Background
End-stage kidney disease (ESKD) is a significant and growing public health problem, associated with high morbidity, mortality and diminished quality of life. ESKD also generates disproportionately high costs to the health care system as patients require renal replacement therapy (RRT), by dialysis or kidney transplantation, for long-term survival [
1]. The Northern Territory (NT) is a federal territory of Australia occupying much of the central and northern part of the country and has a much greater proportion of Aboriginal Australians (30%) than the balance of Australia (3%). The NT has the highest incidence and prevalence of RRT for ESKD among all states and territories in Australia [
2]. The incidence of RRT among Aboriginal Australians is substantially higher than non-Aboriginal Australians, and among Aboriginal Australians the incidence of RRT is highest in the NT [
2].
A limitation for international reporting of the prevalence and incidence of ESKD is that estimates are generally restricted to individuals who receive RRT, as recorded and reported by national or regional data registries [
3]. In Australia, the Australia and New Zealand Dialysis and Transplant Registry (ANZDATA) reports individuals who receive RRT [
2]. This approach will underestimate the true burden of ESKD in the community [
1], missing individuals with diagnosed ESKD but not accessing RRT as well as an uncertain number of undiagnosed cases. One previous Australian study estimated total ESKD incidence by linking ANZDATA data with national mortality data and reported an age-standardised incidence rate, for 2003–2007, for the total NT population of 0.72 per 1000, much higher than the corresponding national rate (0.2 per 1000) [
4]. A separate study, also linked ANZDATA data and national mortality data and reported that for 2009–2013, the age-standardised incidence rate for ESKD was almost five times higher in the Australian Aboriginal population than the non-Aboriginal Australian population (0.95 vs 0.19 per 1000 population) [
1]. To our knowledge, there have been no studies that have estimated total ESKD prevalence in Australia.
The capture-recapture method was originally used in ecological research to quantify populations [
5]. This method is now used in epidemiology to estimate ascertainment level by linking unit level data across multiple data sources [
6,
7]. The capture–recapture method uses the degree of overlap of individuals between sources to estimate the number of undiagnosed cases not registered in any of the sources and hence provides a more complete estimate of the total number of cases in a population [
8,
9]. This approach is particularly relevant for the Aboriginal population for whom anecdotal reports suggest a significant proportion may not be accessing services and as a result have conditions which remain undiagnosed.
The aim of this study was to estimate the total prevalence and incidence of ESKD in NT Aboriginal and non-Aboriginal populations, to examine the demographic characteristics of patients with ESKD, and the proportion of patients with ESKD who did not receive RRT.
Discussion
This study applied capture–recapture methods to unit level data linked across multiple data sources to estimate the number of ESKD cases in the NT Aboriginal and non-Aboriginal populations. This novel approach incorporates patients who are known to have either treated or untreated ESKD as well as an estimate for undiagnosed cases. Our study found age-adjusted prevalence of 18.0 and 1.1 per 1000 for the NT Aboriginal and non-Aboriginal population respectively, while the corresponding age-adjusted incidence was 5.26 and 0.36 per 1000 population. Most previous research has used participation rates for RRT as an indirect measure of the incidence of ESKD, however, not all people with ESKD will commence RRT [
2,
4,
15,
22]. Our ESKD incidence estimates for total NT population of 1.02 per 1000 population however are also higher than the previous Australian study which linked ANZDATA data with national mortality data and estimated a total NT incidence of 0.72 per 1000 population [
4]. The difference will be a combination of increased incidence since 2003–2007 and the addition, in our study, of an estimate for undiagnosed cases. Our study also confirms that the incidence of diagnosed ESKD for the NT Aboriginal population was much greater than national Aboriginal age-standardised estimates for the period 2009–2013 (3.96 vs 0.95 per 1000 population) [
1]. Consistent with the previous report, our study reports earlier onset of ESKD for the Aboriginal population with a substantial excess among those aged 35–64 years, as well as a higher rate among females than males [
2].
Estimated undiagnosed ESKD prevalent and incident Aboriginal cases were a much larger proportion (81% and 82% respectively) of all NT cases than the corresponding proportion of the total NT population (30%) [
20], which is consistent with the high prevalence and incidence in the Aboriginal population [
23‐
26]. The effect of different eGFR cut-off thresholds for the definition of ESKD on the number of undiagnosed Aboriginal cases was substantial. Using the threshold of an eGFR < 15 maintained consistency of a definition of ESKD across administrative and clinical datasets and followed international guidelines, whereas a threshold of an eGFR = <7 (based on the median eGFR at commencement of RRT in Australia) identified fewer patients. However, these differences did not alter a key outcome of the analysis: ESKD incidence and prevalence is much higher among Aboriginal than non-Aboriginal NT residents, even after accounting for undiagnosed cases.
