Development of a Remote Version of the Graded Redefined Assessment of Strength, Sensation, and Prehension (GRASSP): Validity and Reliability

Abstract

Background.

The Graded Redefined Assessment of Strength, Sensation, and Prehension (GRASSP V1.0) was developed in 2010 as a 3-domain assessment for upper extremity function after tetraplegia (domains: Strength, Sensibility, and Prehension). A remote version (rGRASSP) was created in response to the growing needs of the field of Telemedicine.

Objective.

The purpose of this study was to assess the psychometric properties of rGRASSP, establishing concurrent validity and inter-rater reliability.

Methods.

Individuals with tetraplegia (n = 61) completed 2 visits: 1 in-person and 1 remote. The first visit was completed in-person to administer the GRASSP, and the second visit was conducted remotely to administer the rGRASSP. The rGRASSP was scored both by the administrator of the rGRASSP (Examiner 1), and a second assessor (Examiner 2) to establish inter-rater reliability. Agreement between the in-person and remote GRASSP evaluations was assessed using the intraclass correlation coefficient (ICC) and Bland–Altman agreement plots.

Results.

The remote GRASSP demonstrated excellent concurrent validity with the GRASSP (left hand intraclass correlation coefficient (ICC) = .96, right ICC = .96). Concurrent validity for the domains was excellent for strength (left ICC = .96, right ICC = .95), prehension ability (left ICC = .94, right ICC = .95), and prehension performance (left ICC = .92, right ICC = .93), and moderate for sensibility (left ICC = .59, right ICC = .68). Inter-rater reliability for rGRASSP total score was high (ICC = .99), and remained high for all 4 domains. Bland–Altman plots and limits of agreements support these findings.

Conclusions.

The rGRASSP shows strong concurrent validity and inter-rater reliability, providing a psychometrically sound remote assessment for the upper extremity in individuals with tetraplegia.

Introduction

Background

Over half of all spinal cord injuries (SCI) occur at the cervical level resulting in tetraplegia. Most individuals with tetraplegia have impairment of hand and arm function, which greatly impacts their level of independence and quality of life. Therefore, recovery of upper limb function is of high functional value in tetraplegic patients.^1–3 In order to optimize rehabilitation and develop effective interventions, unbiased and reproducible outcome measurements are essential for the SCI population.^4,5 The Graded Redefined Assessment of Strength, Sensation, and Prehension (GRASSP) is a clinical composite measurement tool to assess upper limb function in persons with tetraplegia.⁶ It is a multimodal test comprising 5 subtests for each upper limb: dorsal sensation, palmar sensation (tested with Semmes-Weinstein monofilaments), strength (tested with motor grading of 10 muscles), and prehension (distinguishes scores for qualitative and quantitative grasping). Thus, administration of the GRASSP results in 5 numerical scores that provide a comprehensive profile of upper-limb function. The established interrater and test-retest reliability for all subtests within the GRASSP range from 0.84 to 0.96 and from 0.86 to 0.98, respectively. The GRASSP is approximately 50% more sensitive (construct validity) than the International Standards of Neurological Classification of SCI (ISNCSCI) in defining sensory and motor integrity of the upper limb. The subtests also show concurrence with the Spinal Cord Independence Measure and the Capabilities of Upper Extremity Questionnaire.³ Because of the combined assessment of neurological (strength and sensation) and functional status (prehension), the GRASSP provides a comprehensive appraisal of upper limb function and is increasingly viewed as the “gold standard” outcome measure in clinical trials investigating rehabilitation interventions for the restoration of hand function in SCI.^6,7,8 This is because GRASSP has robust psychometric properties in the SCI population: intraclass correlation coefficients (ICC) ranging from .84 to .98, demonstrated construct validity against the International Standards of Neurological Classification of Spinal Cord Injury (ISNCSCI), and concurrent validity against both the Spinal Cord Independence Measure III and the Capabilities of Upper Extremity Questionnaire.^3,5,6,7,9,10

