Risk factor specific methods
This page describes in detail the methods unique to each risk factor included in the ABDS 2018. It is focused on the calculation of exposure estimates, as this was the aspect of risk estimation most influenced by Australia-specific data. The amount of detail described for each risk factor varies; more detail is included for risk factors for which there were new developments in the ABDS 2018, in particular, dietary risks.
Overarching methods and choices for risk factors describes the overall method used to calculate the PAFs and attributable burden, including the selection of linked diseases, estimation of effect sizes (relative risks), and assumptions for TMREDs (see risk factor specific methods).
The linked diseases and relative risks were sourced from the GBD 2019 or an AIHW review of the literature as described here and in Overarching methods and choices for risk factors. Most TMREDs were also sourced from the GBD 2019, with the exceptions described in the Risk factor-specific methods.
Exposure to risk factors in the lifetime of the individuals in the population can influence the proportion of burden in the reference year. For risk factors such as tobacco use, occupational risks, alcohol use, child abuse and neglect, illicit drug use, and unsafe sex, the burden can continue to exist from past exposure levels. Where evidence of ever being exposed to a risk factor can be linked to current burden, this is included in the analyses and described under the individual risk factor.
For some risk factors, such as overweight (including obesity), current exposure can have an impact on future burden. This is not accounted for in this study as the burden pertains to the reference year.
Not all risk factors are relevant to all population (age and sex) groups. For example, the bulk of the burden from high blood pressure occurs for people aged 25 and over. The choices for population groups and type of burden (fatal or non-fatal) were informed by the GBD 2019 (GBD 2019 Risk Factor Collaborators 2020). The population group for which attributable burden from a given risk factor has been estimated is described in each section.
Also, both fatal and non-fatal burden are relevant for most linked diseases in the study. For others, such as back pain & problems linked to occupational risks, only non-fatal burden has been estimated.
A supplementary table contains detailed definitions, data sources and linked diseases for all risk factors (Table S4.2).
Behavioural risk factors
The burden attributable to dietary risk factors was estimated in people aged 25 and over.
It should be noted that the methods, including the TMREDs, used in the ABDS 2018 to calculate attributable burden due to dietary risk factors do not align with current Australian dietary guidelines as they are used to calculate disease burden (see Dietary risk model parameters in the table below). For information on recommended food choices, see the Australian Dietary Guidelines (NHMRC 2015).
A literature review of the list of dietary risk factors and the methods used in ABDS 2015 was undertaken as part of this study to identify any additional dietary risk factors or update existing methods and linked diseases. The review looked into updates on reviews undertaken to inform dietary guidelines in Australia or internationally, published meta-analyses on dietary risk factors and associated linked diseases, and studies that specifically estimate relative risks of outcomes such as the World Cancer Research Fund International (WCRFI) continuous update project for linked cancers and the GBD 2019 (GBD 2019 Risk Factors Collaborators 2020).
The risk factors included were based on the AIHW review of evidence from the GBD 2019 (which included 15 dietary risk factors) and other systematic reviews from authoritative sources that have also assessed the impact of dietary risk factors on health. These other sources included the NHMRC dietary guidelines evidence paper, which provided scientific evidence for healthier Australian diets (NHMRC 2011). Evidence from the continuous update project by the World Cancer Research Fund International (WCRFI) was used for cancer outcomes (WCRFI 2017). Information on carbohydrates and health (SACN 2015) and the evidence used to create the World Health Organization guideline for sugar intake for adults and children by Moynihan and Kelly (2014) were reviewed while considering the inclusion of diet high in added sugar.
Due to methodological differences, methods for diet high in sodium are discussed in a separate sub-section.
Dietary risks included
The same dietary risk factors were included as in the ABDS 2015.
The risk factors included by the GBD that were not in this study include diet low in fibre and diet low in calcium. These were mediated entirely through diet low in whole grains and diet low in milk, respectively, which were included in this study. To avoid double-counting, diet low in fibre and diet low in calcium were excluded from this study.
The risk factor diet high in trans-fat was excluded from the study as consumption is low in Australia, on average.
The risk factor diet low in omega-3 seafood fatty acids was replaced by diet low in fish and seafood to align this risk factor with the other whole food risk factors, included using evidence from Zheng et al. (2012).
Population attributable fractions estimated using comparative risk assessment
The risk factors estimated using the comparative risk assessment were diet low in fruit, vegetables, wholegrains, legumes, nuts and seeds, milk, fish and seafood, and polyunsaturated fats; and diet high in red meat, processed meat and sugar-sweetened beverages.
The models for the risk factors in GBD 2019 changed to have different levels of exposure which are not associated with increased risk (TMRED). The updated TMRED are listed in Dietary risk model parameters above and indicate for some risk factors levels of consumption above which there is a conferred protective impact against linked diseases. In the ABDS 2015, TMREDs were made up of a range of exposure categories (e.g. 200 to 300g of fruit); however, now the increased risk has been grouped into categories where it is only possible to include a single cut off value for the TMRED (e.g. 300g of fruit).
Exposure estimate
The National Nutrition and Physical Activity Survey (NNPAS) part of the AHS 2011–12 collected food intake data (through a 24-hour recall) from participants for 2 days. As with the ABDS 2015, the amount of each food was adjusted to the usual intake, taking into account reported intake on day 1 and day 2 and using the method developed by the National Cancer Institute. This method was used to estimate the distribution of intake of foods as described in Dietary risk model parameters in the Australian population.
To estimate consumption in 2018, unit record level data from the AHS 2011–12 was adjusted by the percentage change from 2011 to 2018, based on the mean exposure from the National Nutrition Survey 1995 component of NHS 1995 and the mean exposure from the AHS 2011–12 by age. The mean exposure in each year was estimated by mean number of serves per 10,000 kJ, as published by the ABS (2017).
It is important to note that there is significant under-reporting of dietary intake in the AHS 2011–12 (as with all representative dietary surveys) (ABS 2014). There is a tendency for survey respondents to either change their behaviour or misrepresent their consumption (whether consciously or subconsciously) to report a lower energy or food intake. This under‑reporting is unlikely to affect all foods and nutrients equally (that is, ‘unhealthy’ discretionary foods are most likely to be under-reported, and healthy foods, such as fruit and vegetables, are likely to be over-reported). The AIHW was unable to adjust for under‑reporting in the ABDS 2018, except for diet high in sodium.
Table 4.44: Dietary risk model parameters
Risk factor | Diet low in fruit – Average daily consumption of fresh, frozen, cooked, canned, or dried fruits (excluding fruit juices) |
---|---|
Disease outcome | Coronary heart disease, lung cancer, oesophageal cancer, stroke, type 2 diabetes |
TMRED | Consumption of at least 300 g of fruit per day |
National data source | Self-reported from AHS 2011–12 |
Units for effect size calculation | Per 100 g per day intake decrease |
Risk factor | Diet low in legumes – Average daily consumption of fresh, frozen, cooked, canned, or dried legumes |
---|---|
Disease outcome | Coronary heart disease |
TMRED | Consumption of at least 150 g of legumes per day |
National data source | Self-reported from AHS 2011–12 |
Units for effect size calculation | Per 50 g per day intake decrease |
Risk factor | Diet low in milk – Average daily consumption of milk including non-fat, low-fat and full-fat milk, excluding soy milk and other plant derivatives |
---|---|
Disease outcome | Bowel cancer |
TMRED | Consumption of at least 240 g of milk per day |
National data source | Self-reported from AHS 2011–12 |
Units for effect size calculation | Per 60 g per day intake decrease |
Risk factor | Diet low in nuts and seeds – Average daily consumption of nut and seed foods |
---|---|
Disease outcome | Coronary heart disease, type 2 diabetes |
TMRED | Consumption of at least 14 g of nuts and seeds per day |
National data source | Self-reported from AHS 2011–12 |
Units for effect size calculation | Per 7 g per day intake decrease |
Risk factor | Diet low in polyunsaturated fats – Average daily consumption of polyunsaturated fats |
---|---|
Disease outcome | Coronary heart disease |
TMRED | Consumption of polyunsaturated fatty acids at least 8% of total daily energy |
National data source | Self-reported from AHS 2011–12 |
Units for effect size calculation | Per 2% energy from polyunsaturated fat decrease |
Risk factor | Diet high in processed meats – Average daily consumption of meat preserved by smoking, curing, salting, or addition of chemical preservatives |
---|---|
Disease outcome | Bowel cancer, coronary heart disease, type 2 diabetes |
TMRED | Consumption of less than 24 g of processed meat per day |
National data source | Self-reported from AHS 2011–12 |
Units for effect size calculation | Per 25 g per day intake increase |
Risk factor | Diet high in red meat – Average daily consumption of red meat (beef, pork, lamb, and goat) (excluding poultry, fish, eggs and all processed meats) |
---|---|
Disease outcome | Bowel cancer, breast cancer, coronary heart disease, stroke, type 2 diabetes |
TMRED | Consumption of less than 50 g of red meat per day |
National data source | Self-reported from AHS 2011–12 |
Units for effect size calculation
|
Per 50 g per day intake increase |
Risk factor | Diet low in vegetables – Average daily consumption of fresh, frozen, cooked, canned, or dried vegetables, (excluding vegetable juices, legumes and starchy vegetables such as potatoes or corn) |
---|---|
Disease outcome | Coronary heart disease, oesophageal cancer, stroke |
TMRED | Consumption of at least 300 g of vegetables per day |
National data source | Self-reported from AHS 2011–12 |
Units for effect size calculation | Per 100 g per day of vegetable intake decrease |
Risk factor | Diet low in whole grains (including high fibre cereals) – Average daily consumption of wholegrain or higher fibre breads, cereals, rice, pasta, crumpets, muffins, crispbreads, relevant fortified cereals with 1 g of fibre per 10 g of carbohydrate |
---|---|
Disease outcome | Bowel cancer, coronary heart disease, stroke, type 2 diabetes |
TMRED | Consumption of at least 150 g of wholegrains per day |
National data source | Self-reported from AHS 2011–12 |
Units for effect size calculation | Per 50 g per day intake decrease |
Risk factor | Diet high in sugar sweetened beverages – Consumption of beverages with ≥50 kcal per 226.8 g serving, including carbonated beverages, sodas, energy drinks and fruit drinks (excluding 100% fruit and vegetable juices) |
---|---|
Disease outcome | Coronary heart disease, type 2 diabetes |
TMRED | Consumption of less than 60 g of sugar-sweetened beverages per day |
National data source | Self-reported from AHS 2011–12 |
Units for effect size calculation |
Per 60 g per day intake increase |
Risk factor | Diet low in fish and seafood - Average daily consumption of fish and seafood |
---|---|
Disease outcome | Coronary heart disease |
TMRED | Consumption of fish or seafood 100 g per week |
National data source |
Self-reported from AHS 2011–12 (day 1 only) |
Units for effect size calculation | Per 15g per day intake decrease |
Estimates by socioeconomic group
Exposure to dietary risks was estimated from the AHS 2011–12, modelled to 2018, and the difference in the mean estimate in each socioeconomic group quintile as described in Overarching methods and choices for risk factors.
2015, 2011 and 2003 estimates
The analysis for the year 2011 was based on the methods using the AHS 2011–12 data as described earlier in this section on dietary risks.
The exposure to these risk factors over time was calculated by comparing the mean exposure from the NHS 1995 and the mean exposure from the AHS 2011–12 by age. Unit record level data from the AHS 2011–12 were adjusted by the percentage change from 2011 to 2003 and 2015 in these data sources to estimate the distribution of dietary intake in those reference years. This method is the same as was used for the 2018 study.
Indigenous specific estimates
Exposure estimates for each of the 13 dietary risk factors for the Indigenous population was estimated from the AATSIHS 2012–13 using the same methods as used for national estimates. For example, day 1 dietary recall data were used for the micronutrients. For whole foods, the AHS whole food database was used to estimate the average proportion of the whole foods from within each food classified in the AATSIHS 2012–13. Exposure for 2003 and 2018 were estimated by assuming the same trends as in the national population.
Dietary risks mediated through other risk factors - Diet high in sodium
Diet high in sodium was measured by the amount it mediated blood pressure. The methods for this risk factor use comparative risk assessment and are based on the GBD 2019.
The attributable burden for diet high in sodium was calculated from a model of the impact of current sodium consumption on blood pressure levels in Australia. The model estimates the blood pressure distribution of Australians if no sodium above the TMRED was consumed.
