Introduction

Cervical spinal-cord injury (CSCI) is the most common type of spinal-cord injury (SCI) [1, 2]. CSCI has been reported to have a crucial impact on activities of daily living (ADL) owing to severe upper limb dysfunction [3, 4], and various clinical trials [5]. Symptoms of CSCI vary depending on whether the injury is complete or incomplete and the level of injury. Therefore, evaluation of upper-limb dysfunction in CSCI is important for appropriate intervention [6].

The International Standards for Neurological Classification of SCI upper extremity motor score (UEMS); Capabilities of Upper Extremity Test (CUE-T); and Graded Redefined Assessment of Strength, Sensibility, and Prehension (GRASSP) are recommended for evaluation of upper-limb function in individuals with CSCI [6]. These evaluations have been reported to have good reliability, validity, and responsiveness and are used in clinical practice [6,7,8]. UEMS indicates the total manual muscle test score. GRASSP is specific to grasping movements in addition to the neurological findings, and does not include evaluation of bimanual and gross movements. It is difficult to assess upper limb function of the whole CSCI with these evaluations. CUE-T consists of several subtests and requires considerable time for evaluation. However, it evaluates several factors, including fine, gross, and bimanual movements and allows for more detailed performance-based evaluation of upper-limb function in individuals with CSCI [8,9,10].

In clinical practice and trials for stroke, the interpretability of the evaluations has been applied. Severity classification, derived from the Fugl-Meyer Assessment (FMA) for upper limb function evaluation [11], has guided the establishment of adaptation criteria for interventions [12]. Additionally, cutoff values have been calculated for achieving independence in ADL based on assessments of balance ability and upper limb function [13, 14]. Effective improvement in ADL performance can be achieved through the implementation of therapeutic or compensatory interventions for functions that have not met these cutoff values [13, 14]. Therefore, standardized severity classifications and cutoff values in evaluations are useful for establishing intervention details and rehabilitation objectives.

SCI is generally classified as complete or incomplete based on the American Spinal Injury Association (ASIA) Impairment Scale (AIS). However, no classification based on performance evaluations, such as upper-limb function, has been reported. Furthermore, although a prediction model focusing on independence of gait has been validated [15], no prediction model using upper-limb function, which is strongly related to ADL, has been reported. Severity classification and predictors for independence in ADL may be useful because the status of upper-limb function in CSCI is complex.

Therefore, this study aimed to classify the severity of upper-limb dysfunction and calculate the cutoff values for independence in ADL using CUE-T to determine adaptation criteria for interventions, verify effectiveness, and set goals for clinical practice. The severity classifications and cutoff values calculated in this study may facilitate ADL goal setting and intervention selection in clinicians and facilities with limited experience in CSCI. This may contribute to the standardization of upper limb functional rehabilitation for CSCI.

Participants and methods

Participants

This prospective observational study included participants who underwent CSCI rehabilitation at the Chiba Rehabilitation Center in Japan between 2019 and 2023. Subacute CSCI was defined as occurrence of symptoms within 9 months of injury, based on a previous study [16, 17]. The exclusion criteria were as follows: upper-limb dysfunction due to injury other than spinal-cord injury such as fracture, peripheral-nerve injury, or central nervous system disorders such as stroke; difficulty in understanding the examination content due to cognitive or mental dysfunction; and a total CUE-T score of 0.

Data collection

The following data were collected: sex, age, injury duration (days), AIS score, neurological level of injury, traumatic or non-traumatic, CUE-T, UEMS, and Spinal Cord Independence Measure III (SCIM III). CUE-T and other data were collected within 14 days of obtaining consent. Participants were in the subacute phase within 9 months of injury [17], excluding those in the acute phase (within 15 days of injury) [18]. All evaluations were performed by registered occupational therapists trained at Chiba Rehabilitation Center. We used a training video created by Thomas Jefferson University [19]. No blinding was performed during data collection.

Capabilities of Upper Extremity Test

CUE-T is a CSCI-specific upper-limb-function-evaluation tool developed by Tomas Jefferson University [8]. The evaluation items include gross movements, consisting mainly of reaching movements in various directions, and skillful movements (pinch strength, grip strength, speed to manipulate objects), and bimanual movements. Scoring is based on the number of times a specified movement is performed in 30 s, grip strength, and pinch strength, and each item is scored from 0 to 4 points. The total score ranges from 0 to 128 points, including scores for the hand (36 points per side) for fine movements and the side (60 points per side), excluding bimanual movements [8]. Previous studies have reported that CUE-T has good reliability, validity, and responsiveness [8, 10].