However, there are some barriers to completing the GRASSP in routine practice. Individuals must travel to healthcare facilities to participate in an in-person session. People with tetraplegia may face significant barriers to travel such as relying on others, using scheduled transport, or health comorbidities.¹¹ The COVID-19 pandemic worsened these barriers,¹² precipitating a growing need for remote assessments and for the use of telemedicine.¹³ The use of patient-assisted virtual physical examination techniques has also grown in recent years to assess various body systems remotely.^14,15

Currently, virtual assessment of the upper limb function in the SCI population is limited to qualitative data obtained through questionnaires or self-report and non-standardized assessment using items obtained in the home environment.^16,17 Although fully automated quantitative equipment for remote assessment of upper limb function is available (eg. ReJoyce Rehabilitation Workstation),^18,19 high cost and complexity of these devices makes clinical implementation challenging. A standardized remote GRASSP (rGRASSP) does not currently exist, and it would be a valuable tool for clinicians and researchers to assess neurological recovery and hand function in a virtual environment.¹⁶ Therefore, there is a need to develop and validate a standardized rGRASSP instrument without jeopardizing sensitivity and psychometric properties. The rGRASSP would provide immediate access to an affordable outcome measure to reliably assess multiple domains of hand function, essential to clinical trials assessing upper limb function remotely.

During the COVID-19 pandemic related restrictions in 2020, our study team partnered with the original GRASSP Research and Design Team (Neural Outcomes Consulting Inc.) to develop a remote version of GRASSP which can be administered via secured video conferencing using a standardized testing kit provided to participants. The rGRASSP includes the same 3 domains of the original version, with modifications to the 5 subtests to allow for remote testing. The primary objective of this study was to demonstrate the concurrent validity and inter-rater reliability of the newly developed remote GRASSP against the original in-person GRASSP.

Methods

Remote GRASSP Development

The modifications made to GRASSP V1.0 are outlined below and summarized in (Table 1):

• Strength domain assesses the 10 arm and hand muscles corresponding to myotome sites for C5-T1, with each myotome represented by multiple muscles tested as in the GRASSP V1.0. A home assessor was instructed on the standardized Manual Muscle Test (MMT) hand placement and application of isotonic resistance, as described in the GRASSP V1.0. Home assessors were instructed to report application of moderate or maximum resistance to differentiate between MMT grades 4 and 5. If a home assessor was unavailable, the participant was instructed to provide contralateral muscle resistance.

• Sensibility domain assesses 3 dorsal and 3 palmar sites on each hand as in the GRASSP V1.0, corresponding to dermatome sites for C6, C7, and C8. A home assessor applied a cotton swab to the location shown by the test administrator for this subsection. Additionally, sensitivity to light touch was measured on a scale of 0 to 2 (absent, impaired, normal), modeled after the sensation section of the ISNCSCI exam,²⁰ in which a home assessor was instructed to apply a cotton swab to the location indicated by the test administrator.

• Prehension Ability domain evaluates the quality of the cylindrical grasp, lateral pinch, and tip pinch, and was administered identical to the GRASSP V1.0. The home assessor was asked to stabilize the wrist or apply resistance, when necessary, to determine the presence/absence of a tenodesis grasp pattern.

• Prehension Performance domain was reduced to include 4 of the original 6 tasks (pouring water, opening jars, translating pegs, and transferring coins). These items were included in the rGRASSP kit provided to the participant and scored identically to the GRASSP V1.0. The key and nuts/bolts tasks were eliminated from the Prehension Performance domain of the rGRASSP due to difficulty replicating the materials affordably for usage within the home.

Table 1.

Comparison of the GRASSP and Remote GRASSP Components.

