Living Well
Article
The social determinants of being an Indigenous non-smoker Abstract
David P. Thomas
Objective: To examine the association between various social factors and being a non-smoker in a national survey of Aboriginal and Torres Strait Islander people aged 15 and over. Methods: We analysed data from the 2002 National Aboriginal and Torres Strait Islander Social Survey (n=9,400) using logistic regression. Results: About a half (51.2%) of the Aboriginal and Torres Strait Islander
Menzies School of Health Research, Northern Territory; Institute of Advanced Studies, Charles Darwin University, Northern Territory; and Centre for Health and Society, University of Melbourne, Victoria
Viki Briggs Centre for Excellence in Indigenous Tobacco Control, University of Melbourne, Victoria
Ian P. S. Anderson Onemda VicHealth Koori Health Unit and Centre for Health and Society, University of Melbourne, Victoria
population aged 15 years and over
Joan Cunningham
smoked, 33.4% had never smoked, and 15.4% were ex-smokers. Higher socio-economic position (as measured by each of nine variables) was strongly associated with being a non-smoker rather
Menzies School of Health Research, Northern Territory; Institute of Advanced Studies, Charles Darwin University, Northern Territory; and Centre for Health and Society, University of Melbourne, Victoria
than a smoker, after controlling for age and gender. There was a clear income gradient: increasing household income was associated with increasing likelihood of being a non-smoker. Indigenous people who had been arrested in the last five years were 4.5 times less likely to be non-smokers, adjusted for age and gender. Indigenous people who had been removed from their natural family were half as likely to be a non-smoker. Conclusions: Different groups within the Indigenous population have quite different smoking behaviours, although the prevalence of smoking is very high in all groups. The poorest and most socially disadvantaged are the least likely to be non-smokers. Implications: Indigenous tobacco control programs need to consider additional targeting of more disadvantaged groups. Tobacco control programs should work with broader campaigns to ameliorate social disadvantage among Indigenous peoples. Key words: Smoking; smoking cessation; socio-economic factors; Indigenous
T
obacco smoking is an important preventable cause of many Indigenous deaths. For example, an estimated 20% of Aboriginal adult deaths in the Northern Territory (1986-95) were due to smoking.1 About one-half of Aboriginal and Torres Strait Islander adults said they smoked in the three national surveys conducted since 1994.2-4 In contrast to the decline since the early 1970s in smoking among the whole Australian population, there has been no apparent change in Indigenous smoking.5 Most public health researchers now accept that smoking, like other health behaviours, is socially determined; however, the exact mechanisms are disputed. 6 An examination of the 1994 National Aboriginal and Torres Strait Islander Survey identified that the factors most strongly associated with smoking were alcohol use and three indicators of socio-economic
position: education, employment, and home ownership.2 In this paper, we will examine the social patterning of an Indigenous adult being a non-smoker using the 2002 National Aboriginal and Torres Strait Islander Social Survey (NATSISS). Our focus on the predictors of being a non-smoker, rather than of smoking, may seem cumbersome to some tobacco control researchers. However, it follows the approach of other researchers of Aboriginal tobacco and alcohol use and calls for Indigenous health research that investigates factors associated with resilience and wellness.7-9 Unlike the 1994 survey, non-smokers were identified as either ex-smokers or those who had never smoked. So we were able to examine the association of socio-economic position and other factors with having started, stopped and continued to smoke.
