Identification of clinical predictors for functional recovery in patients with thoracolumbar fractures and incomplete spinal cord injury: an internally validated prediction model

TL;DR

This study identifies factors that predict functional recovery in patients with thoracolumbar fractures and incomplete spinal cord injury.

Contribution

A new internally validated prediction model for 12-month functional recovery after thoracolumbar incomplete spinal cord injury is developed.

Findings

Eight independent predictors of functional recovery were identified, including admission AIS grade and decompression timing.

The model demonstrated strong discrimination (AUC 0.87) and satisfactory calibration in internal validation.

Predictive effects were stronger in younger patients and those with early surgery.

Abstract

Accurate prediction of functional recovery after thoracolumbar fracture with incomplete spinal cord injury (SCI) remains challenging. We aimed to identify independent predictors and develop a validated model for 12-month functional outcome.

In this single-center, retrospective cohort study (January 2018–December 2024), consecutive adults (≥18 years) with acute T11-L2 fractures and admission American Spinal Injury Association Impairment Scale (AIS) grade B, C, or D were enrolled. Functional recovery was defined as ≥1 AIS grade improvement plus ≥10-point Spinal Cord Independence Measure version III (SCIM-III) gain at 12 months. Candidate predictors (n = 23) were selected a priori based on literature review and expert consensus. Missing data (<8% per variable) were multiply imputed (m = 20). Multivariable logistic regression with Firth’s correction and backward elimination guided by Akaike Information Criterion was used. Model discrimination [optimism-corrected area under the curve (AUC) 0.87] and calibration (Hosmer-Lemeshow test) were assessed by 1,000 bootstrap resamples. Pre-specified subgroup analyses examined age, AIS grade, surgical timing and lesion length.

Among 1,032 eligible patients, 206 (20.0%) achieved functional recovery. Eight predictors were independently associated: admission AIS grade (OR 4.5 per grade, 95% CI 3.5–5.8), motor score (OR 1.05 per point, 1.03–1.07), intact posterior ligamentous complex (OR 2.3, 1.6–3.2), decompression ≤ 24 h (OR 1.9, 1.4–2.7), non-smoking (OR 1.7, 1.2–2.4), Charlson Comorbidity Index = 0 (OR 1.5, 1.1–2.1), shorter intramedullary T2 lesion length (OR 0.96 per mm, 0.95–0.97) and rehabilitation intensity ≥3 h/day (OR 1.4, 1.0–1.9). The final model demonstrated an optimism—corrected AUC of 0.87 (95% CI: 0.85–0.89) and calibration characteristics with a calibration slope of 1.02, an intercept of 0.01, and a Hosmer–Lemeshow p-value of 0.18 during internal validation. Predictive effects were stronger in younger, AIS B/C patients and when surgery was performed early.

A parsimonious eight-factor model showed robust discrimination and satisfactory calibration in internal validation for 12-month functional recovery after thoracolumbar incomplete SCI, enabling individualized prognostication. External validation in independent multicenter cohorts is required before clinical implementation and treatment decision-making.

Figures