Abstract Objectives Predicting serious infections (SI) in patients with rheumatoid arthritis (RA) is crucial for the implementation of appropriate preventive measures. Here we aimed to identify risk factors for SI and to validate the RA Observation of Biologic Therapy (RABBIT) risk score in real-life settings. Methods A multi-centre, prospective, RA cohort study in Greece. Demographics, disease characteristics, treatments and comorbidities were documented at first evaluation and one year later. The incidence of SI was recorded and compared with the expected SI rate using the RABBIT risk score. Results A total of 1557 RA patients were included. During follow-up, 38 SI were recorded [incidence rate ratio (IRR): 2.3/100 patient-years]. Patients who developed SI had longer disease duration, higher HAQ at first evaluation and were more likely to have a history of previous SI, chronic lung disease, cardiovascular disease and chronic kidney disease. By multivariate analysis, longer disease duration (IRR: 1.05; 95% CI: 1.005, 1.1), history of previous SI (IRR: 4.15; 95% CI: 1.7, 10.1), diabetes (IRR: 2.55; 95% CI: 1.06, 6.14), chronic lung disease (IRR: 3.14; 95% CI: 1.35, 7.27) and daily prednisolone dose ≥10 mg (IRR: 4.77; 95% CI: 1.47, 15.5) were independent risk factors for SI. Using the RABBIT risk score in 1359 patients, the expected SI incidence rate was 1.71/100 patient-years, not different from the observed (1.91/100 patient-years; P = 0.97). Conclusion In this large real-life, prospective study of RA patients, the incidence of SI was 2.3/100 patient-years. Longer disease duration, history of previous SI, comorbidities and high glucocorticoid dose were independently associated with SI. The RABBIT score accurately predicted SI in our cohort.