Clinical Utility of GRASSP in Traumatic Tetraplegia: A Narrative Review and Retrospective Analysis Incorporating Machine Learning with an Explainability Framework

Clinical Utility of GRASSP in Traumatic Tetraplegia: A Narrative Review and Retrospective Analysis Incorporating Machine Learning with an Explainability Framework | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Clinical Utility of GRASSP in Traumatic Tetraplegia: A Narrative Review and Retrospective Analysis Incorporating Machine Learning with an Explainability Framework Sukhvinder Kalsi-Ryan, Yuchen Wang, Danvir Sandhu, Jose Zariffa, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8683387/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 8 You are reading this latest preprint version Abstract Study Design: Narrative Review and Retrospective Post-Hoc Analysis Objectives: 1. Summarize GRASSP literature and provide evidentiary tables to support clinical research decision making 2. Establish enrollment cut-offs for acute and cross-sectional chronic samples 3. Define recovery profiles of GRASSP subtests and typical relationships of impairment to function. Methods: A literature review of existing GRASSP publications was conducted. Two datasets were created from three historical cohorts of data collected for GRASSP development. A longitudinal dataset includes a total of 182 individuals with tetraplegia, each of whom was assessed at 0, 1, 3, and 6 months after injury (728 observations). The cross-sectional chronic dataset includes a total of 254 individuals with tetraplegia. The sample was stratified by AIS at Exam Stage 1. AIS classification of A-B were assigned to the motor complete group and AIS classification of C-D were assigned to the motor incomplete group Analysis: Literature was reviewed and knowledge synthesized. Data was summarized with descriptive statistics. Machine learning was applied to both datasets for prediction. Results: Evidentiary tables summarize three versions of GRASSP and synthesize the literature to define the clinical research utility. Three figures define the discriminative groupings of GRASSP subtest scores along with recovery profiles. Machine learning identifies the predictive qualities of GRASSP Strength. Conclusion: GRASSP use recommendations are meaningful to researchers/clinicians for implementation and sample characteristics of historical data provide insight for researchers into expected outcomes of individuals with traumatic tetraplegia. Health sciences/Medical research/Outcomes research Health sciences/Health care spinal cord injury traumatic tetraplegia measurement properties reliability validity responsiveness minimal clinically important difference minimally detectable difference recovery profiles machine learning functional independence AIS classification cross-sectional chronic spinal cord injury longitudinal analysis upper limb function Figures Figure 1 Figure 2 Figure 3 Brief Informative This paper includes: A summary of three versions of GRASSP and clinical research recommendations that are useful to researchers and clinicians for implementation of GRASSP. Sample characteristics of historical data provide insight for researchers into expected outcomes of individuals with traumatic tetraplegia and additional metrics for future study design in both acute and chronic studies related to traumatic tetraplegia. Introduction Since 2012, the Graded Redefined Assessment of Strength, Sensibility, and Prehension (GRASSP) has been utilized as a primary, secondary, or exploratory measure in numerous trials and studies, including recent interventional work such as the Nogo‑A antibody trial in acute cervical spinal cord injury (SCI) [ 1 ], the Up‑LIFT pivotal trial by ONWARD Medical using ARC‑EX® stimulation in individuals with chronic tetraplegia [ 2 ], and a 2024 case series evaluating temporal interference stimulation [ 3 ]. Although there is a comprehensive body of research detailing the measurement properties and applications of GRASSP, there is now an opportunity to summarize and analyze historical data to guide study design, endpoint formulation, and provide historical values for comparison in interventional studies. There has been approximately 15 years of experience with use of the GRASSP and this information has clinical and research utility. By performing this summative work, we are able to move closer to: 1) Avoiding ceiling and floor effects; 2) Improving detection of clinically meaningful neurological change; 3) Enhancing statistical performance; and 4) Aligning with FDA/EMA expectations for sensitive and reliable endpoints. GRASSP Version 1 (GV1) was released in 2009, combining the Link Hand Function Test and the Tetraplegia Hand Measure [ 4 , 5 ]. The development phase of GV1 spanned from 2006 to 2009, funded by not-for-profit agencies and engaging six individuals on the GRASSP development group. The primary aim was to address the gap in assessing subtle changes in the upper limb following traumatic cervical spinal cord injury. The GRASSP is specifically designed to evaluate upper limb function after traumatic spinal cord injury. Beyond its development, the measurement properties of reliability, validity, and responsiveness were established, making it a highly sought-after tool that met several regulatory criteria for clinical research [ 4 , 6 , 7 ]. At the time, it was the only measure available with established measurement properties, interpretability, and generalizability. Therefore, it was accepted by the research community as it could be implemented in multiple centers, which made it ideal for multi-center and multi-national use. GRASSP attracted interest from industry and was commercialized through licensing by University Health Network. Neural Outcomes Consulting, Inc. was established in 2011 to manage the commercial sales of GRASSP and GRASSP products. The responsiveness properties of GRASSP were published in 2015, based on two complementary studies conducted between 2010 and 2013—one by Kalsi-Ryan et al. in Canada and the other by Velstra et al. in Europe [ 6 , 7 ]. In 2018, Velstra et al. leveraged longitudinal data to identify item redundancies within GRASSP, informing the development of a Version 2 [ 8 ]. The revised version included updated measurement properties, a new scoring manual, and a clear rationale for item reduction [ 9 ]. Approximately a year later, further adaptations were introduced to improve sensitivity to impairments specific to Degenerative Cervical Myelopathy (DCM), resulting in the development of GRASSP Version Myelopathy (GVM). Both the measure and measurement properties of this specialized version were documented in a subsequent development paper [ 10 ]. With a large knowledge base of GRASSP measures, there is no existing summary or illustration of the historical data. GRASSP has been employed in numerous sponsored trials, investigator-driven studies, and academic projects by graduate students, clinicians, and researchers. There is a need for historical summaries to facilitate comparisons with existing studies and to aid in the design of future research. Therefore, the developers have pursued this work to update the body of knowledge related to GRASSP. The aim of this review and post-hoc analysis is to integrate and summarize historical datasets to generate metrics that inform future study design, endpoint formulations, provide historical benchmarks for interventional studies and provide the evidence to support GRASSP as an important or primary endpoint in clinical trials. The objectives were to: Summarize GRASSP literature and provide usable evidentiary tables to support clinical research decision making To determine enrollment thresholds for acute and chronic samples, considering floor and ceiling effects. To define recovery profiles, descriptive statistical values and characterize typical relationships between impairment and function in a standard sample using machine learning. Methods Literature Review : A structured literature search was conducted in PubMed, Scopus, and Google Scholar between May and June 2025. The search strategy employed combinations of keywords such as “GRASSP,” “GRASSP Version 1,” “GRASSP Version 2,” “GRASSP Myelopathy,” “GV1,” “GV2,” “DCM,” “tetraplegia,” “cervical SCI,” “validity,” “reliability,” “responsiveness,” “minimally detectable difference,” and “minimal clinically important difference.” Search parameters were limited to peer-reviewed, English-language articles published between 2010 and 2025, with a focus on studies that evaluated measurement properties in traumatic cervical spinal cord injury (SCI) and degenerative cervical myelopathy (DCM) cohorts. Evidentiary tables were synthesized from the articles reviewed. Data Preparation and Preprocessing : For this analysis, we focus on two derived datasets: the longitudinal dataset and the cross-sectional chronic dataset. Both were constructed by combining and reformatting the aforementioned sources. The longitudinal dataset combines observational data from the 2010–2013 [ 6 ] studies which include observations from Canada and the EMSCI [ 1 , 7 ]. It includes a total of 182 patients, each assessed at 0, 1, 3, and 6 months post-injury, resulting in 728 observations. Note that 0 refers to the baseline assessment that occurs generally between 0–14 days after injury. The cross-sectional chronic dataset includes observations from the 2009 cross-sectional chronic SCI cohort (n = 72) [ 4 ], along with all occurrences from the longitudinal dataset at 6 months post-injury [ 1 , 6 , 7 ]. This results in a combined total of 254 observations. The longitudinal dataset contains repeated observations of the same patients at multiple time points, whereas the cross-sectional chronic dataset includes only a single observation per patient. By utilizing both datasets, we can gain a more comprehensive understanding of patient recovery—capturing not only the recovery profile at a specific time point, but also the trajectory of recovery over a six-month period. In the longitudinal analysis, the sample was grouped based on the American Spinal Injury Association Impairment Scale (AIS) classification at baseline. AIS grades A or B were categorized as motor complete, while those with grades C, D, or E were classified as motor incomplete. Some variables were summed to simplify the analysis. The variable named “str_Total” is the right and left GRASSP strength total, summing the scores of 10 muscle strength tests on both the left and right sides of the body, resulting in a total score ranging from 0 to 100. Scores assessing upper limb tasks from the GRASSP Prehension Performance were combined to create a new variable called “PP_total”. From the spinal cord independence measure (SCIM) [ 15 ] we focus on the self-care subscore which ranges from 0–20 and is the sum of items 1–5. We focus on the SCIM selfcare subscore, as it primarily assesses self-care abilities and upper limb function. In addition, a total SCIM score was computed by summing all 17 items. Distribution of cross-sectional chronic sample - In this study, boxplots were generated for each of the five AIS grades (A, B, C, D, and E) within the cross-sectional chronic dataset to define the differences in distribution of functional scores between groups. Additionally, examining the median values provides insight into the typical functional score for each group. Single muscle Relationship with Function - Pearson correlation coefficients between individual muscle strength scores and functional measures (SCIM and GR-PP scores). In the cross-sectional chronic dataset these were calculated to assess the strength of association between each muscle and overall functional performance. This process was repeated with the longitudinal dataset at two critical time points: 1 month and 6 months post-injury. Recovery Profiles - were generated with the longitudinal dataset at 1, 3, and 6 months post-injury to illustrate the recovery trajectories of patients grouped as motor complete and motor incomplete. Baseline (immediately post-injury) were excluded in this summary as there is inconsistent timing of the baseline. Enrollment Cut-offs for GRASSP strength and prehension performance - Typically, studies use the upper extremity motor score from the International Standards of Neurological Classification for Spinal Cord Injury (ISNCSCI) as a screening tool for eligibility in acute and chronic trials. We ordered all chronic and acute occurences based on UEMS, and then reviewed the GRASSP strength and Prehension Performance scores. For the chronic sample we implanted the MDD values to determine which individuals could continue to show measurable change. For the acute sample we reviewed GRASSP strength and Prehension Performance at endpoint. Based on the endpoint results we were able to identify expanded cut-off scores for the acute sample. Muscles Strength Relationships to SCIM-SS - To identify which of the 10 muscles most strongly influence the self-care abilities of individuals with SCI, we applied machine learning models to both cross-sectional chronic and longitudinal datasets. The target variable was the categorized SCIM self-care score, while the primary inputs were 20 individual muscle strength scores (left and right sides of 10 muscles). AIS was also included as a categorical indicator of motor and sensory function loss. In the longitudinal dataset, patients were grouped by motor complete vs. motor incomplete at 1 month post-injury. The same classification was applied to the cross-sectional chronic dataset, which includes a single observation per patient. Patients in both datasets were further categorized into two functional groups based on SCIM self-care scores: those scoring 0–14 (low independence) and those scoring 15–20 (high independence). Separate models were trained for the cross-sectional chronic and longitudinal datasets. For the cross-sectional chronic dataset, input features included the 20 muscle strength scores and complete/incomplete group. The target variable was the binary self-care score group (low vs. high independence). For the longitudinal dataset, where patients were evaluated at four time points, we used muscle strength scores and complete/incomplete group recorded at 1 month post-injury as inputs, with the self-care score at 6 months post-injury as the target. This setup