Purpose. I determined the standardized mortality ratio (SMR) of rheumatoid arthritis (RA), compared rates with non-inflamatory rheumatic disease (NIRD) patients and accounted for participation biases within patient-report observational data.

Methods. Three data sets were used: 1-CLINIC) 5,896 consecutive patients including 2,078 with RA (22,817 patient-years) seen in a rheumatology clinic in Wichita, KS between 1973 to 2003; 2-REG) 31,766 rheumatology patients, 78% having RA, who completed a brief observational study registration form at their rheumatologist's office between 1998 and 2003; 3-ACTIVE) 22,777-patient subset of REG who enrolled with multiple 6-month questionnaire assessments between 1998 and 2003. All-cause mortality was determined by the US National Death Index (NDI). Rates were compared by calendar-year, age and gender. Cox proportional hazards model including demographic covariates was used to determine the hazard ratio (HR) for having RA versus NIRD. Heckman selection model was used to account for bias of enrollment in ACTIVE.

Results. The SMR (95% CI) for the CLINIC RA group was 2.091 (1.963, 2.228) and the RA HR was 2.02 (1.83, 2.23). The ACTIVE RA group had SMR 1.256 (1.191, 1.325) and HR 1.94 (1.68, 2.23). The non-enrolled/enrolled ratio of SMRs in the REG RA group was 1.324 (1.189, 1.474), and the Heckman model predicted the effect of enrollment to reduce mortality by 16.70%. The adjusted ACTIVE RA group SMR was 1.993 (1.659, 2.388).

Conclusion. Though often seen only as a disabling chronic disease, having RA doubles the risk of death. Observational studies based on patient participation can be used to determine generalizable mortality rates if baseline data is also captured for non-participants or if compared with a NIRD control group.