Clinical indicators associated with pericardial effusion in rheumatoid arthritis: a machine learning-based analysis

Background Pericardial effusion (PE) is a frequent yet underdiagnosed complication of rheumatoid arthritis (RA), with substantial mortality risk. Nevertheless, early detection remains challenging due to nonspecific presentations and the limited feasibility of routine echocardiography. We aimed to identify simple clinical and laboratory features associated with the presence of PE using advanced machine learning (ML) approaches. Methods In this cross-sectional study, 958 RA patients were evaluated, all of whom underwent transthoracic echocardiography for PE detection. A comprehensive set of demographic, clinical, and laboratory variables was analyzed. Eight ML algorithms were developed and evaluated using 45 repeated train-test splits to enhance robustness. Model performance was assessed using area under the precision-recall curve (AUC-PR) and receiver operating characteristic curve (AUC-ROC). Feature importance was determined using SHAP analysis. Results PE was present in 126 patients (13.2%). The random forest (RF) model achieved the best performance (mean AUC-PR, 0.674 [95% CI 0.504–0.824]; mean AUC-ROC, 0.943 [95% CI 0.903–0.968]). The most influential features were palpitation, chest pain, RA disease duration, and age. Patients with PE were older (62.8 ± 12.15 vs. 57.7 ± 12.65 years, p ≤ 0.001), and several clinical features differed significantly between groups (p < 0.05). SHAP analysis demonstrated a consistent positive association of palpitation, chest pain, longer disease duration, and older age with PE detection. Conclusion These findings support the use of readily available clinical features by ML to increase diagnostic suspicion for PE in patients with RA. RA patients presenting with palpitations or chest pain, particularly those of older age or with long-standing disease, may warrant closer consideration for echocardiographic evaluation. Key Points • Pericardial effusion is a common yet often underrecognized and potentially life-threatening complication in patients with rheumatoid arthritis. • This study is the first to apply machine learning techniques to identify simple clinical and laboratory indicators associated with pericardial effusion in rheumatoid arthritis. • Palpitation, chest pain, older age, and longer disease duration were identified as the clinical features most strongly associated with the presence of pericardial effusion. • These findings may help clinicians identify patients in whom suspicion for pericardial effusion should be heightened and echocardiographic evaluation more strongly considered.