Hospitalisation for cirrhosis in Australia: disparities in presentation and outcomes for Indigenous Australians

Demographic and health service characteristics and clinical data were obtained from Queensland Hospital Admitted Patient Data Collection. Date and cause of death were obtained from the Queensland Death Registry.

In Australia, a person is considered Indigenous if they are of Aboriginal and/or Torres Strait Islander descent, identify as Aboriginal and/or Torres Strait Islander, and are accepted as an Aboriginal or Torres Strait Islander person in the community in which they live, or have lived [16]. Indigenous status for an individual may vary across records for the same individual as Indigenous status is defined by self-assessment [17]. Using a validated algorithm for identification of Indigenous status from hospital linked data, patients were coded as Indigenous if at least one of their records within the study period identified them as Indigenous [17].

Patients’ residential postcodes were used to determine patients’ index of relative socioeconomic disadvantage score [18] and remoteness of residence [19]. Code lists for identification of cases, aetiology and cofactors were reviewed by four hepatologists and the Principal Statistical Data Quality Officer, Statistical Services Branch, Queensland Health [7]. Comorbidity was measured using the Charlson Comorbidity Index (Charlson index) [20] using validated coding algorithms [21].

We selected the first hospitalisation record for cirrhosis of each patient (referred to here as index admission) to avoid duplicate counting of individuals. When patients were transferred within the same or to another hospital, we considered this as one hospital stay. Length of stay was calculated by adding all days the patient was an admitted patient during one hospital stay.

Cumulative overall survival estimates by Indigenous status were calculated using the Kaplan–Meier method (log-rank statistic). All cases were followed until date of death, or December 31, 2017, whichever came sooner. The survival time for patients who died on the index admission was counted as half a day. Multivariable Cox regression analysis reported in terms of hazard ratios (HRs) with associated 95% confidence intervals (CIs) was used to assess the differences by Indigenous status with respect to survival. Informed by our previous work [7], we included in the main effects model factors that could influence overall survival, such as patients’ sociodemographic features, disease aetiology, presence of comorbidity and complications of cirrhosis at index admission. We have excluded factors where prevalence of exposure in Indigenous patients was < 10%. In many patients with cryptogenic cirrhosis, non-alcoholic steatohepatitis (NASH) may be the primary cause of cirrhosis, as the diagnostic features of steatohepatitis may disappear as the liver disease progresses [22, 23]. We have therefore combined cryptogenic and non-alcoholic fatty liver disease (NAFLD)/NASH cirrhosis. We have also combined viral hepatitis B and C. When the overall model was statistically significant, a Least Absolute Shrinkage and Selection Operators (LASSO) penalised regression cox proportional hazards model was used to identify the set of variables that had the strongest association with the survival outcome. The LASSO procedure was used due to the high number of predictors (demographic, health service characteristics and clinical data) and potentially complex patterns of collinearity among predictor variables. If a set of variables are highly collinear, the LASSO only picks out one from the set, so it leads to a more parsimonious model. Variables included in the model were checked to ensure that they adhered to the assumption of proportional hazards over time (Shoenfeld residuals). The vce (robust) option was used to obtain robust standard errors for the parameter estimates to control for mild violation of underlying assumptions.

Indigenous patients were significantly younger (346 (44%) had < 50 years vs. 2063 [19%]), and 480 (67%) had no partner (vs. 4880 [48%]) when compared to non-Indigenous patients; Table 1). Three hundred ninety-five Indigenous patients (51%) lived in most disadvantaged areas, and 154 (20%) lived in remote/very remote areas compared 25 and 1% of 10,642 non-Indigenous patients, respectively.

While the most commonly recorded aetiology was alcohol for both groups, 547 (70%) Indigenous patients had alcohol-related cirrhosis compared to 5041 non-Indigenous patients (47%; Table 2). Notably, 48 (6%) of Indigenous patients had Hepatitis B virus (HBV) related cirrhosis (vs. 459 [4%]), and 23 Indigenous patients (3.0%) had NAFLD/NASH (vs. 560 [5%] and 172 cryptogenic (22% vs. 3199 [30%]) cirrhosis compared to their non-Indigenous counterparts.

Indigenous patients had significantly higher Charlson index scores than non-Indigenous patients (p = 0.04). Compared to non-Indigenous patients, Indigenous patients had higher prevalence of diabetes (192 [25%] vs. 2080 [20%]; p < 0.001), chronic renal disease (76 [10%] vs. 631 [6%]; p < 0.001) and congestive heart failure (63 [8%] vs. 523 [5%]; p < 0.001), and lower prevalence of cancer (95 [12%] vs. 1840 [17%]; p < 0.001). Indigenous and non-Indigenous patients were similar with regard to the other diseases listed in the Charlson index. As drug use/abuse is a cofactor for HBV and HCV-related cirrhosis, we have examined these. 29 (4%) Indigenous patients had drug use/abuse listed as co-factors compared to 243 (2%) non-Indigenous patients (p = 0.01).

