In Part 1, geographical classifications available for collation and reporting of health data were discussed.

The demographic characteristics of a population are generally presented before moving into health statistics. The ABS Census of Population and Housing Data includes this information and is available by geography online for the past 4 surveys (2016, 2011, 2006, 2001). The 2016 Census for ABS Statistical Area 3 for Woden Valley in Canberra, Australian Capital Territory (ACT), the location of the Department of Health is shown. ‘Quick Statistics’ provide a one page (scroll down) summary for the approximately 35,000 residents in a table format, and a side by side comparison of each variable to that recorded for the 400,000 inhabitants of the ACT, and just over 24 million people in Australia. While 68% of the residents of the Woden Valley were born in Australia, 25% speak a language other than English at home.

A ‘General Community Profile’ is also generated and consists of an Excel workbook containing 60 spreadsheets of data for the area selected. A range of other Census data products are also free online, including Time Series Profiles across Censuses, DataPacks and TableBuilders.

The 5-yearly Census of Population and Housing data is used to generate the Socio-Economic Indexes for Areas (SEIFAs) which, as the name implies, rank areas in Australia according to relative socio-economic advantage and disadvantage. Each of the four indexes is a summary of a different subset of Census variables with a different focus:

• Index of Relative Socio-economic Disadvantage (IRSD)
• Index of Relative Socio-economic Advantage and Disadvantage (IRSAD)
• Index of Economic Resources (IER)
• Index of Education and Occupation (IEO)

The website includes an interactive map for each index. If we enter Woden Valley ACT on the IRSD Interactive Map,  the SA3 geographic area is not available in the drop down menu. Our selection is limed to SA1, SA2, LGA, POA or SSC, however data cubes with results by decile are available for the wider range of geographic classifications.

Sixteen variables are used in the IRSD index. The percent of people with stated household equivalised income between $1 and $26,999 per year is the strongest indicator of disadvantage. Other associated factors include unemployment, lower levels of education attainment, no internet connection, no car, overcrowded private dwelling and poor English language skills.

Unfortunately, the indexes are not foolproof. Selection of geographic area can impact the results as shown for Woden Valley,  where choice of SA2 displays pockets of Quintile 1, most disadvantaged through to Quintiles 3, 4 and 5, least disadvantaged (on the left) suggesting a different socio-economic situation than when Local Government Area is selected with all areas being displayed as Quintile 5, least disadvantaged (on the right). This re-enforces the need to compare like with like to ensure appropriate interpretation.

 

 

 

 

 

 

Source Tom Cruise meme: Google images (unfortunately, the link goes to one of those surveys)

The newly released AIHW Report on Gastrointestinal Cancers is the first time Australia-wide data, specific to gastrointestinal-tract cancers, has been collated. It is the result of collaboration between the AIHW, all state and territory population-based cancer registries and Cancer Australia.

The report is a fantastic resource as it brings together epidemiological, diagnostic and treatment data in a comprehensive way thus providing a full picture for comparison and interpretation. The explanations regarding data sources in the appendices are also very helpful.

The eight upper and lower gastro-intestinal cancers included are shown in the graph below of cases diagnosed, deaths and relative 5-year survival rates.  Colorectal cancer is the most commonly diagnosed, a function of higher incidence and the National Bowel Cancer Screening Program (NBCSP), followed by pancreatic cancer as the second most commonly diagnosed.  These two cancer types are associated with the highest and lowest, 5-year relative survival rates of all gastrointestinal-tract cancers, at 69% and 9%, respectively.

 

The inclusion of disease Stage at diagnosis enables presentation of data as shown below from the report. Cases of colorectal cancer diagnosed in 2011 and the 5-year relative survival for the period 2011-2016, both reported by disease Stage, clearly demonstrate the impact of early diagnosis on mortality. For people diagnosed with Stage I colorectal cancer, the 5-year survival rate was 99%. Delay diagnosis to Stage IV and this figure becomes 13%.