International standards for neurological classification of SCI

This evaluation is considered the gold standard for SCI [20]. It is based on the neurological parameters of muscle strength and sensation after spinal-cord injury. Based on the AIS score, injury was classified into complete (A), incomplete (B, C, D), or recovered (E). UEMS is the total manual muscle test score of the elbow flexors (C5), wrist extensors (C6), elbow extensors (C7), finger flexors (C8), and finger abductors (T1) and ranges from 0 to 50 points [20].

Spinal cord independence measure III

This ADL evaluation is specific to SCI. SCIM III evaluates “Self-care,” “Respiration and sphincter management,” and “Mobility.” “Self-care” refers to feeding, dressing, bathing, and grooming. “Respiration and sphincter management” refers to breathing, urination and defecation management, and toilet use. “Mobility” refers to the ability to transfer and move. These are scored based on the participants’ performance. In “Self-care,” scores for feeding, bathing, and grooming range from 0 (needing assistance) to 3 (completely independent). Dressing scores range from 0 (needing assistance) to 4 (completely independent). The total score ranges from 0 to 100 points [21].

Statistical analysis

Severity classification

The hierarchical cluster analysis used the total CUE-T score as an independent variable. The degree of dissimilarity between the participants was calculated using the squared Euclidean distance. Ward’s method was used to define the degree of dissimilarity between the clusters. A dendrogram was created from the results of the hierarchical cluster analysis, and a dividing line was drawn based on the dendrogram to determine the appropriate number of clusters.

To analyze whether the clusters were ordered according to severity and the characteristics of each cluster, we verified the differences in the CUE-T subtests, basic information, upper-limb function, and ADL between the clusters. Differences in the CUE-T subtests were compared between clusters using the percentage of each subtest. Ordinal variables such as sex, AIS score, and neurological level of injury were compared using the χ2 test. Interval variables such as age, injury duration (days), and total CUE-T, UEMS, and SCIM III scores were compared using the Kruskal–Wallis test with post-hoc Bonferroni correction. The effect size was determined using φ and Cramer’s V for ordinal variables and r (Z/√N) for interval variables [22]. The criteria were “0.10 (small)”, “0.30 (medium)”, and “0.50 (large)” [23]. The cutoff scores, which are the thresholds between clusters, were defined as the average of the highest and lowest scores of adjacent clusters [11].

Calculation of cutoff value leading to ADL independence

The SCIM III classification system was used to determine the degrees of independence. Dependence was binary: for feeding, bathing (upper and lower body), dressing (upper and lower body), and grooming 0 and 1 points indicated “non-independent,” whereas for feeding and grooming and bathing (upper and lower body) and dressing (upper and lower body) 2 and 3 points and 2, 3, and 4 points, respectively, indicated “independent.” Logistic regression analysis was performed with binary ADL independence as the dependent variable and the total CUE-T score as the independent variable. The cutoff values were calculated by fitting the obtained regression equation to the adjustment equation reported by Terluin et al. [24]. Sensitivity, specificity, and positive and negative predictive values were calculated for these cutoff values. The cutoff values were validated using the statistical assumptions of the logistic regression analysis. Linearity in the logit transformation of continuous variables was verified using the Box–Tidwell test [25], because only one independent variable was input in the analysis. Strong influence outliers were defined as variables with four or more standard deviations [25]. The goodness of fit of the obtained regression equation was verified using the Hosmer–Lemeshow test. Internal validation of the model was performed using logistic regression analysis. The area under the curve (AUC) of the original data model was compared to that of bootstrapping (a random sample of 1000 from the original data set). Sensitivity analysis was performed by calculating and comparing the cutoff values for the AIS-C and AIS-D only the incomplete-injury groups.

Statistical analyses were performed using SPSS Statistics version 29 (IBM, Armonk, New York), the R4.3.1 rms package, and Microsoft Excel 2019 (Microsoft, Redmond, Washington, USA). Statistical significance was set at p < 0.05.

Sample size

The recommended sample size for segmentation studies, including cluster analysis, is at least 70 × the number of variables [26]. The only variable included in this study was CUE-T. Therefore, the sample size was set at ≥70.

Results

Participants

A total of 76 participants were enrolled in this study. Of these, five participants with upper-limb dysfunction due to causes other than SCI or a CUE-T score of 0 were excluded. Finally, 71 participants (60 men and 11 women) were included in the study (Supplement 1). Of these, 9,7,18, and 37 were classified as AIS-A, AIS-B, AIS-C, and AIS-D, respectively (Table 1). The median age of participants was 61.0 years (interquartile range, 49.5–67.0).