All diagnosed patients had to have at least one contact with the health services to be found in one of the health datasets. Australia has a largely publicly funded health care system with universal access to health care. The national health care system allows Australians to access health care, regardless of their personal circumstances, while also giving options for individuals through private sector involvement in delivery and financing. Aboriginal Australians have much lower levels of private health insurance coverage compared to non-Aboriginal Australians (15% vs 51%), are more likely to use public hospital services and have lower rates of elective surgery [
26]. Barriers to accessing care when needed vary between remote and non-remote areas, while 21% and 58% of NT Aboriginal population live in a remote or a very remote area [
20,
27]. A previous NT study indicated that miscommunications and lack of shared understanding between Aboriginal patients and health staff in renal and hospital services often limited patients’ opportunities and capacities to make informed choices about their healthcare. Appropriate information and the associated understanding of the patients and their family are crucial for them to make the important decisions about treatment choice [
28,
29].
While the overall prevalence estimates provide a broad measure for planning appropriate services, the incidence estimates are of even greater concern. An estimated 53 largely middle-aged Aboriginal incident ESKD cases were not identified in the health data, although the awareness of chronic kidney disease amongst clinicians and clinical coders is already high in the NT. This finding supports the need for further measures for timely diagnosis, to reduce premature death and improve access for suitable treatment. Such measures include improved support for health care providers to recognise ESKD, including a need for better risk stratification tools in the high-risk NT Aboriginal population.
A previous study reported that more Aboriginal females than Aboriginal males commence treatment for ESKD [
30], which has been subsequently interpreted as NT Aboriginal women being more likely to start treatment [
31]. Our results indicate that while NT Aboriginal women have a greater prevalence and incidence of ESKD than their male counterparts, they were less likely to access RRT. This finding should be the subject of further investigation. Overall, 28% of NT patients did not receive RRT treatment and the proportion of Aboriginal patients who did not receive RRT treatment (29%) was higher than the non-Aboriginal figure (24%). For each patient, the decision to commence RRT or to have conservative management (no RRT) is complex. Medically assessed suitability for RRT, prognosis, anticipated quality of life, treatment burden, accessibility of health services, and personal choice all play a part in the decision of whether or not to receive RRT treatment [
1,
32]. ESKD has major medical and social implications for Aboriginal people [
33]. Isolation, the need to travel and social dislocation are particular issues for Aboriginal patients from remote communities, which may deter some patients from undertaking RRT. Despite a significant increase since 1999 in the number of remote communities in which dialysis is provided [
34], many areas of high need persist in having poor access to RRT service provision, with hundreds of kilometres between communities and the nearest dialysis unit [
22,
35]. Local policies and clinical pathways to support patients who choose not to take up RRT treatment are being evaluated [
32].
Our study has a number of limitations. First, it is possible that the linkage of the same individuals in different datasets was incomplete. This would have the effect of overestimating prevalence and incidence. However, the unique patient identifier within NT health data has previously been shown to be highly reliable, and a substantial majority of individuals within the samples were linked across data sources [
11]. A second limitation is that capture-recapture methods assume a closed population. Although the NT Aboriginal population is highly mobile between communities, interstate net migration is minimal [
36]. However, the non-Aboriginal population may move either into or out of the NT at a time of declining health or retirement. The similarity of non-Aboriginal results with the national estimate suggests that the effect of interstate migration may not be substantial [
17]. A third limitation is that the study relied on recording of a clinical diagnosis in electronic data systems. Case notes and clinical reviews were not available to validate diagnoses. A fourth limitation is that the definitions for ESKD varied between data sources, for example between a laboratory diagnosis in a primary care setting and the commencement of maintenance dialysis leading to registry notification. Similarly not all patients with an eGFR <15 may need RRT. This variation means that an individual will have a diagnosis in one dataset but may not appear in a second dataset or the documentation may be delayed for several months. These differences will have little effect on the measurement of known cases but may result in an overestimation of undiagnosed cases, as shown in the sensitivity analysis. A final limitation is that only 5.5 years’ data were available for the study. If information had been available for a longer period, some additional cases may have been identified and apparent incident cases may have been recognised as having had an earlier onset. However, using four different data sources reduced the probability of having missed the index episode. Future estimates, based on these data sources, will benefit from the availability of additional years of data.
Acknowledgments
We wish to thank the Registrars of Australia and New Zealand Dialysis and Transplant Registry and the Northern Territory Registry of Births, Deaths and Marriages and Northern Territory Department of Health for their support in providing data for this study.