 

In 2016–17, the gastrointestinal-tract accounted for approximately 20% of cancer diagnoses, 18% of cancer-related hospitalisations and 27% of chemotherapy procedures in Australia. Further detail on the treatment used at which disease Stage, would be a valuable inclusion in the next iteration of this report.

The majority (92%) of the burden of gastrointestinal-tract cancer was due to premature death. The remainder was associated with diagnosis (such as  biopsy) and primary treatment of the cancer (for example, surgery which may include bowel resection). Long-term effects, such as the use of a stoma with a colostomy bag, also contributes to the non-fatal burden from colorectal cancer in the Australian population.

Picture source: ‘The Scream’ original and stylised

The Australian Institute of Health and Welfare (AIHW) report on national health expenditure for the financial year 2016-17 was released last week. The usual on-line and pdf versions are available and the information can also be viewed using a data visualisation tool.

The tool accesses the AIHW Expenditure Database from 1996 through to 2016. Note: these years refer to financial years (the full year covered appears when the cursor is hovered over a data point on graphs produced).

The choice of presentation in constant or current dollars is important when assessing trends. Current prices are the actual amounts spent in the specified year. These will match the dollar amounts in the published reports, however they are not adjusted for inflation. Any comparisons made over different time periods require the use of constant prices that have been ‘deflated’ to a selected reference year. The tool uses the most recent data presented as the reference year. Growth in expenditure when expressed in constant prices is termed ‘real growth’ as opposed to ‘nominal growth’ when current prices are used.

There are 3 chart options: Total Health Expenditure (HE), Recurrent HE by Source of Funds and Recurrent HE by Area of Expenditure.

The Total HE option, as shown below, is limited to a comparison of fund source and area of expenditure by States/Territories.

Source of funds choices are Government (Australian or State/Territory and local) or non-Government (Health insurance, individuals or other).

Each of the five broad area of expenditures can be presented by (detailed areas of expenditure):

1. Hospitals (Private or Public hospital services)
2. Other services (Administration, Aids and appliances, Patient transport services)
3. Primary health care (All other medications, Benefit-paid pharmaceuticals, Community health & other, Dental services, Other health practitioners, Public health, Unreferred medical services)
4. Referred medical services
5. Research

HE on Public Hospital services by the source of funds for NSW is shown above. Although the contribution of Private Health Insurance (PHI) as a proportion rose over 15% from 2012/13 (3.02%) to 2016/17 (3.48%), it is difficult to detect graphically as it is overwhelmed by the Government sources. Nationally, over the same period the proportion increased by approximately 9%.

Finally, HE by Area of Expenditure (above) provides a clear picture of the move of medication costs from Government to individuals over the past decade. Unfortunately, the variable labels continued to be truncated when the graph image was downloaded.

Overall, the data visualisation tool is a useful addition to the reporting of information by the AIHW.

Picture source and worth a read: https://www.mindjet.com/blog/2013/07/visualisation-rocks/

Additional article on data visualisations

The Australian Institute of Health and Welfare (AIHW) has reported for the first time linkage and analysis of data by their Data Linkage Unit from the following six sources:

• BreastScreen Australia registers from the 8 states & territories;
• National Cervical Screening Program  registers from each state/territory;
• the National Bowel Cancer Screening Program Register;
• the Australian Cancer Database;
• the National Death Index; and
• the National HPV (human papillomavirus) Vaccination Program Register.

Privacy was protected as no one person had access to both the identified data and analysis variables. Data linkage was carried out in a step-wise fashion using identifying variables names, sex, date of birth and postcode. Probabilistic data linkage based on the method developed by Fellegi and Sunter (1969) was used.

Retrospective cohort studies were undertaken for breast, cervical and bowel cancer to assess survival for screen-detected compared with non-screen-detected cancers.

