Prognostic influences and prediction model construction for traumatic cervical spinal cord injury

Background : Traumatic cervical spinal cord injury (TCSCI), a critical neurological disorder with rising global incidence, burdens patients physically, psychologically and socially, but has relatively scarce prognostic assessment research compared to therapeutic studies. Objective : This study aims to investigate a novel approach for prognostic assessment in TCSCI. Methods : A retrospective analysis of clinical data was carried out on patients who received surgical treatment at Xi'an Honghui Hospital from March 2016 to May 2023. Nine factors were incorporated in the study: sex, age, hypertension, diabetes mellitus (DM), damage extent, ossification of the posterior longitudinal ligament (OPLL), cervical spinal canal stenosis grading (CSCS), surgical timing, and postoperative complications. The patients' preoperative clinical details, imaging data, and postoperative clinical information were taken into account. A prognostic model was established by statistically analyzing these differential factors and evaluating the sensitivity and specificity for predicting the occurrence of paralysis. The efficacy of the model was evaluated using the area under the curve (AUC) of the receiver operating characteristic (ROC) and the bootstrap resampling method (with 1000 iterations). Results : Patients with DM were 4.009 times more likely to experience postoperative paralysis than those without (95% CI: 1.250 - 12.863, P = 0.020). Individuals with diffuse damage had a 7.912 - fold higher probability of developing paralysis compared to those without diffuse damage (95% CI: 3.466 - 18.063, P < 0.001). Patients with CSCS grade III were 8.862 times more likely to develop paralysis than those with grades 0, I, or II (95% CI: 3.280 - 23.946, P < 0.001). The probability of paralysis in patients with two or more postoperative complications was 4.625 times greater (95% CI: 1.107 - 19.323, P = 0.036) than in those without complications. The probability of paralysis in patients with a surgical time exceeding 3 days was 3.132 times higher (95% CI: 1.325 - 7.407, P = 0.009) than in those with a surgical time within 3 days. Among these factors, CSCS had the most significant impact on the prognosis of paralytic patients, followed by the damage extent. Sex, age, hypertension, and OPLL were not associated with prognosis ( P > 0.05). The AUC of the model was 0.850, and the optimal cutoff value was 0.227, which outperformed the specificity and sensitivity of the ASIA prediction model (0.721, 0.816). Additionally, the calibration curve and decision - curve analysis (DCA) indicated that the model had stable performance and clinical benefits. Conclusion : TCSCI prognosis may be influenced by DM, damage extent, CSCS, surgical timing, and postoperative complication count, with a clinical prognostic model based on these factors showing promising predictive ability for paralysis outcomes.