BACKGROUNDIntraoperative neuromonitoring (IONM) alert is one of the worrying events of kyphosis corrective surgery, which can result in a postoperative neurological deficit. To our knowledge, there is no risk prediction score to predict such events in patients undergoing kyphosis surgery. PURPOSETo develop a new preoperative MRI-based cord morphology classification (CMC) and risk prediction score for predicting IONM alerts in patients with kyphotic deformity. STUDY DESIGNRetrospective analysis of prospectively collected data. PATIENT SAMPLEAbout 114 patients undergoing surgical correction for kyphotic deformity. OUTCOME MEASURESIntraoperative neuromonitoring alerts and postoperative neurological status using AIS grading. METHODSKyphotic deformity patients undergoing posterior spinal fusion were retrospectively reviewed. Based on the morphology of the spinal cord and surrounding CSF in MRI, there are 5 types of cord. Type 1 (normal cord): circular cord with surrounding visible CSF between the cord and the apex, Type 2 (flattened cord): cord with <50% distortion at the apex with obliteration of the anterior CSF; Type 3 (deformed cord): cord with >50% distortion at the apex with complete obliteration of the surrounding CSF; Type 4 (stretched cord): the cord is stretched and atrophied over the apex of the curve. Type 5 (translated cord): horizontal translation of the cord at the apex with buckling collapse of the vertebral column. Preoperative radiographs were used to measure the preoperative sagittal cobbs angle, sagittal deformity angular ratio (S-DAR), sagittal vertical axis (SVA), apex of the curve, and type of kyphosis. Clinical data like the duration of symptoms, clinical signs of myelopathy, neurological status (AIS grade), grade of myelopathy using the mJOA score, and type of osteotomy were documented. Multivariate logistic regression was used to determine the risk factors for IONM alerts and the risk prediction score was developed which was validated with new cohort of 30 patients. RESULTSA total of 114 patients met the inclusion criteria. IONM alerts were documented in 33 patients (28.9%), with full recovery of the signal in 25 patients and a postoperative deficit in 8 patients. Rate of IONM alerts was significantly higher in Type 5 (66%), followed by Type 4 (50%), Type 3 (21.1%), Type 2 (11.1%), and Type 1 (11.1%) (p-value<.001). Based on multiple logistic regression, 7 factors, namely preoperative neurological status, mJOA score≤6, presence of signs of myelopathy, apex of the curve above T5, preoperative sagittal cobbs, S-DAR, and MRI-based CMC, were identified as risk predictors. The value for the risk factors varies from 0 to 4, and the maximum total risk score was 13. The cut-off value of 6 had good sensitivity (84.9%) and specificity (77.8%) indicating a high risk for IONM alerts. The AUC of the predictive model was 0.92, indicating excellent discriminative ability. CONCLUSIONWe developed and validated a risk predictive score that identifies patients at risk of IONM alerts during kyphosis surgery. Identification of such high-risk patients (risk score≥6) helps in proper evaluation and preoperative counselling and helps in providing a proper evidence-based reference for treatment strategies.