Breast cancers were identified on the Australian Cancer Database by ICD-10 code (C50) with date of diagnosis between 1 January 2002 and 31 December 2012 inclusive, for women aged 50–69 at diagnosis. These were linked with available data from BreastScreen registers (from 1 January 2000), and the screening history prior to each cancer used to assign a screening status to each breast cancer. These were: screen-detected cancers; non-screen-detected cancers in screened women;  interval cancers; non-screen-detected cancers in never-screened women. These individuals were then linked with data from the National Death Index to ascertain date of death and cause of death for those who had died by 31 December 2015.

Similar processes were undertaken for cervical and bowel cancers to create a National Screening Data Set. National identification numbers provided by the AIHW Data Linkage Unit to the AIHW Cancer and Screening Unit allowed the formation of a de-identified national cancer screening data set for analysis.

The Australian Cancer Database was linked to the National Screening Data Set using an outer join, meaning that all data from both data sets were retained in the final data set, which then comprised all screened individuals irrespective of whether they were diagnosed with a cancer, and all individuals diagnosed with a cancer irrespective of whether they were screened. This was important to ensure that breast, cervical and bowel cancers that were not detected through screening could contribute to the analyses, and that screening behaviour related to prior cancer diagnosis of any cancer type could be assessed.

Linkage with the National Death Index was the last step in the data linkage process. After linkage, there were 15,238,666 unique individuals in the project. The majority (61.7%) of individuals in the data linkage project had only one type of event (that is, appeared in only one of the data sources), while 24.6% experienced 2 event types, and 12.4% experienced 3 event types.

Three individuals experienced all 6 event types in the project, meaning that 3 individuals had a screening mammogram through BreastScreen, had a cervical cytology, histology or HPV test, were invited to the bowel screening program, were diagnosed with cancer, received at least 1 dose of HPV vaccine, and had died.

Breast cancer
There were 73,440 breast cancers diagnosed in the cohort selected for survival analyses (women aged 50–69 diagnosed 1 January 2002 to 31 December 2012). Women diagnosed with screen-detected breast cancers were less likely to die, and those who did die were less likely to die from breast cancer than women whose breast cancer was not screen-detected, with 50.5% of deaths in women with screen-detected breast cancer due to breast cancer, compared with 74.2% in never-screened women.

Cervical cancer 
There were 6,897 cervical cancers diagnosed in the cohort selected for survival analyses (women aged 20–69 diagnosed 1 January 2002 to 31 December 2012). Women diagnosed with screen-detected cervical cancers were less likely to die, and those who did die were less likely to die from cervical cancer than women whose cervical cancer was not screen-detected, with 67.7% of deaths in women with screen-detected cervical cancer due to cervical cancer, compared with 78.7% in never-screened women.

Never-screened women diagnosed with cervical cancer had the highest risk of cervical cancer mortality. The unadjusted hazard ratio for cervical cancer mortality for women diagnosed with cervical cancer was 0.11, i.e. 89% less chance of dying of cervical cancer than a never screened woman. All cause mortality was also statically significantly lower.

Bowel cancer 
There were 31,427 bowel cancers diagnosed in the cohort selected for survival analyses (people aged 50–69 diagnosed 1 August 2006 to 31 December 2012). Of the diagnosed bowel cancers, 3,316 (10.6%) were screen-detected. The relatively low proportion is due to the need to be invited to have a bowel cancer detected through the screening program (previously limited to those turning 50, 55 or 65). The proportion of bowel cancers that are screen-detected is expected to increase with the introduction of biennial screening for all Australians aged 50–74 from the year 2019 (see previous post). People diagnosed with screen-detected bowel cancers were less likely to die, and those who did die were less likely to die from bowel cancer than people whose bowel cancer was not screen-detected, with 65.8% of deaths in people diagnosed with a screen-detected bowel cancer due to bowel cancer, compared with 78.1% in people never-invited to screen.

Further publications from the project are to follow. This first report clearly shows the value of data linkage, and overwhelmingly, the public health benefit of both Government subsidised screening programs and of participating in them!