Population attributable fraction
This was calculated in 4 steps for 2011:
- The consumption of sodium self-reported on day 1 in the AHS 2011–12 study was adjusted due to under-reporting. The data from dietary recall studies are known to include an under-reporting of the consumption of discretionary foods that are high in sodium (ABS 2014). An adjustment factor was calculated by comparing the mean amount of sodium from 24 hour urinary samples for Australia estimated by the Global Burden of Diseases Nutrition and Chronic Disease Expert Group (Powles et al. 2013) with the mean amount by dietary recall in the AHS 2011–12.
- The prevalence of blood pressure due to high sodium intake was estimated using the effect of sodium consumption on blood pressure. The effect was estimated by an adjustment factor, which varies by age, the presence or absence of hypertension, and race (non-African descent), sourced from the GBD 2016. These adjustment factors were used to calculate the distribution of systolic blood pressure that would be expected from reducing sodium consumption to the TMRED compared with current levels of sodium consumption. Blood pressure was based on measured estimates in the AHS 2011–12.
- These 2 estimates of the distribution of systolic blood pressure (with and without sodium consumption above the TMRED) were used with the methods for the high blood pressure risk factor (including the TMRED, all linked diseases and relative risks) to estimate the PAFs for both of these scenarios.
- Finally, the PAFs for diet high in sodium was estimated using the difference between the PAFs from the 2 scenarios by age and sex.
To estimate the impact of sodium in 2018, the distribution of blood pressure prevalence from the NHS 2017–18 was estimated. To calculate the distribution without the consumption of sodium above the TMRED, the blood pressure of each survey respondent was adjusted by the average adjustment per weighted count in each age, sex and blood pressure category calculated in 2011–12. The PAF for diet high in sodium was then calculated as described here for 2011.
Table 4.45: Diet high in sodium risk model parameters
Risk factor | Diet high in sodium – Consumption of sodium |
---|---|
Disease outcome | High blood pressure-linked diseases: Aortic aneurysm, atrial fibrillation and flutter, cardiomyopathy, chronic kidney disease, coronary heart disease, dementia, hypertensive heart disease, inflammatory heart disease, non-rheumatic valvular disease, peripheral vascular disease, rheumatic heart disease, stroke |
TMRED | 24 hr urinary sodium of 2 g per day |
National data source | Self-reported from AHS 2011–12; adjusted based on urinary sodium estimate (Powles et al. 2013) |
Units for effect size calculation | Per 2.3g per day intake increase |
Estimates by socioeconomic group
The average adjustment factors estimated for sodium intake and blood pressure for 2011 by socioeconomic quintile were applied to the distribution of blood pressure prevalence in the NHS 2017–18.
2015, 2011 and 2003 estimates
The analysis for the year 2011 was based on the methods using the AHS 2011–12 data as described earlier.
The average adjustment factors estimated for sodium intake and blood pressure for 2011 were applied to the distribution of blood pressure prevalence in the NHS 2004–05 and 2014–15 to calculate 2003 and 2015 estimates, respectively.
Indigenous specific estimates
The average adjustment factors estimated for sodium intake and blood pressure for 2011 and 2018 were applied to the distribution of blood pressure prevalence in the NATSIHS 2004–05 and NATSIHS 2018–19 to calculate 2003 and 2018 respectively.
References
ABS 2014. Australian Health Survey: users’ guide, 2011–13: under-reporting in nutrition surveys. ABS cat. no. 4363.0.55.001. Canberra: ABS. Viewed 22 June 2016.
ABS 2017. Australian Health Survey: consumption of food groups from the Australian Dietary Guidelines, 2011–12. ABS cat. no. 4364.0.55.012. Canberra: ABS.
GBD 2019 Risk Factors Collaborators 2020. Global burden of 87 risk factors in 204 countries and territories, 1990–2019: a systematic analysis for the Global Burden of Disease Study 2019. The Lancet 396:1223–249.
Moynihan PJ & Kelly SA 2014. Effect on caries of restricting sugars intake: systematic review to inform WHO Guidelines. Journal of Dental Research 93:8–18.
NHMRC (National Health and Medical Research Council) 2011. A review of the evidence to address targeted questions to inform the revision of the Australian Dietary Guidelines. Canberra: NHMRC.
NHMRC 2015. Australian Dietary Guidelines 1–5. Canberra: NHMRC. Viewed 15 August 2015.
Powles J, Fahimi S, Micha R, Khatibzadeh S, Shi P, Ezzati M et al. 2013. Global, regional and national sodium intakes in 1990 and 2010: a systematic analysis of 24 h urinary sodium excretion and dietary surveys worldwide. BMJ Open 3:e003733.
SACN (Scientific Advisory Committee on Nutrition) 2015. Carbohydrates and health. London: The Stationery Office Limited.
WCRFI (World Cancer Research Fund International) 2017. About the Continuous Update Project. Viewed 1 August 2018.
Zheng J, Huang T, Yu Y, Hu X, Yang B & Li D 2012. Fish consumption and CHD mortality: an updated meta-analysis of seventeen cohort studies. Public Health Nutrition 15(4):725–37.
This risk factor was estimated in people aged 15 and over using direct evidence. It was not possible to estimate the burden due to this risk factor by socioeconomic group as the data were not available.
Population attributable fraction estimated using direct evidence
The entire burden of cervical cancer, chlamydia, gonorrhoea, syphilis and other sexually transmitted infections was attributed to unsafe sex; therefore, a PAF of 1 was used.
PAFs were estimated directly for chronic liver disease, hepatitis B, hepatitis C, HIV/AIDS, and liver cancer from the National Notifiable Diseases Surveillance Scheme data published in annual surveillance reports by The Kirby Institute (The Kirby Institute 2018).
Acute hepatitis B and C
For acute hepatitis B and hepatitis C, the direct PAFs were calculated from estimated proportions of people with newly acquired hepatitis B or hepatitis C infections in 2018 who were exposed to unsafe sex.
Chronic liver disease and liver cancer
Chronic hepatitis C infection
The annual rates of decompensated cirrhosis (chronic liver disease), hepatocellular carcinoma (liver cancer) and liver transplants due to hepatitis C between 2006 and 2015 were published in the 2016 Annual Surveillance Reports (The Kirby Institute 2016). This trend information was used to determine the rate of hepatitis C related morbidity in each reference year (2003, 2011, 2015 and 2018).
To determine the rate of hepatitis C related chronic liver disease and liver cancer due to unsafe sex, data on newly acquired hepatitis C infection in men between the years 2000 and 2013 by exposure type was used a proxy.
The proportion of chronic liver disease and liver cancer due to unsafe sex was estimated by dividing the number of hepatitis C related morbidity cases due to unsafe sex by the total prevalence for liver cancer and chronic liver disease in each reference year.
Chronic hepatitis B infection
There is little data on the proportion of people living with chronic hepatitis B due to unsafe sex; however, there is more data available on the proportion of people living with chronic hepatitis B due to unsafe injecting practices (MacLachlan et al. 2013; O’Sullivan 2004).
Therefore, an indirect method was used to estimate hepatitis C related morbidity due to unsafe sex. The proportion of chronic liver disease and liver cancer due to unsafe sex was estimated by applying an unsafe sex exposure:drug use exposure ratio to the proportion of hepatitis B related chronic outcomes due to unsafe injecting practices in each reference year. Estimates of the number of newly acquired hepatitis B infection in men between 2002 and 2011 by exposure type were used to estimate the unsafe sex exposure:drug use exposure ratio.
Since only a single direct PAF is required for chronic liver disease due to unsafe sex and another for liver cancer due to unsafe sex, the separate PAFs calculated for hepatitis C related and hepatitis B related chronic liver disease and liver cancer due to unsafe sex were summed.
HIV/AIDS
For HIV/AIDS, direct PAFs were calculated from estimated proportions of diagnosed AIDS cases in 2018 with a relevant exposure category (including homosexual contact only, homosexual contact and injecting drug use or heterosexual contact).
Table 4.46: Unsafe sex risk model parameters
Risk factor |
Unsafe sex – Unsafe sex |
---|---|
Disease outcome |
Cervical cancer, chlamydia, chronic liver disease, gonorrhoea, hepatitis B, hepatitis C, HIV/AIDS, liver cancer, syphilis, other sexually transmitted infections |
TMRED |
No unsafe sex |
National data source |
National notifiable disease annual surveillance reports (The Kirby Institute) |
Units for effect size calculation |
All sexually transmitted infections and cervical cancer attributed to unsafe sex HIV/AIDS, hepatitis B and hepatitis C from direct evidence |
2015, 2011 and 2003 estimates
Methods for estimating exposure and calculating the PAFs in 2018 were used to produce 2015, 2011 and 2003 estimates. Data from the NNDSS published in the annual surveillance reports by The Kirby Institute were used to calculate PAFs for unsafe sex (The Kirby Institute 2004, 2012, 2013, 2016).
Indigenous specific estimates
For Indigenous estimates for unsafe sex, the same methods and exposure data sources were used as for national estimates. The quality of Indigenous data in the NNDSS varies by disease and state/territory, and is described in the annual surveillance reports published by the Kirby Institute.
References
MacLachlan JH, Allard N, Towell V & Cowie BC 2013. The burden of chronic hepatitis B virus infection in Australia, 2011. Australian and New Zealand Journal of Public Health 37(5):416–22.
O’Sullivan BG, Gidding HF, Law M, Kaldor JM, Gilbert GL & Dore GJ 2004. Estimates of chronic hepatitis B virus infection in Australia, 2000. Australian and New Zealand Journal of Public Health 28(3):212–16.
The Kirby Institute 2004. HIV, viral hepatitis and sexually transmitted infections in Australia: annual surveillance report 2004. Sydney: The Kirby Institute, UNSW.
The Kirby Institute 2012. HIV, viral hepatitis and sexually transmitted infections in Australia: annual surveillance report 2012. Sydney: The Kirby Institute, UNSW.
The Kirby Institute 2013. National blood-borne virus and sexually transmissible infections surveillance and monitoring report 2013. Sydney: The Kirby Institute, UNSW.
The Kirby Institute 2016. HIV, viral hepatitis and sexually transmissible infections in Australia: Annual surveillance report 2016. Sydney: The Kirby Institute, UNSW.
The Kirby Institute 2018. HIV, viral hepatitis and sexually transmissible infections in Australia: Annual surveillance report 2018. Sydney: The Kirby Institute, UNSW.
The impact of tobacco use captures the burden attributable to current smoking, past smoking (in people aged 30 and over) and exposure to second-hand smoke in the home (in people of all ages). In the GBD 2016, chewing tobacco was added as an exposure to tobacco use. Due to very low prevalence in Australia, chewing tobacco was not included in the ABDS 2018.
Population attributable fraction estimated using comparative risk assessment
Linked diseases and relative risks
Linked diseases and relative risks were sourced from the GBD 2016 (GBD 2016 Risk Factor Collaborators 2017). More detail on the methods are described further in the report Burden of tobacco use in Australia: Australian Burden of Disease Study 2015 (AIHW 2019).
Exposure estimates
The NDSHS 2013 was used to estimate the proportion of the population who are current smokers (5-year lagged). Using these data for current smokers allows for a 5-year lag between exposure and these disease outcomes. Current smoking (5-year lagged) was linked to cardiovascular diseases, diabetes, asthma and respiratory infections. Exposure to current tobacco smoking (5-year lagged) was calculated from the proportion of individuals in the NDSHS 2013 who reported smoking daily, weekly or less than weekly.
The NDSHS 2019 was used to estimate the proportion of non-smokers exposed to environmental tobacco in the home (second-hand smoke).
Due to the much longer lag between smoking and the incidence of cancers and chronic respiratory conditions, as well as consistent reductions in smoking rates over recent decades, the tobacco attributable burden for those disease outcomes cannot be estimated from data on the current or recent prevalence. For these conditions, the ‘smoking impact ratio’ (described by Peto et al. 1992) was used as an indirect method to estimate the accumulated risk from tobacco smoking. Lung cancer mortality in 2018 (by age and sex) from the NMD was compared with lung cancer mortality rates among a cohort of smokers and never-smokers in the United States (Peto et al. 1992). The excess mortality seen in the Australian population, compared with this cohort of non-smokers, is used to determine the proportion of the population living with accumulated tobacco risk. The burden attributable to past smoking was estimated in people aged 40 and over because the small number of lung cancer deaths observed in those aged 30–39 resulted in unreliable PAFs.
