Background and purpose Spondylodiscitis may be a serious disease due to diagnostic delay and inadequate treatment. There is no consensus on when and how to operate. We therefore retrospectively analyzed the outcome of a large series of patients treated either nonoperatively or surgically.Patients and methods Between 1992 and 2000, 163 patients (101 males) were hospitalized due to spondylodiscitis. The mean age was 56 (1-83) years. The infection was located in the cervical spine in 13 patients (8%), in the thoracic spine in 62 patients (38%), at the thoracolumbar junction in 10 patients (6%), and in the lumbo-sacral spine in 78 patients (48%). In 67 patients (41%), no microorganisms were detected. Most of the other patients had Staphylococcus aureus infection (53 patients) and/or Mycobacterium tuberculosis (22 patients). The patients were divided into 3 groups: (A) 70 patients who had nonoperative treatment, (B) 56 patients who underwent posterior decompression alone, and (C) 37 patients who underwent decompression and stabilization.Results At 12-month follow-up, nonoperative treatment (A) had failed in 8/70 patients, who had subsequently been operated. 24/56 and 6/37 had been reoperated in groups B and C, respectively. Group A patients had no neurological symptoms. In group B, 11 had neurological deficits and surgery was beneficial for 5 of them; 4 remained unchanged and 2 deteriorated (1 due to cerebral abscess). 11 patients in group C had altered neurogical deficits, which improved in 9 of them. 20 patients had died during the 1-year follow-up, 3 in hospital, directly related with infection.