Introduction
Spinal cord injury is a common cause of disability, affecting more than 300,000 people in the US [
1]. The majority of injuries occur at the cervical level, which often impairs function of upper extremities and can lead to chronic disability [
2‐
4]. Accurate and sensitive measurement of upper limb function is a critical part of the development and assessment of new therapies to improve recovery after SCI.
The most common method of arm assessment after cSCI involves subjective ordinal scoring of motor function by a skilled examiner [
5‐
7]. While this approach yields rapid and reliable results, ordinal assessment may not always be sensitive to small improvements that can be functionally meaningful [
8,
9]. Objective measurement may provide improved measurement sensitivity, and as sensor technologies have become smaller and more accessible, numerous tools have been developed for assessing arm function after cSCI [
10]. Dynamometers and myometers are widely used to measure isometric force of isolated arm functions in continuous physical units [
11,
12]. Other assessment tools measure position and force during execution of simulated functional tasks [
13,
14]. These systems indeed provide greater precision or a more direct characterization of functional ability than traditional measures, but the continued reliance upon ordinal assessment indicates an unmet clinical need for the development of measurement technologies.
To increase their utility and implementation, measurement systems should be simple to use, facilitate standardized administration, report sensitive, quantitative metrics, and provide reliable longitudinal testing [
15]. We designed a suite of modular rehabilitative devices for objective assessment of various isolated hand and wrist motor functions to address these needs. In addition to the measurement modules, a table-mounted armrest with simple, interchangeable tasks facilitates standardized administration across a wide range of arm impairments [
11]. The system records measurements in continuous physical units, providing unambiguous results across many aspects of function. Furthermore, the devices enable collection of multiple trials to increase sensitivity and reliability of measurements.
In this study, we tested whether these tools could quantify motor impairments in individuals with cSCI, determined measurement detection limits, and examined retest reliability by assessing the same participants 4 months later. We also established concurrent validity by correlating performance on the rehabilitative devices with two common metrics of upper limb function after cSCI, the Graded Refined Assessment of Strength, Sensibility, and Prehension (GRASSP) exam and the Jebsen Hand Function Test (JHFT) [
5]. Our results demonstrate that the system provides reliable measurements over time and that performance correlates with established outcome measures. The results indicate that the novel system can deliver simple and reliable longitudinal evaluation of upper extremity function after cSCI and may provide a framework to assess the efficacy of strategies aimed at improving recovery of upper limb function.
Discussion
Here we provide a characterization of a novel suite of automated devices to measure hand and wrist motor function after cSCI. The system consists of seven distinct tasks that quantify various aspects of isometric force and single joint ROM. We compared measurements made with each of the devices in cSCI and uninjured participants, determined detection limits of each metric for participants with cSCI, evaluated reliability of measures over time, and correlated the metrics with established assessments of quadriplegic upper limb function.
Deficits in strength and range of motion of the arm impair prehension after cSCI [
8,
9,
22,
33]. Accurate and sensitive measurement of functional impairment is critical for determining the efficacy of treatments focused on restoring motor function. The most common methods of arm assessment after cSCI include subjective categorization of isometric force and free range of motion, like in the ASIA and GRASSP exams [
5]. These assessments use ordinal scoring with scales designed to balance measurement sensitivity and reproducibility and consequently demonstrate excellent inter-rater reliability [
7,
34]. Sensor technology is capable of measuring motor function with greater precision, both because of a higher resolution scale and the capacity for a larger number of repeated trials within the same test session, and so numerous technologies have been developed to provide objective measurement of prehension [
10]. Confining measurements to a single degree of freedom of movement limits sources of error, helping to reduce variability and improve sensitivity. Following this rationale, we restricted our measurements to well-controlled movements that together require the majority of the muscles in the hand and forearm.
Overall, the devices provided a robust characterization of impairments after cSCI. As expected, when compared to uninjured controls, cSCI participants were the most severely impaired on tasks that required finger strength, specifically the finger and doorknob isometric force devices [
9]. The force metric that was least impaired compared to controls was wrist extension [
35]. The majority of cSCI participants demonstrated ROM that was comparable to uninjured controls, indicating the isometric tasks provide a more robust assessment of motor deficits. Altogether, these results demonstrate that the devices described in this study are capable of providing accurate measurements across many distinct and essential wrist and finger functions commonly impaired after cSCI.