allowed us to assess how early motor function predicts longer-term self-care outcomes. In this study, four ML algorithms were investigated: random forest [ 16 ], eXtreme gradient boosting (XGBoost) [ 17 ], K-Nearest Neighbour (KNN) [ 18 ], Support vector machines [ 19 ]. We employed grid search [ 20 ] during model training to optimize hyperparameters for each algorithm. For each candidate hyperparameter configuration, model performance was assessed using k-fold cross-validation on the training set, and the configuration with the highest mean cross-validated performance was selected. This procedure ensured that hyperparameter tuning was based solely on training data and was not influenced by the test set. Model performance was then evaluated using standard metrics, including accuracy, precision, recall, F1 score, support, and AUC. The best-performing model was subsequently interpreted using SHAP (Shapley Additive Explanations), which provided both quantitative values and visualizations to explain the contribution of each input feature to the model's predictions [ 21 ]. Results Literature Review : A comprehensive review of the literature on the GRASSP Versions was synthesized into two evidentiary tables to support clinical decision-making and measure selection in SCI and DCM. Table 1 provides a side-by-side comparison of GRASSP versions—GV1, GV2), and GVM, detailing subtest domains, item descriptions, maximum scores, and the equipment required for each version. The table highlights the structural and psychometric evolution from GV1’s comprehensive multi-domain model to GV2’s Rasch-refined iteration and ultimately to the highly time-efficient and targeted GVM, which omits prehension modules not relevant to early-stage DCM detection. While GV1 includes five bilateral subtests (strength, dorsal and palmar sensibility, prehension ability, and prehension performance), GV2 eliminates the dorsal sensibility component and incorporates psychometric refinements [ 12 ]. GVM streamlines the tool further, using only three subtests—strength, palmar sensation, and dexterity—adapted to detect the subtle deficits of DCM with minimal equipment [ 6 ]. Table 1 GRASSP Versions – Subtests, Item Descriptions, and Maximum Scores Subtest Domain GRASSP V1 Max Score (V1) - Per Side GRASSP V2 (GV2) Max Score (V2) - Per Side GRASSP-M Max Score (GVM) - Per Side Equipment Required GRASSP Version 1 Kit — GRASSP Version 1 Kit or Version 2 Kit and Version 2 Manual — GRASSP Version 1 or 2, or Myelopathy Kit and Version Myelopathy Manual. Adapted from original GRASSP kit. No full prehension module required [ 10 ]. — Strength 10 upper limb muscle groups (C5–T1); MRC scale 0–5 [ 4 ] 50 Same 10 muscle groups; transitioned from isotonic to isometric testing for improved standardization [ 9 ] 50 Same as GV2 50 Dorsal Sensibility 3 dorsal hand sites; Semmes-Weinstein monofilaments; scored 0–4[ 4 ] 12 Removed due to low Rasch functional relevance[ 9 ] — Not included[ 10 ] — Palmar Sensibility 3 palmar hand sites; Semmes-Weinstein monofilaments; scored 0–4[ 4 ] 12 Retained with Rasch-optimized 3-category scale[ 9 ] 12 Same as GV2 12 Prehension Ability 3 grasp types (cylindrical, lateral/key, tip-to-tip); scored 0–4[ 4 ] 12 Same grasp types; Rasch-calibrated scoring and standardized instructions[ 9 ] 12 Not included NA Prehension Performance 6 tasks (e.g., pegboard, jar, coin, key turn); scored 0–5[ 4 ] 30 Reduced to 4 tasks (e.g., jar and coin tasks removed); simplified with Rasch hierarchy[ 9 ] 20 Reduced to 1 dexterity task 9 Legend : • MRC : Medical Research Council scale (0 = no contraction, 5 = normal strength) • Rasch : Refers to Rasch analysis, a psychometric method used to refine scoring and item calibration Table 2 outlines the measurement properties of each GRASSP version across reliability, validity, responsiveness, and clinical utility. All versions demonstrate excellent inter- and test retest-reliability and strong construct validity across SCI or DCM severity levels [ 6 , 9 , 10 ]. Responsiveness is particularly well-documented in GV1 and GV2, with standardized response means (SRM) ranging from 0.67 to 1.35 across various subtests and timepoints [ 6 , 9 ]. GV2’s Rasch analysis supported item-level uni-dimensionality and reduced administration time shows are reduced measure without sacrificing measurement sensitivity [ 9 ]. GVM shows validity and discrimination across DCM severity categories, with strong intra- and test retest reliability ICCs for the dexterity subtest (≥ 0.79) [ 10 ]. Minimum detectable differences (MDD) are reported for GV1 and GV2, offering benchmarks for real clinical change. Minimal clinically important differences (MCID) for interpreting meaningful change are also reported for GV1 and GV2 [ 6 , 9 , 12 ]. Although GVM lacks direct MCID reporting, its strong known-group validity and high ICC values suggest sufficient sensitivity to detect clinically relevant impairment gradations in DCM [ 4 ]. Table 2 Evidentiary Summary of GRASSP, Versions and Measurement Qualities GRASSP Version 1 (GV1) GRASSP Version 2 (GV2) GRASSP Version Myelopathy (GVM, GRASSP-M) Purpose To assess upper limb sensorimotor impairment in individuals with cervical spinal cord pathology—both traumatic and degenerative—by quantifying changes in neurological function and hand performance across the full spectrum of severity and recovery, from acute to chronic phases in traumatic SCI and from mild to severe impairment in degenerative cervical myelopathy (DCM). Description GRASSP Version 1 (GV1) assesses five subdomains—strength, dorsal/palmar sensory function, prehension ability, and prehension performance—independently for each hand. It is validated for use across acute and chronic phases of traumatic SCI [ 6 , 11 ]. GRASSP Version 2 (GV2) refines GV1 using Rasch analysis to improve psychometric properties and streamline test items. It preserves the original structure and clinical scope, while enhancing feasibility, precision, and applicability in trials and clinical settings [ 8 , 9 ]. Each hand is assessed separately. GRASSP-Myelopathy (GRASSP-M / GVM) is an adapted version of GV1 tailored to detect early, subtle upper limb deficits in Degenerative Cervical Myelopathy (DCM), such as intrinsic hand muscle weakness, sensory decline, and reduced dexterity. It offers a sensitive, low-burden alternative to broader tools like the mJOA [ 10 ]. ICF Domain Body Structure, Body Function, Activity [ 9 , 11 ] Time to Administer 30–45 minutes per patient [ 6 ] Approximately 15–30 minutes (approximately 15 min less time than GV1) [ 9 ] 10–15 minutes (significantly reduced from GRASSP V1’s 45-minute duration) [ 10 ] Domains, Items and Subtests Sensation • Items : Dorsal and Palmar Sensation • Subtests : Dorsal Sensation (12), Palmar Sensation (12) Strength • Items : C5 - Biceps, Ant Deltoid, C6 - Wrist Extensors, C7 - Triceps, C7 - Group Phallic • Subtests : Strength (50) Prehension • Items : C5 - Shoulder Flexion, Ext Dig, Finger Flexion, Lateral Key, Finger Abduction • Subtests : Prehension Ability (12), Prehension Performance (30) Sensation • Items : Palmar Sensation • Subtests : Palmar Sensation (12) Strength • Items : C5 - Biceps, Ant Deltoid, C7 - Triceps, C7 - Group Phallic • Subtests : Strength (50) Prehension • Items : C5 - Shoulder Flexion, Ext Dig, Finger Flexion, Lateral Key, Finger Abduction • Subtests : Prehension Ability (12), Prehension Performance (20) Sensation • Items : Palmar Sensation • Subtests : Dorsal Sensation (12) Strength • Items : C5 - Biceps, Ant Deltoid, C6 - Wrist Extensors, C7 - Triceps, C7 - Opponens Pollicis, C8 - Finger Flexion, Ext Dig, Flex Pollicis Long, T1 - Finger Abduction, First DI • Subtests : Strength (50) Prehension/Dexterity (exclusive to GVM) • Items : Tip to Tip, Tripod, Nuts • Subtests : Prehension (12) Reliability Excellent inter- and test-retest reliability: • Sensibility ICC = 0.84–0.95 • Strength ICC = 0.95–0.98 • Prehension Ability ICC = 0.95–0.98 • Prehension Performance ICC = 0.93–0.96 Validated in both adults and children (≥ 6 years) [ 6 , 12 , 14 ] Excellent inter- and test-retest reliability across subtests: • Strength: ICC = 0.95–0.98 • Palmar Sensation: ICC = 0.84–0.97 • Prehension Ability: ICC = 0.95–0.98 • Prehension Performance: ICC = 0.96–0.97 [ 9 ] Strength and Sensation – see GV2 • Dexterity Intra-rater ICC: 0.868 (dominant), 0.790 (non-dominant) • Dexterity Inter-rater ICC: 0.869 (dominant), 0.862 (non-dominant) All values > 0.75 indicate strong reproducibility across raters [ 10 ]. Construct Validity • Sensory agreement with ISNCSCI: Kappa = 0.41–0.51 • Motor innervation discordance shows GRASSP detects differences missed by ISNCSCI • Designed to detect discordance due to expanded sites and use of Semmes-Weinstein thresholds [ 4 , 6 , 11 ] All subtests significantly differentiated AIS grades A–D (p < 0.01); Rasch analysis confirmed item fit, uni-dimensionality, and strong internal consistency (Cronbach's α = 0.85–0.97 across domains) [ 2 , 9 ]. Demonstrated strong known-group discrimination across mJOA-defined severity levels (mild, moderate, severe): • Dexterity: p < .001 across all group comparisons • Strength & sensation: significant differences (p < .05 to < .001), particularly in moderate vs. severe cases • Bonferroni correction applied (p ≤ .016) to ensure statistical rigor [ 10 ]. Concurrent Validity Strong correlation with functional measures: • SCIM: r = 0.57–0.68 • SCIM Self-care: r = 0.74–0.79 • CUE/CUE-Q: r = 0.76–0.84 All p < 0.0001; validated in both adult and pediatric SCI populations [ 6 , 12 , 13 , 14 ] Moderate to strong correlations with functional outcome measures: • SCIM: r = 0.53–0.71 • SCIM Self-care: r = 0.72–0.82 • CUE-Q: r = 0.76–0.83 All significant at p < 0.001 [ 9 ] Moderate and significant correlations with mJOA: • Dexterity: r = 0.502–0.533 (p ≤ .05) • Dexterity Time: inverse correlation r = − 0.407 to − 0.469 (p ≤ .05) • Strength & Sensation subtests: r = 0.25–0.48 (p ≤ .05), showing convergent validity with both motor and sensory mJOA subscores [ 10 ]. Responsiveness Demonstrated across subtests: • SRM = 0.67–1.1 in GRASSP-PP comparisons • Compared favorably with CUE-T, though floor effects noted in GRASSP for minimal function [6. 13] Demonstrated meaningful change across timepoints (1mo–6mo and 1mo–12mo): • Sensibility: SRM = 0.84–1.35 • Prehension Performance: SRM = 0.67–1.22 Validated across subgroups (AIS A–D); supports longitudinal monitoring [ 9 ] Not directly measured in original study, but tool demonstrated clear gradation of impairment across mJOA-defined severity levels. Sensitivity to mild DCM makes it well-suited for detecting progression or therapeutic response [ 10 ] MDD Minimum change required to exceed error: • Strength (bilateral): ≥ 7 points • Sensation (bilateral): ≥ 4 points • Prehension Ability: ≥ 4 points • Prehension Performance (bilateral): ≥ 6 points Values based on SEM/SRD and validated at 95% CI [ 6 ] Reported per subtest: • Strength (bilateral): ≥ 7 points • Palmar Sensation (bilateral): ≥ 3 points • Prehension Ability (bilateral): ≥ 4 points • Prehension Performance (bilateral): ≥ 3 points Standard error and SRD also provided for item-level precision [ 9 ] Not reported MCID Derived via Global Rating of Change (GRoC) for acute SCI: • Strength (bilateral): 13–19 points • Prehension Performance (bilateral): 3–12 points • Sensibility showed no consistent MCID Valid only for acute (0–6 months) SCI samples [ 12 ] Not explicitly reported; SRM and MDD suggest values above noise threshold. Interpretation supported by statistical responsiveness [ 9 ] Not reported Year of Copyright 2008 (validated in 2012, 2015, 2022) [ 4 , 11 ] 2019 - Published by Kalsi-Ryan et al. [ 9 ] 2020 - Developed/validated by Kalsi-Ryan et al. as a DCM-specific adaptation of GV1 [ 10 ] Training Clinicians can self-train using manual and eLearning modules; typically requires 2–3 supervised administrations to achieve proficiency. Research Use: Group training is recommended to ensure standardized administration across raters; this may be done via formal sessions or team-based review and calibration. Legend : • ICF Domain : Refers to domains outlined by the WHO’s International Classification of Functioning, Disability and Health (ICF), including body function, structure, and activity limitations. • ICC (Intraclass Correlation Coefficient) : A measure of reliability; values > 0.75 are considered excellent. • Kappa : A statistic reflecting inter-rater agreement beyond chance (0.41–0.60 = moderate, 0.61–0.80 = substantial). • r (Pearson’s Correlation Coefficient) : Measures strength of association between tools (0.3–0.5 = moderate, > 0.5 = strong). • SRM (Standardized Response Mean) : A measure of responsiveness; SRM > 0.8 indicates large sensitivity to change. • SEM (Standard Error of Measurement) : Reflects the precision of individual scores. • SRD (Smallest Real Difference) : The minimum score change required to surpass measurement error with 95% confidence. • MDD (Minimal Detectable Difference) : The smallest difference that can be interpreted as a real change, not measurement variability. • MCID (Minimal Clinically Important Difference) : The smallest change perceived as beneficial by patients, determined via anchors like Global Rating of Change. • mJOA (modified Japanese Orthopaedic Association score) : A validated clinical scale assessing DCM severity. • SCIM (Spinal Cord Independence Measure) : A functional scale for daily living tasks in SCI. • CUE / CUE-Q : Capabilities of Upper Extremity Questionnaire; a patient-reported measure of upper limb function in SCI. • AIS Grades : American Spinal Injury Association Impairment Scale levels (A–E) for classifying spinal cord injury severity. • Bonferroni Correction : A statistical adjustment used when multiple comparisons are made, to control for Type I error. Distribution of Cross Sectional Sample : In the boxplot, we can see how functional scores vary by AIS level, where AIS A represents the most severe impairment and E the least. Across all four metrics—muscle strength, SCIM self-care, UEMS and GRASSP-PP there is a consistent trend: scores generally increase from AIS A to AIS E, indicating better motor and functional performance with less severe impairment. Notably, AIS E individuals consistently achieve near-maximal scores, while AIS A individuals show the