Review of 2021 Census Topics

Redacted textAssociate Professor, Menzies Centre for Health Policy, School of Public Health, The University of Sydney, and School of Public Health and Community Medicine, University of New South Wales, Redacted text, Professor, Melbourne School of Population and Global Health, University of Melbourne
Redacted text Professor, Associate Pro Vice-Chancellor (Academic Partnerships), College of Science, Health and Engineering, La Trobe University
Redacted text, University of Melbourne Laureate Professor, Rowden-White Chair of Global Health and Burden of Disease Measurement, and Director, Global Burden of Disease Group, Director, Bloomberg Initiative for Civil Registration and Vital Statistics
Redacted text Senior Statistician, Cancer Research Division, Cancer Council NSW, Sydney Redacted text Doctoral NHMRC and Brawn Career Development Fellow, School of Medicine & Public Health, Faculty of Health & Medicine, University of Newcastle, NSW, Australia
Redacted text, Senior Post-Doctoral Research Fellow, Faculty of Health, School of Public Health and Social Work, Queensland University of Technology, Brisbane
Redacted text Professor, Head, Tobacco Control Research, Menzies School of Health Research, Darwin
Redacted text Professor and Head, School of Public Health, The University of Sydney.

We propose that questions ascertaining the smoking status of census participants be considered for addition to the Australian 2021 Census.

Recent estimates from the Australian Institute of Health and Welfare report (table shown below) indicate that smoking is responsible for about 1 in 8 of all deaths in Australia (12.8%, 18,762 deaths) in 2011. It is clear tobacco use is the topmost (and most clear-cut single preventable) risk factor in Australia. This is followed by all dietary risks (12.1% of all deaths), blood pressure (9.9%), and physical inactivity (7.8%), where prevention for each is more complex. Tremendous progress has been achieved in reducing smoking rates to 14.5%. While Australia relied on classic epidemiological survey research methods to measure smoking prevalence, risk and use these data in policy and advocacy to reduce smoking rates from say 35% to 15%, the strategies to reduce smoking even further to e.g. 5% need far greater precision than ever before to inform policy and program development.

As eloquently also argued by Thomas and Scollo while national smoking prevalences are currently at 14.5% and declining, there are a number of significant Australian minority groups which have burdens of disease and smoking rates that are double of more the national average. This includes populations in rural and remote areas, Aboriginal and Torres Strait Island populations, those in lower socioeconomic strata and certain migrant groups. These marginalised populations have significantly higher smoking rates and poorer quitting outcomes than those who are not. , Each of these groups comprises less than 5-10% of the national population, are sometimes overlapping, nationally dispersed and therefore difficult to sample using traditional survey methods.
There are numerous lifestyle surveys from e.g. ABS that include questions on smoking but again, they are small and of low coverage. Such finely targeted estimates of smoking prevalence and risks are rarely available from national sample-based surveys and can only become available using the national census.
Importantly, with Australian initiatives to link census data to mortality records, a unique opportunity exists to develop local precision public health initiatives to estimate the smoking burden in small subgroups of Australian society, and better still, to measure the level of mortality, relative risks and proportions attributed to smoking in each of these groups. While cohort studies such as the 45 and Up Study (the largest in Australia, n=267,000, comprising 10% of NSW, 3% of national population ≥45 years) are effective for measuring relative risks of overall deaths in relation to smoking in Australian populations, they are limited in their ability to provide estimates of the prevalence of smoking and proportions of individual causes of deaths caused by smoking in such small subgroups. This is because participants in cohort studies tend to be more health conscious generally and, more specifically, smoke at lower rates than the general population. New Zealand has since 1981 collected smoking information from census participants linked this information to mortality records, and has provided a rich literature using direct, and arguably less biased data on the burden of smoking in that country. Australia relies on indirect, counterfactual estimations of the national smoking burden, which are based on several assumptions therefore prone to variation and critique.

Modest questions on smoking, ascertaining status of never, ex- and current smokers, have been routinely asked in several household Australian surveys and have been accurately collected with low margins of error and little difficulty. Two questions are used in the New Zealand census to obtain three categories of smoker: never, ex-, current. (In the 1981 census number of cigarettes per day was deemed too inaccurate to justify the space.) Of course, such a simple two-question approach may be prone to error, but significant statistical methods have now evolved to make valid comparisons using bias analysis and error correction to retrieve smoking-health associational estimates closer to the truth – e.g. as was done in these two New Zealand studies using cohort analyses of census-smoking-mortality/cancer data. ,

In the New Zealand Census, the response rate for the smoking question is greater than 95%

Provided the questions on smoking are modest, the household head will be able to answer the questions efficiently, i.e. without having to interrogate household members or look up family medical records, files etc. As per question 3 and 4 above.

