Résumé
Background
The spleen is frequently injured in blunt abdominal trauma, and proper management depends on accurate CT assessment using the AAST-OIS scoring system. However, the reproducibility of CT scoring in trauma patients remains uncertain.
Purpose
To assess the reproducibility of AAST-OIS scoring for high-grade splenic injuries (grades 3, 4, and 5) without vascular anomalies on initial CT scans and to evaluate the clinical impact of misclassification, particularly in relation to hemorrhagic complications.
Methods
This study is an ancillary analysis of the SPLASH-randomized clinical trial, which involved 6 Level 1 trauma centers in France. Initial CT scans were scored by the trauma center’s radiology team and then independently reviewed by two expert radiologists. Interobserver agreement as assessed between the trauma center radiologists and the expert reviewers, as well as between the two independent experts. The study also examined whether underestimation of injury severity influenced early clinical outcomes.
Results
The agreement between trauma center radiologists and expert reviewers was moderate, with kappa coefficients of 0.493 using linear weighting and 0.511 using quadratic weighting, whereas inter-expert agreement was excellent (kappa: 0.905 linear, 0.917 quadratic). The primary source of discordance was the underestimation of parenchymal devascularization, which frequently led to misclassification of grade 4 injuries as grade 3. hile overall complications on Day 5 did not significantly differ between correctly classified and misclassified patients (
p
= 0.0593), patients whose injuries were underestimated (false negatives) had a threefold higher risk of hemorrhagic complications compared to the rest of the cohort (
p
= 0.0351).
Conclusion
This study highlights the moderate reproducibility of AAST-OIS scoring in emergency trauma settings, emphasizing the need for standardized evaluation protocols to improve diagnostic consistency. Given the higher risk of hemorrhagic complications in underestimated cases, increased awareness of parenchymal devascularization and improved radiological training are essential to optimizing patient outcomes. The integration of AI-based decision-support tools and structured reporting templates may enhance real-time injury classification and reduce observer variability.