Components of the GRASSP (max score = 116)				Components of the rGRASSP (max score = 94)
Details	Equipment	Description	Scoring	Details	Equipment	Description	Scoring
Strength
MMT-4 arm and 7 hand muscles	None	Muscle grade assessed by therapist	/50 per side	MMT-4 arm and 7 hand muscles	None	Home assessor instructed to test each muscle and grade	/50 per side
		Right and left side scored separately				Right and left side scored separately
Sensibility
3 palmar and 3 dorsal test sites for each hand	Semmes Weinstein Monofilaments	Monofilaments applied to all test locations	/24 per side	3 palmar and 3 dorsal test sites for each hand	Cotton swab	Home assessor instructed to apply cotton swab to all test locations	/12 per side
		Right and left side scored separately				Right and left side scored separately
Prehension ability
3 grasps rated 0 to 4 (cylindrical grasp, lateral key pinch, and tip to tip pinch)	None	Participant performs each grasp and is rated by therapist	/12 per side	3 grasps rated 0 to 4 (cylindrical grasp, lateral key pinch, and tip to tip pinch)	None	Participant performs each grasp and is rated by therapist, home assessor instructed to provide resistance as needed	/12 per side
Prehension performance
6 tasks rated 0 to 5	GRASSP V1 materials: bottle cup, 2 jars, 9 pegs and pegboard, key and lock, 4 different sized nuts and screws	Participant performs all 6 tasks for each hand separately while seated at table	/30 per side	4 tasks rated 0 to 5	Bottle Cup 2 jars, 9 pegs, and pegboard coins and coin slot	Participant performs all 4 tasks for each hand separately while seated at table	/20 per side

A standardized rGRASSP kit, consisting of items necessary to complete the Sensibility and Prehension Performance subtests, was provided to all participants in order to complete the remote assessment. These items included 3 to 5 cotton swabs for the Sensibility testing; and a 16.9 oz plastic water bottle (19.5 cm circumference), 2 jars, with metal lids 1 slotted, 4 coins, and a wooden pegboard and pegs for the Prehension Performance domain. Except for the water bottle all test materials are the same dimensions as the original GRASSP V1.0 kit. Additionally, participants were asked to provide a sheet of paper for completion of the Prehension Ability domain. In order to ensure standardization of the rGRASSP kit, all materials except for a sheet of paper were provided to the participants by the research team prior to the rGRASSP session. In addition a revised and modified manual for the assessor was established to ensure instruction to the patient and caregiver were consistent, along with an instruction booklet for the patient to have at home. All test materials were modified from the original GRASSP V1.0.

Recruitment

The rGRASSP study protocol was approved by the Northwestern University Institutional Review Board (study ID# STU00213683). All participants were required to provide a written informed consent and were recruited from the Shirley Ryan Ability Lab’s research registry, Midwest Regional SCI Care System’s SCI Model Systems registry, and recruitment flyers.

Potential participants were prescreened via phone or in-person to determine eligibility prior to study enrollment. The inclusion criteria for this study consisted of the following: history of a non-progressive spinal cord injury, inclusive of levels C1-T1, and at least 6 months post-injury, age 18 to 75 years, and able to provide informed consent. The exclusion criteria consisted of the following: orthopedic injuries or surgeries that would affect upper extremity function, comorbid traumatic brain injury or other neurological impairments that would affect cognition, or Botox injections within the last 3 months in the upper extremity. Additionally, the following vulnerable populations were excluded from enrollment: adults unable to consent, unless accompanied by a legally authorized representative, minors, pregnant women, and prisoners.

Study Design

Participants were asked to complete 2 visits: 1 in-person and 1 remote visit. The first visit was completed in-person to administer the GRASSP V1.0, and the second visit was conducted remotely to administer the rGRASSP. The rGRASSP visit was scheduled within 7 to 10 days of the in-person assessment. The rGRASSP visit was conducted using the rGRASSP kit and the assistance of a family member, friend, or caregiver aiding as a home assessor. The home assessors were not provided any formal training prior to the home assessment.

For the strength domain, the home assessor was instructed to stabilize the limb and provide light resistance during ROM (Grade 4 MMT) and heavy resistance during ROM (Grade 5 MMT). If a home assessor was not present, the participant was asked to use the contralateral limb to provide light resistance during ROM (Grade 4 MMT) and heavy resistance during ROM (Grade 5 MMT). For the sensibility domain, the home assessor was instructed to fluff the cotton swab and lightly touch areas of the hand as instructed by the clinician. If a home assessor was not present, the sensibility portion was not assessed. For the prehension ability domain, the home assessor was asked to place a paper in the participants hand and/or provide manual resistance to assess the strength of the grasp pattern. If a home assessor was not present, the participant was asked to use the contralateral limb to place the paper and/or provide manual resistance. For the prehension performance domain, the home assessor was asked to assist with the set up and stabilization of testing materials. If a home assessor was not present, the participant was asked to set up the materials and use the contralateral limb to stabilize testing materials.