Table 4.47: Tobacco risk model parameters
Risk factor |
Tobacco use – Second-hand smoke |
---|---|
Disease outcome |
Breast cancer, coronary heart disease, influenza, lower respiratory infections, lung cancer, otitis media, stroke, type 2 diabetes |
TMRED |
No tobacco use |
National data source |
NDSHS 2019 |
Units for effect size calculation |
Proportion of the population exposed to second-hand smoke |
|
|
Risk factor |
Tobacco use – Current smoking (5-year lagged) |
Disease outcome |
Age-related macular degeneration, aortic aneurysm, asthma, atrial fibrillation & flutter, back pain & problems, cataract & other lens disorders, coronary heart disease, dementia, gallbladder & biliary diseases, gastroduodenal disorders, hypertensive heart disease, lower respiratory infections, multiple sclerosis, other cardiovascular diseases, peripheral vascular disease, rheumatoid arthritis, stroke, type 2 diabetes |
TMRED |
No tobacco use |
National data source |
NDSHS 2013 |
Units for effect size calculation |
Proportion of the population who smoked 5 years ago |
|
|
Risk factor |
Tobacco use – Smoking impact ratio |
Disease outcome |
Acute lymphoblastic leukaemia, acute myeloid leukaemia, bladder cancer, bowel cancer, breast cancer, cervical cancer, chronic lymphocytic leukaemia, chronic myeloid leukaemia, COPD, kidney cancer, laryngeal cancer, lip & oral cavity cancer, liver cancer, lung cancer, nasopharynx cancer, oesophageal cancer, other leukaemias, other respiratory diseases, pancreatic cancer, prostate cancer, stomach cancer |
TMRED |
No tobacco use |
National data source |
NMD |
Units for effect size calculation |
Lung cancer mortality rate; Peto et al. 1992 |
Estimates by socioeconomic group
Exposure estimates by socioeconomic group were based directly from the same data source as the national exposure estimates.
2015, 2011 and 2003 estimates
The NDSHS 2010 was used to estimate the proportion of the population who are current (5‑year lagged) smokers for 2015. The NDSHS 2016 was used to estimate the proportion of non‑smokers exposed to second-hand smoke. The NMD 2015 was used to estimate lung cancer mortality.
National exposure estimates for 2011 and 2003 were calculated from the earlier iterations of the same surveys used for the 2015 estimates and followed the same method.
Indigenous specific estimates
The same general methods were used to estimate exposure to tobacco use in the Indigenous population. However, there were some differences in the data sources used to estimate the proportion of the Indigenous population who are current and former smokers, as well as the proportion of non-smokers exposed to environmental tobacco in the home.
Due to the small Aboriginal and Torres Strait Islander sample in the NDSHS (about 460 respondents in 2010), estimates of tobacco exposure were not considered reliable when broken down by age, sex and smoking status. Instead, the National Aboriginal and Torres Strait Islander Social Surveys (NATSISS) (2002 and 2008) were used for 2003 and 2011 estimates, respectively, and the Australian Aboriginal and Torres Strait Islander Health Survey (AATSIHS) 2012–13 was used for 2018 estimates. While the earlier two survey dates do not directly align with the 5-year lagged smoking prevalence used for national estimates, analysis of Indigenous smoking rates in consecutive ABS Indigenous health and social surveys (2001, 2002, 2004–05, 2008, 2011–12) showed no discernible trends up to the AATSIHS 2011–12. Therefore, the choice of the 2002 and 2008 NATSISS surveys is likely to have had little impact on the proportion of the population exposed.
Similar to national estimates, a smoking impact ratio was used as an indirect method to estimate the accumulated risk from tobacco smoking for cancers and respiratory diseases.
References
AIHW 2019. Burden of tobacco use in Australia: Australian Burden of Disease Study 2015. Cat. no. BOD 20. Canberra: AIHW.
GBD 2016 Risk Factors Collaborators 2017. Global, regional, and national comparative risk assessment of 84 behavioural, environmental and occupational, and metabolic risks or clusters of risks, 1990–2016: a systematic analysis for the Global Burden of Disease Study 2016. The Lancet 390:1345–422.
Peto R, Boreham J, Lopez AD, Thun M & Heath C 1992. Mortality from tobacco in developed countries: indirect estimation from national vital statistics. The Lancet 339(8804):1268–78.
The impact of illicit drug use was estimated in people aged 15 and over. The burden attributable to this risk factor was calculated as described in detail in the AIHW publication Impact of alcohol and illicit drug use on the burden of disease and injury in Australia: Australian Burden of Disease Study 2011 (AIHW 2018).
Population attributable fraction estimated using direct evidence
Unsafe injecting practices
PAFs for the linked diseases for unsafe injecting practices (chronic liver disease, hepatitis B hepatitis C, HIV/AIDS and liver cancer) were calculated from the NNDSS data published in the annual surveillance reports by The Kirby Institute (The Kirby Institute 2018).
HIV/AIDS
For HIV/AIDS, direct PAFs were calculated from the estimated proportion of diagnosed AIDS cases in 2018 who were exposed to unsafe injecting practices with or without homosexual contact.
Acute hepatitis B and C
For acute hepatitis B and hepatitis C, the direct PAFs were calculated from the estimated proportion of newly acquired hepatitis B or hepatitis C infections in 2018 who were exposed to unsafe injecting practices with or without homosexual contact.
Chronic liver disease and liver cancer
Chronic hepatitis C infection
The rates of decompensated cirrhosis (chronic liver disease), hepatocellular carcinoma (liver cancer) and liver transplants due to hepatitis C are published in the annual surveillance reports by The Kirby Institute. These were multiplied by the earliest year of exposure data estimates available to determine the proportion of hepatitis C related morbidity due to unsafe injecting practices.
The proportion of chronic liver disease and liver cancer due to unsafe injecting practices was then estimated by quantifying the rate of hepatitis C related morbidity from the total prevalence for liver cancer and chronic liver disease in 2018.
Chronic hepatitis B infection
The Kirby Institute reported that 5.7% of people living with chronic hepatitis B in 2017 and 2015 had acquired this condition through unsafe injecting practices (The Kirby Institute 2016, 2018). This is similar to Australian estimates reported by other published studies for the years 2011 (5.7%) and the year 2000 (4.7%) (MacLachlan et al. 2013; O’Sullivan 2004).
The proportion of these chronic outcomes being chronic liver disease or liver cancer was then estimated using total disease prevalence data from the ABDS 2018.
Accidental poisoning
The direct PAFs for accidental poisoning linked to specific illicit drugs was estimated using the number of deaths due to accidental poisoning with a mention of each drug type compared with the total number of accidental poisoning deaths in 2018 in the NMD. These methods are described in more detail in the section on alcohol use. The PAFs were also applied to non-fatal burden due to accidental poisoning.
Illicit drug dependence
All of the burden due to drug use disorders (including amphetamine, cannabis, cocaine, opioid and other illicit drug use disorders) was attributable to illicit drug use (a PAF of 1).
Population attributable fraction estimated using comparative risk assessment
Exposure estimates
There are 2 types of exposure to drug use estimated for the risk factor illicit drug use: drug dependence and driving under the influence of illicit drugs. Estimates of the exposure to drug dependence are sourced from prevalence estimates for the relevant drug use disorder from the ABDS 2018. Exposure to drug dependence—not drug use—was used in this study.
Exposure to driving under the influence of illicit drugs was estimated from the 2016 NDSHS (as this question was not available in the 2019 NDSHS)—specifically, the proportion of the population that responded yes to the question: ‘In the last 12 months did you undertake the activity—drove a motor vehicle—while under the influence of or affected by illicit drugs?’ However, these data do not provide details on the type of drug used while driving and are likely to be an underestimate.
The type of drug used while driving was sourced by the relative prevalence of the use of different drugs self-reported in the NDSHS. This data source was used as a source of drug type in preference to roadside drug testing, as it included a full range of illicit drugs associated with driving impairment and was not impacted by the ability to measure the presence of the drug in saliva tests.
Table 4.48: Illicit drug use risk model parameters
Risk factor |
Cannabis use – Cannabis dependence |
---|---|
Disease outcome |
Anxiety disorders, depressive disorders, schizophrenia |
TMRED |
No illicit drug use |
National data source |
ABDS 2018 |
Units for effect size calculation |
Prevalence of illicit drug use disorders |
|
|
Risk factor |
Cannabis use – Driving under the influence of cannabis |
Disease outcome |
Road traffic injuries—motorcyclists and road traffic injuries—motor vehicle occupants |
TMRED |
No illicit drug use |
National data source |
NDSHS 2016 |
Units for effect size calculation |
Prevalence of driving under the influence of illicit drugs |
|
|
Risk factor |
Cannabis use – Cannabis use and dependence |
Disease outcome |
Accidental poisoning |
TMRED |
No illicit drug use |
National data source |
NMD |
Units for effect size calculation |
Direct evidence |
|
|
Risk factor |
Amphetamine, cocaine and opioid use – Amphetamine, cocaine and opioid use or dependence |
Disease outcome |
Suicide & self-inflicted injuries |
TMRED |
No illicit drug use |
National data source |
ABDS 2018 |
Units for effect size calculation |
Prevalence of illicit drug use disorders |
|
|
Risk factor |
Amphetamine, cocaine and opioid use – Driving under the influence of amphetamine, cocaine or opioids |
Disease outcome |
Road traffic injuries—motorcyclists and road traffic injuries—motor vehicle occupants |
TMRED |
No illicit drug use |
National data source |
NDSHS 2016 |
Units for effect size calculation |
Prevalence of driving under the influence of illicit drugs |
|
|
Risk factor |
Amphetamine, cocaine and opioid use – Amphetamine or opioid use and dependence |
Disease outcome |
Accidental poisoning |
TMRED |
No illicit drug use |
National data source |
NMD |
Units for effect size calculation |
Direct evidence |
|
|
Risk factor |
Amphetamine, cannabis cocaine, opioid and other illicit drug use – Illicit drug dependence |
Disease outcome |
Drug use disorders (excluding alcohol) |
TMRED |
No illicit drug use |
National data source |
ABDS 2018 |
Units for effect size calculation |
Direct evidence |
|
|
Risk factor |
Unsafe injecting practices – Unsafe injecting practices |
Disease outcome |
Chronic liver disease, hepatitis B, hepatitis C, HIV/AIDS, liver cancer |
TMRED |
No unsafe injecting practices |
National data source |
National notifiable disease annual surveillance reports (The Kirby Institute) |
Units for effect size calculation |
Direct evidence |
Estimates by socioeconomic group
The data source used for the national estimates as described above also provided data by socioeconomic status, except for unsafe injecting practices for which these data were not available. The national PAFs were used for each socioeconomic group for diseases linked to unsafe injecting practices.
2015, 2011 and 2003 estimates
The burden attributable to illicit drug use in 2015 was estimated using the NDSHS 2016 and The Kirby Institute data as described in Impact of alcohol and illicit drug use on the burden of disease and injury in Australia: Australian Burden of Disease Study 2011 (AIHW 2018), using the same methods as for 2018.
The burden attributable to illicit drug use in 2011 and 2003 was estimated using the NDSHS 2010 and 2004 and The Kirby Institute data for 2011 and 2003, respectively, using the same methods as for 2018.
Indigenous specific estimates
For Indigenous risk factor estimates for drug use, the same data sources and methods were used as for national estimates. The quality of Indigenous data in the NNDSS varies by disease and state/territory, and is described in the annual surveillance reports published by the Kirby Institute. For drug driving data, due to a lack of suitable data for Indigenous Australians, the prevalence of drug driving was assumed to be the same as for the general population.
References
AIHW 2018. Impact of alcohol and illicit drug use on the burden of disease and injury in Australia: Australian Burden of Disease Study 2011. Cat. no. BOD 19. Canberra: AIHW.
MacLachlan JH, Allard N, Towell V & Cowie BC 2013. The burden of chronic hepatitis B virus infection in Australia, 2011. Australian and New Zealand Journal of Public Health 37(5):416–22.
O’Sullivan BG, Gidding HF, Law M, Kaldor JM, Gilbert GL & Dore GJ 2004. Estimates of chronic hepatitis B virus infection in Australia, 2000. Australian and New Zealand Journal of Public Health 28(3):212–16.
The Kirby Institute 2016. HIV, viral hepatitis and sexually transmissible infections in Australia: Annual surveillance report 2016. Sydney: The Kirby Institute, UNSW.
The Kirby Institute 2018. HIV, viral hepatitis and sexually transmissible infections in Australia: Annual surveillance report 2018. Sydney: The Kirby Institute, UNSW.
The burden attributable to this risk factor was calculated as described in detail in the AIHW publication Impact of alcohol and illicit drug use on the burden of disease and injury in Australia: Australian Burden of Disease Study 2011 (AIHW 2018). Note that the risk factor is alcohol use while alcohol use disorders is a linked disease.