greatest variability and lowest median scores. The trend is especially pronounced in the total SCIM and muscle strength scores, emphasizing the relationship between AIS level and functional ability. The recovery profiles illustrate the patients in motor complete versus motor incomplete across four functional metrics: total muscle strength, GRASSP-PP, SCIM self-care sub score, and UEMS. Across all time points (1, 3, and 6 months), patients in the motor incomplete group consistently show higher performance compared to those in the motor complete. Furthermore, all metrics demonstrate a clear upward trend over time in both groups, indicating continuous functional improvement during the 6-month period. See Fig. 1 . Enrollment Cutoffs : In the motor incomplete samples of the chronic and acute samples, we found using the GRASSP strength as an enrolment cut-off allowed for a larger and more inclusive sample. In the chronic group (n = 133), using the UEMS of 40 as a cut-off rendered 46.6% (62/133) of the sample eligible, while using GRASSP Strength of 89 as a cut-off rendered 90% (120/133) of the sample eligible. If strength assessment is used to determine outcome, this is feasible, as the MDD is 7, thus a true clinical change can be determined without a ceiling effect. In the acute group we conducted a similar analysis by selecting at baseline based on UEMS and then determining at 6 months if there was or was not a ceiling effect with GRASSP strength. We found from a sample of 87, if using a UEMS cut-off of 28 or less at baseline, 60% (53/87) of the sample would be eligible for enrollment. If we consider the GRASSP strength endpoint remaining at 90 or less at 6 months, then the enrollment UEMS can be as high as 46 and this makes 96% (84/87) of the sample eligible at baseline. Therefore, GRASSP historical cohorts can be useful in establishing more inclusive cut-offs which avoid ceiling effects. Muscle relationships to function, illustrated in heatmaps provide a visual summary of how the strength of relationships between specific muscles and overall functional performance changes during the recovery process. By comparing the correlation patterns at these two time points, we can identify which muscles demonstrate a consistent association with functional outcomes and which may become more or less relevant as recovery progresses. This temporal perspective offers valuable insights into the dynamic nature of muscle-function relationships following spinal cord injury, potentially informing targeted rehabilitation strategies and clinical decision-making. The correlations between each of the 10 individual muscle strength scores and functional outcomes were computed and visualized in heatmaps at two time points: 1 month and 6 months post-injury. Warmer colors (reds) represent stronger positive correlations, while cooler colors (blues) indicate weaker correlations. Overall, the correlations tend to increase from 1 month to 6 months in both groups, with the motor complete group showing a more pronounced rise—several muscles exhibit strong correlations (above 0.5) by 6 months. The motor incomplete group also demonstrates improvement over time, though the correlations remain more moderate and consistent across muscles. Notably, Triceps consistently displays strong correlations with all functional outcomes across both groups and time points, suggesting it may serve as a key predictor of recovery. See Fig. 2 . The heatmaps provide some indicators of muscle strength to function and can be interpreted for comparison. Machine learning : Fig. 3 a-b defines the best model performances on the test sets for both the cross-sectional chronic and longitudinal datasets. For the cross-sectional chronic dataset, the KNN model achieved the highest test accuracy of 90%. We observed that both KNN and XGBoost demonstrated relatively high and consistent performance across five evaluation metrics compared to the other models. However, due to the advantage of SHAP for interpretability and the difficulty of interpreting KNN in high-dimensional data, we selected XGBoost as the final model for the cross-sectional chronic dataset. For the longitudinal dataset, all four models achieved similar test accuracies of approximately 80%. Notably, XGBoost attained the highest AUC score among them. Therefore, we also chose XGBoost as the final model for interpreting the longitudinal dataset. The SHAP bar plots (Fig. 3 c-d) and beeswarm plots (Fig. 3 e-f) illustrate the feature contributions of the XGBoost models fitted on the cross-sectional chronic and longitudinal datasets. In both datasets, the categorical variable AIS_Group , which indicates the patient’s AIS classification, emerged as the most influential feature in model predictions. Patients with AIS grades motor complete were more likely to have low self-care independence , while those with AIS grades motor incomplete tended to exhibit higher levels of independence . Regarding muscle strength scores, there were slight differences in feature importance across the two datasets. In the cross-sectional chronic dataset, Extensor Digitorum DIII had the strongest influence on self-care ability, followed by Opponens Pollicis and Anterior Deltoid . In the longitudinal dataset , Opponens Pollicis was the top contributor, followed by First Dorsal Interosseous and Extensor Digitorum DIII . Notably, the muscles Extensor Digitorum DIII , Opponens Pollicis , Anterior Deltoid , Wrist Extensors , and First Dorsal Interosseous consistently ranked among the top features in both datasets, further confirming their potential role in improving patients’ self-care ability. Discussion This study synthesizes existing literature to provide accessible, consolidated data for clinicians and researchers when evaluating the utility of GRASSP outcome measures. A comprehensive summary of the three versions is presented, allowing for quick reference and informed decision-making regarding outcome measure selection. The table outlines the structural features of each tool, including the subtests, item breakdowns, scoring structure, and equipment requirements. In parallel, meaurement properties—such as reliability, validity, responsiveness, and clinical interpretability—are summarized to help users quickly identify which version of GRASSP aligns best with their specific research or clinical application. Where more in-depth information is required, full methodological and validation details can be accessed in the referenced literature. These findings support all three versions of GRASSP as valid, reliable, and clinically applicable tools, with each optimized for specific neurological populations and assessment contexts. GV1 and GV2 are suited for comprehensive bilateral impairment tracking in traumatic SCI. They both offer psychometric rigor and efficiency for clinical trials and rehabilitation settings. GV2 having a reduced set of hand assessments is better suited for studies/trials investigating global change after SCI. GVM provides a streamlined, targeted approach for early identification and monitoring of upper limb deficits in DCM. This study also analyzes historical data to define normative profiles and functional stratifications based on AIS grade. By distinguishing where individuals fall within the cross-sectional chronic spectrum, this analysis provides a valuable reference for determining whether a patient's presentation aligns with expected recovery benchmarks. The longitudinal component of the data captures recovery trajectories over the first six months post-injury, offering a robust natural history framework against which treatment effects or cohort differences can be evaluated The boxplots of GRASSP scores provide a clear and discriminative depiction of functional differences across AIS grades A through E. The longitudinal recovery profiles derived from observational follow-up illustrate how individuals progress over time, enabling comparison against expected recovery patterns. The data support the identification of cut-off GRASSP strength scores that may inform enrollment criteria for both chronic and acute clinical studies; metrics with fewer ceiling and floor effects are particularly valuable as more credible clinical trial endpoints. Finally, the dataset facilitates the identification of muscle groups whose recovery is most strongly associated with functional improvement, helping to refine rehabilitation targeting and outcome evaluation. Perhaps most critically, the study highlights the relationships between individual muscle groups and overall functional independence. These findings carry important clinical implications: for instance, when rehabilitation resources are limited, understanding which muscles contribute most significantly to self-care outcomes enables clinicians to prioritize their interventions. Targeting key muscle groups—identified through both statistical correlation and machine learning—can potentially maximize functional gains and efficiency in upper limb rehabilitation for individuals with tetraplegia. Conclusions GV1, GV2 and GVM are all sound measures with rigorous development which can be implemented in clinical settings, academic research settings and in regulated industry sponsored trials. This work summarizes the robustness of the measure and can be used to substantiate the decision for GRASSP use in different trial types and studies. A measure with an available reference profile such as we have presented with GRASSP in this paper, substantiates it’s use and requirement as an important measure in the study of tetraplegia across the spectrum of clinical and research activity in the field. Declarations Acknowledgements: We would like to acknowledge the funding for this work: University of Toronto, Data Sciences Institute, Dana and Christopher Reeve Foundation, Praxis (Rick Hanson Institute), Ontario Neurotrauma Foundation, Craig Neilsen Foundation and Canadian Institutes for Health Research for the project and trainee funding. We would also like to acknowledge the collaboration with the European Multi-Center SCI Study and our ongoing work to further develop the GRASSP. Data Availability: Upon request with an adequate Data Sharing Agreement. References Weidner N, Abel R, Maier D, Röhl K, Röhrich F, Baumberger M, et al.; Nogo Inhibition in Spinal Cord Injury Study Group. Safety and efficacy of intrathecal antibodies to Nogo-A in patients with acute cervical spinal cord injury: A randomised, double-blind, multicentre, placebo-controlled, phase 2b trial. Lancet Neurol. 2025;24(1):42–53. https://doi.org/10.1016/S1474-4422(24)00447-2 Moritz C, Field-Fote EC, Tefertiller C, van Nes I, Trumbower R, Kalsi-Ryan S, et al. Non-invasive spinal cord electrical stimulation for arm and hand function in chronic tetraplegia: A safety and efficacy trial. Nat Med. 2024;30(5):1276–1283. https://doi.org/10.1038/s41591-024-02940-9 Cheng TT, Dore S, Fischer I, Tomycz ND. Upper extremity motor improvement with temporal interference stimulation in spinal cord injury: A pilot case series. J Neurosurg Spine. 2024;40:431–437. https://doi.org/10.3171/2023.12.SPINE231058 Kalsi-Ryan S, Beaton D, Curt A, Duff S, Popovic MR, Rudhe C, et al. The Graded Redefined Assessment of Strength Sensibility and Prehension: Reliability and validity. J Neurotrauma. 2012;29:905–914. https://doi.org/10.1089/neu.2010.1504 GRASSP. GRASSP development. (n.d.) https://grassptest.com/development/ . Accessed 23 Nov 2025. Kalsi-Ryan S, Beaton D, Ahn H, Askes H, Drew B, Curt A, et al. Responsiveness, sensitivity and minimally detectable difference of the Graded and Redefined Assessment of Strength, Sensibility, and Prehension, Version 1.0 (GRASSP V1). J Neurotrauma. 2016;33:307–314. https://doi.org/10.1089/neu.2015.4217 Velstra IM, Curt A, Frotzler A, Abel R, Kalsi-Ryan S, Rietman JS, et al. Changes in strength, sensibility and prehension in acute cervical spinal cord injury: European multicenter responsiveness study of the GRASSP. Neurorehabil Neural Repair. 2015;29(8):755–766. https://doi.org/10.1177/1545968314565466 Velstra IM, Fellinghauer C, Abel R, Kalsi-Ryan S, Rupp R, Curt A. The Graded and Redefined Assessment of Strength, Sensibility, and Prehension Version 2 provides interval measure properties. J Neurotrauma. 2018;35(6):854–863. https://doi.org/10.1089/neu.2017.5195 Kalsi-Ryan S, Chan C, Verrier M, Curt A, Fehlings M, Bolliger M, et al. The graded redefined assessment of strength, sensibility and prehension version 2 (GV2): Psychometric properties. J Spinal Cord Med. 2019;42(S1):S149–S153. https://doi.org/10.1080/10790268.2019.1616950 Kalsi-Ryan S, Riehm LE, Tetreault L, Martin AR, Teoderascu F, Massicotte E, et al. Characteristics of upper limb impairment related to degenerative cervical myelopathy: Development of a sensitive hand assessment (Graded Redefined Assessment of Strength, Sensibility, and Prehension Version Myelopathy). Neurosurgery. 2020;86(3):E292–E299. https://doi.org/10.1093/neuros/nyz499 Kalsi-Ryan S, Verrier M, Curt A, Fehlings M. Development of the Graded Redefined Assessment of Strength, Sensibility and Prehension (GRASSP): Reviewing measurement specific to the upper limb in tetraplegia. J Neurosurg Spine. 2012;17(1 Suppl):65–76. https://doi.org/10.3171/2012.6.AOSPINE1258 Kalsi-Ryan S, Balbinot G, Wang JZ, Abel R, Bolliger M, Curt A, et al. Minimal clinically important difference of Graded Redefined Assessment of Strength, Sensibility, and Prehension Version 1 in acute cervical traumatic spinal cord injury. J Neurotrauma. 2022;39:1645–1653. https://doi.org/10.1089/neu.2021.0500 Marino RJ, Sinko R, Bryden A, Backus D, Chen D, Nemunaitis GA, Leiby BE. Comparison of responsiveness and minimal clinically important difference of the Capabilities of Upper Extremity Test (CUE-T) and the Graded Redefined Assessment of Strength, Sensibility and Prehension (GRASSP). Top Spinal Cord Inj Rehabil. 2018;24(3):227–238. https://doi.org/10.1310/sci2403-227 Mulcahey MJ, Thielen CC, Dent K, Sinko R, Sadowsky C, Martin R, et al. Evaluation of the graded redefined assessment of strength, sensibility and prehension (GRASSP) in children with tetraplegia. Spinal Cord. 2018;56(6):581–588. https://doi.org/10.1038/s41393-018-0084-0 Catz A, Greenberg E, Itzkovich M, Bluvshtein V, Ronen J, Gelernter I. A new instrument for outcome assessment in rehabilitation medicine: Spinal cord injury ability realization measurement index. Archives of physical medicine and rehabilitation. 