To explore the clinical characteristics and surgical treatment strategies of delayed spinal cord injury (SCI) caused by atypical compression of old thoracolumbar fracture. Between January 2011 and June 2018, 32 patients with delayed SCI caused by atypical compression of old thoracolumbar fracture who met the inclusion criteria were admitted and divided into group A (20 cases, underwent anterior subtotal vertebral body resection+titanium mesh reconstruction+screw rod internal fixation) and group B (12 cases, underwent posterior 270° ring decompression of vertebral canal+titanium mesh reconstruction+screw rod internal fixation) according to the different operation approaches. There was no significant difference between the two groups in age, gender, cause of injury, fracture segment, disease duration, preoperative American Spinal Injury Association (ASIA) classification, and preoperative back pain visual analogue scale (VAS) score, lumbar Japanese Orthopaedic Association (JOA) score, kyphosis angle, and vertebral canal occupational ratio ( P>0.05). The incision length, operation time, intraoperative blood loss, complications, and bone fusion time of reconstructed vertebrae were recorded and compared between the two groups; the kyphosis angle, back pain VAS score, and lumbar JOA score were used to evaluate the effectiveness. Except that the incision length in group A was significantly shorter than that in group B ( t=-4.865, P=0.000), there was no significant difference in intraoperative blood loss and operation time between the two groups ( P>0.05). There was no deaths or postoperative paraplegia cases in the two groups, and no deep infection or skin infection occurred. There was 1 case of cerebrospinal fluid leakage, 1 case of inferior vena cava injury, and 1 case of chyle leakage in group A. No serious complications occurred in group B. There was no significant difference in the incidence of complications between the two groups ( P=0.274). All 32 patients were followed up 12-61 months, with an average of 20.8 months. The follow-up time for groups A and B were (19.35±5.30) months and (23.25±12.20) months respectively, and the difference was not significant ( t=-1.255, P=0.219). The reconstructed vertebrae in all cases obtained bony fusion postoperatively. The fusion time of groups A and B were (8.85±2.27) months and (8.50±2.50) months respectively, and the difference was not significant ( t=0.406, P=0.688). The kyphosis angle, back pain VAS score, and lumbar JOA score of the two groups at each time point after operation and last follow-up were significantly improved when compared with preoperatively ( P<0.05); the lumbar JOA score was further improved with time postoperatively ( P<0.05), while the kyphosis angle and the VAS score of back pain remained similarly ( P>0.05). Comparison of kyphosis angle, back pain VAS score, and lumbar JOA score between the two groups at various time points postoperatively showed no significant difference ( P>0.05). At last follow-up, the JOA score improvement rate in groups A and B were 83.87%±0.20% and 84.50%±0.14%, respectively, and the difference was not significant ( t=-0.109, P=0.914); the surgical treatment effects of the two groups were judged to be significant. In the later stage of treatment of old thoracolumbar fractures, even mild kyphosis and spinal canal occupying may induce delayed SCI. Surgical correction and decompression can significantly promote the recovery of damaged spinal cord function. Compared with anterior approach surgery, posterior approach surgery has the advantages of less trauma, convenient operation, and fewer complications, so it can be preferred.