Population attributable fraction calculated with direct evidence
In the GBD study, the linked diseases chronic liver disease due to alcohol and liver cancer due to alcohol were entirely attributed to alcohol use, and no relative risks were published for use in the comparative risk assessment approach. In the ABDS 2018, chronic liver disease and liver cancer were not broken down to this level. The PAFs for chronic liver disease were estimated from the proportion represented by chronic liver disease due to alcohol of all chronic liver disease burden, as estimated for Australia by the GBD 2019. The same method was used to estimate the PAFs for liver cancer. The burden of alcohol dependence (the linked disease) was entirely attributed to alcohol use (the risk factor).
Direct evidence was used to derive the PAFs for accidental poisoning linked to alcohol use, using the mention of specific drugs recorded in the NMD 2018 as described by the AIHW (2018).
Population attributable fraction estimated using comparative risk assessment
Exposure estimates
The proportions of the Australian population who are current drinkers, former drinkers or never drank alcohol were sourced from self-reported data in the NDSHS 2019. However, the amount of alcohol self-reported to be consumed by current drinkers in this and other surveys is known to be an underestimate of actual consumption (Rehm et al. 2010).
To overcome this, alcohol sales data were used to inflate the survey estimates. The total volume of alcohol sold in Australia was sourced from the apparent consumption of alcohol (ABS 2019). In the ABDS 2018, self-reported daily consumption from the NDSHS, by age and sex, was inflated to match alcohol sales data in each reference year, based on the methods described by Rehm et al. (2010).
The proportion of self-reported lifetime abstainers and ex-drinkers from the NDSHS was assumed to be correct. Among current drinkers, the mean number of standard drinks self‑reported per day was converted into litres of self-reported alcohol consumption for that year. In 2018, the inflation factor was estimated to be 1.46.
Following methods used in Rehm et al. (2010) and in the GBD 2010, 80% of the alcohol available nationally was assumed to have been consumed (Lim et al. 2012). Only a proportion (80%) of alcohol sold in Australia was used, because the total figure includes alcohol discarded due to changes in stocks, alcohol consumed by overseas travellers, alcohol that has been stored or cellared, and alcohol that has been used to prepare food or discarded as waste.
The adjusted litres of alcohol consumed nationally were distributed among self-reported current drinkers using a 2-parameter gamma distribution, which has been found to be the best model to shift the distribution of survey data to fit sales data (Rehm et al. 2010). While this approach brings self-reported alcohol consumption in line with known alcohol sales, a limitation is that it assumes the degree of under-reporting of alcohol consumption is uniform across all age and sex groups. This distribution was used to estimate the proportion of the population who consumed alcohol in categories relevant to the relative risks.
Table 4.49: Alcohol use risk model parameters
Risk factor |
Alcohol use – former drinkers |
---|---|
Disease outcome |
Atrial fibrillation & flutter, bowel cancer, breast cancer, coronary heart disease, epilepsy, hypertensive heart disease, laryngeal cancer, lower respiratory infections, lip & oral cavity cancer, nasopharynx cancer, oesophageal cancer, other oral cavity & pharynx cancers, pancreatitis, stroke |
TMRED |
No alcohol use |
National data source |
NDSHS 2019 |
Units for effect size calculation |
Former drinker |
|
|
Risk factor |
Alcohol use – Average daily alcohol consumption by current drinkers |
Disease outcome |
Atrial fibrillation & flutter, bowel cancer, breast cancer, coronary heart disease, drowning, epilepsy, falls, fire, burns and scalds, homicide and violence, hypertensive heart disease, laryngeal cancer, lip & oral cavity cancer, lower respiratory infections, nasopharynx cancer, oesophageal cancer, other land transport injuries, other oral cavity & pharynx cancers, other unintentional injuries, pancreatitis, road traffic injuries (RTI)—motor vehicle occupants, RTI—motorcyclists, RTI—pedal cyclists, RTI—pedestrians, stroke |
TMRED |
No alcohol use |
National data source |
NDSHS 2019 apparent consumption of alcohol data; |
Units for effect size calculation |
Average consumption of pure alcohol (g per day) |
|
|
Risk factor exposure |
Alcohol use – Alcohol use and dependence |
Disease outcome |
Alcohol use disorders, accidental poisoning, liver cancer, chronic liver disease |
TMRED |
No alcohol use |
National data source |
NMD; GBD 2019 |
Units for effect size calculation |
Direct evidence |
|
|
Risk factor |
Alcohol use – Alcohol dependence |
Disease outcome |
Suicide & self-inflicted injuries |
TMRED |
No alcohol use |
National data source |
ABDS 2018 |
Units for effect size calculation |
Prevalence alcohol use disorders |
Estimates by socioeconomic group
Exposure estimates by socioeconomic group were based directly from the same data source as the national exposure estimates.
2015, 2011 and 2003 estimates
Exposure estimates for 2015 were calculated using data from the NDSHS 2016 and alcohol sales data for 2015, while exposure for 2011 and 2003 were calculated using data from the NDSHS 2010 and 2004 and alcohol sales data for 2011 and 2003, respectively. These followed the method for estimating exposure used for 2018. Direct PAFs were calculated using the method for 2018, which were based on the GBD 2019 estimates for 2015, 2011 and 2003.
Indigenous-specific estimates
Due to the small Aboriginal and Torres Strait Islander sample in the NDSHS (about 460 respondents in 2010), estimates were not considered reliable when broken down by age, sex and amount of alcohol consumed. Instead, the NATSIHS 2004–05, AATSIHS 2012–13 and NATSIHS 2018–19 were used. As alcohol excise, sales and import figures published by the ABS represent a single national figure, it is not possible to calculate Indigenous-specific factors to correct for under-reporting. Therefore, national factors were applied to Indigenous estimates from the NATSIHS and AATSIHS.
References
AIHW 2018. Impact of alcohol and illicit drug use on the burden of disease and injury in Australia: Australian Burden of Disease Study 2011. Australian Burden of Disease Study series no. 17. Cat. no. BOD 19. Canberra: AIHW.
ABS 2019. Apparent alcohol consumption, Australia, 2017–18. ABS cat. no. 4307.0.55.001. Canberra: ABS. Viewed 10 October 2020.
Lim SS, Vos T, Flaxman AD, Danaei G, Shibuya K, Adair-Rohani H et al. 2012. A comparative risk assessment of burden of disease and injury attributable to 67 risk factors and risk factor clusters in 21 regions, 1990–2010: a systematic analysis for the Global Burden of Disease Study 2010. Lancet 380(9859):2224–60.
Rehm J, Kehoe T, Gmel G, Stinson F, Grant B & Gmel G 2010. Statistical modelling of volume of alcohol exposure for epidemiological studies of population health: the US example. Population Health Metrics 8(3):1–12.
Burden due to physical inactivity was estimated in people aged 20 and over.
Population attributable fraction estimated by comparative risk assessment
Exposure estimates
Population exposure to physical inactivity was treated as a categorical variable. The categories describe the range of total activity per week, as measured by the total metabolic equivalent of tasks (METs). This measure encompasses the rate of energy expenditure, with one (1) MET equivalent to 1 kcal/kg/hr, which is about the energy expended in sitting. The higher the MET, the greater the energy expended. The calculation of METs requires the input of:
- time spent undertaking the activity in 1 week (T)
- intensity score for that specific activity (I).
The total MET for each activity is calculated as:
In this study, the total MET score describes the total rate of energy expended across 4 activity domains: leisure, transportation, occupational, and household. The categories included:
- Fewer than 600 METs per week
- 600–1,199 METs per week
- 1,200–1,799 METs per week
- 1,800–2,399 METs per week
- 2,400–2,999 METs per week
- 3,000–3,599 METs per week
- 3,600–4,199 METs per week
- 4,200 METs and over per week.
These categories align with relative risks provided by the GBD 2019 and were used in the ABDS 2018.
The METs for leisure, walking for transport and occupational activity were estimated from the trend in METs reported in successive health surveys, including the NHS 2001, the NHS 2004–05, the NHS 2007–08, the AHS 2011–12, the NHS 2014–15 and the NHS 2017–18. The change over time was used to adjust estimates of exercise in the NHS 2017–18 to represent all other reference years. The number of adjusted self‑reported minutes spent in each activity per week was multiplied by the intensity scores as provided by the AHS 2011–12 to calculate the total MET for each individual in the survey.
The AHS 2011–12 and the NHS 2017–18 do not provide information on the time spent and the intensity of activity due to household chores, so this was obtained from alternative data sources. The time taken on specific household chores was obtained from the ABS Time Use Survey 2006 and this estimate was used in all reference years (ABS 2008). This survey provides detailed information on daily activity patterns of people in Australia and the time allocated to different activities. The time spent undertaking household chores (excluding meal and drink preparation) by sex in 10-year age groups was extracted and multiplied by the conservative intensity of 3.0. The calculated METs by age and sex were added to the calculated METs from remaining domains to provide the total MET.
Average time spent gardening and strengthening and toning were estimated by age and sex using the National Nutrition and Physical Activity Survey (NNPAS) as part of the AHS 2011–12. A trend for the proportion of individuals doing at least some strengthening and toning was estimated using successive health surverys (AHS 2011–12, the NHS 2014–15 and the NHS 2017–18). This trend was used to randomly allocate time spent strengthening and toning to individuals who responded having said they completed at least one day of the activity in the past week. As informing a trend wasn’t possible for gardening (due to data limitations between surveys), proportions of those doing any gardening from the NNPAS in 2011 were applied to all reference years with average time spent gardening then being randomly allocated by age and sex.
Prevalence was estimated from the proportion of people within each activity category once the METs from each domain were summed.
Table 4.50: Physical inactivity risk model parameters
Risk factor |
Physical inactivity – Metabolic equivalent of task (METs) |
---|---|
Disease outcome |
Breast cancer, bowel cancer, coronary heart disease, dementia, type 2 diabetes, stroke, uterine cancer |
TMRED |
All adults experience average 4200 metabolic equivalent of task (METs) per week (highly physically active) |
National data source |
AHS 2011–12; NHS 2017–18 |
Units for effect size calculation |
METs of less than 600, 600–1,999, 1,200–1,799, 1,800–2,399, 2,400–2,999, 3,000–3,599, 3,600–4,199 |
Estimates by socioeconomic group
Exposure estimates by socioeconomic group were based directly from the same data source as the national exposure estimates.
2015, 2011 and 2003 estimates
The number of total METs in 2015, 2011 and 2003 was estimated using the same trend analyses used to estimate METs in 2018. The NHS 2017–18 data were adjusted based on this trend to represent these METs in earlier reference years. Average METs for household chores were the same as in 2018 as no further data were available.
Indigenous specific estimates
Exposure estimates of physical inactivity for the Indigenous population was estimated from the NATSIHS 2018–19 (for 2018 estimates), AATSIHS 2012–13 (for 2011 estimates) and the NATSIHS 2004–05 (for 2003 estimates). It was not possible to adjust these estimates to include stretching and gardening, as this information was not available from the relevant surveys used.
References
ABS 2008. How Australians use their time, 2006. ABS cat. no. 4153.0 Canberra: ABS. Viewed 2 July 2017.
The burden of intimate partner violence was estimated in women aged 15 and over.
The burden was estimated as described further in the report Examination of the burden of disease of intimate partner violence against women in 2011: Final report (Ayre et al. 2016).
This risk factor was estimated in women only as the evidence in the literature used to inform the linked diseases and relative risks was not available for men (AIHW unpublished, Ayre et al. 2016; GBD 2019 Risk Factor Collaborators 2020).
Population attributable fraction estimated with direct evidence
Homicide and violence linked to intimate partner violence was estimated using direct evidence from the National Homicide Monitoring Program (NHMP) for fatal burden, which estimated that 58% of homicides in females were due to an intimate partner in 2018.
Non-fatal burden from homicide and violence due to an intimate partner was estimated directly from the NHMD, using the proportion of hospitalisations (with any principal diagnosis) with an external cause related to assault by an intimate partner (ICD-10-AM codes X85–Y09 with a fifth digit of 0).
Population attributable fraction estimated with comparative risk assessment
Exposure estimates
Exposure to intimate partner violence data were sourced from the PSS 2016 (ABS 2017). It was based on survey respondents aged 18 and over who self-reported intimate partner violence from a cohabiting partner from the age of 15 onwards.
Multiple definitions of exposure to intimate partner violence exist to reflect the complexity of violence against women. This study has been able to include emotional, physical and sexual intimate partner violence by a cohabiting current or previous intimate partner. It was not possible to estimate violence by a non-cohabiting current or previous intimate partner. This is because the PSS 2016 did not include an estimate of emotional abuse by non-cohabiting partners (ABS 2017).