Length of hospital stay varied significantly by Indigenous status, with fewer Indigenous patients having one-day admissions (108 [14%] vs. 3056 [29%]; p < 0.001). The median length of hospital stay was 7 days (interquartile range (IQR) 3–13) compared to 5 days (IQR 1–11) for their non-Indigenous counterparts. As an inpatient, Indigenous patients frequently accessed multidisciplinary health services (dietician, physiotherapist, social worker, pharmacist and occupational therapist). A higher proportion of Indigenous patients utilised the service of a dietitian (292 [38%] vs. 3500 [33%]; p = 0.01) and social worker (208 [27%] vs. 2425 [23%]; p = 0.01), and a higher proportion of non-Indigenous were assessed by a pharmacist (108 Indigenous [14%] vs. 1,776 non-Indigenous [17%]; p = 0.04) and occupational therapist (104 [13%] vs. 1759 [17%]; p = 0.02). Only 4% of patients utilised hospital-based palliative care services, with no difference by Indigenous status (p = 0.24).

Once discharged, Indigenous patients with cirrhosis were more likely to be readmitted than non-Indigenous patients, whether for liver disease or other reasons. Indigenous Australians had 3 times the rate of non-cirrhosis-related admissions compared to non-Indigenous patients (IRR = 3.04 95%CI 2.98–3.10) and a higher rate of cirrhosis-related admissions (IRR = 1.35 95%CI 1.29–1.41). Overall, the rate of readmissions was 86.3 admissions/10,000 PYAR, 21.1/10,000 PYAR for cirrhosis-related readmissions and 65.2/10,000 PYAR for admissions not related to cirrhosis (Table 3).

At the end of the follow-up period 61% of Indigenous patients had died compared to 57% of non-Indigenous patients (p = 0.03). A higher proportion of Indigenous patients died in hospital, rather than not (e.g. died at home, at a hospice), compared to non-Indigenous patients (74% vs. 70%, p = 0.05).

The median time from index admission to death was 1.68 years (IQR 0.32–4.25) overall, 1.47 years (IQR 0.24–3.81) for Indigenous patients with cirrhosis and 1.70 years (IQR 0.33–4.29) for non-Indigenous patients. Across 2- and 5-year survival estimates, Indigenous patients have a significantly lower survival compared to non-Indigenous patients (Fig. 1). The probability of overall survival was 17% (95%CI 12.5–22.6) and 27% (95%CI 25.6–28.2), respectively (p = 0.002). These disparities were reflected in the unadjusted hazard rate, which was 16% higher for Indigenous patients compared to their non-Indigenous counterparts (HR = 1.16 95%CI 1.06–1.27).

In univariable analysis of overall survival, all factors examined were significantly associated with survival (Table 4). Age, variation in referral source (emergency presentations), and alcohol-related cirrhosis had the strongest influence in the adjusted difference in survival between the two groups. In multivariable analysis, adjusting for the younger age of Indigenous patients increased the HR to 1.30 (95%CI 1.18–1.42). Adding referral source (adj-HR = 1.03 95%CI 0.93–1.13), and alcohol-related cirrhosis (adj-HR = 1.08 95%CI 0.99–1.19), one at a time, explained most of the survival deficit. Addition of other variables (demographic and clinical factors, and health service use), one at a time, led to slight changes in the hazard ratio for Indigenous status (data not shown). The final (full model) adjusted hazard ratio was 1.08 (95%CI 0.96–1.20). There were 13 independent factors significantly associated with survival. The highest adjusted HRs were for having cancer (adj-HR = 2.59 95%CI 2.26–2.97), followed by age ≥ 50 years (adj-HR = 1.57 95%CI 1.45–1.69), emergency presentation (adj-HR = 1.56 95%CI 1.46–1.67), ascites (adj-HR = 1.47 95%CI 1.37–1.57), and Charlson index> 0 (adj-HR = 1.46 95%CI 1.36–1.56).

As Indigenous Australians are overrepresented in rural/remote and socio-disadvantaged areas [24], we examined if there was interaction between Indigenous status and place of residence in relation to survival. There was no evidence of interaction between Indigenous status and rurality of residence (p = 0.17) or between Indigenous status and residence in the ‘most disadvantaged’ area (p = 0.27).