechanisms underlying the changes in horizontal and vertical forces during the BMHF test with age.

A reliable measure of manual skill in the form of dynamic force control would be more sensitive than current clinical tests in detecting changes after surgery or therapeutic interventions. The BMHF test may facilitate diagnosis at young ages and detect functional changes caused by treatment.

The BMHF test uses a familiar toy with hidden sensors to quantify a child’s typical manual performance objectively. The test provides both force- and time-based assessment of hand function, with the former being more sensitive to the precision requirement of the task. We show excellent interrater reliability and concurrent validity of both test measures in typically developing children.

We thank the children and their parents for their support and Fred J. Rose for his assistance in fabricating the BMHF apparatus. The work reported in this article was primarily supported by a research grant to Pranav J. Parikh from the National Institutes of Health (NIH) Center for Smart Use of Technologies to Assess Real-World Outcomes (C-STAR) at the Shirley Ryan AbilityLab, which was awarded under Grant P2CHD101899. This work was also supported by NIH/National Institute of Child Health and Human Development Grant R25HD106896 and the University of Houston CLASS Research Progress Grant to Pranav J. Parikh.