Preoperative Dermatomal Somatosensory Evoked Potentials in Risk Prediction of Postoperative Neurological Deficit After Thoracic Spine Surgery: A Retrospective Cohort Study

Background Postoperative neurological deficit represents one of the most serious complications following thoracic spine surgery. Traditional risk assessment primarily relies on clinical and imaging variables, whereas the predictive value of electrophysiological indicators in preoperative risk stratification remains insufficiently explored. Dermatomal somatosensory evoked potentials (DSEP) directly reflect the integrity of sensory conduction pathways and may provide functional information beyond conventional structural imaging. This study aimed to develop and internally validate a clinically applicable prediction model integrating electrophysiological parameters. Methods A total of 508 patients who underwent thoracic decompression surgery were retrospectively included. Collected variables comprised age, preoperative Japanese Orthopaedic Association (JOA) score, number of compressed levels, T2-weighted signal changes, and DSEP parameters including latency, amplitude, and number of abnormal segments. Multivariable logistic regression models were constructed, including a clinical model, a clinical–imaging model, a clinical–electrophysiological model, and a combined model. Internal validation was performed using stratified five-fold cross-validation. Model discrimination was assessed using receiver operating characteristic (ROC) curves and the area under the curve (AUC). Ninety-five percent confidence intervals were estimated using bootstrap resampling. Results Among the 508 patients, 107 (21.1%) developed postoperative neurological deficit. Multivariable analysis identified age (OR = 1.03), preoperative JOA score (OR = 0.67), number of compressed levels (OR = 1.84), and maximal N1 latency (OR = 1.13) as independent predictors. In cross-validated model evaluation, the clinical model demonstrated limited discriminative ability (AUC = 0.58, 95% CI 0.519–0.64). The clinical–imaging model showed modest improvement (AUC = 0.625, 95% CI 0.561–0.683). Incorporation of electrophysiological parameters substantially improved prediction performance, with the clinical–electrophysiological model achieving an AUC of 0.736 (95% CI 0.688–0.789). The combined model integrating clinical, imaging, and electrophysiological variables showed the highest overall performance (AUC = 0.742, 95% CI 0.689–0.792). Conclusions Preoperative DSEP parameters, particularly maximal N1 latency, significantly improve prediction of postoperative neurological deficit after thoracic spine surgery. Integration of electrophysiological and clinical variables may enhance perioperative risk stratification and support individualized surgical decision-making.