2004;85(3):399–404. Breiman L. Random forests. Mach Learn. 2001;45(1):5–32. https://doi.org/10.1023/A:1010933404324 Chen T, Guestrin C. XGBoost: A scalable tree boosting system (arXiv:1603.02754). arXiv. 2016. https://doi.org/10.48550/arXiv.1603.02754 Cunningham P, Delany SJ. k-Nearest Neighbour Classifiers: 2nd Edition (with Python examples) (arXiv:2004.04523). arXiv. 2020. https://doi.org/10.48550/arXiv.2004.04523 Hearst MA, Dumais ST, Osuna E, Platt J, Scholkopf B. Support vector machines. IEEE Intell Syst Their Appl. 1998;13(4):18–28. https://doi.org/10.1109/5254.708428 Claesen M, De Moor B. Hyperparameter search in machine learning (arXiv:1502.02127 [cs.LG]). arXiv. 2015. https://doi.org/10.48550/arXiv.1502.02127 Lundberg SM, Erion GG, Lee S-I. Consistent individualized feature attribution for tree ensembles (arXiv:1802.03888 [cs.LG]). arXiv. 2019. https://doi.org/10.48550/arXiv.1802.03888 Additional Declarations There is a duality of interest Cite Share Download PDF Status: Under Review Version 1 posted Review # 2 received at journal 27 Feb, 2026 Reviewer # 2 agreed at journal 10 Feb, 2026 Reviewer # 1 agreed at journal 09 Feb, 2026 Reviewers invited by journal 09 Feb, 2026 Editor assigned by journal 29 Jan, 2026 Submission checks completed at journal 29 Jan, 2026 First submitted to journal 27 Jan, 2026 Unknown event 26 Jan, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8683387","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":588201678,"identity":"af069c8a-1992-493c-9e94-3d69cabc429d","order_by":0,"name":"Sukhvinder Kalsi-Ryan","email":"data:image/png;base64,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","orcid":"https://orcid.org/0000-0003-2332-5986","institution":"University Health Network","correspondingAuthor":true,"prefix":"","firstName":"Sukhvinder","middleName":"","lastName":"Kalsi-Ryan","suffix":""},{"id":588201679,"identity":"c1fa8cf1-ead8-4e52-8140-e8407c2a3844","order_by":1,"name":"Yuchen Wang","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Yuchen","middleName":"","lastName":"Wang","suffix":""},{"id":588201680,"identity":"1a28b346-9e8b-4fef-bb51-53502739d246","order_by":2,"name":"Danvir Sandhu","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Danvir","middleName":"","lastName":"Sandhu","suffix":""},{"id":588201681,"identity":"a0c66d9e-601e-4531-8595-6a41bd5e0c99","order_by":3,"name":"Jose Zariffa","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Jose","middleName":"","lastName":"Zariffa","suffix":""},{"id":588201682,"identity":"18cbbe7e-fb56-4fec-9d4a-704a62434aea","order_by":4,"name":"Armin Curt","email":"","orcid":"","institution":"University Hospital Balgrist","correspondingAuthor":false,"prefix":"","firstName":"Armin","middleName":"","lastName":"Curt","suffix":""}],"badges":[],"createdAt":"2026-01-24 03:20:25","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8683387/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8683387/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":102378203,"identity":"9063fc8a-7b1d-4faa-b5a9-bca461845790","added_by":"auto","created_at":"2026-02-11 06:01:07","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":134064,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDistributions and Recovery Trajectories of Key Metrics by AIS Group.\u003cbr\u003e\n \u003c/strong\u003e(a–d) Box plots of each metric in the cross-sectional dataset, stratified by AIS level. Red dots indicate mean values. (e–h) Longitudinal recovery curves for each metric from 1 to 6 months post-injury, shown separately for patients in AIS groups AB and CDE in the historical dataset. The shaded area around the line represents the 95% confidence interval, indicating the uncertainty associated with the estimated trend.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-8683387/v1/7e7b6ebec91f5d05b0a86bb2.png"},{"id":102378201,"identity":"d43fcb28-fd52-48fe-904c-a247209dff5c","added_by":"auto","created_at":"2026-02-11 06:01:07","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":183795,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eHeatmaps of Muscle-Measurement Correlations Across AIS Levels and Timepoints.\u003c/strong\u003e\u003cbr\u003e\n (a, b) Heatmaps showing the correlations between 10 individual muscles and the SCIM Self-Care score, SCIM Total score, and PP Total score in the cross-sectional dataset, stratified by AIS groups AB and CDE. (c–f) Correlation heatmaps computed from the historical longitudinal dataset at 1 month and 6 months post-injury, also stratified by AIS group. In the heatmap, color intensity reflects the magnitude of the correlation, with warmer colors indicating stronger positive correlations and cooler colors representing weaker or negative correlations.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-8683387/v1/fa4abfd00e9f25df901a60e1.png"},{"id":102378205,"identity":"9d4e1686-ad30-4116-94d2-c5e1dd0e186c","added_by":"auto","created_at":"2026-02-11 06:01:07","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":168693,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eMachine Learning Model Performance and SHAP-Based Feature Interpretability.\u003c/strong\u003e\u003cbr\u003e\n(a, b) Discrimination performance of classification models on the test sets of the cross-sectional and historical longitudinal datasets, respectively. (c, d) Bar plots ranking the importance of 21 indicators based on SHAP values in the cross-sectional and longitudinal datasets. (e, f) SHAP value distributions showing the impact of each feature on the model output in the cross-sectional and longitudinal datasets.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-8683387/v1/7c1d81dda4c97ddb33e03036.png"},{"id":102398250,"identity":"a105af13-e409-4332-b3bf-d94a24ed6969","added_by":"auto","created_at":"2026-02-11 10:21:55","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1904872,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8683387/v1/d3a4dc21-d91a-415c-b705-85f72f2efaf6.pdf"}],"financialInterests":"There is a duality of interest","formattedTitle":"Clinical Utility of GRASSP in Traumatic Tetraplegia: A Narrative Review and Retrospective Analysis Incorporating Machine Learning with an Explainability Framework","fulltext":[{"header":"Brief Informative","content":"\u003cp\u003eThis paper includes: A summary of three versions of GRASSP and clinical research recommendations that are useful to researchers and clinicians for implementation of GRASSP. Sample characteristics of historical data provide insight for researchers into expected outcomes of individuals with traumatic tetraplegia and additional metrics for future study design in both acute and chronic studies related to traumatic tetraplegia.\u0026nbsp;\u003c/p\u003e"},{"header":"Introduction","content":"\u003cp\u003eSince 2012, the Graded Redefined Assessment of Strength, Sensibility, and Prehension (GRASSP) has been utilized as a primary, secondary, or exploratory measure in numerous trials and studies, including recent interventional work such as the Nogo‑A antibody trial in acute cervical spinal cord injury (SCI) [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e], the Up‑LIFT pivotal trial by ONWARD Medical using ARC‑EX\u0026reg; stimulation in individuals with chronic tetraplegia [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e], and a 2024 case series evaluating temporal interference stimulation [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. Although there is a comprehensive body of research detailing the measurement properties and applications of GRASSP, there is now an opportunity to summarize and analyze historical data to guide study design, endpoint formulation, and provide historical values for comparison in interventional studies. There has been approximately 15 years of experience with use of the GRASSP and this information has clinical and research utility. By performing this summative work, we are able to move closer to: 1) Avoiding ceiling and floor effects; 2) Improving detection of clinically meaningful neurological change; 3) Enhancing statistical performance; and 4) Aligning with FDA/EMA expectations for sensitive and reliable endpoints.\u003c/p\u003e \u003cp\u003eGRASSP Version 1 (GV1) was released in 2009, combining the Link Hand Function Test and the Tetraplegia Hand Measure [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. The development phase of GV1 spanned from 2006 to 2009, funded by not-for-profit agencies and engaging six individuals on the GRASSP development group. The primary aim was to address the gap in assessing subtle changes in the upper limb following traumatic cervical spinal cord injury. The GRASSP is specifically designed to evaluate upper limb function after traumatic spinal cord injury. Beyond its development, the measurement properties of reliability, validity, and responsiveness were established, making it a highly sought-after tool that met several regulatory criteria for clinical research [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. At the time, it was the only measure available with established measurement properties, interpretability, and generalizability. Therefore, it was accepted by the research community as it could be implemented in multiple centers, which made it ideal for multi-center and multi-national use. GRASSP attracted interest from industry and was commercialized through licensing by University Health Network. Neural Outcomes Consulting, Inc. was established in 2011 to manage the commercial sales of GRASSP and GRASSP products.\u003c/p\u003e \u003cp\u003eThe responsiveness properties of GRASSP were published in 2015, based on two complementary studies conducted between 2010 and 2013\u0026mdash;one by Kalsi-Ryan et al. in Canada and the other by Velstra et al. in Europe [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. In 2018, Velstra et al. leveraged longitudinal data to identify item redundancies within GRASSP, informing the development of a Version 2 [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. The revised version included updated measurement properties, a new scoring manual, and a clear rationale for item reduction [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. Approximately a year later, further adaptations were introduced to improve sensitivity to impairments specific to Degenerative Cervical Myelopathy (DCM), resulting in the development of GRASSP Version Myelopathy (GVM). Both the measure and measurement properties of this specialized version were documented in a subsequent development paper [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eWith a large knowledge base of GRASSP measures, there is no existing summary or illustration of the historical data. GRASSP has been employed in numerous sponsored trials, investigator-driven studies, and academic projects by graduate students, clinicians, and researchers. There is a need for historical summaries to facilitate comparisons with existing studies and to aid in the design of future research. Therefore, the developers have pursued this work to update the body of knowledge related to GRASSP.\u003c/p\u003e \u003cp\u003eThe aim of this review and post-hoc analysis is to integrate and summarize historical datasets to generate metrics that inform future study design, endpoint formulations, provide historical benchmarks for interventional studies and provide the evidence to support GRASSP as an important or primary endpoint in clinical trials. The objectives were to:\u003c/p\u003e \u003cp\u003e \u003col\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eSummarize GRASSP literature and provide usable evidentiary tables to support clinical research decision making\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eTo determine enrollment thresholds for acute and chronic samples, considering floor and ceiling effects.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eTo define recovery profiles, descriptive statistical values and characterize typical relationships between impairment and function in a standard sample using machine learning.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003c/ol\u003e \u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003e \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eLiterature Review\u003c/span\u003e: A structured literature search was conducted in PubMed, Scopus, and Google Scholar between May and June 2025. The search strategy employed combinations of keywords such as \u0026ldquo;GRASSP,\u0026rdquo; \u0026ldquo;GRASSP Version 1,\u0026rdquo; \u0026ldquo;GRASSP Version 2,\u0026rdquo; \u0026ldquo;GRASSP Myelopathy,\u0026rdquo; \u0026ldquo;GV1,\u0026rdquo; \u0026ldquo;GV2,\u0026rdquo; \u0026ldquo;DCM,\u0026rdquo; \u0026ldquo;tetraplegia,\u0026rdquo; \u0026ldquo;cervical SCI,\u0026rdquo; \u0026ldquo;validity,\u0026rdquo; \u0026ldquo;reliability,\u0026rdquo; \u0026ldquo;responsiveness,\u0026rdquo; \u0026ldquo;minimally detectable difference,\u0026rdquo; and \u0026ldquo;minimal clinically important difference.\u0026rdquo; Search parameters were limited to peer-reviewed, English-language articles published between 2010 and 2025, with a focus on studies that evaluated measurement properties in traumatic cervical spinal cord injury (SCI) and degenerative cervical myelopathy (DCM) cohorts. Evidentiary tables were synthesized from the articles reviewed.\u003c/p\u003e \u003cp\u003e \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eData Preparation and Preprocessing\u003c/span\u003e: For this analysis, we focus on two derived datasets: the longitudinal dataset and the cross-sectional chronic dataset. Both were constructed by combining and reformatting the aforementioned sources. The longitudinal dataset combines observational data from the 2010\u0026ndash;2013 [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e] studies which include observations from Canada and the EMSCI [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. It includes a total of 182 patients, each assessed at 0, 1, 3, and 6 months post-injury, resulting in 728 observations. Note that 0 refers to the baseline assessment that occurs generally between 0\u0026ndash;14 days after injury. The cross-sectional chronic dataset includes observations from the 2009 cross-sectional chronic SCI cohort (n\u0026thinsp;=\u0026thinsp;72) [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e], along with all occurrences from the longitudinal dataset at 6 months post-injury [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. This results in a combined total of 254 observations. The longitudinal dataset contains repeated observations of the same patients at multiple time points, whereas the cross-sectional chronic dataset includes only a single observation per patient. By utilizing both datasets, we can gain a more comprehensive understanding of patient recovery\u0026mdash;capturing not only the recovery profile at a specific time point, but also the trajectory of recovery over a six-month period.