An essential feature of any assessment is its ability to detect significant changes outside of measurement error. The MDD is necessary to determine an assessment scale’s capacity to detect real changes when only a single sample is available per test session, as is the case for the GRASSP and JHFT assessments [
21]. In contrast, the devices provided continuous data and enabled rapid collection of repeated trials within the same test session. To characterize the capacity of the devices to detect real changes in performance, we compared two distinct measures of within-subject variability: (1) the MDD, which is based on the variance of changes in individuals’ mean performance between test sessions, and (2) the average variance of all 20 trials acquired during both test sessions. For participants with cSCI, we found the 95% confidence detection limit based on variance was highly correlated with and reliably smaller than the MDD, consistent with the notion that sampling more trials increases the probability of identifying significant changes in performance. Furthermore, the acquisition of multiple trials per test session allows within-subject comparison, which increases statistical power.
Consistent, stable assessment is critical for longitudinal studies that rely on evaluating recovery over time. Overall, we find that the devices provide good test-retest reliability. All of the ROM measures, as well as a composite score that took into account performance on all devices, demonstrated excellent reliability. The force measures were less reliable, likely due to the relatively smaller measurement spread and larger CVs. These results suggest that the devices developed here may be useful for long-term studies aimed at improving hand and wrist function after cSCI.
Previous cSCI studies suggest that isolated components of upper limb strength can significantly contribute to task performance [
8,
9,
36]. The metrics of hand and wrist function collected with the devices correlate well with gold-standard assessments used in SCI rehabilitation studies. The composite score was very highly correlated with the GRASSP score for the right arm. Of the four GRASSP subscores, quantitative prehension was best correlated with the composite score, demonstrating a strong relationship between execution of basic motor functions and functional task performance (Additional file
1: Table S2). Additionally, the composite score was well-correlated with the JHFT test. However, the correlation was slightly weaker than that observed for GRASSP scores, primarily driven by a more bimodal distribution of JHFT scores compared to a more even distribution of GRASSP and composite device scores. The strong correlations between the composite score and both gold-standard assessments suggests that relatively limited, simple assessment with the devices can be used to accurately gauge hand and wrist function after cSCI.
The assessment devices described in this study were developed to provide a number of advantages over existing tests of function for rehabilitation studies. One key advantage is the simple measurement of continuous data in physical units, including angle, torque, and linear force. This allows direct comparison of effect size and magnitude of changes in performance, a benefit over assessments that rely on ordinal scores. The use of continuous physical values also largely mitigates any variance in tests that rely on a subjective scoring system, which would facilitate comparison of data across multiple sites in a trial. Moreover, unlike ordinal assessments, there is no performance ceiling or necessary stratification, which may make the devices useful for measuring hand and wrist function in other populations with varying degrees of impairment, such as stroke patients.
Measurement with the devices is simple and relatively rapid. Data collection with the entire suite of devices took approximately twenty minutes to collect ten repeats of each movement in each direction, a total of 140 trials. This testing duration is comparable to or slightly shorter than that typically required for GRASSP and JHFT, which respectively require approximately thirty or fifteen minutes to complete for each hand. While the purpose of this study was to collect comprehensive data on all devices, selecting a subset of devices based on residual upper limb function and reducing the number of repeats to match the desired statistical power could potentially speed data collection further. Finally, the system is compact in size and constructed from low-cost components. This raises the potential for the devices to be packaged for home use. The simple data stream collected with the devices could also be easily implemented into a video game architecture to increase engagement and promote user compliance [
37‐
39].
While the devices provide a number of advantages for quantitative rehabilitation studies, one disadvantage is the absence of direct measurement of sensory function, which would need to be supplemented with an additional assessment. Another disadvantage is the restriction to movement about a single joint. Given the complexity and fine motor control of prehension, this restriction to a small number of single joint motions fails to capture the full range of hand and wrist dysfunction. However, constraining the complexity of movement simplifies and improves measurement capabilities [
9,
17]. Future studies are required to directly correlate device measures with functional outcomes, including ability to perform activities of daily living.