Rheumatoid arthritis (RA) is a chronic autoimmune disorder characterized by persistent synovial inflammation, potentially leading to joint destruction, systemic organ involvement, and unfavorable long‑term prognosis. Accurate prognostic stratification is essential for personalized treatment and improved patient outcomes. However, existing prognostic tools for RA have notable limitations: traditional disease activity scores such as DAS28 focus on short‑term symptom evaluation rather than long‑term prognostic prediction, whereas current risk models rarely integrate systemic inflammatory markers and comorbidity‑related indicators, resulting in suboptimal predictive performance. This study aimed to develop and validate a comprehensive prognostic risk scoring model for RA patients incorporating clinical, laboratory, and comorbidity variables to enhance long‑term prognostic precision and clinical applicability.

We retrospectively enrolled 568 eligible patients with RA consecutively treated at multiple centers between 2019 and 2024. The cohort was randomly assigned to a training set (70%, n = 398) and an internal validation set (30%, n = 170). The primary composite endpoint included all‑cause mortality, severe disease recurrence (persistent DAS28 > 5.1), or major complications (cardiovascular events, progression of RA‑associated interstitial lung disease) within 24 months of follow‑up. Prognostic factors were screened using LASSO regression, and independent predictors were identified by multivariate Cox proportional hazards regression. A prognostic risk scoring model was constructed based on regression coefficients, incorporating the systemic inflammatory response index (SIRI), anti‑citrullinated protein antibody (ACPA) titer, C‑reactive protein (CRP), DAS28, history of heart failure, and forced vital capacity percentage (FVC%). Model performance was assessed by discrimination (C‑index, AUC), calibration plots, and decision curve analysis (DCA) in the training and internal validation sets. External validation was performed in an independent cohort of 156 RA patients from a separate center to evaluate generalizability.

Six independent prognostic factors for the composite endpoint were identified: SIRI, ACPA titer, CRP, DAS28, history of heart failure, and FVC% (all P < 0.05). The prognostic model exhibited favorable discriminative capacity, with a C‑index of 0.812 (95% CI: 0.765–0.859) and AUC of 0.825 in the training set, a C‑index of 0.796 (95% CI: 0.731–0.861) and AUC of 0.803 in the internal validation set, and a C‑index of 0.788 (95% CI: 0.719–0.857) and AUC of 0.795 in the external validation set. Calibration curves showed excellent consistency between predicted and observed risks in all cohorts. DCA confirmed that the model yielded favorable clinical net benefit across a wide range of threshold probabilities, superior to single predictors and traditional disease activity scores.

We established and validated a novel multidimensional prognostic risk scoring model for RA integrating systemic inflammation, autoantibodies, disease activity, comorbidities, and organ function. The model demonstrated good discrimination, calibration, and external generalizability, enabling effective identification of high‑risk patients. This tool provides a practical and reliable approach for prognostic evaluation and individualized management in clinical practice.