\u003c/p\u003e \u003cp\u003eIn the longitudinal analysis, the sample was grouped based on the American Spinal Injury Association Impairment Scale (AIS) classification at baseline. AIS grades A or B were categorized as motor complete, while those with grades C, D, or E were classified as motor incomplete. Some variables were summed to simplify the analysis. The variable named \u003cb\u003e\u0026ldquo;str_Total\u0026rdquo;\u003c/b\u003e is the right and left GRASSP strength total, summing the scores of 10 muscle strength tests on both the left and right sides of the body, resulting in a total score ranging from 0 to 100. Scores assessing upper limb tasks from the GRASSP Prehension Performance were combined to create a new variable called \u003cb\u003e\u0026ldquo;PP_total\u0026rdquo;.\u003c/b\u003e From the spinal cord independence measure (SCIM) [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e] we focus on the self-care subscore which ranges from 0\u0026ndash;20 and is the sum of items 1\u0026ndash;5. We focus on the SCIM selfcare subscore, as it primarily assesses self-care abilities and upper limb function. In addition, a total SCIM score was computed by summing all 17 items.\u003c/p\u003e \u003cp\u003e \u003cb\u003eDistribution of cross-sectional chronic sample\u003c/b\u003e - In this study, boxplots were generated for each of the five AIS grades (A, B, C, D, and E) within the cross-sectional chronic dataset to define the differences in distribution of functional scores between groups. Additionally, examining the median values provides insight into the typical functional score for each group.\u003c/p\u003e \u003cp\u003e \u003cb\u003eSingle muscle Relationship with Function\u003c/b\u003e - Pearson correlation coefficients between individual muscle strength scores and functional measures (SCIM and GR-PP scores). In the cross-sectional chronic dataset these were calculated to assess the strength of association between each muscle and overall functional performance. This process was repeated with the longitudinal dataset at two critical time points: 1 month and 6 months post-injury.\u003c/p\u003e \u003cp\u003e \u003cb\u003eRecovery Profiles -\u003c/b\u003e were generated with the longitudinal dataset at 1, 3, and 6 months post-injury to illustrate the recovery trajectories of patients grouped as motor complete and motor incomplete. Baseline (immediately post-injury) were excluded in this summary as there is inconsistent timing of the baseline.\u003c/p\u003e \u003cp\u003e \u003cb\u003eEnrollment Cut-offs for GRASSP strength and prehension performance\u003c/b\u003e - Typically, studies use the upper extremity motor score from the International Standards of Neurological Classification for Spinal Cord Injury (ISNCSCI) as a screening tool for eligibility in acute and chronic trials. We ordered all chronic and acute occurences based on UEMS, and then reviewed the GRASSP strength and Prehension Performance scores. For the chronic sample we implanted the MDD values to determine which individuals could continue to show measurable change. For the acute sample we reviewed GRASSP strength and Prehension Performance at endpoint. Based on the endpoint results we were able to identify expanded cut-off scores for the acute sample.\u003c/p\u003e \u003cp\u003e \u003cb\u003eMuscles Strength Relationships to SCIM-SS -\u003c/b\u003e To identify which of the 10 muscles most strongly influence the self-care abilities of individuals with SCI, we applied machine learning models to both cross-sectional chronic and longitudinal datasets. The target variable was the categorized SCIM self-care score, while the primary inputs were 20 individual muscle strength scores (left and right sides of 10 muscles). AIS was also included as a categorical indicator of motor and sensory function loss. In the longitudinal dataset, patients were grouped by motor complete vs. motor incomplete at 1 month post-injury. The same classification was applied to the cross-sectional chronic dataset, which includes a single observation per patient. Patients in both datasets were further categorized into two functional groups based on SCIM self-care scores: those scoring 0\u0026ndash;14 (low independence) and those scoring 15\u0026ndash;20 (high independence).\u003c/p\u003e \u003cp\u003eSeparate models were trained for the cross-sectional chronic and longitudinal datasets. For the cross-sectional chronic dataset, input features included the 20 muscle strength scores and complete/incomplete group. The target variable was the binary self-care score group (low vs. high independence). For the longitudinal dataset, where patients were evaluated at four time points, we used muscle strength scores and complete/incomplete group recorded at 1 month post-injury as inputs, with the self-care score at 6 months post-injury as the target. This setup allowed us to assess how early motor function predicts longer-term self-care outcomes. In this study, four ML algorithms were investigated: random forest [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e], eXtreme gradient boosting (XGBoost) [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e], K-Nearest Neighbour (KNN) [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e], Support vector machines [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. We employed grid search [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e] during model training to optimize hyperparameters for each algorithm. For each candidate hyperparameter configuration, model performance was assessed using k-fold cross-validation on the training set, and the configuration with the highest mean cross-validated performance was selected. This procedure ensured that hyperparameter tuning was based solely on training data and was not influenced by the test set. Model performance was then evaluated using standard metrics, including accuracy, precision, recall, F1 score, support, and AUC. The best-performing model was subsequently interpreted using SHAP (Shapley Additive Explanations), which provided both quantitative values and visualizations to explain the contribution of each input feature to the model's predictions [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e].\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eLiterature Review\u003c/span\u003e: A comprehensive review of the literature on the GRASSP Versions was synthesized into two evidentiary tables to support clinical decision-making and measure selection in SCI and DCM.\u003c/p\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e provides a side-by-side comparison of GRASSP versions\u0026mdash;GV1, GV2), and GVM, detailing subtest domains, item descriptions, maximum scores, and the equipment required for each version. The table highlights the structural and psychometric evolution from GV1\u0026rsquo;s comprehensive multi-domain model to GV2\u0026rsquo;s Rasch-refined iteration and ultimately to the highly time-efficient and targeted GVM, which omits prehension modules not relevant to early-stage DCM detection. While GV1 includes five bilateral subtests (strength, dorsal and palmar sensibility, prehension ability, and prehension performance), GV2 eliminates the dorsal sensibility component and incorporates psychometric refinements [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. GVM streamlines the tool further, using only three subtests\u0026mdash;strength, palmar sensation, and dexterity\u0026mdash;adapted to detect the subtle deficits of DCM with minimal equipment [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e].\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eGRASSP Versions \u0026ndash; Subtests, Item Descriptions, and Maximum Scores\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"7\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSubtest Domain\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eGRASSP V1\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMax Score (V1) - Per Side\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eGRASSP V2 (GV2)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eMax Score (V2) - Per Side\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eGRASSP-M\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eMax Score (GVM) - Per Side\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eEquipment Required\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eGRASSP Version 1 Kit\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026mdash;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eGRASSP Version 1 Kit or Version 2 Kit and Version 2 Manual\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026mdash;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eGRASSP Version 1 or 2, or Myelopathy Kit and Version Myelopathy Manual. Adapted from original GRASSP kit. No full prehension module required [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e].\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e\u0026mdash;\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eStrength\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e10 upper limb muscle groups (C5\u0026ndash;T1); MRC scale 0\u0026ndash;5 [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e50\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSame 10 muscle groups; transitioned from isotonic to isometric testing for improved standardization [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e50\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eSame as GV2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e50\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eDorsal Sensibility\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3 dorsal hand sites; Semmes-Weinstein monofilaments; scored 0\u0026ndash;4[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRemoved due to low Rasch functional relevance[\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026mdash;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eNot included[\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e\u0026mdash;\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003ePalmar Sensibility\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3 palmar hand sites; Semmes-Weinstein monofilaments; scored 0\u0026ndash;4[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRetained with Rasch-optimized 3-category scale[\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eSame as GV2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e12\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003ePrehension Ability\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3 grasp types (cylindrical, lateral/key, tip-to-tip); scored 0\u0026ndash;4[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSame grasp types; Rasch-calibrated scoring and standardized instructions[\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eNot included\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eNA\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003ePrehension Performance\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e6 tasks (e.g., pegboard, jar, coin, key turn); scored 0\u0026ndash;5[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e30\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eReduced to 4 tasks (e.g., jar and coin tasks removed); simplified with Rasch hierarchy[\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e20\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eReduced to 1 dexterity task\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e9\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"7\" nameend=\"c7\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eLegend\u003c/b\u003e:\u003c/p\u003e \u003cp\u003e\u0026bull; \u003cb\u003eMRC\u003c/b\u003e: Medical Research Council scale (0\u0026thinsp;=\u0026thinsp;no contraction, 5\u0026thinsp;=\u0026thinsp;normal strength)\u003c/p\u003e \u003cp\u003e\u0026bull; \u003cb\u003eRasch\u003c/b\u003e: Refers to Rasch analysis, a psychometric method used to refine scoring and item calibration\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e outlines the measurement properties of each GRASSP version across reliability, validity, responsiveness, and clinical utility. All versions demonstrate excellent inter- and test retest-reliability and strong construct validity across SCI or DCM severity levels [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. Responsiveness is particularly well-documented in GV1 and GV2, with standardized response means (SRM) ranging from 0.67 to 1.35 across various subtests and timepoints [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. GV2\u0026rsquo;s Rasch analysis supported item-level uni-dimensionality and reduced administration time shows are reduced measure without sacrificing measurement sensitivity [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. GVM shows validity and discrimination across DCM severity categories, with strong intra- and test retest reliability ICCs for the dexterity subtest (\u0026ge;\u0026thinsp;0.79) [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. Minimum detectable differences (MDD) are reported for GV1 and GV2, offering benchmarks for real clinical change. Minimal clinically important differences (MCID) for interpreting meaningful change are also reported for GV1 and GV2 [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. Although GVM lacks direct MCID reporting, its strong known-group validity and high ICC values suggest sufficient sensitivity to detect clinically relevant impairment gradations in DCM [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e].