Key Points:
For patients with ESKD treated with hemodialysis, the causes of death reported by the United States Renal Data System and the National Death Index show substantial disagreement.
In particular, the proportion of sudden cardiac death was almost two-fold higher in the United States Renal Data System (42%) compared with the National Death Index (22%).
Background: Cause-specific mortality data from the United States Renal Data System (USRDS) form the basis for identifying cardiovascular disease (CVD), specifically sudden cardiac death (SCD), as the leading cause of death for patients on dialysis. Death certificate data from the National Death Index (NDI) is the epidemiological standard for assessing causes of death for the US population. The cause of death has not been compared between the USRDS and the NDI.
Methods: Among 39,507 adults starting dialysis in the United States, we identified 6436 patients who died between 2003 and 2009. We classified the cause of death as SCD, non-SCD CVD, cancer, infection, and others and compared the USRDS data with those from the NDI.
Results: The median age at the time of death was 70 years, 44% were female, and 30% were non-Hispanic Black individuals. The median time from dialysis initiation to death was 1.2 years. Most of the deaths occurred in hospital (N=4681, 73%). The overall concordance in cause of death between the two national registries was 42% (κ=0.23; 95% confidence interval, 0.22 to 0.24). CVD, including SCD and non-SCD CVD, accounted for 67% of deaths per the USRDS but only 52% per the NDI; this difference was mainly driven by the larger proportion of SCD in the USRDS (42%) versus the NDI (22%). Of the 2962 deaths reported as SCD by the USRDS, only 35% were also classified as SCD by the NDI. Out-of-hospital deaths were more likely to be classified as SCD in the USRDS (60%) versus the NDI (29%), compared with in-hospital deaths (41% in the USRDS; 25% in the NDI).
Conclusions: Significant discordance exists in the causes of death for patients on dialysis reported by the USRDS and the NDI. Our findings underscore the urgent need to integrate NDI data into the USRDS registry and enhance the accuracy of cause-of-death reporting.