We used a cross-sectional observational psychometric study design with repeated assessments to investigate the feasibility, validity, and reliability of the ProMeTo.

We recruited children and young people with neuromotor impairments due to UMN lesions (e.g., CP, acquired brain injury) from the inpatient and outpatient Swiss Children's Rehab clinic of the University Children's Hospital Zurich. Inclusion criteria were as follows: age 5–19 years, the ability to sit for 30 min with or without back support, and the ability to stand and walk a few steps with or without support. Exclusion criteria were severe visual impairment, lower limb surgery or botulinum toxin injection within the previous six months, inability to communicate pain or discomfort or to follow simple short instructions, and non-compliance. We also recruited TD peers with the following exclusion criteria: any neurological diagnosis, severe visual impairment, developmental coordination disorders, or a diagnosis of attention deficit hyperactivity disorder.

According to the consensus-based standards for selecting health measurement instruments (COSMIN) guidelines, we aimed to recruit 50 children with UMN lesions and 50 TD peers (11). We recorded the Functional Mobility Scale (FMS) score for the distance of 500 m (in community-based settings) as a descriptive measure of the participant's walking ability (12, 13). All children and young people agreed verbally to participate. Parents and adolescents aged 14 years and above also signed an informed consent form. The Cantonal Ethics Committee of Zurich (BASEC-Nr. PB_2021-01373) approved this study, and we followed the good clinical practice guidelines. The study protocol is available on clinicaltrial.gov (trial identifier: NCT05405881).

The measurements took place in a therapy room in the Swiss Children's Rehab or at the practice where the child attends therapy. At the first visit, one assessor used the ProMeTo (test 1) to measure proprioceptive function in the participants. At the second visit, the same assessor repeated the tests (test 2) approximately ten days later. For inter-rater reliability, a sub-sample of the children repeated the assessment after a 5-minute break with a different assessor (assessor A and B). We conducted the repeated measurements for the ProMeTo only within the group of children with UMN lesions.

We used CE-certified Shimmer inertial measurement units (©Shimmer, Dublin, Ireland) to assess the joint angles. We created an application using the game software Unity (version 2020.4.6f1, Unity Technologies, San Francisco, USA) to guide the assessors through the tests step by step. Figure 1A shows the test setup and sensor application, while Figure 1B shows the calibration position. Back support was provided if needed using a foam pad or assistance from another person. The assessors held the leg distally with predefined grips (Figure 2). The grips were defined to avoid pressure or tension in the applied direction of movement. The assessors used a medium force, comparable to a handshake (around 2.5 N) (14).

The test started with the JMS, followed by the JPS, DPS, and the visual control test. The child played a short game of 25 s between the tests. In the game, the child steered a wizard flexing or extending their knee to collect coins. The game provided fun and allowed the children to have an active break.

For the JMS, the assessor moved the limb at a speed of 5–15°/s in one direction. To avoid moving too fast or slow, the assessor had visual feedback on the convertible notebook to guide the speed. The assessor made one test movement of each joint and used the same terminology as on the digital buttons on the smart phone, to verify the child understood the task. The child was asked to confirm their perceived direction of movement through the digital buttons on a smart phone as fast as possible (Figure 1C). We recorded the percentage of correctly identified directions out of four trials and the angle until detection (°) for each trial.

For the JPS, the assessor moved the limb to a particular position (criterion position) and held this position for three seconds before moving back to the starting position (15). The limb was then slowly (5–15°/s) moved through the whole range by the assessor, and we asked the child to press the digital “Stop” button on the smart phone when they thought their limb had reached the criterion position (Figure 1D). In contrast, for the DPS, we asked the child to actively move their limb to the position that they considered to be the criterion position and press the digital button to confirm (Figure 1E). For each joint, four criterion positions (four trials) for JPS and DPS were assessed; two criterion positions in internal and two in external hip rotation, two positions greater than 90° knee flexion and two less than 90°; and two in ankle dorsiflexion and two in plantarflexion. To prevent selecting a criterion position at the end of the joint range, we assessed the available range of motion for each child by evaluating each joint movement prior to the JPS and DPS tests. The assessor selected the criterion position within the intermediate 80% of the individual range of motion. For JPS, the assessor passively moved the limb twice through the full range and for the DPS, the child actively moved the leg through their active range of motion. For each criterion position, we recorded the difference in degrees (°) between criterion position and the confirmed position (delta) for the JPS and DPS. For both tests, the assessors explained the tasks to the children and let them try out one time to try the test.