\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eEvidentiary Summary of GRASSP, Versions and Measurement Qualities\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eGRASSP Version 1 (GV1)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eGRASSP Version 2 (GV2)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eGRASSP Version Myelopathy (GVM, GRASSP-M)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePurpose\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003eTo assess upper limb sensorimotor impairment in individuals with cervical spinal cord pathology\u0026mdash;both traumatic and degenerative\u0026mdash;by quantifying changes in neurological function and hand performance across the full spectrum of severity and recovery, from acute to chronic phases in traumatic SCI and from mild to severe impairment in degenerative cervical myelopathy (DCM).\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDescription\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003eGRASSP Version 1 (GV1) assesses five subdomains\u0026mdash;strength, dorsal/palmar sensory function, prehension ability, and prehension performance\u0026mdash;independently for each hand. It is validated for use across acute and chronic phases of traumatic SCI [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eGRASSP Version 2 (GV2) refines GV1 using Rasch analysis to improve psychometric properties and streamline test items. It preserves the original structure and clinical scope, while enhancing feasibility, precision, and applicability in trials and clinical settings [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e].\u003c/p\u003e \u003cp\u003e\u003cb\u003eEach hand is assessed separately.\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eGRASSP-Myelopathy (GRASSP-M / GVM) is an adapted version of GV1 tailored to detect early, subtle upper limb deficits in Degenerative Cervical Myelopathy (DCM), such as intrinsic hand muscle weakness, sensory decline, and reduced dexterity. It offers a sensitive, low-burden alternative to broader tools like the mJOA [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e].\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eICF Domain\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003eBody Structure, Body Function, Activity [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eTime to Administer\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e30\u0026ndash;45 minutes per patient [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eApproximately 15\u0026ndash;30 minutes (approximately 15 min less time than GV1) [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e10\u0026ndash;15 minutes (significantly reduced from GRASSP V1\u0026rsquo;s 45-minute duration) [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eDomains, Items and Subtests\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003eSensation\u003c/b\u003e\u003c/p\u003e \u003cp\u003e\u0026bull; \u003cb\u003eItems\u003c/b\u003e: Dorsal and Palmar Sensation\u003c/p\u003e \u003cp\u003e\u0026bull; \u003cb\u003eSubtests\u003c/b\u003e: Dorsal Sensation (12), Palmar Sensation (12)\u003c/p\u003e \u003cp\u003e\u003cb\u003eStrength\u003c/b\u003e\u003c/p\u003e \u003cp\u003e\u0026bull; \u003cb\u003eItems\u003c/b\u003e: C5 - Biceps, Ant Deltoid, C6 - Wrist Extensors, C7 - Triceps, C7 - Group Phallic\u003c/p\u003e \u003cp\u003e\u0026bull; \u003cb\u003eSubtests\u003c/b\u003e: Strength (50)\u003c/p\u003e \u003cp\u003e\u003cb\u003ePrehension\u003c/b\u003e\u003c/p\u003e \u003cp\u003e\u0026bull; \u003cb\u003eItems\u003c/b\u003e: C5 - Shoulder Flexion, Ext Dig, Finger Flexion, Lateral Key, Finger Abduction\u003c/p\u003e \u003cp\u003e\u0026bull; \u003cb\u003eSubtests\u003c/b\u003e: Prehension Ability (12), Prehension Performance (30)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003eSensation\u003c/b\u003e\u003c/p\u003e \u003cp\u003e\u0026bull; \u003cb\u003eItems\u003c/b\u003e: Palmar Sensation\u003c/p\u003e \u003cp\u003e\u0026bull; \u003cb\u003eSubtests\u003c/b\u003e: Palmar Sensation (12)\u003c/p\u003e \u003cp\u003e\u003cb\u003eStrength\u003c/b\u003e\u003c/p\u003e \u003cp\u003e\u0026bull; \u003cb\u003eItems\u003c/b\u003e: C5 - Biceps, Ant Deltoid, C7 - Triceps, C7 - Group Phallic\u003c/p\u003e \u003cp\u003e\u0026bull; \u003cb\u003eSubtests\u003c/b\u003e: Strength (50)\u003c/p\u003e \u003cp\u003e\u003cb\u003ePrehension\u003c/b\u003e\u003c/p\u003e \u003cp\u003e\u0026bull; \u003cb\u003eItems\u003c/b\u003e: C5 - Shoulder Flexion, Ext Dig, Finger Flexion, Lateral Key, Finger Abduction\u003c/p\u003e \u003cp\u003e\u0026bull; \u003cb\u003eSubtests\u003c/b\u003e: Prehension Ability (12), Prehension Performance (20)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003eSensation\u003c/b\u003e\u003c/p\u003e \u003cp\u003e\u0026bull; \u003cb\u003eItems\u003c/b\u003e: Palmar Sensation\u003c/p\u003e \u003cp\u003e\u0026bull; \u003cb\u003eSubtests\u003c/b\u003e: Dorsal Sensation (12)\u003c/p\u003e \u003cp\u003e\u003cb\u003eStrength\u003c/b\u003e\u003c/p\u003e \u003cp\u003e\u0026bull; \u003cb\u003eItems\u003c/b\u003e: C5 - Biceps, Ant Deltoid, C6 - Wrist Extensors, C7 - Triceps, C7 - Opponens Pollicis, C8 - Finger Flexion, Ext Dig, Flex Pollicis Long, T1 - Finger Abduction, First DI\u003c/p\u003e \u003cp\u003e\u0026bull; \u003cb\u003eSubtests\u003c/b\u003e: Strength (50)\u003c/p\u003e \u003cp\u003e\u003cb\u003ePrehension/Dexterity (exclusive to GVM)\u003c/b\u003e\u003c/p\u003e \u003cp\u003e\u0026bull; \u003cb\u003eItems\u003c/b\u003e: Tip to Tip, Tripod, Nuts\u003c/p\u003e \u003cp\u003e\u0026bull; \u003cb\u003eSubtests\u003c/b\u003e: Prehension (12)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eReliability\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eExcellent inter- and test-retest reliability:\u003c/p\u003e \u003cp\u003e\u0026bull; Sensibility ICC\u0026thinsp;=\u0026thinsp;0.84\u0026ndash;0.95\u003c/p\u003e \u003cp\u003e\u0026bull; Strength ICC\u0026thinsp;=\u0026thinsp;0.95\u0026ndash;0.98\u003c/p\u003e \u003cp\u003e\u0026bull; Prehension Ability ICC\u0026thinsp;=\u0026thinsp;0.95\u0026ndash;0.98\u003c/p\u003e \u003cp\u003e\u0026bull; Prehension Performance ICC\u0026thinsp;=\u0026thinsp;0.93\u0026ndash;0.96\u003c/p\u003e \u003cp\u003eValidated in both adults and children (\u0026ge;\u0026thinsp;6 years) [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eExcellent inter- and test-retest reliability across subtests:\u003c/p\u003e \u003cp\u003e\u0026bull; Strength: ICC\u0026thinsp;=\u0026thinsp;0.95\u0026ndash;0.98\u003c/p\u003e \u003cp\u003e\u0026bull; Palmar Sensation: ICC\u0026thinsp;=\u0026thinsp;0.84\u0026ndash;0.97\u003c/p\u003e \u003cp\u003e\u0026bull; Prehension Ability: ICC\u0026thinsp;=\u0026thinsp;0.95\u0026ndash;0.98\u003c/p\u003e \u003cp\u003e\u0026bull; Prehension Performance: ICC\u0026thinsp;=\u0026thinsp;0.96\u0026ndash;0.97 [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eStrength and Sensation \u0026ndash; see GV2\u003c/p\u003e \u003cp\u003e\u0026bull; Dexterity Intra-rater ICC: 0.868 (dominant), 0.790 (non-dominant)\u003c/p\u003e \u003cp\u003e\u0026bull; Dexterity Inter-rater ICC: 0.869 (dominant), 0.862 (non-dominant)\u003c/p\u003e \u003cp\u003eAll values\u0026thinsp;\u0026gt;\u0026thinsp;0.75 indicate strong reproducibility across raters [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e].\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eConstruct Validity\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u0026bull; Sensory agreement with ISNCSCI: Kappa\u0026thinsp;=\u0026thinsp;0.41\u0026ndash;0.51\u003c/p\u003e \u003cp\u003e\u0026bull; Motor innervation discordance shows GRASSP detects differences missed by ISNCSCI\u003c/p\u003e \u003cp\u003e\u0026bull; Designed to detect discordance due to expanded sites and use of Semmes-Weinstein thresholds [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eAll subtests significantly differentiated AIS grades A\u0026ndash;D (p\u0026thinsp;\u0026lt;\u0026thinsp;0.01); Rasch analysis confirmed item fit, uni-dimensionality, and strong internal consistency (Cronbach's α\u0026thinsp;=\u0026thinsp;0.85\u0026ndash;0.97 across domains) [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e].\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eDemonstrated strong known-group discrimination across mJOA-defined severity levels (mild, moderate, severe):\u003c/p\u003e \u003cp\u003e\u0026bull; Dexterity: p \u0026lt; .001 across all group comparisons\u003c/p\u003e \u003cp\u003e\u0026bull; Strength \u0026amp; sensation: significant differences (p \u0026lt; .05 to \u0026lt; .001), particularly in moderate vs. severe cases\u003c/p\u003e \u003cp\u003e\u0026bull; Bonferroni correction applied (p \u0026le; .016) to ensure statistical rigor [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e].\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eConcurrent Validity\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eStrong correlation with functional measures:\u003c/p\u003e \u003cp\u003e\u0026bull; SCIM: r\u0026thinsp;=\u0026thinsp;0.57\u0026ndash;0.68\u003c/p\u003e \u003cp\u003e\u0026bull; SCIM Self-care: r\u0026thinsp;=\u0026thinsp;0.74\u0026ndash;0.79\u003c/p\u003e \u003cp\u003e\u0026bull; CUE/CUE-Q: r\u0026thinsp;=\u0026thinsp;0.76\u0026ndash;0.84\u003c/p\u003e \u003cp\u003eAll p\u0026thinsp;\u0026lt;\u0026thinsp;0.0001; validated in both adult and pediatric SCI populations [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eModerate to strong correlations with functional outcome measures:\u003c/p\u003e \u003cp\u003e\u0026bull; SCIM: r\u0026thinsp;=\u0026thinsp;0.53\u0026ndash;0.71\u003c/p\u003e \u003cp\u003e\u0026bull; SCIM Self-care: r\u0026thinsp;=\u0026thinsp;0.72\u0026ndash;0.82\u003c/p\u003e \u003cp\u003e\u0026bull; CUE-Q: r\u0026thinsp;=\u0026thinsp;0.76\u0026ndash;0.83\u003c/p\u003e \u003cp\u003eAll significant at p\u0026thinsp;\u0026lt;\u0026thinsp;0.001 [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eModerate and significant correlations with mJOA:\u003c/p\u003e \u003cp\u003e\u0026bull; Dexterity: r\u0026thinsp;=\u0026thinsp;0.502\u0026ndash;0.533 (p \u0026le; .05)\u003c/p\u003e \u003cp\u003e\u0026bull; Dexterity Time: inverse correlation r = \u0026minus;\u0026thinsp;0.407 to \u0026minus;\u0026thinsp;0.469 (p \u0026le; .05)\u003c/p\u003e \u003cp\u003e\u0026bull; Strength \u0026amp; Sensation subtests: r\u0026thinsp;=\u0026thinsp;0.25\u0026ndash;0.48 (p \u0026le; .05), showing convergent validity with both motor and sensory mJOA subscores [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e].\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eResponsiveness\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDemonstrated across subtests:\u003c/p\u003e \u003cp\u003e\u0026bull; SRM\u0026thinsp;=\u0026thinsp;0.67\u0026ndash;1.1 in GRASSP-PP comparisons\u003c/p\u003e \u003cp\u003e\u0026bull; Compared favorably with CUE-T, though floor effects noted in GRASSP for minimal function [6. 13]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eDemonstrated meaningful change across timepoints (1mo\u0026ndash;6mo and 1mo\u0026ndash;12mo):\u003c/p\u003e \u003cp\u003e\u0026bull; Sensibility: SRM\u0026thinsp;=\u0026thinsp;0.84\u0026ndash;1.35\u003c/p\u003e \u003cp\u003e\u0026bull; Prehension Performance: SRM\u0026thinsp;=\u0026thinsp;0.67\u0026ndash;1.22\u003c/p\u003e \u003cp\u003eValidated across subgroups (AIS A\u0026ndash;D); supports longitudinal monitoring [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNot directly measured in original study, but tool demonstrated clear gradation of impairment across mJOA-defined severity levels. Sensitivity to mild DCM makes it well-suited for detecting progression or therapeutic response [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMDD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMinimum change required to exceed error:\u003c/p\u003e \u003cp\u003e\u0026bull; Strength (bilateral): \u0026ge; 7 points\u003c/p\u003e \u003cp\u003e\u0026bull; Sensation (bilateral): \u0026ge; 4 points\u003c/p\u003e \u003cp\u003e\u0026bull; Prehension Ability: \u0026ge; 4 points\u003c/p\u003e \u003cp\u003e\u0026bull; Prehension Performance (bilateral): \u0026ge; 6 points\u003c/p\u003e \u003cp\u003eValues based on SEM/SRD and validated at 95% CI [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eReported per subtest:\u003c/p\u003e \u003cp\u003e\u0026bull; Strength (bilateral): \u0026ge; 7 points\u003c/p\u003e \u003cp\u003e\u0026bull; Palmar Sensation (bilateral): \u0026ge; 3 points\u003c/p\u003e \u003cp\u003e\u0026bull; Prehension Ability (bilateral): \u0026ge; 4 points\u003c/p\u003e \u003cp\u003e\u0026bull; Prehension Performance (bilateral): \u0026ge; 3 points\u003c/p\u003e \u003cp\u003eStandard error and SRD also provided for item-level precision [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNot reported\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMCID\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDerived via Global Rating of Change (GRoC) for acute SCI:\u003c/p\u003e \u003cp\u003e\u0026bull; Strength (bilateral): 13\u0026ndash;19 points\u003c/p\u003e \u003cp\u003e\u0026bull; Prehension Performance (bilateral): 3\u0026ndash;12 points\u003c/p\u003e \u003cp\u003e\u0026bull; Sensibility showed no consistent MCID\u003c/p\u003e \u003cp\u003eValid only for acute (0\u0026ndash;6 months) SCI samples [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNot explicitly reported; SRM and MDD suggest values above noise threshold. Interpretation supported by statistical responsiveness [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNot reported\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYear of Copyright\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2008 (validated in 2012, 2015, 2022) [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2019 - Published by Kalsi-Ryan et al. [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2020 - Developed/validated by Kalsi-Ryan et al. as a DCM-specific adaptation of GV1 [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTraining\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003eClinicians can self-train using manual and eLearning modules; typically requires 2\u0026ndash;3 supervised administrations to achieve proficiency.\u003c/p\u003e \u003cp\u003eResearch Use: Group training is recommended to ensure standardized administration across raters; this may be done via formal sessions or team-based review and calibration.