Derivation and Validation of a Clinical Prediction Rule for Upper Limb Functional Outcomes After Traumatic Cervical Spinal Cord Injury

IMPORTANCE: Traumatic cervical spinal cord injury (SCI) can result in debilitating paralysis. Following cervical SCI, accurate early prediction of upper limb recovery can serve an important role in guiding the appropriateness and timing of reconstructive therapies. OBJECTIVE: To develop a clinical prediction rule to prognosticate upper limb functional recovery after cervical SCI. DESIGN, SETTING, AND PARTICIPANTS: This prognostic study was a retrospective review of a longitudinal cohort study including patients enrolled in the National SCI model systems (SCIMS) database in US. Eligible patients were 15 years or older with tetraplegia (neurological level of injury C1-C8, American Spinal Cord Injury Association [ASIA] impairment scale [AIS] A-D), with early (within 1 month of SCI) and late (1-year follow-up) clinical examinations from 2011 to 2016. The data analysis was conducted from September 2021 to June 2022. MAIN OUTCOMES AND MEASURES: The primary outcome was a composite of dependency in eating, bladder management, transfers, and locomotion domains of functional independence measure at 1-year follow-up. Each domain ranges from 1 to 7 with a lower score indicating greater functional dependence. Composite dependency was defined as a score of 4 or higher in at least 3 chosen domains. Multivariable logistic regression was used to predict the outcome based on early neurological variables. Discrimination was quantified using C statistics, and model performance was internally validated with bootstrapping and 10-fold cross-validation. The performance of the prediction score was compared with AIS grading. Data were split into derivation (2011-2014) and temporal-validation (2015-2016) cohorts. RESULTS: Among 2373 patients with traumatic cervical SCI, 940 had complete 1-year outcome data (237 patients [25%] aged 60 years or older; 753 men [80%]). The primary outcome was present in 118 patients (13%), which included 92 men (78%), 83 (70%) patients who were younger than 60 years, and 73 (62%) patients experiencing AIS grade A SCI. The variables significantly associated with the outcome were age (age 60 years or older: OR, 2.31; 95% CI, 1.26-4.19), sex (men: OR, 0.60; 95% CI, 0.31-1.17), light-touch sensation at C5 (OR, 0.44; 95% CI, 0.44-1.01) and C8 (OR, 036; 95% CI, 0.24-0.53) dermatomes, and motor scores of the elbow flexors (C5) (OR, 0.74; 95% CI, 0.60-0.89) and wrist extensors (C6) (OR, 0.61; 95% CI, 0.49-0.75). A multivariable model including these variables had excellent discrimination in distinguishing dependent from independent patients in the temporal-validation cohort (C statistic, 0.90; 95% CI, 0.88-0.93). A clinical prediction score (range, 0 to 45 points) was developed based on these measures, with higher scores increasing the probability of dependency. The discrimination of the prediction score was significantly higher than from AIS grading (change in AUC, 0.14; 95% CI, 0.10-0.18; P < .001). CONCLUSIONS AND RELEVANCE: The findings of this study suggest that this prediction rule may help prognosticate upper limb function following cervical SCI. This tool can be used to set patient expectations, rehabilitation goals, and aid decision-making regarding the appropriateness and timing for upper limb reconstructive surgeries.