The final assessment was the visual control test. To investigate whether the child's motor skills influenced the outcome of the DPS, we used a test with visual feedback as a control test (CT; Figure 1F). The child could see the movement of their limbs on an avatar on the convertible notebook. A target marker on the screen indicated the critical angle that the child had to reach. For each single test, a repetition was possible if the child or the assessor were distracted.

We assessed the feasibility of the ProMeTo by recording testing time, technical issues, and participant feedback on pain, concentration required, and fatigue using a 0–10 visual analogue scale. Verbal feedback and the assessor's assessment using the same scale were also recorded, as well as the child's understanding of instructions (16).

For the JMS, the percentage value for each joint was calculated based on the correct recognition of the movement direction across four trials, i.e., 100% if the movement direction was correctly recognised in all four trials. We also averaged the angle at which the child recognised the movement direction over the four trials for each joint. For the JPS and DPS, the average delta (criterion position—confirmed position) from the four trials was calculated for each joint.

In addition to the deltas for each joint (hip, knee, and ankle), we also calculated the mean value over the three joints, reflecting the overall proprioceptive error of the leg. We calculated these mean values separately for each modality, child, and the more and less affected leg.

To calculate the proprioceptive component score, the CT was subtracted by the DPS. Statistical analyses were performed using SPSS version 27 (IBM SPSS Statistics, Chicago, IL). Data distribution was evaluated using Shapiro-Wilk tests and visual inspection of the Q-Q plots. For statistical tests, alpha was set to 0.05. Participant characteristics and feasibility data were analysed descriptively.

The results of the first ProMeTo of the children with UMN lesions were used for the statistics for discriminative and convergent validity. For discriminative validity, differences between the two groups were analysed with the Mann-Whitney-U-test. For the convergent validity, Spearman correlation coefficients (r_s) were calculated to quantify the association between the mean values of the three proprioception modalities, and between the proprioceptive component and the JPS. We used the following benchmarks: 0–0.25 (no or little relationship), 0.25–0.50 (fair), 0.50–0.75 (moderate to good), 0.75–1.00 (very good to excellent) (17).

Test-retest reliability and inter-rater reliability were calculated for each joint and the mean value of the three joints per modality using Intraclass Correlation Coefficients (ICCs) and 95% Confidence Intervals (CIs), using the two-way random effect model ICC (2,1) (18). For the interpretation, the following benchmark ICC values were used: lower than 0.25 (poor reliability); 0.26–0.49 (low reliability); 0.50–0.69 (moderate reliability); 0.70–0.89 (high reliability); and higher than 0.90 (very high reliability) (19). The absolute reliability, the standard error of measurement (SEM), and the Smallest Detectable Change (SDC) were calculated with the following formulae: SEM = SD (Standard Deviation) √(1-ICC) and SDC = 1.96x√2xSEM (20). Additionally, we applied Bland-Altman plots to check for systematical bias and the limits of agreements (95% CI) for the test-retest and interrater reliability (21).

The three tests lasted, on average, 17 min (SD 7.5 min; range: 9–54 min) for the children with UMN lesions and 15 min (SD 3.1 min; range: 8–22 min) for the TD peers. This difference was not significant (p = 0.12). All children with UMN lesions were able to perform the JMS test, but four did not correctly identify the movement in any of the three joints. These four children were also not able complete the JPS and DPS tests due to their low level of proprioception in their legs. Their age ranged between 6 and 12 years, and their diagnoses were CP (two with GMFCS level III and one with GMFCS level IV), and one child with acquired brain lesion due to a stroke.

The children with UMN lesions generally understood and accepted the ProMeTo assessments well (Figure 3). Additionally, two children could not perform the DPS of the ankle of their more affected leg. For these two children, the DPS mean scores were calculated based on the hip and knee outcomes.

Overall, executing the tests was relatively easy for the assessors. For a few children who showed voluntary active resistance against the movement, performing the JMS and JPS tests (2 and 3 children, respectively) was particularly difficult. For the TD peers, the tests were generally easy to understand and execute. In addition, the assessors reported the children's high level of comprehension (Supplementary Figure S1).

Verbal feedback from children with UMN lesions indicated varying experiences: some found it fun (n = 5), easy to understand (n = 2), and enjoyable (n = 4), while others found it boring (n = 2), slightly tiring (n = 2), or uncomfortable due to factors like cold feet or feeling hot under the cuffs with the sensors (n = 3). The TD peers generally found it fun (n = 5) but noted difficulties in controlling foot movement (n = 6) or maintaining leg internal rotation (n = 4). Furthermore, they also mentioned that they felt warm under the cuffs (n = 2). Only a few technical issues were reported during all the tests, such as occasional loss of connection to the mobile phone (n = 4) and slow mobile phone response (n = 2).