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"4\" nameend=\"c4\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eLegend\u003c/b\u003e:\u003c/p\u003e \u003cp\u003e\u0026bull; \u003cb\u003eICF Domain\u003c/b\u003e: Refers to domains outlined by the WHO\u0026rsquo;s International Classification of Functioning, Disability and Health (ICF), including body function, structure, and activity limitations.\u003c/p\u003e \u003cp\u003e\u0026bull; \u003cb\u003eICC (Intraclass Correlation Coefficient)\u003c/b\u003e: A measure of reliability; values\u0026thinsp;\u0026gt;\u0026thinsp;0.75 are considered excellent.\u003c/p\u003e \u003cp\u003e\u0026bull; \u003cb\u003eKappa\u003c/b\u003e: A statistic reflecting inter-rater agreement beyond chance (0.41\u0026ndash;0.60\u0026thinsp;=\u0026thinsp;moderate, 0.61\u0026ndash;0.80\u0026thinsp;=\u0026thinsp;substantial).\u003c/p\u003e \u003cp\u003e\u0026bull; \u003cb\u003er (Pearson\u0026rsquo;s Correlation Coefficient)\u003c/b\u003e: Measures strength of association between tools (0.3\u0026ndash;0.5\u0026thinsp;=\u0026thinsp;moderate, \u0026gt; 0.5\u0026thinsp;=\u0026thinsp;strong).\u003c/p\u003e \u003cp\u003e\u0026bull; \u003cb\u003eSRM (Standardized Response Mean)\u003c/b\u003e: A measure of responsiveness; SRM\u0026thinsp;\u0026gt;\u0026thinsp;0.8 indicates large sensitivity to change.\u003c/p\u003e \u003cp\u003e\u0026bull; \u003cb\u003eSEM (Standard Error of Measurement)\u003c/b\u003e: Reflects the precision of individual scores.\u003c/p\u003e \u003cp\u003e\u0026bull; \u003cb\u003eSRD (Smallest Real Difference)\u003c/b\u003e: The minimum score change required to surpass measurement error with 95% confidence.\u003c/p\u003e \u003cp\u003e\u0026bull; \u003cb\u003eMDD (Minimal Detectable Difference)\u003c/b\u003e: The smallest difference that can be interpreted as a real change, not measurement variability.\u003c/p\u003e \u003cp\u003e\u0026bull; \u003cb\u003eMCID (Minimal Clinically Important Difference)\u003c/b\u003e: The smallest change perceived as beneficial by patients, determined via anchors like Global Rating of Change.\u003c/p\u003e \u003cp\u003e\u0026bull; \u003cb\u003emJOA (modified Japanese Orthopaedic Association score)\u003c/b\u003e: A validated clinical scale assessing DCM severity.\u003c/p\u003e \u003cp\u003e\u0026bull; \u003cb\u003eSCIM (Spinal Cord Independence Measure)\u003c/b\u003e: A functional scale for daily living tasks in SCI.\u003c/p\u003e \u003cp\u003e\u0026bull; \u003cb\u003eCUE / CUE-Q\u003c/b\u003e: Capabilities of Upper Extremity Questionnaire; a patient-reported measure of upper limb function in SCI.\u003c/p\u003e \u003cp\u003e\u0026bull; \u003cb\u003eAIS Grades\u003c/b\u003e: American Spinal Injury Association Impairment Scale levels (A\u0026ndash;E) for classifying spinal cord injury severity.\u003c/p\u003e \u003cp\u003e\u0026bull; \u003cb\u003eBonferroni Correction\u003c/b\u003e: A statistical adjustment used when multiple comparisons are made, to control for Type I error.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eDistribution of Cross Sectional Sample\u003c/span\u003e: In the boxplot, we can see how functional scores vary by AIS level, where AIS A represents the most severe impairment and E the least. Across all four metrics\u0026mdash;muscle strength, SCIM self-care, UEMS and GRASSP-PP there is a consistent trend: scores generally increase from AIS A to AIS E, indicating better motor and functional performance with less severe impairment. Notably, AIS E individuals consistently achieve near-maximal scores, while AIS A individuals show the greatest variability and lowest median scores. The trend is especially pronounced in the total SCIM and muscle strength scores, emphasizing the relationship between AIS level and functional ability. The recovery profiles illustrate the patients in motor complete versus motor incomplete across four functional metrics: total muscle strength, GRASSP-PP, SCIM self-care sub score, and UEMS. Across all time points (1, 3, and 6 months), patients in the motor incomplete group consistently show higher performance compared to those in the motor complete. Furthermore, all metrics demonstrate a clear upward trend over time in both groups, indicating continuous functional improvement during the 6-month period. See Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eEnrollment Cutoffs\u003c/span\u003e: In the motor incomplete samples of the chronic and acute samples, we found using the GRASSP strength as an enrolment cut-off allowed for a larger and more inclusive sample. In the chronic group (n\u0026thinsp;=\u0026thinsp;133), using the UEMS of 40 as a cut-off rendered 46.6% (62/133) of the sample eligible, while using GRASSP Strength of 89 as a cut-off rendered 90% (120/133) of the sample eligible. If strength assessment is used to determine outcome, this is feasible, as the MDD is 7, thus a true clinical change can be determined without a ceiling effect. In the acute group we conducted a similar analysis by selecting at baseline based on UEMS and then determining at 6 months if there was or was not a ceiling effect with GRASSP strength. We found from a sample of 87, if using a UEMS cut-off of 28 or less at baseline, 60% (53/87) of the sample would be eligible for enrollment. If we consider the GRASSP strength endpoint remaining at 90 or less at 6 months, then the enrollment UEMS can be as high as 46 and this makes 96% (84/87) of the sample eligible at baseline. Therefore, GRASSP historical cohorts can be useful in establishing more inclusive cut-offs which avoid ceiling effects. Muscle relationships to function, illustrated in heatmaps provide a visual summary of how the strength of relationships between specific muscles and overall functional performance changes during the recovery process. By comparing the correlation patterns at these two time points, we can identify which muscles demonstrate a consistent association with functional outcomes and which may become more or less relevant as recovery progresses. This temporal perspective offers valuable insights into the dynamic nature of muscle-function relationships following spinal cord injury, potentially informing targeted rehabilitation strategies and clinical decision-making.\u003c/p\u003e \u003cp\u003eThe correlations between each of the 10 individual muscle strength scores and functional outcomes were computed and visualized in heatmaps at two time points: 1 month and 6 months post-injury. Warmer colors (reds) represent stronger positive correlations, while cooler colors (blues) indicate weaker correlations. Overall, the correlations tend to increase from 1 month to 6 months in both groups, with the motor complete group showing a more pronounced rise\u0026mdash;several muscles exhibit strong correlations (above 0.5) by 6 months. The motor incomplete group also demonstrates improvement over time, though the correlations remain more moderate and consistent across muscles. Notably, Triceps consistently displays strong correlations with all functional outcomes across both groups and time points, suggesting it may serve as a key predictor of recovery. See Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e. The heatmaps provide some indicators of muscle strength to function and can be interpreted for comparison.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eMachine learning\u003c/span\u003e: Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ea-b defines the best model performances on the test sets for both the cross-sectional chronic and longitudinal datasets. For the cross-sectional chronic dataset, the KNN model achieved the highest test accuracy of 90%. We observed that both KNN and XGBoost demonstrated relatively high and consistent performance across five evaluation metrics compared to the other models. However, due to the advantage of SHAP for interpretability and the difficulty of interpreting KNN in high-dimensional data, we selected XGBoost as the final model for the cross-sectional chronic dataset.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFor the longitudinal dataset, all four models achieved similar test accuracies of approximately 80%. Notably, XGBoost attained the highest AUC score among them. Therefore, we also chose XGBoost as the final model for interpreting the longitudinal dataset.\u003c/p\u003e \u003cp\u003eThe SHAP bar plots (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ec-d) and beeswarm plots (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ee-f) illustrate the feature contributions of the XGBoost models fitted on the cross-sectional chronic and longitudinal datasets.\u003c/p\u003e \u003cp\u003eIn both datasets, the categorical variable \u003cb\u003eAIS_Group\u003c/b\u003e, which indicates the patient\u0026rsquo;s AIS classification, emerged as the most influential feature in model predictions. Patients with AIS grades motor complete were more likely to have \u003cb\u003elow self-care independence\u003c/b\u003e, while those with AIS grades motor incomplete tended to exhibit \u003cb\u003ehigher levels of independence\u003c/b\u003e.\u003c/p\u003e \u003cp\u003eRegarding muscle strength scores, there were slight differences in feature importance across the two datasets. In the cross-sectional chronic dataset, \u003cb\u003eExtensor Digitorum DIII\u003c/b\u003e had the strongest influence on self-care ability, followed by \u003cb\u003eOpponens Pollicis\u003c/b\u003e and \u003cb\u003eAnterior Deltoid\u003c/b\u003e. In the \u003cb\u003elongitudinal dataset\u003c/b\u003e, \u003cb\u003eOpponens Pollicis\u003c/b\u003e was the top contributor, followed by \u003cb\u003eFirst Dorsal Interosseous\u003c/b\u003e and \u003cb\u003eExtensor Digitorum DIII\u003c/b\u003e. Notably, the muscles \u003cb\u003eExtensor Digitorum DIII\u003c/b\u003e, \u003cb\u003eOpponens Pollicis\u003c/b\u003e, \u003cb\u003eAnterior Deltoid\u003c/b\u003e, \u003cb\u003eWrist Extensors\u003c/b\u003e, and \u003cb\u003eFirst Dorsal Interosseous\u003c/b\u003e consistently ranked among the top features in both datasets, further confirming their potential role in improving patients\u0026rsquo; self-care ability.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis study synthesizes existing literature to provide accessible, consolidated data for clinicians and researchers when evaluating the utility of GRASSP outcome measures. A comprehensive summary of the three versions is presented, allowing for quick reference and informed decision-making regarding outcome measure selection. The table outlines the structural features of each tool, including the subtests, item breakdowns, scoring structure, and equipment requirements. In parallel, meaurement properties\u0026mdash;such as reliability, validity, responsiveness, and clinical interpretability\u0026mdash;are summarized to help users quickly identify which version of GRASSP aligns best with their specific research or clinical application. Where more in-depth information is required, full methodological and validation details can be accessed in the referenced literature.\u003c/p\u003e \u003cp\u003eThese findings support all three versions of GRASSP as valid, reliable, and clinically applicable tools, with each optimized for specific neurological populations and assessment contexts. GV1 and GV2 are suited for comprehensive bilateral impairment tracking in traumatic SCI. They both offer psychometric rigor and efficiency for clinical trials and rehabilitation settings. GV2 having a reduced set of hand assessments is better suited for studies/trials investigating global change after SCI. GVM provides a streamlined, targeted approach for early identification and monitoring of upper limb deficits in DCM.\u003c/p\u003e \u003cp\u003eThis study also analyzes historical data to define normative profiles and functional stratifications based on AIS grade. By distinguishing where individuals fall within the cross-sectional chronic spectrum, this analysis provides a valuable reference for determining whether a patient's presentation aligns with expected recovery benchmarks. The longitudinal component of the data captures recovery trajectories over the first six months post-injury, offering a robust natural history framework against which treatment effects or cohort differences can be evaluated The boxplots of GRASSP scores provide a clear and discriminative depiction of functional differences across AIS grades A through E. The longitudinal recovery profiles derived from observational follow-up illustrate how individuals progress over time, enabling comparison against expected recovery patterns. The data support the identification of cut-off GRASSP strength scores that may inform enrollment criteria for both chronic and acute clinical studies; metrics with fewer ceiling and floor effects are particularly valuable as more credible clinical trial endpoints. Finally, the dataset facilitates the identification of muscle groups whose recovery is most strongly associated with functional improvement, helping to refine rehabilitation targeting and outcome evaluation.\u003c/p\u003e \u003cp\u003ePerhaps most critically, the study highlights the relationships between individual muscle groups and overall functional independence. These findings carry important clinical implications: for instance, when rehabilitation resources are limited, understanding which muscles contribute most significantly to self-care outcomes enables clinicians to prioritize their interventions. Targeting key muscle groups\u0026mdash;identified through both statistical correlation and machine learning\u0026mdash;can potentially maximize functional gains and efficiency in upper limb rehabilitation for individuals with tetraplegia.\u003c/p\u003e"},{"header":"Conclusions","content":"\u003cp\u003eGV1, GV2 and GVM are all sound measures with rigorous development which can be implemented in clinical settings, academic research settings and in regulated industry sponsored trials. This work summarizes the robustness of the measure and can be used to substantiate the decision for GRASSP use in different trial types and studies. A measure with an available reference profile such as we have presented with GRASSP in this paper, substantiates it\u0026rsquo;s use and requirement as an important measure in the study of tetraplegia across the spectrum of clinical and research activity in the field.