Purpose <p>Thoracic spinal tuberculosis (TSTB) can lead to vertebral destruction, spinal instability, and neurological deficits. Paraplegia is one of its most severe manifestations, yet a concise and clinically interpretable framework for assessing neurological severity at presentation remains lacking.</p> Methods <p>We retrospectively reviewed 148 patients with thoracic spinal tuberculosis who underwent surgery at a tertiary referral centre between 2019 and 2024. Clinical, laboratory, and imaging variables associated with paraplegia at presentation were evaluated using logistic regression analysis. A clinical-imaging severity stratification framework was subsequently developed using routinely available preoperative variables. Model discrimination and calibration were assessed using receiver operating characteristic analysis and bootstrap resampling, and LASSO regression was performed as a sensitivity analysis.</p> Results <p>Among 148 patients, 54 (36.5%) presented with paraplegia. In complete-case multivariable analysis (n = 113), lower body mass index (BMI) (OR 0.79, 95% CI 0.67–0.92; P = 0.004), longer duration of neurological impairment (OR 1.06 per week, 95% CI 1.01–1.11; P = 0.010), and higher compression grade (OR 2.47, 95% CI 1.50–4.07; P &lt; 0.001) were independently associated with paraplegia. A framework incorporating age, BMI, duration of neurological impairment, and compression grade demonstrated acceptable discrimination (AUC 0.903, 95% CI 0.85–0.96) and calibration within the development cohort. Internal validation showed generally consistent performance.</p> Conclusions <p>In surgically treated patients with thoracic spinal tuberculosis, prolonged neurological symptom duration, greater compression severity, and lower BMI were independently associated with paraplegia at presentation. The proposed framework, based on routinely available variables, may support structured assessment of neurological severity in patients with thoracic spinal tuberculosis.</p>