Sara Lener, Raphael Gmeiner, Anto Abramovic, Anna Koller, Christoph Orban, Sebastian Hartmann, Alexandra Grob, Victor Staartjes, Ann Kathrin Joerger, Carolin Albrecht, Bernhard Meyer, Felipa Lange, Andreas Kramer, Florian Ringel, Christopher Marvin Jesse, Ralph T Schär, Christoph Bettag, Veit Rohde, Insa K Janssen, Karl Schaller, Julia Onken, Aminaa Sanchin, Peter Vajkoczy, Imran Shah, Rebecca Sutherland, Andreas K Demetriades, Claudius Thomé
This large multicenter registry study provides the first external validation of the MSI-20, confirming reproducibility and generalizability. The score demonstrated fair discrimination and a clear risk gradient across increasing score categories. Given its simplicity and high negative predictive value, the MSI-20 may serve as an adjunctive tool to support risk communication, multidisciplinary discussion, patient counseling, and preoperative optimization in spinal infection.
OBJECTIVE: Spinal infection (SI) is associated with substantial morbidity and mortality, and optimal treatment strategies remain debated, particularly in medically fragile patients. The Mortality in Spinal Infection (MSI-20) score was developed as the first dedicated prognostic tool for mortality risk estimation, but external validation in large cohorts has been lacking. This study aimed to externally validate the MSI-20 in a large multicenter registry cohort and assess its performance, clinical thresholds, and generalizability.
METHODS: This retrospective multi-institutional registry study included 1,122 adult patients with clinically, radiologically, and laboratory-confirmed SI treated at 6 tertiary referral centers between 2010 and 2023. The multicenter registry design enabled robust analysis of this relatively rare outcome. MSI-20 scores were calculated according to the original definition. Predictive performance was evaluated using receiver operating characteristic (ROC) analysis with 95% confidence intervals. Mortality across score strata and center-wise performance were analyzed, with exploratory assessment of additional baseline predictors.
RESULTS: Mean age was 66.4±13.1 years, and overall mortality was 18.3%. The MSI-20 demonstrated fair discriminative ability (area under the curve [AUC], 0.68; 95% confidence interval [CI], 64-72). The optimal ROC threshold was 3.5; a cutoff ≥4 yielded sensitivity 0.49 and specificity 0.74 (F1 score, 0.38). Mortality increased progressively, approaching 50% (47.4%) at scores ≥9. Performance was consistent across centers (AUC, 0.60-0.75). Exploratory univariable analyses did not identify additional baseline variables with consistent strong associations beyond the MSI-20 components.
CONCLUSION: This large multicenter registry study provides the first external validation of the MSI-20, confirming reproducibility and generalizability. The score demonstrated fair discrimination and a clear risk gradient across increasing score categories. Given its simplicity and high negative predictive value, the MSI-20 may serve as an adjunctive tool to support risk communication, multidisciplinary discussion, patient counseling, and preoperative optimization in spinal infection.