K DANSIE1,2, M SYPEK1, S MCDONALD1,2,3
1ANZDATA, Adelaide, South Australia; 2University of Adelaide, Adelaide, South Australia; 3Central Northern Adelaide Renal and Transplantation Service, South Australia
Aim: To compare the date and cause of death (CoD) recorded by ANZDATA with the National Death Index (NDI) and explore the differences when disagreement exists.
Background: ANZDATA uses a bespoke coding system for CoD. For researchers utilising ANZDATA it is important to understand the differences in CoD coding to the NDI and how this may affect comparisons with other reports.
Methods: Data linkage was performed between ANZDATA and NDI for all deaths in the period 1980-2013. CoD were classified at ICD chapter level. Overall and chapter specific agreement was assessed using the Kappa statistic.
Results: 28,583 patients were included in the cohort. 95.3% of ANZDATA reported deaths fell within +/- 3 days of the date of death reported by the NDI. Circulatory death was the most common CoD in both databases (ANZDATA 48.2%, NDI 32.3%). Overall agreement at ICD chapter level of primary CoD was poor (35.9%, Kappa 0.22). Agreement was best for malignancy (kappa 0.71) followed by digestive (kappa 0.34) then circulatory (kappa 0.31) causes.
When there was disagreement on primary CoD these were most commonly coded as genitourinary (34.9%) and endocrine (25.0%) in NDI, and circulatory (38.9%) and withdrawal (24.0%) in ANZDATA. Where there was a ‘withdrawal from dialysis’ code listed as the CoD in ANZDATA, NDI CoD was most commonly coded as genitourinary (37%) or endocrine (21.1%).
Conclusions: There is poor agreement in primary CoD between ANZDATA and NDI, which is in part explained by the absence of diabetes and renal failure as CoD in ANZDATA and the absence of ‘withdrawal from dialysis’ in NDI. These differences should be appreciated when comparing reports that utilise these two data sources.