Liver-related mortality among people with hepatitis B and C: Evaluation of definitions based on linked healthcare administrative datasets.

Abstract

Routinely collected and linked healthcare administrative datasets could be used to monitor mortality among people with hepatitis B (HBV) and C (HCV). This study aimed to evaluate the concordance in records of liver-related mortality among people with an HBV or HCV notification, between data on hospitalization for end-stage liver disease (ESLD) and death certificates. In New South Wales, Australia, HBV and HCV notifications (1993-2017) were linked to hospital admissions (2001-2018), all-cause mortality (1993-2018) and cause-specific mortality (1993-2016) datasets. Hospitalization for ESLD was defined as a first-time hospital admission due to decompensated cirrhosis (DC) or hepatocellular carcinoma (HCC). Consistency of liver death definition of mortality following hospitalization for ESLD was compared with two death certificate-based definitions of liver deaths coded among primary and secondary cause-specific mortality data, including ESLD-related (deaths due to DC and HCC) and all-liver deaths (ESLD-related and other liver-related causes). Of 63,292 and 107,430 individuals with an HBV and HCV notification, there were 4478 (2.6%) post-ESLD hospitalization deaths, 5572 (3.3%) death certificate liver disease deaths and 2910 (1.7%) death certificate ESLD deaths. Between 2001 and 2016, among HBV post-ESLD hospitalization deaths (n= 891), 63% (562) had death certificate ESLD recorded, and 83% (741) had death certificate liver disease recorded. Between 2001 and 2016, among HCV post-ESLD hospitalization deaths (n= 3587), 58% (2082) had death certificate ESLD recorded, and 87% (3135) had death certificate liver disease recorded. At least one-third of death certificates with DC and HCC as cause of death had no mention of HBV, HCV or viral hepatitis. Our study identified limitations in estimating and tracking HBV and HCV liver disease mortality using death certificate-based data only. The optimum data for this purpose is either ESLD hospitalisations with vital status information or a combination of these with cause-specific death certificate data.