Background: Organ damage is a critical determinant of poor prognosis and increased mortality in systemic lupus erythematosus (SLE).

Objectives: To develop a de-novo machine learning-based model to predict early organ damage in SLE patients.

Methods: A cohort of 912 SLE patients was analyzed, using classification criteria, non-criteria features and BILAG index items to predict organ damage within five years since SLE diagnosis. Organ damage was assessed by the SLICC/ACR Damage Index (SDI). Patients were randomly divided into a training (70%) and test (30%) set. The Least Absolute Shrinkage and Selection Operator (LASSO)-logistic regression method with optimal hyper-parameter λ selected via 10-fold cross-validation was used to construct the predictive models. The sensitivity, specificity, accuracy and Area Under the Receiver-Operating-Characteristic Curve (AUC) were evaluated in the test set. The optimal cut-off values were determined using the Youden index. Calibration curve of the model with the best performance was generated using the CalibrationCurves R package.

Results: The model incorporating all items from the classification criteria, BILAG items, and excluding ANA exhibited the highest predictive accuracy for early organ damage, with an accuracy of 0.773, sensitivity of 0.535, and specificity of 0.856. Significant predictors identified in the model included BILAG items related to neurologic, cardiovascular involvement and the presence of antiphospholipid antibodies. Conversely, the presence of SLICC 2012-defined arthritis, the skin involvement and low complement levels were associated with reduced risk of early organ damage. The model demonstrated adequate discrimination ability, with a C-statistic of 0.76.

Conclusion: Machine learning models incorporating both classification criteria and non-criteria features show promise in predicting early organ damage in SLE. Further validation in larger, external cohorts is necessary to improve risk prediction and early intervention in SLE patients.

REFERENCES: NIL.

Acknowledgements: NIL.

Disclosure of Interests: None declared.

© The Authors 2025. This abstract is an open access article published in Annals of Rheumatic Diseases under the CC BY-NC-ND license ( http://creativecommons.org/licenses/by-nc-nd/4.0/ ). Neither EULAR nor the publisher make any representation as to the accuracy of the content. The authors are solely responsible for the content in their abstract including accuracy of the facts, statements, results, conclusion, citing resources etc.