The rGRASSP visit was administered via Webex (a secure video conferencing platform). To establish inter-rater reliability, the rGRASSP administration was video recorded and assessed by a second examiner at a later time-point. All examiners administering the GRASSP and rGRASSP were licensed physical and occupational therapists with experience administering the GRASSP V1.0. Examiners were trained on standardized procedures to administer the rGRASSP, including cueing to optimize camera angles and instructing home assessors on hand placement, correct application of resistance and joint stabilization, as necessary.

The reliability of rGRASSP versus GRASSP was assessed using Intra Class Correlation, a common statistical test used to assess consistency and reliability of quantitative measurements made by different instruments or observers but measuring the same quantity. Sample size was determined based on the desired precision of ICC estimate. In a study by Kalsi-Ryan et al the inter-rater and test-retest ICCs were above 0.80 for all but one of the GRASSP domains (sensation) and ranged from 0.84 to 0.98.⁶ Assuming true ICC = .80, a sample size of n = 55 evaluable participants who completed both assessments would provide a 95% one-sided confidence interval with a lower bound 0.703, which is within 0.1 of ICC = .80, providing sufficient precision. Greater precision would be achieved for higher ICC’s. This calculation was based on estimating the ICC using a mixed effects model with test type (GRASSP vs rGRASSP) as the fixed effect and subject as the random effect using PASS 2020 software. Assuming a 10% drop-out, we planned to enroll at least 62 participants.

Statistical Analysis

The demographics and clinical characteristics of the study participants were summarized using standard descriptive statistics. Because maximum GRASSP and rGRASSP scores differed for sensibility, prehension performance, and the total score, each participant’s score was divided by the maximum possible score and thus converted into percentage of the maximum score for all scales. For example, the maximum GRASSP and rGRASSP total scores were 116 and 94, respectively, and a participant with a score of 90 would receive 90/116 = 77.6% for GRASSP and 90/94 = 95.7% on rGRASSP. Patterns of agreement were visually examined using Bland–Altman plots.²¹ Bland–Altman plots allow us to assess the bias and magnitude of disagreement between measurements by plotting the difference between 2 measurements versus their average. Left and right measures were analyzed separately. We compared the following: (1) GRASSP versus rGRASSP (left and right-side measures), and (2) rGRASSP Examiner 1 versus rGRASSP Examiner 2 (left and right-side measures). A 2-way random effects model with the GRASSP score as the outcome, test type (in-person or remote) as the fixed effect, and random effects for subject and rater was fitted, and ICC was calculated based on this model. This model also provides an estimated difference between the 2 test types. Based on the 95% confident interval of the ICC estimate, values less than 0.5, between 0.5 and 0.75, between 0.75 and 0.9, and greater than 0.90 are indicative of poor, moderate, good, and excellent reliability, respectively.²² Agreement between multiple raters evaluating rGRASSP using the recorded session were analyzed similarly. Analyses were preformed using R v. 4.0.2, including the “psych” package for ICC analyses²³ and “blandr” package for Bland–Altman plots.²⁴

Results

Between Nov 24, 2020 to Dec 31, 2021, the study enrolled 61 participants from 3 different sites (n = 53 from Shirley Ryan Ability Lab, n = 7 from Hines VA Hospital, IL and n = 1 from Kessler Institute for Rehabilitation, New Jersey). Demographics and SCI-related characteristics of the study participants are presented in Table 2. Study participants were predominantly male (70%) and ranged in age from 20 to 79 years (mean 49 ± 15 years). Most participants were AIS D Classification (59%), with Motor Vehicle Accidents identified as the primary cause of SCI injury (36%). There was a wide range of variability in time since SCI (0.6–43.3 years). Among the 61 participants, 2 did not have access to a home assessor.

Table 2.