Table 4.51: Intimate partner violence risk model parameters
Risk factor |
Intimate partner violence – Physical, sexual, emotional abuse from a cohabitating partner |
---|---|
Disease outcome |
Anxiety disorders, alcohol use disorders, early pregnancy loss, depressive disorders, homicide and violence, suicide and self-inflicted injuries |
TMRED |
No exposure to intimate partner violence |
National data source |
ABS Personal Safety Survey 2016; National Homicide Monitoring Program |
Units for effect size calculation |
Ever been exposed to intimate partner violence since the age of 15 years (prevalence) |
Estimates by socioeconomic group
Exposure estimates by socioeconomic group were based directly from the same data source as the national exposure estimates.
2015, 2011 and 2003 estimates
The burden due to intimate partner violence in 2015 was estimated using data also from the PSS 2016 (ABS 2017), NHMD hospitalisations in 2015 and the NHMP 2012–2014 (Bryant & Bricknell 2017).
Intimate partner violence burden in 2011 was estimated using data from the PSS 2012 (ABS 2013), NHMD hospitalisations in 2011 and the National Homicide Monitoring Program 2010–2012 (Bryant & Cussen 2015).
Burden due to intimate partner violence in 2003 was estimated using data from the PSS 2005 (ABS 2006), NHMD hospitalisations in 2003 and the National Homicide Monitoring Program annual report 2003–04 (Mouzos 2005). Prevalence of emotional abuse in 2003 was based on the PSS 2012, assuming no trends, as it was not estimated in the PSS 2005 (ABS 2006).
Indigenous specific estimates
For fatal burden due to homicide and violence, direct evidence for Indigenous women was used from the National Homicide Monitoring Program. In 2010–2012, 65% of Indigenous female homicides were classified as perpetrated by an intimate partner (Bryant & Cussen 2015); while for 2003, this was assumed to be 59% based on estimates from 2006–07 (Dearden & Jones 2008).
For the remaining burden, the ABS Personal Safety Survey 2012 did not include an Indigenous identifier, so indirect methods were used to estimate Indigenous exposure to intimate partner violence. A rate ratio of 3.1 was applied to national exposure estimates (AIHW & NIAA 2020). This rate ratio is based on age-standardised rates for 12-month prevalence of physical or threatened violence victimisation reported by females aged 15 years and over, from the 2014 General Social Survey (for national estimates) and the 2014–15 NATSISS (for Indigenous estimates). The same rate ratio was applied to the national exposure estimates to derive Indigenous exposure for both 2003 and 2011.
References
ABS 2006. Personal safety, Australia, 2005 (reissue). ABS cat. no. 4906.0. Canberra: ABS. Viewed 22 March 2018.
ABS 2013. Personal safety, Australia, 2012. ABS cat. no. 4906.0. ABS: Canberra. Viewed 15 August 2015.
ABS 2017. Personal Safety Survey, Australia: user guide, 2016. ABS cat. no. 4906.0.55.003. Canberra: ABS. Viewed 15 March 2018.
AIHW unpublished. Health outcomes of violence: A review of data sources and evidence. Report to the Australian Government Department of Social Services.
AIHW & NIAA (National Indigenous Australians Agency) 2020. Aboriginal and Torres Strait Islander Health Performance Framework: Measure 2.10 Community safety. Canberra: AIHW. Accessed 11 August 2021.
Ayre J, Lum On M, Webster K & Moon L 2016. Examination of the burden of disease of intimate partner violence against women in 2011: final report. Sydney: Australian National Research Organisation for Women’s Safety.
Bryant W & Bricknell S 2017. Homicide in Australia, 2012–13 to 2013–14: National Homicide Monitoring Program report. Statistical report no. 02. Canberra: Australian Institute of Criminology. Viewed 19 September 2018.
Bryant W & Cussen T 2015. Homicide in Australia, 2010–11 to 2011–12: National Homicide Monitoring Program report. Monitoring report no. 23. Canberra: Australian Institute of Criminology. Viewed 22 June 2016.
Dearden J and Jones W 2008. Homicide in Australia: 2006-07 National Homicide Monitoring Program annual report (monitoring reports no. 1). Canberra: Australian Institute of Criminology. Accessed 17 November 2020.
GBD 2019 Risk Factors Collaborators 2020. Global burden of 87 risk factors in 204 countries and territories, 1990–2019: a systematic analysis for the Global Burden of Disease Study 2019. The Lancet 396:1223–49.
Mouzos J 2005. Homicide in Australia: 2003–2004 National Homicide Monitoring Program (NHMP) annual report. Canberra: Australian Institute of Criminology. Viewed 23 August 2016.
Child abuse & neglect included emotional, physical, sexual abuse and neglect. The burden of child abuse & neglect was estimated in people aged 5 and over. The burden for this risk factor was calculated as described in the study by Moore et al. (2015). It was not possible to estimate this risk factor by socioeconomic group as this was not reported by Moore et al. (2015).
Moore et al. (2015) identified Australian studies from the period of 2005 to 2015 that estimated exposure to the different types of abuse in different age groups and by sex. They also identified from the literature an estimate of the proportion of cases that had multiple types of abuse. Using these data together, they estimated the proportion of the Australian population with all types of abuse and neglect, with only a single type, and with different types of abuse in combination.
Moore et al. (2015) also reviewed the literature and identified linked diseases and relative risks for child abuse & neglect. Using these data and comparative risk assessment methodology, they estimated the PAF of each linked disease by age and sex for Australia.
Population attributable fraction
Exposure estimates
Exposure and PAFs were estimated by Moore et al. (2015) by age and sex. The same PAFs were applied to each reference year of this study: 2003, 2011, 2015 and 2018.
Table 4.52: Child abuse & neglect risk model parameters
Risk factor |
Child abuse & neglect – Physical, sexual and emotional abuse and neglect |
---|---|
Disease outcome |
Anxiety disorders, depressive disorders, suicide & self-inflicted injuries |
TMRED |
No child abuse and neglect |
National data source |
Moore et al. 2015 |
Units for effect size calculation |
Prevalence of childhood abuse and neglect |
Indigenous specific estimates
Child protection data on victims of sexual, emotional, physical and neglect abuse were used to estimate the relative difference in the prevalence of child abuse between the Indigenous and the total Australian populations, based on the Overcoming Indigenous Disadvantage 2020 tables (SCRGSP 2020). The ratio was based on nationwide data and applied to national prevalence estimates. PAFs were calculated.
References
Moore SE, Scott JG, Ferrari AJ, Mills R, Dunne MP, Erskine HE et al. 2015. Burden attributable to child maltreatment in Australia. Child Abuse & Neglect 48:208–20.
SCRGSP (Steering Committee for the Review of Government Service Provision) 2020. Overcoming Indigenous Disadvantage: Key Indicators 2020. Canberra: Productivity Commission.
The burden for bullying victimisation was estimated in people aged 10 to 24. It was not possible to estimate burden by socioeconomic group as exposure data by socioeconomic quintile was not available.
Population attributable fraction estimated by comparative risk assessment
Prevalence and relative risks were taken from a recent systematic review and meta-analysis of bullying victimisation among Australian children and adolescents.
Jadambaa et al. (2019a) identified a number of studies estimating exposure of Australian children to bullying victimisation between 1991 and 2015. Various types of bullying were identified, including both traditional and cyberbullying. A meta-analysis was conducted estimating that 15% of Australian children and adolescents were exposed to bullying within the past 12 months. However, this estimate did not distinguish between the types of bullying with the authors noting a strong overlap between the two types, even with cyberbullying only more recently gaining prominence.
This prevalence was applied to relative risks estimated via another meta-analysis (Jadambaa et al 2019b) using the comparative risk assessment methodology to estimate PAFs for anxiety disorders and depressive disorders. The same PAFs were applied to each reference year of this study.
Table 4.53: Bullying victimisation risk model parameters
Risk factor |
Bullying victimisation - Exposure to bullying within the past 12 months |
---|---|
Disease outcome |
Anxiety disorders, depressive disorders |
TMRED |
No bullying victimisation |
National data source |
Jadambaa et al. 2019a, Jadambaa et al. 2019b |
Units for effect size calculation |
Prevalence of bullying victimisation |
Indigenous specific estimates
This risk factor was not estimated for the Indigenous population due to a lack of suitable data to measure exposure. Investigations are underway as to how this measure may be included in future studies.
References
Jadambaa A, Thomas HJ, Scott JG, Graves N, Brain D & Pacella R 2019a. Prevalence of traditional bullying and cyberbullying among children and adolescents in Australia: A systematic review and meta-analysis. Australian & New Zealand Journal of Psychiatry 53:878–888.
Jadambaa A, Thomas HJ, Scott JG, Graves N, Brain D & Pacella R 2019b. The contribution of bullying victimisation to the burden of anxiety and depressive disorders in Australia. Epidemiology and Psychiatric Sciences 29:1–23.
Metabolic/biomedical risk factors
The burden due to overweight (including obesity) was estimated in people aged 5 and over. The methods used for this risk factor are described in detail in the AIHW publication Impact of overweight and obesity as a risk factor for chronic conditions: Australian Burden of Disease Study (AIHW 2017).
Population attributable fraction
Exposure estimates
Age- and sex-specific data were extracted in the finest possible increments from a continuous high body mass distribution for the Australian population based on measurements of height and weight from the NHS 2017–18. For children and adolescents aged 5–14, age- and sex-specific BMI cut-off levels indicating overweight (including obesity) were derived from the study by Cole et al. (2000).
Relative risks
The relative risks used were largely based on those published by the GBD 2019, including atrial fibrillation & flutter, cataract, non-Hodgkin lymphoma and multiple myeloma. Other relative risks were based on work by the AIHW (AIHW 2017). For dementia and gallbladder and bile duct disease, relative risks from the GBD 2019 were used instead of relative risks from the AIHW as they were based on a more recent meta-analysis.
Table 4.54: Overweight (including obesity) risk model parameters
Risk factor |
Overweight, obese – Body mass index BMI |
---|---|
Disease outcome |
Acute lymphoblastic leukaemia, acute myeloid leukaemia, asthma, atrial fibrillation & flutter, back pain & problems, bowel cancer, breast cancer, cataract & other lens disorders, chronic kidney disease, chronic lymphocytic leukaemia, chronic myeloid leukaemia, coronary heart disease, dementia, gallbladder and bile duct disease, gallbladder cancer, gout, hypertensive heart disease, kidney cancer, liver cancer, myeloma, non-Hodgkin lymphoma, oesophageal cancer, osteoarthritis, other leukaemias, ovarian cancer, pancreatic cancer, stroke, thyroid cancer, type 2 diabetes, uterine cancer |
TMRED |
Body mass index between 20 and 25 BMI |
National data source |
NHS 2017–18 |
Units for effect size calculation |
Per 5 BMI |
Estimates by socioeconomic group
It was not possible to aggregate risk factor exposure data by socioeconomic group with acceptable RSEs; therefore, exposure was estimated based on the difference in the mean estimate in each quintile as described in Overarching methods and choices for risk factors.
2015, 2011 and 2003 estimates
Exposure for 2011 and 2015 were estimated as described above, using data from the AHS 2011–12 and NHS 2014–15, respectively.
For people aged 20 and over, prevalence by BMI category, age and sex was estimated for the time-point 2003 by using the trends in the prevalence of the distribution of BMI from the 3 successive health surveys (the NHS 2007–08, the AHS 2011–12 and the NHS 2014–15) as described in AIHW 2017.
For people aged 5–19, prevalence by BMI category, age and sex was estimated for the time-point 2003, using the NHS 2007–08. The estimate of prevalence of obesity in people aged 5–19 decreased slightly from the NHS 2007–08 to the AHS 2011–12, and from the AHS 2011–12 to the NHS 2014–15, but these differences were not statistically significant. Due to this, the trend from the 3 successive health surveys (the NHS 2007–08, the AHS 2011–12 and the NHS 2014–15) were not considered accurate for this age group when compared with the 1995 National Nutrition Survey estimates.
Indigenous specific estimates
Exposure for 2018 was estimated as described for the national estimates and was based on measurements of height and weight from the NATSIHS 2018–19.
Exposure for 2011 was estimated as the distribution of body mass index in Indigenous Australians from the AATSIHS 2012–13. The 2003 estimates were calculated using the same method as described for national estimates by comparing the trend in mean body mass index in 2011 to that estimated for Indigenous Australians in the 2003 Indigenous Australian Burden of disease study (Vos et al. 2007).
The 2003 estimates were based on data from the 2001 NATSIHS, which used measured height and weight information to estimate mean body mass index for Indigenous Australians living in remote areas. The relative difference between self-reported and measured body mass index were assumed to be the same in Indigenous Australians living in remote and non-remote areas, and was applied to the mean self-reported body mass index for Indigenous Australians living in non-remote areas (Vos et al. 2007).
