Discharge diagnoses following hospital admissions in Denmark are recorded in the Danish National Patient Registry (NPR). Such routine hospitalization records may serve as useful research tools in epidemiological studies. The aim of the study was to provide measures of the validity and completeness of rheumatoid arthritis (RA) diagnoses recorded in the NPR. We identified medical records for 217 patients recorded as having RA in the NPR between 1977 and 2001. Using two definitions of RA (clinically confirmed RA and fulfilment of the American College of Rheumatology (ACR) 1987 diagnostic criteria for RA), a rheumatologist assessed the proportion of RA diagnoses recorded in the NPR that could be confirmed by scrutiny of the original medical records. The completeness of RA diagnoses in the NPR was estimated by a two-sample capture-recapture method. Overall, 59% of the 217 RA diagnoses in the NPR were confirmed by information in the medical records. However, major differences were seen according to characteristics of the underlying hospital registrations. Generally, RA diagnoses were most often confirmed for patients registered as inpatients and for patients with more than one hospital registration with RA. Specifically, only 42% of patients with one RA registration from a rheumatology department were confirmed as having RA. In contrast, 91% of patients treated at a rheumatology department and having three or more hospital registrations with RA were confirmed as having RA. The completeness of the NPR with respect to RA satisfying the ACR 1987 classification criteria was estimated to 26%. Our conclusion is that with careful attention to the limitations in the data, discharge diagnoses for patients with records of RA in the Danish NPR can be used for epidemiological research purposes; however, our findings prompt general carefulness when using non-audited registries for research in RA.