Demographic and Clinical Characteristics of Participants (N = 61).

Characteristics	N = 61a (%)
Cause of SCI
Motor vehicle accident	22 (36)
Fall	13 (21)
Act of violence (gunshot/knife wound)	7 (11)
Sports or recreation	10 (16)
Medical/surgical complication	1 (1.6)
Other	8 (13)
AIS
A	6 (9.8)
B	7 (11)
C	12 (20)
D	36 (59)
Gender
Female	18 (30)
Male	43 (70)
Age
Median	52
Mean (SD)	49 (15)
Range	20, 79
Race
White	39 (64)
Black	18 (30)
Asian	2 (3.3)
Other	0 (0)
Declined to answer	1 (1.6)
Missing	1 (1.6)
Ethnicity
Hispanic or Latino	6 (9.8)
Not Hispanic or Latino	48 (79)
Declined to answer	6 (9.8)
Missing	1 (1.6)
Level of education
No high school diploma	2 (3.3)
High school graduate or equivalent (eg, GED)	14 (23)
Some college credit, no degree	13 (21)
Associate degree	4 (6.6)
Bachelor’s degree	17 (28)
Master’s/doctorate/professional degree	11 (18)

Abbreviations: SCI, spinal cord injury; AIS, American Spinal Injury Association Impairment Scale; SD, standard deviation.

^an (%).

The description of the mean values for the GRASSP, rGRASSP Examiner 1, and rGRASSP Examiner 2 are provided in Table 3 below.

Table 3.

Descriptive Statistics of GRASSP and Remote GRASSP. Mean Values for the GRASSP, Remote GRASSP Examiner 1 (rGRASSP 1), and Remote GRASSP Examiner 2 (rGRASSP2) are Provided for Both Left and Right Hand.

Domain	Left						Right
	GRASSP		rGRASSP 1		rGRASSP 2		GRASSP		rGRASSP 1		rGRASSP 2
	Mean	SD	Mean	SD	Mean	SD	Mean	SD	Mean	SD	Mean	SD
Total	67.9	23.4	69.3	22.8	68.8	23.2	68.7	23.5	68.9	22.5	68.6	22.1
Strength	68.8	24.1	69.5	23.6	69.5	23.5	69.3	23.7	69.7	23.1	69.3	22.6
Dorsal + Palmar	72.6	27.2	64.1	26.6	62.9	27.1	73.7	26.9	63.7	27.5	63.8	26.5
Dorsal	69.4	29.7	68.6	25.3	67.7	25.6	72.0	28.3	67.6	29.7	67.7	28.7
Palmar	75.8	26.7	59.5	32.4	58.2	33.5	75.5	27.2	59.8	31.7	59.9	31.2
Prehension ability	67.4	30.4	70.3	32.1	66.8	33.7	67.1	30.2	70.4	30.4	68.3	31.6
Prehension performance	64.2	25.8	70.5	24.8	71.9	25.8	68.1	26.7	72.8	24.9	72.2	24.8

Summaries are percentages of total score.

Abbreviation: SD, standard deviation.

Validity of the rGRASSP

Overall, the remote GRASSP demonstrated excellent concurrent validity with the GRASSP. The ICC and 95% confidence interval (CI) values comparing GRASSP and rGRASSP are reported in Table 4.

Table 4.

Intraclass Correlation and 95% Confidence Interval Comparing the GRASSP and Remote GRASSP (Validity), and Comparing the Remote GRASSP Examiner 1 and Remote GRASSP Examiner 2 (Reliability) for Both the Left and Right Hand.

Scale	ICC_left	CI_left	ICC_right	CI_right
Validity
Total	0.96	(0.94–0.97)	0.96	(0.94–0.98)
Strength	0.96	(0.93–0.97)	0.95	(0.92–0.97)
Sensibility	0.59	(0.41–0.73)	0.68	(0.47–0.8)
Prehension ability	0.94	(0.91–0.96)	0.95	(0.92–0.97)
Prehension performance	0.92	(0.82–0.96)	0.93	(0.88–0.96)
Reliability
Total	0.99	(0.98–0.99)	0.99	(0.98–0.99)
Strength	0.99	(0.99–1)	0.99	(0.99–0.99)
Sensibility	1.00	(0.99–1)	1.00	(1–1)
Prehension ability	0.95	(0.92–0.97)	0.97	(0.95–0.98)
Prehension performance	0.95	(0.92–0.97)	0.94	(0.9–0.96)

Abbreviations: ICC, intraclass correlation coefficient; CI, confidence interval.