References
AIHW 2017. Impact of overweight and obesity as a risk factor for chronic conditions: Australian Burden of Disease Study. Australian Burden of Disease Study series no. 11. Cat. no. BOD 12. Canberra: AIHW.
Cole TJ, Bellizzi MC, Flegal KM & Dietz WH 2000. Establishing a standard definition for child overweight and obesity worldwide: international survey. British Medical Journal 320:1240–3.
Vos T, Barker B, Stanley L & Lopez AD 2007. The burden of disease and injury in Aboriginal and Torres Strait Islander peoples 2003. Brisbane: University of Queensland.
The burden attributable to high blood pressure was estimated in people aged 25 and over.
Population attributable fraction
Exposure estimates
Age- and sex-specific data were extracted in the finest possible increments from a continuous systolic blood pressure distribution for the Australian population based on blood pressure measurements from the NHS 2017–18 (ABS 2019).
Table 4.55: High blood pressure risk model parameters
Risk factor |
High blood pressure – Systolic blood pressure |
---|---|
Disease outcome |
Aortic aneurysm, atrial fibrillation & flutter, cardiomyopathy, chronic kidney disease, coronary heart disease, dementia, hypertensive heart disease, inflammatory heart disease, non-rheumatic valvular disease , peripheral vascular disease, rheumatic heart disease, stroke |
TMRED |
Systolic blood pressure between 110–115 mmHg |
National data source |
NHS 2017–18 |
Units for effect size calculation |
Per 10 mmHg of systolic blood pressure increase |
Estimates by socioeconomic group
Exposure to high blood pressure by socioeconomic group was based on the same data source as for the national exposure estimates. It was not possible to aggregate risk factor exposure data by socioeconomic group to generate acceptable RSEs; therefore, exposure was estimated based on the difference in the mean estimate in each quintile as described in Overarching methods and choices for risk factors.
2015, 2011 and 2003 estimates
Exposure data for 2015 was sourced from the NHS 2014–15 using the same method as for 2018. For 2011, data were sourced directly from the AHS 2011–12.
For 2003 estimates, the exposure to high blood pressure in 2003 was calculated by comparing the mean exposure from the Australian Diabetes, Obesity and Lifestyle Study (AusDiab) 1999–2000 and the mean exposure from the AHS 2011–12 by age and sex (Begg et al. 2007). Record level data from the AHS 2011–12 were adjusted by the percentage change in the mean from 2011 to 2003. The adjusted unit record data were used to estimate the distribution of exposure to high blood pressure in 2003.
Indigenous specific estimates
Exposure for 2018 was estimated using the same methods as described for the national estimates using systolic blood pressure distribution extracted from the NATSIHS 2018–19.
Exposure for 2011 was estimated as the distribution of blood pressure in Indigenous Australians from the AATSIHS 2012–13. The 2003 estimates were calculated using the same method as described for national estimates by comparing the trend in mean exposure in 2011 to exposure estimated in the 2003 Indigenous Australian burden of disease study (Vos et al. 2007).
The 2003 estimates were based on data published in the 2003 Australian burden of disease study (Vos et al. 2007). These data are from relatively small studies covering 2 regions (the DRUID study by Cunningham et al. 2006; Wang & Hoy 2003), in which it was assumed that the measured systolic blood pressure mean and standard deviations were representative of Indigenous Australians living in non-remote and remote areas.
References
ABS 2019. National Health Survey: users’ guide, 2017–18. ABS cat. no. 4363.0. Canberra: ABS. Viewed 21 August 2017.
Begg S, Vos T, Barker B, Stevenson C, Stanley L & Lopez AD 2007. The burden of disease and injury in Australia 2003. Cat. no. PHE 82. Canberra: AIHW.
Cunningham J, O'Dea K, Dunbar T, Weeramanthri T, Zimmet P & Shaw J 2006. Study protocol--diabetes and related conditions in urban indigenous people in the Darwin, Australia region: aims, methods and participation in the DRUID Study. BMC Public Health 6:8.
Vos T, Barker B, Stanley L & Lopez AD 2007. The burden of disease and injury in Aboriginal and Torres Strait Islander peoples 2003. Brisbane: University of Queensland.
Wang Z & Hoy WE 2003. Hypertension, dyslipidemia, body mass index, diabetes and smoking status in Aboriginal Australians in a remote community. Ethnicity & disease 13(3):324–30.
The burden attributable to high cholesterol was estimated in people aged 25 and over. In ABDS 2018, the risk factor was updated to be as a measure of low-density lipoprotein (LDL) cholesterol to align with the methods for GBD 2019 (GBD 2019 Risk Factors Collaborators 2020).
Population attributable fraction
Exposure estimates
Age- and sex-specific data were extracted in the finest possible increments from a continuous measured LDL cholesterol distribution for the Australian population from the AHS 2011–12.
The exposure to high cholesterol in 2018 was calculated by comparing the mean exposure of total cholesterol from the AusDiab 1999–2000 and the mean exposure from the AHS 2011–12 by age and sex (Begg et al. 2007). Record level data from the AHS 2011–12 were adjusted by the percentage change in the mean that would be expected between the years 2011 to 2018. The adjusted unit record data were used to estimate the distribution of exposure to high cholesterol in 2018.
Table 4.56: High cholesterol risk model parameters
Risk factor |
High cholesterol – Low-density lipoprotein (LDL) cholesterol |
---|---|
Disease outcome |
Coronary heart disease, stroke |
TMRED |
LDL cholesterol between 0.7–1.3 mmol/L |
National data source |
AHS 2011–12 |
Units for effect size calculation |
Per 1 mmol/L of LDL cholesterol increase |
Estimates by socioeconomic group
Exposure to high cholesterol was estimated using data from the AHS 2011–12, modelled to 2018, and the difference in the mean estimate in each quintile as described in Overarching methods and choices for risk factors.
2015, 2011 and 2003 estimates
The prevalence of total cholesterol for 2011 was estimated using data from the AHS 2011–12.
The same trend described here for 2018 was used to estimate prevalence in total cholesterol in 2015 and 2003.
Indigenous specific estimates
Exposure in 2003, 2011 and 2018 was estimated as the distribution of total blood cholesterol levels in Indigenous Australians from the AATSIHS 2012–13.
References
Begg S, Vos T, Barker B, Stevenson C, Stanley L & Lopez AD 2007. The burden of disease and injury in Australia 2003. Cat. no. PHE 82. Canberra: AIHW.
GBD 2019 Risk Factors Collaborators 2020. Global burden of 87 risk factors in 204 countries and territories, 1990–2019: a systematic analysis for the Global Burden of Disease Study 2019. The Lancet 396:1223–49.
The burden attributable to high blood plasma glucose was estimated in people of all ages. The risk factor includes estimates of the burden due to intermediate hyperglycaemia and diabetes. Burden due to this risk factor was not estimated for the 2003 reference year as there were no data on trends of blood plasma glucose between 2003 and 2011.
Population attributable fraction using direct evidence
All types of diabetes were entirely attributable to high blood plasma glucose (PAF of 1) as high blood plasma glucose is a diagnostic criteria for all types of diabetes.
Chronic kidney disease due to high blood plasma glucose
The method for attributing the amount of chronic kidney disease due to diabetes is based on the GBD 2019 and involves a 2-step approach:
- The proportion of the GBD cause ‘chronic kidney disease due to diabetes’ of the total GBD cause ‘chronic kidney disease’ in the GBD 2019 (14%) was used to estimate the direct PAF of chronic kidney disease due to high blood plasma glucose (GBD 2019 Risk Factors Collaborators 2020).
-
Exposure to high blood plasma glucose is linked to the remaining 86% of chronic kidney disease not attributed in step 1 as described later in this section. Part of this remaining proportion (the GBD causes ‘chronic kidney disease due to hypertension, glomerulonephritis or other and unspecified causes’) is attributed to high blood plasma glucose, using the comparative risk assessment method.
Population attributable fraction using comparative risk assessment
Exposure estimates
Exposure to high plasma glucose included 2 parts: the population distribution of blood plasma glucose levels (continuous risk model) and the proportion of the population with diabetes (categorical risk model). Each of these exposures was linked to different diseases (see High blood plasma glucose risk model parameters below).
To estimate and report the burden attributable by intermediate hyperglycaemia and diabetes, the continuous distribution of high blood plasma glucose was divided into the following categories:
- exposure to 4.9 to 6.9 mmol/L high plasma glucose was attributable to intermediate hyperglycaemia. This range was defined by the GBD TMRED of 4.9 mmol/L and expert advice for the 6.9 mmol/L cut-off
- burden due to blood plasma glucose of 7 mmol/L or more was attributable to diabetes in addition to the attributable burden estimated from exposure to diabetes.
High blood plasma glucose
Age- and sex-specific data were extracted in the finest possible increments from a continuous fasting blood plasma glucose distribution for the Australian population from the AHS 2011–12. As no data were available to inform trends, this estimate was also applied in 2015 and 2018.
Diabetes
The prevalence of diabetes was based on the prevalence of type 1, type 2 and other diabetes in 2018. All types of diabetes are included because people exposed to all types of diabetes are at risk of the disease outcomes identified, and the risk factor is modifiable.
Table 4.57: High blood plasma glucose risk model parameters
Risk factor |
Intermediate hyperglycaemia; diabetes – High fasting plasma glucose |
---|---|
Disease outcome |
Chronic kidney disease, coronary heart disease, stroke |
TMRED |
Blood plasma glucose 4.8–5.4 mmol/L |
National data source |
AHS 2011–12 |
Units for effect size calculation |
Per 1 mmol/L of fasting plasma glucose increase |
|
|
Risk factor |
Diabetes – Diabetes prevalence |
Disease outcome |
Bladder cancer, bowel cancer, breast cancer, cataract & other lens disorders, chronic kidney disease, coronary heart disease, dementia, glaucoma, liver cancer, lung cancer, ovarian cancer, pancreatic cancer, peripheral vascular disease |
TMRED |
No diabetes |
National data source |
ABDS 2018 |
Units for effect size calculation |
Prevalence of type 1, type 2 and other diabetes |
|
|
Risk factor |
Diabetes – Direct PAFs |
Disease outcome |
Chronic kidney disease, type 2 diabetes, type 1 diabetes, other diabetes |
TMRED |
No diabetes |
National data source |
GBD 2019 |
Units for effect size calculation |
Direct evidence |
Estimates by socioeconomic group
Exposure estimates by socioeconomic group were calculated directly from the same data source as for the national exposure estimates.
For high blood plasma glucose, it was not possible to aggregate risk factor exposure data by socioeconomic group with acceptable RSEs; therefore, exposure was estimated based on the difference in the mean estimate in each quintile as described in Overarching methods and choices for risk factors.
2015, 2011 estimates
The prevalence of high blood plasma glucose in 2011 was estimated using measured data from the AHS 2011–12. As mentioned above, these estimates were also applied for 2015.
It was not possible to estimate this risk factor in 2003 because there were no data available to estimate the trend in high blood plasma glucose.
Indigenous specific estimates
Exposure in 2011 and 2018 was estimated as the distribution of fasting plasma glucose levels in Indigenous Australians from the AATSIHS 2012–13.
References
GBD 2019 Risk Factors Collaborators 2020. Global burden of 87 risk factors in 204 countries and territories, 1990–2019: a systematic analysis for the Global Burden of Disease Study 2019. The Lancet 396:1223–49.
These PAFs were estimated using direct evidence for people of all ages. Iron deficiency anaemia is the only disease linked to iron deficiency and was 100% attributable to this risk factor (PAF of 1). The method was the same in all 4 years.
Table 5.58: Iron deficiency risk model parameters
Risk factor |
Iron deficiency |
---|---|
Disease outcome |
Iron deficiency anaemia |
TMRED |
No Iron deficiency anaemia |
National data source |
n.a. |
Units for effect size calculation |
All of iron deficiency anaemia is attributable |
Estimates by socioeconomic group
An estimate by socioeconomic group was not included to match other risk factors where exposure does not change by socioeconomic group.
Indigenous specific estimates
The same methods as for the national population were used.
Burden due to low birthweight & short gestation was estimated in people of all ages. The risk factor represents the combined impact of being born of low weight and prematurely and not as separate risk factors. Due to data limitations, this risk factor was only estimated for the reference year 2018.
Population attributable fraction estimated by comparative risk assessment
Exposure estimates
Exposure estimates were obtained using the National Perinatal Data Collection (NPDC), which contains data on all live births and stillbirths of at least 20 weeks gestation or 400 grams birthweight, and National Perinatal Mortality Data Collection (NPMDC) which contains data on all stillbirths and neonatal deaths of at least 20 weeks gestation or 400 grams birthweight. The number of deaths was obtained within categories of birthweight and gestational age, and also disaggregated by whether they occurred within the early or late neonatal period so as to correspond with relative risks provided by the GBD 2019 (see Table 4.59).
