Factors Associated with Recording of Rheumatoid Arthritis on Death Certificates

Background/Purpose: Death certificates can be used to study mortality due to a particular disease. However, rheumatoid arthritis (RA) often remains unreported in death certificates. We sought to determine to what extent RA is underreported and what demographic and clinical characteristics could predict mention of RA in the death certificate.

Methods: Between 1996 and 2009, we recruited 1,328 patients with RA that met the American College of Rheumatology criteria. Patients were followed prospectively. A rheumatologist assessed clinical characteristics of RA at each evaluation, including number of tender, swollen and deformed joints, presence of rheumatoid nodules, as well as Steinbroker classification and Charlson comorbidity index. Joint damage was determined by Sharp score, using hand radiographs taken from the most recent visit prior to death. Deaths were identified through family members, friends, neighbors, other physicians, obituaries and public death databases. We obtained state-issued death certificates and mapped causes of death to ICD9 codes. Standard bivariate analyses were conducted comparing patients with and without RA on the death certificate. A multivariable logistic regression model was performed to determine what variables were associated with recording RA.

Results: By December 2013, 323 deaths had occurred during 8,326 person-years of observation, for a mortality rate of 3.8 per 100 person-years [95% confidence interval (CI) (3.4, 4.3)]. Of the 308 death certificates we received, 61 (19.8%) mentioned RA on the death certificate. Only two of them recorded RA as the immediate cause of death. Bivariate analysis revealed that a greater number of deformities (mean ±SD = 17.6±9.0 vs. 14.4±9.4; P=0.016), higher Sharp score (mean ±SD = 192±132 vs. 138±112; P=0.010) and lower socioeconomic status (mean ±SD = 42.4±21 vs. 48.3±19; P=0.041) were each associated with recording RA on the death certificate. Place of death, presence of rheumatoid nodules, having health insurance or an autopsy were not associated with recording RA. Multivariable analyses revealed that an increased number of deformed joints [odds ratio (95% confidence interval) = 1.04 (1.00, 1.07); P=0.04], less comorbidity [OR (95%CI) = 0.87 (0.76, 0.99); P=0.048] and having a certified physician sign the certificate [OR (95%CI) = 3.56 (1.16, 10.8); P=0.026] were associated with listing RA on the death certificate.

Conclusion: In this RA cohort, a diagnosis of RA was not listed in the death certificate in 80% of patients who died. Patients with fewer comorbidities and more joint deformities were more likely to have RA reported. Studies that rely on death certificates may underestimate the mortality of RA and be biased toward patients with more severe RA and less comorbidities.

Figure 1. Percent death certificates reporting RA according to the number of deformed joints.