Figure 1 (top panel) displays the Bland–Altman plot comparing the in-person and remote GRASSP using the % total score, for both the left and right sides. The dashed lines represent the average difference (bias; middle dashed line) and the 95% limits of agreement (LoA, defined as ±1.96*standard deviation of the differences; outer dashed lines). The average difference was 0.11% (95% LoA: −12.6%, 12.8%) for the left side, and 1.2% (95% LoA: −10.7%, 13.2%) for the right side, suggesting limited bias and agreement within about ±12% points.

Figure 1.

Bland–Altman plots comparing GRASSP and remote GRASSP.

Bland–Altman plots comparing the in-person and remote GRASSP on each domain, for both the left and right sides, are also presented in Figure 1. For the Strength domain, the bias (limits of agreement) was 1.1% (−12.8%, 15.0%) for the left side, and 1.8% (−12.9%, 16.5%) for the right side. For the Sensibility domain, the bias was 10.0% (−35.6%, 55.7%) for the left side, and 11.4% (−28.3%, 51.2%) for the right side. Note that some of the limits of agreement are >30% and are therefore not shown. For Prehension Ability, the bias was −1.1% (−21.9%, 19.7%) for the left side, and −1.7% (−20.8%, 17.3%) for the right side. For Prehension Performance, the bias was −5.4% (−22.0%, 11.2%) for the left side, and −3.8% (−21.1%, 13.6%) for the right side.

Inter-Rater Reliability of rGRASSP

Comparison rGRASSP total scores for Examiner 1 versus Examiner 2 demonstrated high inter-rater reliability, with ICC = .99 for both the right and left sides (95% CI [0.98, 0.99] for both sides). The ICC and CI values comparing rGRASSP Examiner 1 and rGRASSP Examiner 2 for each domain are reported in Table 4.

Figure 2 (top panel) illustrates reliability by using Bland–Altman plots comparing the percent total scores of rGRASSP Examiner 1 versus rGRASSP Examiner 2. The bias (limits of agreement) was 0.22% (−6.4%, 6.9%) for the left side, and 0.60% (−6.3%, 7.5%) for the right side. The Bland–Altman plots for the individual domains for rGRASSP Examiner 1 versus rGRASSP Examiner 2 are also displayed in Figure 2. The bias (limits of agreement) for each domain were as follows: Strength, −0.20% (−4.9%, 4.5%) for the left side, and 0.45% (−5.4%, 6.3%) for the right side; Sensibility, 0.17% (−5.1%, 5.4%) for the left side, and 0.17% (−2.2%, 2.5%) for the right side; Prehension Ability, 3.4% (−15.7%, 22.5%) for the left side, and 2.4% (−12.3%, 17.2%) for the right side; Prehension Performance, −1.8% (−17.2%, 13.6%) for the left side, and 0.29% (−17.1%, 17.7%) for the right side.

Figure 2.

Bland–Altman plots comparing remote GRASSP examiner 1 and remote GRASSP examiner 2.

Discussion

In this study, we demonstrated that the rGRASSP is a valid and reliable measure for GRASSP, and therefore provides a solution to the barriers that exist in comprehensive remote assessment of the upper extremity function in the SCI patients. In our study, the rGRASSP demonstrated robust psychometric properties against the GRASSP V1.0, highlighting the value of a virtual assessment that can garner statistically comparable results to the “gold standard” primary outcome measurement in clinical trials for the SCI population. Concurrent validity was particularly high in the Strength, Prehension Ability, and Prehension Performance domains. Additionally, our study suggests that the rGRASSP could produce comparable results when the same participant was assessed by 2 examiners. These strong psychometrics of the rGRASSP serves to enhance clinical and research evaluation of neurological recovery and hand function in individuals with tetraplegia.