Table 4.59: Categories of exposure to gestation and birthweight and TMREDs from GBD 2019 study
Gestation (weeks) |
Birthweight (grams) |
[0, 24) |
0–499, 500–999 |
[24, 26) |
500–999 |
[26, 28) |
500–999, 1000–1499 |
[28, 30) |
500–999, 1000–1499, 1500–1999, 2000–2499, 2500–2999, 3000–3499 |
[30, 32) |
500–999, 1000–1499, 1500–1999, 2000–2499, 2500–2999, 3000–3499, 3500–3999 |
[32, 34) |
1000–1499, 1500–1999, 2000–2499, 2500–2999, 3000–3499, 3500–3999 |
[34, 36) |
1000–1499, 1500–1999, 2000–2499, 2500–2999, 3000–3499, 3500–3999, 4000–4499 |
[36, 37) |
1000–1499, 1500–1999, 2000–2499, 2500–2999, 3000–3499, 3500–3999, 4000–4499 |
[37, 38) |
1000–1499, 1500–1999, 2000–2499, 2500–2999, 3000–3499, 3500–3999, 4000–4499 |
[38, 40) |
1000–1499, 1500–1999, 2000–2499, 2500–2999, 3000–3499, 3500–3999, 4000–4499 |
[40, 42) |
1500–1999, 2000–2499, 2500–2999, 3000–3499, 3500–3999, 4000–4499 |
TMRED | |
[38, 40] | 3000–3499, 3500–3999, 4000–4499 |
[40-42] | 3000–3499, 3500–3999, 4000–4499 |
Source: GBD 2019.
PAFs were estimated using the comparative risk assessment method using NPDC and NPMDC exposure data for the reference year 2018. Relative risks and linked diseases were obtained from the GBD 2019 though not all were deemed appropriate within the Australian context following expert advice. Pre-term birth and low birthweight complications was the only linked disease that was attributed entirely to the risk factor and applied to people of all ages.
Table 4.60: Low birthweight & short gestation risk model parameters
Risk factor |
Low birthweight & short gestation – Birthweight, gestational age |
---|---|
Disease outcome |
Birth trauma & asphyxia, haemophilus influenza type-B, lower respiratory infections, meningococcal disease, neonatal infections, other disorders of infancy, other gastrointestinal diseases, other meningitis and encephalitis, otitis media, pneumococcal disease, pre-term birth & low birthweight complications, rotavirus, salmonellosis, sudden infant death syndrome, upper respiratory infections |
TMRED |
Birthweight ≥ 3000 g and gestational age ≥ 38 weeks |
National data source |
NPDC and NPMDC 2018 |
Units for effect size calculation |
Prevalence of neonatal deaths by birthweight and gestational age categories |
Estimates by socioeconomic group
Exposure estimates by socioeconomic group were calculated directly from the same data source as for the national exposure estimates.
Indigenous specific estimates
For Indigenous estimates for low birthweight & short gestation, the same methods and exposure data sources were used as for national estimates.
The burden attributable to impaired kidney function was estimated in people aged 25 and over.
Population attributable fraction
Exposure estimates
Chronic kidney disease stages 1–3
Age- and sex-specific data were extracted in the finest possible increments from the estimate of stages 1, 2 and 3 chronic kidney disease for the Australian population from the AHS 2011–12.
To estimate prevalence in the year 2018, the AIHW analysis of trends in stages 3–5 chronic kidney disease prevalence from the 1999–2000 AusDiab compared with the AHS 2011–12 in the broad age groups was used (AIHW 2018). The age and sex distribution for stage 3 chronic kidney disease were further refined using the age and sex of people who were hospitalised for stage 3 chronic kidney disease (N18.3) in 2018.
Chronic kidney disease stages 4–5
The prevalence of stage 4 and 5 (end-stage) chronic kidney disease was estimated as the prevalence for the relevant sequelae (stage 4 chronic kidney disease, end-stage chronic kidney disease treated with dialysis or transplant) for the cause chronic kidney disease in the ABDS 2018. The methods for these sequelae are described for the cause chronic kidney disease.
Table 4.61: Impaired kidney function risk model parameters
Risk factor |
Chronic kidney disease stage 1–3 – stages 1–2 |
---|---|
Disease outcome |
Coronary heart disease, dementia, gout, peripheral vascular disease, stroke |
TMRED |
No chronic kidney disease |
National data source |
AHS 2011–12 |
Units for effect size calculation |
Prevalence of chronic kidney disease stages 1–2 |
|
|
Risk factor |
Chronic kidney disease stage 1–3 – stage 3 |
Disease outcome |
Coronary heart disease, dementia, gout, peripheral vascular disease, stroke, chronic kidney disease |
TMRED |
No chronic kidney disease |
National data source |
AHS 2011–12 |
Units for effect size calculation |
Prevalence of chronic kidney disease stage 3 |
|
|
Risk factor |
Chronic kidney disease stage 4–5 |
Disease outcome |
Coronary heart disease, dementia, gout, peripheral vascular disease, stroke, chronic kidney disease |
TMRED |
No chronic kidney disease |
National data source |
AHS 2011–12; ANZDATA for ABDS 2018 |
Units for effect size calculation |
Prevalence of chronic kidney disease stage 4–5 |
Estimates by socioeconomic group
Exposure to stages 1–3 chronic kidney disease by socioeconomic group was estimated using data from the AHS 2011–12, modelled to 2018 and grouped into broad age- and sex‑groups.
Exposure of stages 4–5 chronic kidney disease by socioeconomic group was sourced as described for the cause chronic kidney disease.
2015, 2011 and 2003 estimates
Chronic kidney disease stages 1–3
The prevalence of stages 1, 2 and 3 chronic kidney disease for 2011 was estimated using data from the AHS 2011–12. To estimate prevalence in the years 2003 and 2015, the AIHW analysis of trends in stages 3–5 chronic kidney disease prevalence from the 1999–2000 AusDiab compared with the AHS 2011–12 in the broad age groups was used (AIHW 2018).
Chronic kidney disease stages 4–5
The prevalence of stages 4–5 chronic kidney disease was sourced as described for the cause chronic kidney disease for the ABDS 2018 (see Disease-specific methods - morbidity).
Indigenous specific estimates
Exposure to stages 1–3 chronic kidney disease for Indigenous Australians was estimated using data from the AATSIHS 2012–13, modelled to 2003 and 2018 and grouped into broad age- and sex‑groups.
Exposure of stages 4–5 chronic kidney disease for Indigenous Australians was sourced as described for the cause chronic kidney disease.
References
AIHW 2018. Chronic kidney disease prevalence among Australian adults over time. Cardiovascular, diabetes and chronic kidney disease series no. 6. Cat. no. CDK 6. Canberra: AIHW.
The burden attributable to low bone mineral density was measured in people aged 40 and over. Exposure by socioeconomic group was not estimated due to lack of available data.
Population attributable fraction
Exposure estimates
Self-reported prevalence of osteoporosis underestimates the true community prevalence of the condition, as many individuals with low bone mineral density display no overt symptoms and are therefore undiagnosed.
Exposure data were sourced from the 2001–06 wave of the Geelong Osteoporosis Study (Henry et al. 2010). Mean bone mineral density at the femoral neck, by age and sex, was used to model exposure distributions, assuming a normal distribution and following methods described by Sànchez-Riera et al. (2014).
Table 4.62: Low bone mineral density risk model parameters
Risk factor |
Low bone mineral density |
---|---|
Disease outcome |
Falls |
TMRED |
95th percentile of the Third National Health and Nutrition Examination Survey (NHANES-III) cohort by age (Looker et al. 2012) |
National data source |
Geelong Osteoporosis Study (Barwon Health) |
Units for effect size calculation |
Standardised bone mineral density at the femoral neck |
2015, 2011 and 2003 estimates
Methods for estimating exposure and calculating the PAFs for the 2015, 2011 and 2003 reference year were the same as those used for 2018.
Indigenous specific estimates
Standardised bone mineral density measurements at the femoral neck were not available for the Indigenous population. National rates of low bone mineral density, by age and sex, were applied to the Indigenous population to calculate Indigenous exposure estimates. This approach was supported by the same rates of self-reported osteoporosis for the Indigenous and national populations reported in the AATSIHS 2012–13 and the AHS 2011–12.
References
Henry MJ, Pasco JA, Korn S, Gibson JE, Kotowicz MA & Nicholson GC 2010. Bone mineral density reference ranges for Australian men: Geelong Osteoporosis Study. Osteoporosis International 21(6):909–17.
Looker AC, Borrud LG, Hughes JP, Fan B, Shepherd JA, Melton LJ et al. 2012. Lumbar spine and proximal femur bone mineral density, bone mineral content, and bone area: United States 2005–2008. National Center for Health Statistics, Vital Health Stat 11(251):1-132.
Sànchez-Riera L, Carnahan E, Vos T, Veerman L, Norman R, Lim SS et al. 2014. The global burden attributable to low bone mineral density. Annals of Rheumatic Diseases 73:1635–45.
Environmental risk factors
The fatal burden attributable to air pollution was measured by concentration of particulate matter
2.5 μg/m3 (PM2.5) in Australia in people of all ages. It was not possible to estimate exposure to this risk factor in 2011 or 2003 because comparable exposure data were not available. An estimate by socioeconomic group was not included to match other risk factors where exposure does not change by socioeconomic group.
Population attributable fraction
Exposure estimates
PM2.5 are particles suspended in the air with a diameter in a specified size range, 0–2.5 microns. Average annual PM2.5 data by mesh block geography from the Centre for Air pollution, energy and health Research (CAR) was estimated using satellite-calibrated ground monitoring stations for 2015 and 2018 and was provided via personal communication (L Knibbs 2020, pers. comm., 24 July 2020).
Geographic correspondence files were used to convert this satellite-modelled data by mesh blocks to SA2 geography. This was then aggregated and population-weighted to estimate national exposure to air pollution by age and sex.
Satellite modelling is limited in that it measures ambient air pollution levels rather than actual exposure to air pollution, but has the advantage over previous methods using monitoring stations only in that estimates are based on measurements from larger areas of Australia and are calibrated by ground monitoring stations. However, there are the same issues in that there can be variation in estimated levels of air pollution and actual levels experienced by the population. There is also likely to be a substantial amount of variation between sites in the amount of time that people generally spend outside, being exposed to air pollution.
Table 4.63: Air pollution risk model parameters
Risk factor |
Air pollution – Particulate matter (2.5 µg/m3) |
---|---|
Disease outcome |
COPD, coronary heart disease, lower respiratory infections, lung cancer, stroke, type 2 diabetes |
TMRED |
2.4–5.9 μg/m3 (PM2.5) |
National data source |
Satellite-based model data |
Units for effect size calculation |
Daily maximum atmospheric particulate matter (PM2.5) |
Indigenous specific estimates
For Indigenous estimates for air pollution, the same methods and exposure data sources were used as for national estimates though aggregated exposure data was weighted to the Indigenous population instead.
The burden attributable to sun exposure was estimated in people of all ages using direct evidence. The direct PAFs used here are a proportion of current burden due to past and current sun exposure in the population. An estimate by socioeconomic group was not included to match other risk factors where exposure does not change by socioeconomic group.
Population attributable fractions using direct evidence
The PAFs for sun exposure were calculated by collaborating experts Robyn Lucas and Fan Xiang from the National Centre for Epidemiology and Population Health at the Australian National University. The melanoma PAFs appropriate for Australia were advised to be the upper estimate of 0.9 from the global study on the burden of disease from solar ultraviolet radiation (Lucas et al. 2006). The squamous cell carcinoma and basal cell carcinoma PAFs were calculated using the comparative risk assessment approach, based on levels of ultraviolet exposure in Australia (F Xiang 2015, pers. comm., 11 November 2015).
Table 4.64: High sun exposure risk model parameters
Risk factor |
Sun exposure |
---|---|
Disease outcome |
Melanoma, non-melanoma skin cancer |
TMRED |
No health outcomes from sun exposure |
National data source |
Lucas et al. 2006; F Xiang 2015, pers. comm., 11 November 2015 |
Units for effect size calculation |
Direct evidence |
2015, 2011 and 2003 estimates
The same PAFs were used in 2015, 2011 and 2003 as they were not specific to 2018 but based on latitude.
Indigenous specific estimates
The burden from high sun exposure was not estimated for the Indigenous population as it was not possible to account for the impact of differences in skin melanin levels.