Forty-five children with UMN lesions and 50 TD peers could be included for discriminative and convergent validity testing. In one child with UMN lesions, the JPS data for the more affected leg and the DPS data for the more and less affected leg were missing due to a loss of connection to the sensors; therefore, this analysis included 44 children with UMN lesions.

The children with UMN lesions had significantly higher test values (i.e., lower proprioceptive function) than their TD peers for all tests and all joints (Figure 4, more affected leg; Supplementary Figure S2, less affected leg).

Correlation coefficients between the mean values of JMS and JPS, and between JMS and DPS were moderate to good for both the more (r_s 0.53–0.73; p < 0.001) and less affected side (r_s 0.50–0.79; p < 0.001; Figure 5). The correlation between the JPS and DPS for both the more and less affected sides was good to excellent. Also, the relationship between the proprioceptive component (DPS subtracted by the CT) and the JPS was moderate to good for both the more and less affected sides (Figure 5).

Forty-six children completed the ProMeTo twice (the first test and the second test). Four children did not correctly identify the direction of joint movement in both tests. Therefore, we included 42 children in the test-retest reliability analyses for the JMS. For the JPS and DPS, two children's records of the second test were missing (n = 1 with CP, classified GMFC level IV, n = 1 status after brain tumour); therefore, we included 40 children in these analyses.

The relative reliability was high for the more and less affected side, except for the JMS for the ankle on the less affected side, which was moderate (Table 2). Absolute reliability, expressed by SDC, lay below 10°, except for the JMS (Table 2). The Bland-Altman plots represent the bias for the more affected leg for JMS of 0.6° (95% CI −6.7° to 7.9°), and for the JPS 0.3° (95% CI −2.9° to 3.5°), and for the DPS −0.1° (95% CI −3.7° to 3.6°), and comparable results for the agreement of the less affected leg (Supplementary Figure S3).

The data from sixteen children were used for the inter-rater reliability analysis. Relative reliability was high for all modalities (Supplementary Table S1). Absolute reliability, expressed by SDC, was below 10°, except for the JMS (Supplementary Table S1). The Bland-Altman plots represent the bias for the more affected leg for JMS of −3.0° (95% CI −11.3° to 5.4°), and for the JPS 0.1° (95% CI −2.8° to 3.0°), and the DPS −0.4° (95% CI −3.3° to 2.6°), and comparable results for the agreement of the less affected leg (Supplementary Figure S3).

The Shimmer sensors are anti-gravity sensors and had an effect on the choice of direction of movement, particularly for the hip (rotation instead of flexion and extension). Overall, the methodological quality of this psychometric study is “fair” according to the COSMIN guidelines due to the moderate sample size (11). We assessed 41 children twice for test-retest reliability and only 16 for intra-rater reliability. Therefore, the generalisability of these reliability results needs to be cautiously interpreted. Another limitation is that most children with CP (70%) were classified as GMFCS I, meaning their motor skills are generally good. Despite this high proportion of children with good motor skills in the UMN lesion group, the difference in proprioception function between the two groups was statistically significant.

As proprioception is multimodal and uses tactile information, our results can partly be influenced by how the assessor held and moved the limb (24) or the selected position of the criterion positions. Considering the force of the grip and the position, we standardised this as much as possible to avoid the tactile input, which would indicate the direction and range of joint movement. Still, we cannot exclude the fact that the children used tactile input in their strategy to perform the test. Considering the criterion positions, one suggestion for improvement could be to standardize the testing angle within a predefined intermediate range of motion, more precisely on a specific angle, to account for factors affecting proprioceptive acuities, such as receptor sensitivity or ligament and muscle tension. Such a protocol, however, would take more time because the therapist needs to position the joint at the exact angles, which might affect the feasibility and compliance of the child. In our study, the criterion angle differed between participants and between the repeated tests. Despite these differences' criterion angles, the reliability was still high. Further, we only reported the results for each joint and each leg and not for each set criterion angle, as we randomly selected these for each participant.

As not all the data were normally distributed, we applied nonparametric tests. Even though we calculated ICCs for relative reliability we also applied the ICC on the log-transformed data and the results were of the same level of interpretation as the ICCs of the original data.