Development and validation of a serological marker-based nomogram for risk stratification of spinal brucellosis

Spinal involvement is a major determinant of impaired quality of life in patients with brucellosis and may even be life-threatening. Efficient and cost-effective serological testing may enable earlier identification of spinal brucellosis.

To develop and validate a nomogram based on routine serological indicators for individualized prediction of spinal brucellosis risk.

We retrospectively collected clinical data from 427 patients with brucellosis admitted to the General Hospital of Ningxia Medical University between September 2021 and September 2025. Patients were classified into a non-spinal brucellosis group (n = 359) and a spinal brucellosis group (n = 68) according to the presence of spondylitis. Participants were randomly split into a training set and an internal validation set at a 7:3 ratio. After variable selection in the training set, independent predictors were identified using multivariable logistic regression and incorporated into a nomogram. Model performance was comprehensively assessed using the area under the receiver operating characteristic curve (AUC), calibration curves, the Brier score, and multiple classification metrics. Decision curve analysis (DCA) was used to evaluate clinical net benefit. SHAP was applied to interpret variable contributions, restricted cubic splines (RCS) were used to examine nonlinear relationships, and prespecified subgroup analyses were performed.

The final model included four pre-treatment serological markers, namely platelet count (PLT), platelet-to-lymphocyte ratio (PLR), interleukin-4 (IL-4), and Serum ferritin. The model showed moderate discrimination in the training and validation sets (AUC = 0.762, 95% CI: 0.692-0.831; and AUC = 0.664, 95% CI: 0.521-0.807, respectively), showing overall moderate calibration, with Brier scores of 0.121 and 0.125 in the training and validation cohorts, respectively. DCA indicated stable net benefit across reasonable threshold probabilities. SHAP analysis identified PLR as the strongest contributor to prediction. RCS analysis suggested linear associations for PLT, PLR, and serum ferritin, whereas IL-4 showed nonlinear relationships with risk. Subgroup analyses demonstrated generally consistent effect directions, but a significant interaction between IL-4 and diabetes status was observed.

This nomogram enables individualized risk assessment for spondylitis among patients with brucellosis and may serve as a practical tool for preliminary screening and MRI decision-making in clinical practice.