Prevalence of Diabetes Mellitus Associated Chronic Kidney Disease in Ethnically Diverse Backgrounds in Western Sydney


 Background: Past research has demonstrated that ethnically diverse backgrounds have a greater risk of Diabetes Mellitus and/or Chronic Kidney Disease (CKD), which both contribute significantly to the disease burden placed on the Australian Healthcare system. The purpose of this study was to investigate the differences in DM associated CKD between ethnicities in Western Sydney, NSW Australia, which is known to be an ethnically diverse geographical region.Method: Using data from Blacktown and Mt Druitt Hospitals and a retrospective approach, individuals with diabetes and CKD were identified based on HbA1c and eGFR results. Using univariate analysis and past research, a logistic regression modelling was carried out on the data to identify relationships between ethnicity and DM associated CKD.Results: Out of 73,001 observations, 24,009 individuals were included in the analysis with 3,934 individuals with HbA1c consistent with diabetes and kidney function at eGFR at Stage 2 and above. 47% were female and 53% were males and the median age was 76 ± 12 years. Using a forward-method model building process, ethnicity was compared to a reference of Caucasian. The model showed that Pacific Islanders have the most risk (OR: 5.63, p < 0.001, CI: 4.78–6.65) compared to other ethnicities. Age has an 11% increased risk per year (OR: 1.11, p < 0.001, CI: 1.1–1.11), men were at a 53% greater risk (OR: 1.53, p < 0.001, CI: 1.14–1.67) and Aboriginal/Torres Strait Islanders were 76% more likely to have DM associated CKD (OR: 1.76, p < 0.001, CI: 1.33–2.32). Increased socioeconomic status resulted in a 11% decreased risk in DM associated CKD (OR: 0.82, p < 0.001, CI: 0.79–0.86) and smoking status, interestingly, also has an 18% lower risk of DM associated CKD (OR: 0.81, p = 0.005, CI: 0.71–0.92)Conclusion: The results show a clear difference in risk between ethnicities in DM associated CKD and its associated risk factors. These differences should be accounted for when considering interventions for at risk communities.

past research, a logistic regression modelling was carried out on the data to identify relationships between ethnicity and DM associated CKD.
Results: Out of 73,001 observations, 24,009 individuals were included in the analysis with 3,934 individuals with HbA1c consistent with diabetes and kidney function at eGFR at Stage 2 and above. 47% were female and 53% were males and the median age was 76 ± 12 years. Using a forward-method model building process, ethnicity was compared to a reference of Caucasian. The model showed that Paci c Islanders have the most risk (OR: 5.63, p < 0.001, CI: 4.78-6.65) compared to other ethnicities. Age has an 11% increased risk per year (OR: 1.11, p < 0.001, CI: 1.1-1.11), men were at a 53% greater risk (OR: 1.53, p < 0.001, CI: 1.14-1.67) and Aboriginal/Torres Strait Islanders were 76% more likely to have DM associated CKD (OR: 1.76, p < 0.001, CI: 1.33-2.32). Increased socioeconomic status resulted in a 11% decreased risk in DM associated CKD (OR: 0.82, p < 0.001, CI: 0.79-0.86) and smoking status, interestingly, also has an 18% lower risk of DM associated CKD (OR: 0.81, p = 0.005, CI: 0.71-0.92) Conclusion: The results show a clear difference in risk between ethnicities in DM associated CKD and its associated risk factors. These differences should be accounted for when considering interventions for at risk communities.