Table 1 Characteristics of all participants.
Full size table

Severity classification

Severity was classified into four categories using CUE-T. The categories were in the order of upper-limb function and ADL ability. Based on a dendrogram (Supplement 2) derived from the hierarchical cluster analysis, participants were classified into four clusters as follows: Cluster 1, severe dysfunction; Cluster 2, severe to moderate dysfunction; Cluster 3, moderate to mild dysfunction; and Cluster 4, mild dysfunction. Comparisons of the basic information, upper-limb function, ADL, and CUE-T scores between the clusters are shown in Table 2. Regarding basic information, differences between clusters were observed only for AIS. Significant differences in CUE-T, UEMS, SCIM III, and SCIM III Self-care were observed between all clusters, except between Clusters 3 and 4. The effect size (r) was “large” for all cluster pairs (Supplement 3), and scores did not overlap between clusters. The thresholds between Clusters 1 and 2, 2 and 3, and 3 and 4 were 19, 61, and 98 points, respectively.

Table 2 Analysis of all clusters.
Full size table

A subtest was used to compare the clusters (Fig. 1). In Cluster 1 (severe dysfunction), a good response was obtained for “Pull” (pulling a pan on the table). In Cluster 2 (severe-to-moderate dysfunction), high scores were obtained for “Reach Forward” and “Push,” (extend arm forward in the air or on a table), and a good response was obtained for “Wrist Up” (repetitive flexion and extension of the wrist joint). In Cluster 3 (moderate-to-mild dysfunction), high scores were obtained for “Lift Up” (lifting weight bimanually), “Container,” and “Acquire Release” (grip and release using finger movements). In Cluster 4 (mild dysfunction), high scores were obtained in all subtests.

Fig. 1: Characteristics of the individual subtests of the four severity classifications.
figure 1

This diagram shows the characteristics of the CUE-T subtest in four categories. Rt, Right; Lt, Left.

Full size image

Calculation of cutoff values for independence in ADL

Logistic regression analysis yielded significant regression equations for all items (p < 0.05). Linearity in the logit transformation of the continuous variables was confirmed, and no strong influence outlier with four or more standard deviations was identified. The results of the Hosmer–Lemeshow test confirmed the goodness of fit (p > 0.05).

The cutoff values calculated using an adjustment formula with regression equations were as follows: feeding, 37 points; bathing the upper body, 91 points; bathing the lower body, 90 points; dressing the upper body, 82 points; dressing the lower body, 81 points; grooming, 60 points.

For all cutoff values, the sensitivity, specificity, positive predictive value and negative predictive value were 0.73–0.96, 0.83–0.98, 0.67–0.96, and 0.83–0.96, respectively (Table 3). In the internal validation, AUC of the model for all cut-off values and that of the model calculated using bootstrapping were similar (Table 3). For the group with incomplete injuries (AIS C-D), the cutoff values calculated using sensitivity analysis were similar (Supplement 4).

Table 3 Cutoff values, sensitivity, specificity, positive predictive value, negative predictive value, and AUC.
Full size table

Discussion

The severity using the CUE-T was classified into four categories. These were in order of upper limb function and ADL ability. We calculated cutoff values leading to ADL (Feeding, Dressing, Bathing, Grooming) independence. We created a figure (Fig. 2) combined these results. Feeding and Grooming are included in the Cluster 2, Dressing and Bathing are included in the Cluster 3. These findings may serve in using the CUE-T results to assess the status of upper limb function and stage the setting of goals for ADL using the upper limb.

Fig. 2: Diagram combining severity classification and cutoff values.
figure 2

Four clusters and their corresponding ADL cutoff values.

Full size image

Severity classification

Based on the CUE-T scores, the severity of CSCI upper-limb dysfunction was classified into four categories using hierarchical cluster analysis. In a previous study, a reanalysis was performed by removing duplicate scores between clusters [11]. We analyzed all data, and the results were adopted because no duplicate scores between clusters were observed. The total Functional Independence Measure score ranges from 18 to 126 points and is classified into three categories [27]. This suggests that classification into four clusters was appropriate because the total CUE-T score is higher than the total Functional Independence Measure score.

Significant differences in CUE-T scores between the clusters were observed, and the effect size was “large.” Analysis of the characteristics of each cluster revealed significant differences between clusters in UEMS, which evaluates upper-limb function and SCIM III and SCIM III Self-care, which evaluate ADL, and the effect size was “large.” The CUE-T subtests were analyzed for each cluster. Cluster 1 was characterized by “Pull” movement of the elbow flexors. Cluster 2 was characterized by “Push” movement of the elbow extensors and “Wrist Up”‘ of the wrist extensors. Cluster 3 was characterized by the grip and release movements of the finger extensors and flexors. These results correspond to those for the key muscles in ISNCSCI [20], in the order of degree of injury. A previous study showed good correlation between CUE-T and UEMS [10], implying a link between the CUE-T score and key muscle-related movements. Therefore, we believe that this severity classification reflects the order of difficulty in ADL and neurological findings.