References
Lucas R, McMichael T, Smith W & Armstrong B 2006. Solar ultraviolet radiation: global burden of disease from solar ultraviolet radiation. Environmental burden of disease series. Geneva: World Health Organization.
Occupational exposures and hazards captured the impact of exposure to 13 carcinogens (asbestos, arsenic, benzene, beryllium, cadmium, chromium, diesel engine exhaust, second‑hand smoke, formaldehyde, nickel, polycyclic aromatic hydrocarbons, silica and sulphuric acid), asthmagens, noise, ergonomic stressors, injury, particulate matter, and gases and fumes in the workplace.
Population attributable fraction from direct evidence
The PAFs for injuries were estimated directly from data collected by Safe Work Australia. For all other disease outcomes, the PAFs were estimated from exposure to working in various industries or occupations.
All pneumoconiosis was attributable to occupational exposure as informed by expert advice (T Driscoll 2015, pers. comm., 24 June 2016). As per the disease group methods, pneumoconiosis was split into its component seqeulae of silicosis, asbestosis and other pneumoconiosis for ABDS 2018.
For injuries, direct evidence was sourced from Safe Work Australia, including data on the number of deaths occurring at work (Safe Work Australia 2019) and the number of workers’ compensation injury claims annually (Safe Work Australia 2020). Counts of deaths and injuries, with some disaggregation by age, sex and nature or external cause of injury, were used to directly calculate PAFs.
The PAFs for fatal burden were estimated by the number of deaths occurring at work compared with the total number of deaths due to injuries in the broader population.
The data for non-fatal burden are limited in that compensation claims will capture only injuries that require more than 1 week away from work and are fairly severe. They will also not include people who are self-employed. These PAFs were estimated for people aged 15 and over.
The PAFs for non-fatal burden were estimated by the number of injuries reported at work in 2018 from Safe Work Australia (2019) divided by the incidence of admitted and non-admitted hospitalisations and emergency department presentations in the NHMD in 2018.
Population attributable fraction by comparative risk assessment
Exposure estimates
To estimate the number of people working in Australia—the economically active population—by age, sex and industry or occupation, was estimated from the Labour Force Survey (ABS 2020).
Industry
Exposure to working in certain types of industry was linked to various cancers, hearing loss and COPD (see Occupational exposure risk model parameters below). This is because working in these industries is known to expose a proportion of the workforce to carcinogens, noise, particulate matter, gases and fumes as estimated by the Carcinogen Exposure (CAREX) research project (Kauppinen et al. 2000).
The working population was distributed across 9 broad industry types (agriculture, hunting, forestry and fishing; mining and quarrying; wholesale, retail trade, restaurants and hotels; manufacturing; electricity, gas and water; transport, storage and communication; construction; finance, insurance, real estate and business services; community, social and personal services) based on the 2016 Census of Population and Housing.
A severity distribution from the GBD 2010 was applied to obtain the proportion of people working in these industries exposed to high and low levels of noise, and to high and low levels of particulate matter, gases and fumes. The PAFs were calculated for people aged 15–74.
To account for the latency period between exposure and the symptoms of cancer, an ‘occupational turnover rate’ was applied to the number of people working in these industries. The occupational turnover rate adjusts for annual worker turnover, mortality rates and past trends by industry, to estimate past exposure to carcinogens in each industry. These factors are based on trends observed in the United Kingdom.
Data from the Carcinogen Exposure (CAREX) research project produces estimates of the proportion of workers in each industry who will be exposed to specific carcinogens (Kauppinen et al. 2000). These proportions, which are based on data from the European Union and Canada, are then applied for each of the industries described earlier. The PAFs for carcinogens were calculated for people aged over 15.
Occupation
Exposure to types of occupations was linked to asthma and low back pain (see Occupational exposure risk model parameters below). This is because working in these occupations is known to expose a proportion of the workforce to asthmagens and ergonomic stressors.
The number of working people was apportioned by 8 broad occupational groups (professional, technical and related workers; administrative and managerial workers; clerical and related workers; sales workers; service workers; agricultural, animal husbandry and forestry workers; fishermen and hunters; production and related workers; transport equipment operators and labourers) based on the 2016 Census of Population and Housing (ABS 2017).
Exposure to working in these occupations was used to estimate the PAFs in people aged 15–64 and no severity distribution was applied.
Table 4.65: Occupational exposure risk model parameters
Risk factor |
Occupational exposures and hazards – Occupational injuries |
---|---|
Disease outcome |
Drowning; falls; fire, burns and scalds; homicide and violence; road traffic injuries—motor vehicle occupants; road traffic injuries—motorcyclists; other unintentional injuries; other land transport injuries |
TMRED |
No occupational injuries |
National data source |
Work-related Traumatic Injury Fatalities, Australia 2018; Workers Compensation Statistics 2017–18 |
Units for effect size calculation |
Direct evidence: number of workplace fatalities and the number of workers compensation claims for injuries |
|
|
Risk factor |
Occupational exposures and hazards – Occupational exposure to benzene or formaldehyde |
Disease outcome |
Acute lymphoblastic leukaemia, acute myeloid leukaemia, chronic lymphocytic leukaemia, chronic myeloid leukaemia, other leukaemias, nasopharyngeal cancer |
TMRED |
No occupational exposure to benzene or formaldehyde |
National data source |
Census of Population and Housing 2016; ABS Labour force survey, April 2020 |
Units for effect size calculation |
Distribution of the labour force by broad industry type |
|
|
Risk factor |
Occupational exposures and hazards – Occupational exposure to arsenic, beryllium, cadmium, chromium, diesel engine exhaust, polycyclic aromatic hydrocarbons, nickel, second-hand smoke, silica |
Disease outcome |
Lung cancer |
TMRED |
No occupational exposure to arsenic, beryllium, cadmium chromium, diesel engine exhaust, polycyclic aromatic hydrocarbons, nickel, second-hand smoke, silica |
National data source |
Census of Population and Housing 2016; ABS Labour force survey, April 2020 |
Units for effect size calculation |
Distribution of the labour force by broad industry type |
|
|
Risk factor |
Occupational exposures and hazards – Occupational exposure to asbestos, silicone and other particulate matter |
Disease outcome |
Asbestosis, silicosis, other pneumoconiosis |
TMRED |
No occupational exposure to asbestos, silicone and other particulate matter |
National data source |
GBD 2019 |
Units for effect size calculation |
Direct evidence |
|
|
Risk factor |
Occupational exposures and hazards – Occupational exposure to sulphuric acid |
Disease outcome |
Laryngeal cancer |
TMRED |
No occupational exposure to sulphuric acid |
National data source |
Census of Population and Housing 2016; ABS Labour force survey, April 2020 |
Units for effect size calculation |
Distribution of the labour force by broad industry type |
|
|
Risk factor |
Occupational exposures and hazards – Occupational exposure to trichloroethylene |
Disease outcome |
Kidney cancer |
TMRED |
No occupational exposure to trichloroethylene |
National data source |
Census of Population and Housing 2016; ABS Labour force survey, April 2020 |
Units for effect size calculation |
Distribution of the labour force by broad industry type |
|
|
Risk factor |
Occupational exposures and hazards – Occupational exposure to particulate matter, gas and fumes |
Disease outcome |
COPD |
TMRED |
No occupational exposure to particulate matter, gas and fumes |
National data source |
Census of Population and Housing 2016; ABS Labour force survey, April 2020 |
Units for effect size calculation |
Distribution of the labour force by broad industry type |
|
|
Risk factor |
Occupational exposures and hazards – Occupational exposure to asbestos |
Disease outcome |
Laryngeal cancer, lung cancer, mesothelioma, ovarian cancer |
TMRED |
No occupational exposure to asbestos |
National data source |
Census of Population and Housing 2016; ABS Labour force survey, April 2020 |
Units for effect size calculation |
Distribution of the labour force by broad industry type |
|
|
Risk factor |
Occupational exposures and hazards – Occupational exposure to noise |
Disease outcome |
Hearing loss |
TMRED |
Background noise exposure |
National data source |
Census of Population and Housing 2016; ABS Labour force survey, April 2020 |
Units for effect size calculation |
Distribution of the labour force by broad industry type |
|
|
Risk factor |
Occupational exposures and hazards – Occupational exposure to asthmagens |
Disease outcome |
Asthma |
TMRED |
Background asthmagen exposure |
National data source |
Census of Population and Housing 2016; ABS Labour force survey, April 2020 |
Units for effect size calculation |
Distribution of the labour force by broad industry type |
|
|
Risk factor |
Occupational exposures and hazards – Occupational ergonomic factors |
Disease outcome |
Back pain and problems |
TMRED |
No occupational exposure to ergonomic factors causing back pain and problems |
National data source |
Census of Population and Housing 2016; ABS Labour force survey, April 2020 |
Units for effect size calculation |
Distribution of the labour force by broad industry type |
Estimates by socioeconomic group
The estimate of the economically active population by socioeconomic group was adjusted, based on the proportion of the population in each quintile not in the labour force. The proportion in each industry and occupation group was estimated from the same data source as for the national exposure estimates. National estimates for occupational injury were used for each quintile. Exposure by socioeconomic group was not estimated due to lack of available data.
2015, 2011 and 2003 estimates
Methods for estimating exposure and calculating the PAFs in 2018 were followed for 2015 and 2011 estimates. The working population was estimated from the Labour Force Survey (ABS 2003, 2011, 2018) and disaggregated by occupation and industry using the 2016, 2011 and 2006 Census of Population and Housing (ABS 2017).
Indigenous specific estimates
The attributable burden in Aboriginal and Torres Strait Islander population was calculated in the same way as for the national population with the following changes.
The estimates of the number of Indigenous Australians working were sourced from the labour force survey (ABS 2020). The national estimates of the working population include long-term unemployed people, as they make up only a small proportion of the national population (1.3% in 2003 and 1.0% in 2011) (ABS 2011). As long-term unemployed people represent a much higher proportion of the Indigenous population (5.7% in 2003 and 6.0% in 2011) (AIHW analysis of the NATSISS 2002 and AATSIHS 2012–13), the estimate of economically active Indigenous population was adjusted down by the difference between these rates in each year.
Estimates of the number of Indigenous Australians working in 2003 were sourced from the Labour Force Survey 2006 (ABS 2007). These estimates were broken down by occupation and industry using estimates from the 2001 Census of Population and Housing.
National PAFs were used to estimate attributable burden due to carcinogens for Indigenous Australians, because the occupational turnover rates used in this calculation are not appropriate for the Indigenous population.
The Safe Work Australia data sets do not include an Indigenous identifier, so the direct evidence sourced from these publications was not available for the Indigenous population. Instead, an Indigenous to non-Indigenous rate ratio was calculated for all injury hospitalisations with an ICD-10-AM activity code of U73 (‘While working for income’), by sex. This ratio was applied to the national exposure rates to derive Indigenous exposure estimates for injuries.
References
ABS (Australian Bureau of Statistics) 2003. Labour force, Australia, June 2003. ABS cat. no. 6202.0. Canberra: ABS. Viewed 22 June 2016.
ABS 2007. Labour force characteristics of Aboriginal and Torres Strait Islander Australians, experimental estimates from the Labour Force Survey, 2006. ABS cat. no. 6287.0. Canberra: ABS.
ABS 2011. Labour force, Australia, June 2011. ABS cat. no. 6202.0. Canberra: ABS. Viewed 22 June 2016.
ABS 2017. Census of Population and Housing: TableBuilder Pro, Australia, 2016. ABS cat. no. 2073.0 Canberra: ABS. Viewed 25 September 2018.
ABS 2018. Labour force, Australia, Jan 2018. ABS cat. no. 6202.0. Canberra: ABS. Viewed 25 September 2018.
ABS 2020. Labour force, Australia, Apr 2020. ABS cat. no. 6202.0. Canberra: ABS. Viewed 25 September 2018.
Kauppinen T, Toikkanen J, Pedersen D, Young R, Ahrens W, Boffetta P et al. 2000. Occupational exposure to carcinogens in the European Union. Occupational and Environmental Medicine 57(1):10–18.
Safe Work Australia 2019. Work-related traumatic injury fatalities, Australia 2018. Canberra: Safe Work Australia.
Safe Work Australia 2020. Australian Workers’ Compensation Statistics 2017–18. Canberra: Safe Work Australia.
The burden from unimproved sanitation was not estimated for the non-Indigenous population due to lack of available exposure data, and was assumed to be close to 0.
Indigenous specific estimates
Exposure was estimated from the NATSIHS 2018–19 (for 2018) and the AATSIHS 2012–13 (for 2011). The estimate was based the number of Indigenous Australians living in the households that self-reported not having working sewerage facilities.