Background
Diabetes Mellitus (DM) is a disease characterised by the body's inability to regulate blood sugar levels correctly (1) . This can lead to a range of medical complications and co-morbidities such as heart disease, nerve injury, retinopathy and also kidney disease.
Chronic Kidney Disease (CKD) rates are rising around the world and it is a growing global health problem (2,3) . With diabetes as a major global health burden and one of the major non-communicable diseases, the microvascular complications of diabetes compounds CKD. With DM increasing in prevalence in Australia, CKD will also correspondingly rise over the coming years contributing to the burden on the healthcare system. In 2012, CKD accounted for approximately $4.1 billion in treatment costs, indirect costs and subsidies. Dialysis alone for CKD can cost more than $50,000 per person per year (4) .
Renal function in individuals with diabetes slowly deteriorates over time due to changes in the physical structure of the glomerulus in the kidneys. Initially, renal impairment may not even impact kidney function; proteinuria may be the only indication of dysfunction. This, incipient nephropathy may persist for some years before overt nephropathy develops, leading to Chronic Kidney Disease (CKD) due to DM.
Over 80% of individuals with Type 1 DM can develop some degree of CKD and about 20% of individuals with Type 2 DM will progress to the point of requiring dialysis or a kidney transplant. Even with early diagnosis of diabetes and good management and control, CKD will manifest in a number of patients.
The development and progression of DM and/or CKD is complex with many contributing factors including socio-economic ones. Lifestyle and accessibility to health dietary options play a major role in in treatment, progression and management of both diseases.
Past studies (5)(6)(7) within Australia and around the world have demonstrated that ethnically diverse populations are more susceptible to the development DM and CKD when compared to Caucasian or the native populations of other countries. Western Sydney is known to have an ethnically diverse population and identifying risk factors unique to this area would assist in planning and targeting future intervention programs or allocation of health resources to at risk communities (8)(9)(10) .
This study focuses on individuals living in the Western Sydney region who have presented to the Emergency Departments of Blacktown and Mt Druitt Hospitals. Western Sydney has an acknowledged higher rate of diabetes than other regions in the city, related to socioeconomic and other factors. Previous research has demonstrated a higher rate of diabetes than o cial gures suggest in this area (11) . The study population will be drawn from these individuals and will be, representative of the geographical area. The study will focus on CKD caused by DM and the effect of ethnicity and other socioeconomic factors on the prevalence of CKD within the study population.
It is hypothesised that Ethnicity, as measured by an individual's country of birth, will have a signi cant impact on the prevalence of CKD within the Western Sydney Region. The study will also aim to: 1. Describe the correlation between Diabetes and CKD within the study population 2. Provide insight to ethnicity and its role in DM caused CKD in Western Sydney Methodology Study Design This is a cross-sectional retrospective review of 73,001 presentations to Blacktown or Mt Druitt Emergency Departments, from mid-2016 to 2018, inclusive. These patients were administered a blood test as part of their presentation and the HbA1c (glycated haemoglobin A1c) was recorded into the database, which shall be referred to as the HbA1c database. The database also recorded various demographic data and the estimated Glomerular Filtration Rate (eGFR) for each presentation and other Page 4/16 variables. The database is the product of a testing regime set up as routine care for patients attending the two ED's and has been running since 2016. The full testing protocol has been published elsewhere (11) .

Outcome of Interest
The outcome of interest is patients with potential kidney dysfunction and diabetes. Individuals with diabetes were identi ed based on HbA1c values and kidney dysfunction was de ned as Stage 2 or higher based on eGFR. Those that meet these criteria are referred to as individuals with DM associated CKD for the purposes of this paper. The outcome variable is a coded binary variable for analysis.

Inclusion/Exclusion Criteria
Patients were included in the study if they had an HbA1c result and an eGFR result, along with the ethnicity recorded in the HbA1c database.
Patients' repeat presentations to the emergency department were also recorded into the database but these were excluded from the study so as to keep only unique presentations.
Any patients who didn't have their ethnicity recorded, i.e. not admitted into the hospital, were also excluded from the study database.

Study Population
The following diagram shows the inclusion and exclusion of patients in the study.

Co-variates
The HbA1c database recorded a number of different variables for each patient. For the purposes of this study, only the follow variables were considered.

Variables
The eGFR result will be used to differentiate patients into stages of kidney disease as per the World Health Organisation (WHO) staging of renal impairment as per Table 1. Glycaemic categorise were allocated as per Table 2.  SEIFA is measured in single unit increments, with a bell curve distribution centred over 1000 as a median value of IRSAD for a given region. For ease of interpretation, SEIFA was converted from single unit increments to 100 unit increments for analysis. A higher IRSAD score indicates a greater relative advantage verses disadvantage for an area. A lower score indicates greater disadvantage verses advantage.
The BMI variable contained many missing variables, with only approximately 12, 000 BMIs' recorded in the database and as such was not included in the subsequent analysis. A total of 21 missing Hb1Ac observations were also found in the database, which were not included in the analysis.
Smoking status was determined by using the ICD-10 codes F17 and Z72 which cover Nicotine Dependence and Tobacco use respectively. No differentiation was made between current and past smokers for this analysis.

Model Building Process
Variables were selected based on past literature suggesting correlation between DM and CKD. Past studies have found signi cant correlation between diabetes, ethnicity, age, SES status, marital status and gender, smoking status and Aboriginal/Torres Strait Islander (A/TSI) (7,(12)(13)(14)(15)(16) . These variables were individually tested for correlation with the outcome variable with signi cance set at p<0.05. A forward method model building process was used for the analysis.
Patients were grouped together by ethnic and geographic majority as described in Table 3. Certain ethnicities such as South America for example are an amalgamation of patients whose country of birth is in the South Americas. This was done to balance the groups as certain countries were represented by very few patients. Likewise, a Caucasian ethnicity is represented by the Australian/North American/European group in Table 3 and is used as the base ethnicity for analysis.