Regarding the participants’ demographics, significant differences in AIS scores were observed between clusters. However, the proportion of individuals with AIS-D was the highest in Clusters 2, 3, and 4. Among ISNCSCI, AIS-D has the best functional status except AIS-E [20]. Thus, the high proportion of AIS-D in multiple clusters suggests that this severity classification reflects the function of the upper limb as a whole in CSCI, including both complete and incomplete injuries. Recently, the number of incomplete CSCI has been increasing [1, 2]. We believe that an analysis focused on incomplete injuries would be more useful in the future, because this study included participants with complete and incomplete injuries.

In stroke, upper limb dysfunction is classified based on performance evaluations such as FMA is utilized in clinical trials [11]. Classifying severity based on performance has the advantage of facilitating the development of specific interventions. ISNCSCI is the gold standard for SCI evaluation, and AIS is widely used [20]. AIS focuses on the neurological aspects and does not include a performance evaluation [20]. The performance status in individuals with incomplete CSCI varies [16], and classification using AIS alone is difficult. Particularly, upper limb function in CSCI is complex and important factor with high hope for recovery [3, 4], and a more detailed classification is necessary. CUE-T can be used to evaluate upper-limb function in individuals with CSCI [8, 10]. Therefore, a severity classification combining CUE-T and ISNCSCI may be more beneficial for determining the upper-limb function and setting the criteria for interventions in individuals with CSCI.

Cutoff values for independence in ADL

The cutoff values of CUE-T scores (total 128 points) for independence in ADL ranged from 37 to 91 points. These cutoff values were the lowest for feeding and the highest for bathing. Except for feeding and grooming, this order of difficulty was different from that in a previous study based on the Rasch analysis of SCIM III [21]. However, the previous study included all individuals with SCI, whereas this study included only individuals with CSCI. Therefore, we consider this to be a new finding because focusing only on the upper-limb function is difficult in a study that includes all types of SCI.

The cutoff values for dressing and bathing were higher. The cutoff values were higher for the upper body in bathing and dressing. Cluster 3, which states that bathing and dressing are independent, is characterized by “Lift Up” (lifting both upper limbs upward). It has been reported that upper limbs lifting movements are limited in individuals with CSCI [28]. We believe this result indicates the difficulty in ADL movements that require lifting the upper limbs. Additionally, it is suggested that upper and lower dressing requires balance ability which makes it more difficult [29].

However, some of the cutoff values had a sensitivity, specificity, negative predictive value, and positive predictive value <80%. The positive predictive value for bathing upper body was 67%. It is assumed that ADL, include many factors, such as balance ability [14, 29]. Furthermore, age-related body composition, motivation, and cognition may influence ADL. In future analyses, including factors other than upper-limb function is necessary to create a tool to accurately predict ADL in individuals with CSCI.

In this study, the number of participants in the independent and non-independent groups based on each SCIM III Self-care item was different. Thus, the cutoff values were determined using an adjustment formula [24]. If the number of participants in either of the two groups separated by an external anchor is <50% of the total number of participants, the proposed method, which corrects for participant bias, is more accurate [24]. In the internal validation, the AUC for bootstrapping and that for the original data did not differ, and the reproducibility of the calculated cutoff values using the model was satisfactory. In the sensitivity analysis, the incomplete-injury group showed similar results. This suggests that the cut-off values calculated in this study are effective.

CUE-T evaluates “activities” in the International Classification of Functioning, Disability, and Health [6]. We proposed a suitable tool for predicting independence in ADL. Currently, there is no such tool in CSCI upper limb function, and the established cutoff values may be beneficial in clinical practice.

Limitations

This study collected data from a single center. Hence, the results may not reflect the findings in the general population. The results of the severity classification and calculated cutoff values may vary depending on the population. In addition, the study included participants within 9 months of injury, and both complete and incomplete injuries were analyzed. ADL varies depending on the duration of the injury and whether the injury is complete or incomplete. Therefore, it is necessary to clarify the duration and severity of the injury in detail. Moreover, in this study, we analyzed only objective data, and we believe that analysis using subjective evaluations is necessary in the future. Finally, data collection was not blinded. This may have introduced bias, and should be addressed in future studies.

Conclusion

In this study, we established a severity classification for upper-limb dysfunction in CSCI and calculated the cutoff values for independence in ADL using CUE-T. The classification reflected the severity of upper-limb dysfunction and ADL in CSCI. The cutoff values showed good results in the internal validation and sensitivity analysis and were effective. Our findings may help apply the CUE-T scores for evaluation of upper-limb function and staging goals for ADL using the upper limb in clinical trials and practice.