Importance of Remote Assessments in SCI

The rGRASSP is a valuable contribution to a nascent field of psychometrically strong assessments that can provide researchers and clinicians with valuable, objective information in the field of telemedicine. Even before the global COVID-19 pandemic, the feasibility of telemedicine was growing. By 2018, 88.2% of people with SCI had access to internet-based devices.²⁵ When COVID-19 came to the forefront, clinician and patient capacity to deliver remote therapy was accelerated by sheer necessity.²⁶ Emergency expansion of the Medicare authorization for telemedicine during the COVID-19 pandemic facilitated transitioning what was formerly in-person care to telemedicine, as well as creation of new programs built exclusively for telemedicine.¹⁷ Telemedicine is now a more viable option for assessment and treatment administration than ever before.²⁷

Specifically in the SCI population, where there are many barriers to transportation¹² and locations for SCI-specialized quality care are often far-removed from the patients, telemedicine is especially salient.²⁸ Research regarding the efficacy of telemedicine in the SCI population suggests it is feasible and effective.^29,30 While the body of research about telemedicine intervention grows, limited progress has been made in regards to valid remote assessments for arm and hand function, especially within the SCI population. The rGRASSP serves to address this key gap.

Concurrent Validity of rGRASSP to GRASSP V1.0

This study provides strong evidence for the concurrent validity of the rGRASSP with an overall ICC of 0.96. The validity of the domains varied but was excellent for Strength, Prehension, and Prehension Performance. The Sensibility domain of rGRASSP demonstrated moderate validity compared to the GRASSP V1.0. The Sensibility domain of the rGRASSP may be limited in the concurrent validity because it consists of the greatest deviation in testing method from the GRASSP V1.0. The Sensibility domain of the GRASSP V1.0 consists of key test locations that represented significant anatomical levels of sensory innervation and functionally key areas of the hand were selected and tested with Semmes-Weinstein monofilaments (SWM). The use of SWM is a well-established sensory testing approach that demonstrates excellent psychometric properties.³¹ The SWM could not be used in the home kits due to the cost of the monofilaments and the clinical training that is needed to administer the assessment.

An alternate method of testing Sensibility, using cotton swabs to perform light touch, was chosen instead of the SWM. This testing method is similar to the light touch sensory component of International Standards for Neurological Classification of Spinal Cord Injury (ISNCSCI) Classification examination, where cotton swabs are applied to dermatome sites of the upper extremity. Since 1982, the ISNCSCI has been used to classify sensation of SCI through pinprick and light touch scores. The ISNCSCI and the American Spinal Injury Association (ASIA) Impairment Scale (AIS) have been referred to as the gold standard of SCI classification.³² The cotton swab technique requires minimal training and is cost effective. Despite the moderate validity of the Sensibility domain due to the deviation in testing method, the use of cotton swabs to perform sensory testing for spinal cord injuries has proven to be a reliable method in people with SCI.³³

Inter-Rater Reliability of rGRASSP

This study provides strong evidence of inter-rater reliability with the total scores for rGRASSP Examiner 1 versus rGRASSP Examiner 2, yielding an ICC = .99 for the total rGRASSP score. The Strength domain yielded a 0.99 ICC for both hands. While the Sensibility domain yielded the highest reliability with a ICC of 1.00 for both hands, this scale has the smallest maximum score of all domains (ie, 12) so there is less room for disagreement between raters. The Prehension Ability domain yielded an ICC of ≥0.95, and the Prehension Performance domain yielded an ICC of ≥0.94 for both hands. Similar to the ICC, the Bland–Altman plots indicated high reliability, especially for the Sensibility and Strength domains. The individual domains indicate that there is negligible bias and differences between the mean percentage scores for rGRASSP Examiner 1 versus rGRASSP Examiner 2. These results show excellent interrater reliability and strengthens the psychometric properties of the scale. Real time inter-rater reliability performed by naïve experts needs to be included in future studies.