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAcknowledgements:\u003c/h2\u003e \u003cp\u003eWe would like to acknowledge the funding for this work: University of Toronto, Data Sciences Institute, Dana and Christopher Reeve Foundation, Praxis (Rick Hanson Institute), Ontario Neurotrauma Foundation, Craig Neilsen Foundation and Canadian Institutes for Health Research for the project and trainee funding. We would also like to acknowledge the collaboration with the European Multi-Center SCI Study and our ongoing work to further develop the GRASSP.\u003c/p\u003e\u003ch2\u003eData Availability:\u003c/h2\u003e \u003cp\u003eUpon request with an adequate Data Sharing Agreement.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eWeidner N, Abel R, Maier D, R\u0026ouml;hl K, R\u0026ouml;hrich F, Baumberger M, et al.; Nogo Inhibition in Spinal Cord Injury Study Group. Safety and efficacy of intrathecal antibodies to Nogo-A in patients with acute cervical spinal cord injury: A randomised, double-blind, multicentre, placebo-controlled, phase 2b trial. Lancet Neurol. 2025;24(1):42\u0026ndash;53. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/S1474-4422(24)00447-2\u003c/span\u003e\u003cspan address=\"10.1016/S1474-4422(24)00447-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMoritz C, Field-Fote EC, Tefertiller C, van Nes I, Trumbower R, Kalsi-Ryan S, et al. Non-invasive spinal cord electrical stimulation for arm and hand function in chronic tetraplegia: A safety and efficacy trial. Nat Med. 2024;30(5):1276\u0026ndash;1283. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41591-024-02940-9\u003c/span\u003e\u003cspan address=\"10.1038/s41591-024-02940-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCheng TT, Dore S, Fischer I, Tomycz ND. Upper extremity motor improvement with temporal interference stimulation in spinal cord injury: A pilot case series. J Neurosurg Spine. 2024;40:431\u0026ndash;437. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3171/2023.12.SPINE231058\u003c/span\u003e\u003cspan address=\"10.3171/2023.12.SPINE231058\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKalsi-Ryan S, Beaton D, Curt A, Duff S, Popovic MR, Rudhe C, et al. The Graded Redefined Assessment of Strength Sensibility and Prehension: Reliability and validity. J Neurotrauma. 2012;29:905\u0026ndash;914. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1089/neu.2010.1504\u003c/span\u003e\u003cspan address=\"10.1089/neu.2010.1504\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGRASSP. GRASSP development. (n.d.) \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://grassptest.com/development/\u003c/span\u003e\u003cspan address=\"https://grassptest.com/development/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Accessed 23 Nov 2025.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKalsi-Ryan S, Beaton D, Ahn H, Askes H, Drew B, Curt A, et al. Responsiveness, sensitivity and minimally detectable difference of the Graded and Redefined Assessment of Strength, Sensibility, and Prehension, Version 1.0 (GRASSP V1). J Neurotrauma. 2016;33:307\u0026ndash;314. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1089/neu.2015.4217\u003c/span\u003e\u003cspan address=\"10.1089/neu.2015.4217\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVelstra IM, Curt A, Frotzler A, Abel R, Kalsi-Ryan S, Rietman JS, et al. Changes in strength, sensibility and prehension in acute cervical spinal cord injury: European multicenter responsiveness study of the GRASSP. Neurorehabil Neural Repair. 2015;29(8):755\u0026ndash;766. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/1545968314565466\u003c/span\u003e\u003cspan address=\"10.1177/1545968314565466\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVelstra IM, Fellinghauer C, Abel R, Kalsi-Ryan S, Rupp R, Curt A. The Graded and Redefined Assessment of Strength, Sensibility, and Prehension Version 2 provides interval measure properties. J Neurotrauma. 2018;35(6):854\u0026ndash;863. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1089/neu.2017.5195\u003c/span\u003e\u003cspan address=\"10.1089/neu.2017.5195\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKalsi-Ryan S, Chan C, Verrier M, Curt A, Fehlings M, Bolliger M, et al. The graded redefined assessment of strength, sensibility and prehension version 2 (GV2): Psychometric properties. J Spinal Cord Med. 2019;42(S1):S149\u0026ndash;S153. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/10790268.2019.1616950\u003c/span\u003e\u003cspan address=\"10.1080/10790268.2019.1616950\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKalsi-Ryan S, Riehm LE, Tetreault L, Martin AR, Teoderascu F, Massicotte E, et al. Characteristics of upper limb impairment related to degenerative cervical myelopathy: Development of a sensitive hand assessment (Graded Redefined Assessment of Strength, Sensibility, and Prehension Version Myelopathy). Neurosurgery. 2020;86(3):E292\u0026ndash;E299. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1093/neuros/nyz499\u003c/span\u003e\u003cspan address=\"10.1093/neuros/nyz499\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKalsi-Ryan S, Verrier M, Curt A, Fehlings M. Development of the Graded Redefined Assessment of Strength, Sensibility and Prehension (GRASSP): Reviewing measurement specific to the upper limb in tetraplegia. J Neurosurg Spine. 2012;17(1 Suppl):65\u0026ndash;76. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3171/2012.6.AOSPINE1258\u003c/span\u003e\u003cspan address=\"10.3171/2012.6.AOSPINE1258\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKalsi-Ryan S, Balbinot G, Wang JZ, Abel R, Bolliger M, Curt A, et al. Minimal clinically important difference of Graded Redefined Assessment of Strength, Sensibility, and Prehension Version 1 in acute cervical traumatic spinal cord injury. J Neurotrauma. 2022;39:1645\u0026ndash;1653. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1089/neu.2021.0500\u003c/span\u003e\u003cspan address=\"10.1089/neu.2021.0500\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMarino RJ, Sinko R, Bryden A, Backus D, Chen D, Nemunaitis GA, Leiby BE. Comparison of responsiveness and minimal clinically important difference of the Capabilities of Upper Extremity Test (CUE-T) and the Graded Redefined Assessment of Strength, Sensibility and Prehension (GRASSP). Top Spinal Cord Inj Rehabil. 2018;24(3):227\u0026ndash;238. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1310/sci2403-227\u003c/span\u003e\u003cspan address=\"10.1310/sci2403-227\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMulcahey MJ, Thielen CC, Dent K, Sinko R, Sadowsky C, Martin R, et al. Evaluation of the graded redefined assessment of strength, sensibility and prehension (GRASSP) in children with tetraplegia. Spinal Cord. 2018;56(6):581\u0026ndash;588. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41393-018-0084-0\u003c/span\u003e\u003cspan address=\"10.1038/s41393-018-0084-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCatz A, Greenberg E, Itzkovich M, Bluvshtein V, Ronen J, Gelernter I. A new instrument for outcome assessment in rehabilitation medicine: Spinal cord injury ability realization measurement index. Archives of physical medicine and rehabilitation. 2004;85(3):399\u0026ndash;404.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBreiman L. Random forests. Mach Learn. 2001;45(1):5\u0026ndash;32. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1023/A:1010933404324\u003c/span\u003e\u003cspan address=\"10.1023/A:1010933404324\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen T, Guestrin C. XGBoost: A scalable tree boosting system (arXiv:1603.02754). arXiv. 2016. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.48550/arXiv.1603.02754\u003c/span\u003e\u003cspan address=\"10.48550/arXiv.1603.02754\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCunningham P, Delany SJ. k-Nearest Neighbour Classifiers: 2nd Edition (with Python examples) (arXiv:2004.04523). arXiv. 2020. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.48550/arXiv.2004.04523\u003c/span\u003e\u003cspan address=\"10.48550/arXiv.2004.04523\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHearst MA, Dumais ST, Osuna E, Platt J, Scholkopf B. Support vector machines. IEEE Intell Syst Their Appl. 1998;13(4):18\u0026ndash;28. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1109/5254.708428\u003c/span\u003e\u003cspan address=\"10.1109/5254.708428\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eClaesen M, De Moor B. Hyperparameter search in machine learning (arXiv:1502.02127 [cs.LG]). arXiv. 2015. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.48550/arXiv.1502.02127\u003c/span\u003e\u003cspan address=\"10.48550/arXiv.1502.02127\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLundberg SM, Erion GG, Lee S-I. Consistent individualized feature attribution for tree ensembles (arXiv:1802.03888 [cs.LG]). arXiv. 2019. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.48550/arXiv.1802.03888\u003c/span\u003e\u003cspan address=\"10.48550/arXiv.1802.03888\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"spinal-cord","isNatureJournal":false,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"sc","sideBox":"Learn more about [Spinal Cord](http://www.nature.com/sc/)","snPcode":"41393","submissionUrl":"https://mts-sc.nature.com/cgi-bin/main.plex","title":"Spinal Cord","twitterHandle":"@journalsci","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"spinal cord injury, traumatic tetraplegia, measurement properties, reliability, validity, responsiveness, minimal clinically important difference, minimally detectable difference, recovery profiles, machine learning, functional independence, AIS classification, cross-sectional chronic spinal cord injury, longitudinal analysis, upper limb function","lastPublishedDoi":"10.21203/rs.3.rs-8683387/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8683387/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eStudy Design: \u003c/strong\u003eNarrative Review and Retrospective Post-Hoc Analysis\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eObjectives:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e1. Summarize GRASSP literature and provide evidentiary tables to support clinical research decision making\u003c/p\u003e\n\u003cp\u003e2. Establish enrollment cut-offs for acute and cross-sectional chronic samples\u003c/p\u003e\n\u003cp\u003e3. Define recovery profiles of GRASSP subtests and typical relationships of impairment to function.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods: \u003c/strong\u003eA literature review of existing GRASSP publications was conducted. Two datasets were created from three historical cohorts of data collected for GRASSP development. A longitudinal dataset includes a total of 182 individuals with tetraplegia, each of whom was assessed at 0, 1, 3, and 6 months after injury (728 observations). The cross-sectional chronic dataset includes a total of 254 individuals with tetraplegia. The sample was stratified by AIS at Exam Stage 1. AIS classification of A-B were assigned to the motor complete group and AIS classification of C-D were assigned to the motor incomplete group\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eAnalysis:\u003c/em\u003e Literature was reviewed and knowledge synthesized. Data was summarized with descriptive statistics. Machine learning was applied to both datasets for prediction.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults: \u003c/strong\u003eEvidentiary tables summarize three versions of GRASSP and synthesize the literature to define the clinical research utility. Three figures define the discriminative groupings of GRASSP subtest scores along with recovery profiles. Machine learning identifies the predictive qualities of GRASSP Strength.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusion: \u003c/strong\u003eGRASSP use recommendations are meaningful to researchers/clinicians for implementation and sample characteristics of historical data provide insight for researchers into expected outcomes of individuals with traumatic tetraplegia.\u003c/p\u003e","manuscriptTitle":"Clinical Utility of GRASSP in Traumatic Tetraplegia: A Narrative Review and Retrospective Analysis Incorporating Machine Learning with an Explainability Framework","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-02-11 06:00:57","doi":"10.21203/rs.3.rs-8683387/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"editorInvitedReview","content":"This content is not available.","date":"2026-02-27T20:31:33+00:00","index":2,"fulltext":"This content is not available."},{"type":"reviewerAgreed","content":"This content is not available.","date":"2026-02-11T00:33:52+00:00","index":2,"fulltext":"This content is not available."},{"type":"reviewerAgreed","content":"This content is not available.","date":"2026-02-09T20:36:45+00:00","index":1,"fulltext":"This content is not available."},{"type":"reviewersInvited","content":"","date":"2026-02-09T10:55:45+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-01-29T18:18:54+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-01-29T18:16:24+00:00","index":"","fulltext":""},{"type":"submitted","content":"Spinal Cord","date":"2026-01-27T20:56:40+00:00","index":"","fulltext":""},{"type":"checksFailed","content":"","date":"2026-01-26T17:38:44+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"spinal-cord","isNatureJournal":false,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"sc","sideBox":"Learn more about [Spinal Cord](http://www.nature.com/sc/)","snPcode":"41393","submissionUrl":"https://mts-sc.nature.com/cgi-bin/main.plex","title":"Spinal Cord","twitterHandle":"@journalsci","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"5bd3f74f-1342-4251-997e-8b8e7f56887c","owner":[],"postedDate":"February 11th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":62572192,"name":"Health sciences/Medical research/Outcomes research"},{"id":62572193,"name":"Health sciences/Health care"}],"tags":[],"updatedAt":"2026-02-11T06:00:57+00:00","versionOfRecord":[],"versionCreatedAt":"2026-02-11 06:00:57","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8683387","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8683387","identity":"rs-8683387","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}