Abstract

OBJECTIVE The escalating incidence of de novo spinal infections poses a substantial neurological impact on patients. This has prompted a growing interest in discerning which patients would derive greater benefit from medical as opposed to surgical management of these occurrences. The authors assessed the predictive applicability of the Spinal Infection Treatment Evaluation (SITE) score in discerning between surgical intervention and medical management. This assessment represents the first external validation of the SITE score conducted in a cohort of patients with de novo spinal infections. METHODS A comprehensive retrospective chart review was conducted to identify patients diagnosed with de novo spinal infections (osteomyelitis, discitis, or epidural abscess) at a tertiary center between July 1, 2004, and March 31, 2023. All necessary data for calculating the SITE score were collected for each patient. Surgical intervention was advised for patients scoring 0–8 or exhibiting acute plegia or bladder or bowel dysfunction and optional for those scoring 9–12; medical treatment was recommended for patients scoring 13–15. Predictability of the score was scrutinized using the area under the curve (AUC) of the receiver operating characteristic (ROC) curve. RESULTS Among 194 identified patients, the mean ± SD age was 65.96 ± 13.66 years and 58% were men. Stratification of patients based on medical and surgical management revealed that 27% underwent medical treatment alone and 73% required surgical intervention. In the medical group, 72.2% of patients were neurologically intact compared to 50% in the surgical group (p = 0.006). Surgically managed patients exhibited a higher incidence of spinal stenosis with impingement of the spinal cord, with or without deformity, when compared to nonsurgical patients (38.6% vs 22.2%, p = 0.04). Additionally, surgically managed patients had a lower mean ± SD SITE score (7.16 ± 2.39 vs 8.2 ± 2.33, p < 0.005) and were more likely to have multilevel infection than patients who underwent medical management (59.3% vs 33.3%, p < 0.001). When patients were categorized on the basis of SITE score, the sensitivity of the score (using a threshold of 8) to predict surgical management was 68.6% and specificity was 59.3%. According to ROC curve, the SITE score exhibited an AUC of 0.66. CONCLUSIONS Validation of the SITE score could not accurately predict medical versus surgical management in a tertiary center cohort of patients with de novo spinal infections. Further multicenter studies incorporating additional variables and larger cohorts are imperative to develop an optimal predictive tool.