To investigate the long-term association of type 2 diabetes mellitus (T2DM) with severe infection and hospitalization in patients with rheumatoid arthritis (RA), and to evaluate the incremental predictive value of T2DM for severe infection risk beyond conventional clinical factors, with the aim of improving risk stratification and integrated management in RA.

This was a single-center retrospective real-world cohort study based on nearly 20 years of electronic medical record data from Xiangya Second Hospital. Outpatient and inpatient records, laboratory results, medication orders, and imaging reports were integrated. Adult patients aged 18 years or older with RA were included, with RA defined by diagnostic codes combined with rheumatology specialist clinical diagnosis. The index date was defined as the first date on which inclusion criteria were met and key covariates were complete. Patients were classified into T2DM and non-T2DM groups according to baseline diabetes status. The primary outcome was the first severe infection during follow-up, defined as an infection requiring emergency department care or hospitalization. Secondary outcomes included first all-cause hospitalization, infection-related hospitalization, and readmission. Demographic characteristics, RA-related features, including rheumatoid factor, anti-cyclic citrullinated peptide antibody, erythrocyte sedimentation rate, and C-reactive protein, treatment exposures, including glucocorticoids, conventional disease-modifying antirheumatic drugs, biologics, and Janus kinase inhibitors, as well as major comorbidities, were collected. Kaplan-Meier analysis was used to estimate cumulative incidence, and Cox proportional hazards models were applied to assess the associations of T2DM with severe infection and hospitalization. Inverse probability of treatment weighting based on propensity scores was used to control for confounding. Prediction models for severe infection were further developed, and model discrimination and calibration were compared before and after adding T2DM to assess its incremental predictive value. Sensitivity analyses were conducted using alternative definitions of T2DM and across different treatment eras.

A total of 9,842 patients with RA were included, of whom 2,137 (21.7%) had baseline T2DM. The median follow-up duration was 5.3 years. Compared with patients without T2DM, those with T2DM were older and had a higher comorbidity burden (both P<0.001). During follow-up, the incidence rate of severe infection was higher in the T2DM group than in the non-T2DM group (4.9 vs 3.7 per 100 person-years). Hospitalization rates were also higher in the T2DM group (15.6 vs 11.2 per 100 person-years, P<0.001). Kaplan-Meier analysis showed significantly higher cumulative risks of severe infection and hospitalization among patients with T2DM (log-rank P<0.001). In multivariable Cox models, T2DM was independently associated with an increased risk of severe infection (hazard ratio [HR] 1.24, 95% confidence interval [CI] 1.10-1.39) and hospitalization (HR 1.41, 95% CI 1.28-1.55). These findings remained consistent after inverse probability weighting. In prediction models for severe infection, the addition of T2DM improved model discrimination from 0.71 to 0.73, with good calibration. Sensitivity analyses yielded consistent results.

In this long-term real-world cohort, T2DM was an independent predictor of severe infection and hospitalization in patients with RA. Beyond traditional risk factors, T2DM provided additional predictive value for severe infection risk and may help identify high-risk patients in clinical practice. Enhanced risk stratification and integrated management should be considered for patients with RA and coexisting T2DM to improve outcomes.