Limitations

Although the rGRASSP provides many benefits to the growing field of remote assessment and telemedicine, there are limitations that exist when performing remote assessments in the SCI population. One important limitation is the use of a home assessor. The rGRASSP required a family member, caregiver, or acquaintance to serve as a home assessor when completing the Strength and Sensibility domain. Availability of a home assessor is also helpful for the setup of the Prehension Ability and Prehension Performance domains. Since the home assessors typically do not have a clinical background, their knowledge of medical terminology and anatomical landmarks are limited, presenting challenges for rGRASSP Examiner 1 during cueing of the home assessor. Additionally, the rGRASSP relies on the home assessor’s interpretation of levels of resistance (moderate resistance and maximum resistance) when applying force to score grade 4 and 5 Manual Muscle Test (MMT) of the Strength domain.

In the event that a home assessor is not present, the Sensibility domain could not be assessed. The Strength domain could be assessed without a home assessor, using the participant’s contralateral side for resistance. However, the usage of the contralateral side presents new challenges. The usage of the contralateral side requires the participant to have adequate strength in the contralateral arm supplying resistance. Similar to using a home assessor, the accuracy of the Strength domain is also dependent on the interpretation of levels of resistance when applying force to score a grade 4 and 5 MMT of the Strength domain.

A second limitation is that our study was limited to chronic SCI population as our inclusion criteria required study participants to be at least 6 months post injury. At the chronic stage, most SCI patients have stabilized in terms of medical complexity, independence with activities of daily living, and participation in allied health services. Alternatively, individuals classified as of the acute/sub-acute SCI population may still be experiencing complications such as bradycardia, hypotension, pain, or spasticity.³⁴ Due to the frequency and complexity of these complications, the rGRASSP is likely not an applicable assessment for the acute/sub-acute SCI population. In this study, we first performed the in-person version of the GRASSP assessment to complete the necessary informed consent and medical history screening procedures, as well as provide participants with the remote GRASSP kit. However, performing the in-person GRASSP before the remote version may have familiarized participants with the assessment procedures, potentially confounding the results of the remote assessment.

Practical Considerations for Using the rGRASSP

We have established a new methodology to administer an existing assessment remotely that measures hand function after SCI and demonstrates excellent validity and reliability. Among the benefits of the rGRASSP are its affordability and accessibility. The investigators were able to produce a prototype of the rGRASSP kit using items obtained from online retailers. For this study, standardized kits were purchased from Neural Outcomes Consulting Inc. and provided to study participants after attending in-person GRASSP assessment. Participants were also provided with a shipping box, return shipping label, and a list of FedEx shipping locations near their home. Despite the affordability and accessibility of the rGRASSP, it is important to examine the challenges within the process of distributing and receiving the rGRASSP kits from participants. The rGRASSP kits must be sanitized, assembled, and a prepaid shipment label must be obtained prior to the participant’s scheduled session. This can have clinical implications, as a fair amount of planning is needed to ensure all equipment is ready for the participant. Measures are also needed to obtain the materials after the evaluation is completed. Since access to transportation remains a barrier for the SCI population, returning the kits to a local shipping location can present a challenge. The rGRASSP will be commercially available, with components to address some of the limitations.

Another consideration is the need for access to technology and technological competency in the SCI population. Over the course of this study, many participants faced a variety of technological issues, including finding optimal camera angle, maintaining adequate sound, difficulty logging in, or downloading the video conferencing application. Ultimately, the choice to implement the rGRASSP requires planning days ahead of the session. By contrast, assessments that require observation or patient-reported scales can be used with an immediacy that is often needed. Finally, since the total score of the rGRASSP is different from the GRASSP, it is recommended that the percentage of the maximum score for all scales be used for comparison between the 2 assessments. This will allow for a more accurate comparison.

Conclusions

Our study demonstrates that the rGRASSP is a reliable adjuvant and/or alternative to GRASSP V1.0 for remote assessment of hand function in individuals with SCI. Moreover, rGRASSP of the same person can be reliably completed by different rehabilitation professionals. Thus, this tool can be used to facilitate monitoring of hand function in a clinical, community, or research setting, particularly during the COVID-19 pandemic when traditional in-clinic assessment options are limited for individuals with SCI.