Yes. While the overall smoking prevalence is Australia is low by international standards and decreasing, the smoking prevalence in some marginalised sub-groups remain at epidemic levels. For example, ABS estimates indicate nearly half of Aboriginal and Torres Straight Island living remotely are daily smokers in 2014/15. As the prevalence of smoking continues to fall, the need for ongoing data to estimate prevalence and change over time in smaller more vulnerable groups of the population will increase. For example, in New Zealand for Māori populations, smoking prevalence data are available by tribe/Iwi, which helps add momentum to providing precise local data to enhance tobacco control and monitor results in such fine categories. These data also facilitate the modelling of the effects of introducing different smoking cessation interventions (such as increasing tobacco tax) in population subgroups. Further, having consistent questions like those asked in the New Zealand census will allow for future comparability.
Because of the lengthy time lag between smoking and disease onset, only now are the full hazards of smoking beginning to emerge particularly in women in whom the smoking epidemic tends to lag a decade or more behind that of men: lung cancer for example is now the leading cause of cancer-related death in women and men in Australia. The existence of clinically important and previously under-recognised sex differences in how smoking influences health outcomes are only now emerging because of the maturity of the smoking epidemic in women. Additional questions on smoking-related behaviour and ongoing monitoring of the situation, will provide much needed information to more fully characterise sex differences in smoking-related harm in the Australian population.

There is a pervasive view that information such as smoking is routinely and unbiasedly collected via e.g. hospital systems. Published research on this topic shows this to be incorrect. Smoking is occasionally collected in Australian hospital information systems at patient intake but such questions are usually asked when a patient is admitted for diseases that are thought by the doctor to be directly linked to smoking, e.g. respiratory, lung and heart disease. Acute (dead on arrival, emergency department) coronary / stroke events would be missed, and so would several conditions that are partially caused by smoking e.g. digestive cancers, female reproductive organ cancers, and possibly diabetes. Even if these records are linked to mortality, this approach will cause serious bias in estimations of all deaths and the relative importance of diseases caused by smoking. Information on smoking gathered via GP clinical information systems has not yet been validated in Australia. Obtaining independent information on smoking behaviour (irrespective of someone’s health status) by means of a census provides an unbiased way by which this information can be gathered in order to obtain reliable, representative and national results.

Australian Institute of Health and Welfare, Australian Burden of Disease Study: Impact and causes of illness and death in Australia 2011. Australian Burden of Disease Study series no. 3. BOD 4 Canberra: AIHW; 2016. Available from: https://www.aihw.gov.au/reports-statistics/health-conditions-disability-deaths/burden-of-disease/overview

Thomas DP, Scollo M. Should a smoking question be added to the Australian 2021 census? Aust N Z J Public Health. 2018 Apr 26. doi: 10.1111/1753-6405.12788.

Greenhalgh, EM, Bayly, M, & Winstanley, MH. 1.8 Trends in prevalence of smoking by country of birth. In Scollo, MM and Winstanley, MH [editors]. Tobacco in Australia: Facts and issues. Melbourne: Cancer Council Victoria; 2017. Available from http://www.tobaccoinaustralia.org.au/chapter-1-prevalence/1-8-trends-in-prevalence-of-smoking-by-country-of-

Greenhalgh, EM, van der Sterren, A, Knoche, D, & Winstanley, MH 8.7 Morbidity and mortality caused by smoking among Aboriginal peoples and Torres Strait Islanders. In Scollo, MM and Winstanley, MH [editors]. Tobacco in Australia: Facts and issues. Melbourne: Cancer Council Victoria; 2016. Available from http://www.tobaccoinaustralia.org.au/8-7-morbidity-and-mortality-caused-by-smoking-amon

45 and Up Study Collaborators, Banks E, Redman S, Jorm L, Armstrong B, Bauman A, Beard J, Beral V, Byles J, Corbett S, Cumming R, Harris M, Sitas F, Smith W, Taylor L, Wutzke S, Lujic S. Cohort profile: the 45 and up study. Int J Epidemiol. 2008 Oct;37(5):941-7.

Sitas F, O’Connell DL, Jamrozik K, Lopez A. Smoking questions on the Australian death notification forms: adopting international best practice? Med J Australia, 2009; 191: 166-168

Blakely T, Barendregt JJ, Foster RH, et al. The association of active smoking with multiple cancers: national census-cancer registry cohorts with quantitative bias analysis. Cancer Causes Control 2013;24:1243–55.

Blakely T, Disney G, Valeri L, et al. Socio-economic and tobacco mediation of ethnic inequalities in mortality over time: Repeated census-mortality cohort studies, 1981 to 2011. Epidemiology 2018;In press

Ball J, Stanley J, Wilson N, Blakely T, Edwards R. Smoking prevalence in New Zealand from 1996–2015: a critical review of national data sources to inform progress toward the smokefree 2025 goal. NZ Med J. 2016 Aug 5;129(1439):11-22

Australian Bureau of Statistics. 4714.0 - National Aboriginal and Torres Strait Islander Social Survey, 2014–15. 2016. Available from: http://www.abs.gov.au/AUSSTATS/abs@.nsf/DetailsPage/4714.02014-15?OpenDocument
Huxley RR, Woodward M. Cigarette smoking as a risk factor for coronary heart disease in women compared with men: a systematic review and meta-analysis of prospective cohort studies. Lancet 2011; 378:1297-1305.

Havard A, Jorm LR, Lujic S. Risk adjustment for smoking identified through tobacco use diagnoses in hospital data: a validation study. PLoS One. 2014 Apr 15;9(4):e95029.