Linking Patterns of Intraoperative Neuromonitoring (IONM) Alerts to the Odds of a New Postoperative Neurological Deficit: Analysis of 27,808 Cervical Spine Procedures From a National Multi-institutional Database.

Abstract

Retrospective review of a national multi-institutional database of 27,808 extradural cervical spine procedures performed between January 2017 and May 2021. Characterize intraoperative neuromonitoring alerts by the patterns of modalities and nerves/muscles involved and quantify risk of new-onset neurological deficit for patients with a primary diagnosis of myelopathy, stenosis, or radiculopathy. Phenotyping alert patterns and linking those patterns with risk is needed to facilitate clinical decision-making. Cases with alerts were categorized by patterns of modalities or nerves/muscles involved, and alert status at closure. Unadjusted odds ratios (ORs) for new-onset neurological deficit were calculated. A mixed-effects logistic regression model controlling for demographic and operative factors, with random intercepts to account for clustering in outcomes by surgeon and surgical neurophysiologist was also used to calculate ORs and probabilities of neurological deficit. There was significantly increased risk of a new neurological deficit for procedures involving posterior compared with anterior approaches (OR: 1.82, P=0.001) and procedures involving three levels compared with one (OR: 2.17, P=0.001). Odds of a deficit were lower for patients with radiculopathy compared with myelopathy (OR: 0.69, P=0.058). Compared with cases with no alerts, those with unresolved Spinal Cord alerts were associated with the greatest elevation in risk (OR: 289.05) followed by unresolved C5-6 Nerve Root (OR: 172.7), C5-T1 Nerve Root/Arm (OR: 162.89), C7 Nerve Root (OR:84.2), and C8-T1 Nerve Root alerts (OR:75.49, all P<0.001). Significant reductions in risk were seen for resolved Spinal Cord, C5-6 Nerve Root, and C8-T1 nerve alerts. Overall, unresolved motor evoked potential and somatosensory evoked potential alerts were associated with the greatest elevation in risk (OR:340.92) followed by unresolved motor evoked potential-only (OR:140.6) and unresolved somatosensory evoked potential-Only alerts (OR:78.3, all P<0.001). These relationships were similar across diagnostic cohorts. Risk elevation and risk mitigation after an intraoperative neuromonitoring alert during surgery is dependent on the type and pattern of alert.