Predicting the occurrence of complications following corrective cervical deformity surgery: Analysis of a prospective multicenter database using predictive analytics

Abstract

We developed a predictive model to describe risk factors for complications in cervical deformity surgeries. Cervical deformity (CD) surgical patients are growing in number, but remain under-studied in the literature. CD was defined as at least one of the following: C2–C7 Cobb >10°, CL >10°, cSVA >4 cm, CBVA >25°. Patient demographics and clinical data were assessed as risk factors for medical/surgical complications using multivariate regression models. 123 patients underwent CD surgery (60.6 yrs, 60.8% F). The most common complications were neurologic (24.4%), dysphagia (13.0%), cardiopulmonary (11.4%), infection (9.7%). 51 (41.5%) of patients experienced a medical complication and 73 (59.3%) had a surgical complication. An overall complication was predicted with high accuracy (AUC = 0.79) by the following combinations of factors: higher baseline EQ5D pain and lower baseline EQ5D anxiety/depression scores, and higher cervical and global SVA. A medical complication can be predicted by male gender, baseline mJOA score, and cervical SVA (AUC = 0.770). A surgical complication can be predicted by higher estimated blood loss, lower anxiety scores, and larger global SVA (AUC = 0.739). 64.2% of patients undergoing cervical deformity correction sustained any complication. While the most reliable predictor of the occurrence of a complication involved a cluster of risk factors, a radiographic baseline sagittal parameter of cervical SVA was the strongest isolated predictor for complications across categories. Although these findings are specific to a cervical population with moderate to severe deformities, collectively they can be utilized for pre-operative risk assessment and patient education.