Machine Learning Approach for Predicting Systemic Lupus Erythematosus in an Oman-Based Cohort

Area of Science:

• Medical informatics
• Machine learning in healthcare
• Rheumatology

Background:

• Systemic lupus erythematosus (SLE) diagnosis can be challenging, particularly in early stages.
• Accurate and timely diagnosis is crucial for effective patient management and improved outcomes.
• Omani patients represent a specific demographic for studying rheumatologic diseases.

Purpose of the Study:

• To develop and validate a machine learning (ML) framework for predicting SLE in Omani patients.
• To identify key clinical and demographic features predictive of SLE.
• To provide explainable AI (XAI) insights for clinical decision support.

Main Methods:

• Utilized electronic health records of 219 Omani patients (138 with SLE, 81 with other rheumatologic diseases) from 2006-2019.
• Employed recursive feature selection and the CatBoost classification algorithm for SLE prediction.
• Applied SHAP explainer for model interpretability and validated findings with rheumatologists.

Main Results:

• The CatBoost model achieved a high predictive performance with an AUC of 0.95 and 92% sensitivity.
• SHAP analysis identified alopecia, renal disorders, acute cutaneous lupus, hemolytic anemia, and patient age as significant predictors.
• The explainable framework provided validated reasoning for individual SLE predictions.

Conclusions:

• An effective, explainable ML framework for SLE prediction in Omani patients was successfully designed and validated.
• This tool can aid clinicians in early SLE detection, facilitating timely interventions.
• The framework has the potential to improve patient outcomes through early management of SLE.