AUDITING DATA QUALITY OF ANZDATA CAUSE OF DEATH REPORTING IN DIALYSIS PATIENTS
ALIA N TUN ISMAIL1, NEHAL CHAVAN1, ERIC H K AU2,3,4, SARAH SO1,2 1Department of Renal Medicine, Nepean Hospital, Sydney, NSW, Australia2Faculty of Medicine and Health, The University of Sydney, Sydney, NSW, Australia3Department of Renal Medicine, The Alfred Hospital, Melbourne, VIC, Australia4Central Clinical School, Monash University, Melbourne, VIC, Australia
Abstract
Aim: We aimed to assess agreement between cause of death (COD) records for dialysis patients as reported to the Australian and New Zealand Dialysis and Transplant Registry (ANZDATA), and electronic medical records (EMR) from a single renal service.
Background: ANZDATA contains information on patients receiving dialysis or transplant in Australia and New Zealand, and is a valuable source of data for epidemiological research. Few studies exist which audit the accuracy of ANZDATA records in comparison with EMR.
Methods: We conducted a retrospective cohort study. This included adult dialysis patients under the Western Renal Services unit who died between 1 January 2016 and 31 December 2021. Two independent medical assessors manually evaluated COD by examining EMR and coding these according to ANZDATA’s COD. COD were then recoded into categories according to ANZDATA (cardiovascular, withdrawal, infection, cancer, other). Kappa statistics assessed agreement between the determined COD and ANZDATA records.
Results: Of 709 deaths, 440 (63.3%) were male and 260 (36.7%) were female. 429 (60.5%) died between 2019-2021 and 280 (39.5%) between 2016-2018. 539 patients had records available. Agreement between assessors and ANZDATA for individual COD was poor (kappa 0.39). Agreement improved when comparing COD categories (kappa 0.53). There was no significant difference in agreement between males and females for COD categories (kappa 0.55 vs 0.50 respectively), or between patients who died from 2016-2018 vs 2019-2021 (kappa 0.51 vs 0.55 respectively).
Conclusion: There is poor agreement between individual COD as assessed by EMR and ANZDATA. This improved to moderate agreement when analysing COD categories, suggesting that broader categories are more accurate when utilising death data from ANZDATA. A larger audit is necessary to validate our findings.