Results
Legend: Row Percentages () Discussion Table 5 above shows the rates of patients at each eGFR stage by ethnicity. The base ethnicity of Australian/Caucasian contains the most number of people at each stage of kidney function, followed by Indian/Subcontinent, Asian, Arabic, African and lastly, South American.
From the results in Table 4, Paci c Islanders are most at risk, when compared to the base group, with more than 5 times the risk of having DM associated CKD, when adjusting for the other variables. African ethnicities were twice (OR: 2.18 p < 0.001) as likely followed by Asian (OR: 1.91, p < 0.001) and Indian/Subcontinent (OR: 1.89, p < 0.001) ethnicities.
Marital status had no signi cant correlation with DM associated CKD once included in the multivariate analysis, therefore, was not reported. While marital status has been found to correlate with diabetes in other studies, this has not been replicated in this study. Age was also correlated with an increase in DM associated CKD, with an 11% increase, per year, controlling for other variables. Conversely DM associated CKD has an 18% reduction, per 100 unit increase in SEIFA, controlling for other variables, indicating that access to improved socioeconomic resources can reduce DM associated CKD prevalence.
Gender also has a signi cant effect with men having a 53% increased risk of DM associated CKD when compared with women, controlling for other variables. Finally, Aboriginal/Torres Strait Islander individuals seem to have a 76% greater risk of DM associated CKD, controlling for other variables. This signi cant risk could be due to the unique socio-economic and cultural factors associated Aboriginal communities and individuals. Some examples are lower levels of access/utilisation of resources such as education, health services and a higher rate of detrimental health behaviours such as smoking and alcohol consumption (17) .
Smoking was also interestingly, associated with a 19% reduction in the risk of DM associated CKD when compared to individuals who had never smoked. This is odd as smoking is a major risk factor of both diabetes and kidney disease, therefore an increased risk is to be expected. This could be explained by considering that most individuals with diabetes and renal impairment are given medical advice to quit smoking. This could result in a lower odds ratio associated with smoking, indicating possible survivor bias (16) . Alternatively, access to better and more frequent medical care could also explain the results, again, indicating survivor bias. The maps in Fig. 1 also show that ethnicity is also highly varied around the Western Sydney regions with large localised populations within close geographical proximity to Blacktown hospital. This highlights the diverse multicultural community that these hospitals serve and the challenges it brings. Understanding the needs of the community and its needs will allow for better provision of healthcare and allocation of resources.
These study's strengths are the large sample size that powers the study with over 24,000 individuals included in the analysis. This large sample size also suggests a good representation of the target population in Western Sydney. The results are thus more generalizable for Western Sydney and are very relevant for planning of targeted health interventions in the region. The database was also exible, with clinical information for analysis and socio-economic data as well, allowing for a more in depth modelling.
One of the main limitations of the study is using country of birth as a surrogate measure for ethnicity.
Immigrants have risks from their country of origin due to differences in diet lifestyle factors. This risk difference gradually diminishes as successive generations as they naturalise to the local population, but still can persist within close-knit ethnic communities (12) . The Caucasian ethnicity coding used for our analysis did not take this into account as it grouped together any individuals born in Australia, disregarding the birthplace of their parents, which are major source of cultural risk factors. Linguistically diverse backgrounds were also not considered as part of this study. The lack of certain ethnicities is also an issue such as the South American (1%). It is uncertain if this underrepresentation is re ective of the data or of the actual population of these ethnicities within the Western Sydney region.
The second limitation in this study is the reliance on a single biochemical measurement to de ne diabetes and CKD. Whilst elevated HbA1c levels may indicate diabetes, low levels can be present in diabetics with good glucose control and misclassi cation. Similarly, low eGFR may represent CKD but is not diagnostic and can be due to transient conditions. The causality between concurrent diabetes and CKD has not been con rmed for this study and should not be assumed. BMI and physical activity levels have not been controlled for in this study.

Conclusion
Certain ethnicities are more at risk from DM associated CKD than others and the study highlights this difference in epidemiology between ethnicities in the Western Sydney and some of the associated risk factors. The increased risk of DM associated CKD with certain ethnicities show that interventions targeting CKD directly or even diabetes and lifestyle changes should be tailored to better suit the target demographic. Cultural and language barriers may reduce the e cacy of self-management of chronic conditions and interventions that address ethnic differences as part of the program have more success that ones that do not (18) . The ndings of this study can be used as a starting point for planning and directing more targeted interventions for the communities most at risk.