Australians. (Aust N Z J Public Health. 2008; 32:110-16) doi:10.1111/j.1753-6405.2008.00185.x
110
Submitted: September 2007
Revision requested: January 2008
Accepted: February 2008
Correspondence to: Dr David Thomas, Menzies School of Health Research, PO Box 41096, Casuarina, Northern Terrritory 0811. Fax: (08) 8927 6187; e-mail:
[email protected]
AUSTRALIAN AND NEW ZEALAND JOURNAL OF PUBLIC HEALTH © 2008 The Authors. Journal Compilation © 2008 Public Health Association of Australia
2008 vol. 32 no. 2
Living Well
Social determinants of being an Indigenous non-smoker
Methods National Aboriginal and Torres Strait Islander Social Survey We analysed data from the National Aboriginal and Torres Strait Islander Social Survey (NATSISS), conducted by the Australian Bureau of Statistics (ABS) from August 2002 to April 2003. Information was collected from 9,400 Aboriginal and Torres Strait Islander people aged 15 years and over.3 The survey used two multi-stage sampling strategies. In a random sample of discrete Indigenous communities, up to three Indigenous residents (aged 15 years and over) were sampled from a random sample of dwellings. Elsewhere the sampling strategy was based on Census Collection Districts rather than discrete Indigenous communities. Response rates were about 80% in both samples. A paper questionnaire was used in the first sample with slightly different questions (albeit based on the same concepts) than the computer-assisted interviewing in the second sample. We expect that these different methods had minimal impact on most of the variables we analysed. Visitors and people living in non-private dwellings were excluded from the sample. The ABS estimated that 4% of the Indigenous population was living in non-private dwellings, such as hotels, hostels, hospitals, caravan parks and prisons, in December 2002.3 The survey identified four categories of smoking status: daily smokers, occasional smokers, ex-smokers and never-smokers. Daily and occasional smoker categories were combined as current smokers; only 2.6% of the sample was occasional smokers. There were only 70 missing values for smoking status. The use of self-reports of smoking status has recently been validated in the Aboriginal population.10 In our analyses, we compared non-smokers with current smokers, ex-smokers with current smokers, and never-smokers with ever-smokers. We examined the association of smoking status with the responses to other questions in the survey. We concentrated on the variables included in the ABS report of the survey, including education, income, financial stress, work, health, housing, justice, transport, information technology, culture and language, removal from family, participation in community activities, support and stressors.3 As in the report we used ‘equivalised household gross weekly income’, which enables income comparisons adjusted for the size and composition (e.g. number of adults and children) of households. After discussion with Indigenous tobacco control workers, we also examined variables not discussed in the ABS report: the number of Indigenous dependents in the household; difficulty communicating with service providers in English; and whether the principal source of income was government pensions or allowances (including Community Development Employment Projects, the Indigenous work-for-dole scheme). Two further factors were included that had been identified as associated with Indigenous smoking in earlier research: marital status and whether there were non-Indigenous as well as Indigenous household members.2 We did not include variables about substance abuse 2008 vol. 32 no. 2
other than alcohol because of problems with the way these questions were asked in the discrete communities sample.11 We examined the effect of the nine factors that broadly describe socio-economic position (income, education, employment, source of income, financial stress, home ownership, transport access, computer use and Internet use) on being a non-smoker. Then we examined the effect of adjusting for this subset of these nine factors on the association of all the other examined factors with being a non-smoker.
Statistical analyses All analyses were performed using STATA Version 8.0 via the ABS’s Remote Access Data Laboratory.12 Under this arrangement, researchers do not have direct access to the dataset, but instead submit statistical code to the ABS, which runs the statistical commands and returns the results to researchers. All analyses used the expansion factor (or person weight) for each respondent to adjust for the disproportionate sampling of some groups. This means that we can report estimates of statistics in the total Indigenous population aged 15 and over rather than just from the sample.13 Confidence intervals for all statistics were calculated using the 250 replicate weights for each person generated by the ABS.13 These replicate weights required the use of the SVR module of STATA commands written by Nick Winter (available using the command: search svr, net). Unlike the examination of the 1994 ABS survey,2 we do not report results for logistic regression models adjusted for all factors. We only report logistic regression models adjusted for age and gender, and for the socio-economic position variables as well as age and gender. This use of such a hierarchical (rather than an all-in) approach to statistical modelling enables the adjustment for confounding without also inappropriately adjusting for mediating variables (which would falsely reduce the size of the effect of distal causes).14 The category with most respondents for each variable was used as the reference category in the regression models. Where appropriate, we collapsed categories in variables with more than two categories (e.g. categories with small numbers or similar concepts and similar smoking results). We examined whether there was any interaction with age and gender in the models adjusted for age and gender. There was statistically significant interaction with gender for seven variables, with eight for age, and one for both. These interactions led to associations of different magnitude – but not direction – for the different age or gender groups. As these details do not greatly alter the interpretation of results they are not reported here, but are available from the first author.
Results About half (51.2%) of the Aboriginal and Torres Strait Islander population aged 15 years and over smoked, one-third (33.4%) had never smoked, and one in six (15.4%) were ex-smokers (see Table 1). Women were more likely than men to be non-smokers and more likely to have never smoked; men and women were equally likely
AUSTRALIAN AND NEW ZEALAND JOURNAL OF PUBLIC HEALTH © 2008 The Authors. Journal Compilation © 2008 Public Health Association of Australia
111
Thomas et al.
Article
to be ex-smokers. Smoking rates increased with age and then fell in those over 45 and 55 years. Indigenous people living in remote areas were less likely than those in non-remote areas to be a non-smoker rather than a smoker (OR 0.82 [95% CI 0.70-0.96]) and less likely to have quit than still smoke (OR 0.64 [95% CI 0.52-0.79]), but there was no difference in the proportions of those who had never smoked. While Indigenous men in remote areas were much less likely than those in non-remote areas to be a non-smoker (OR 0.66 [95% CI 0.54-0.80]), there was no such difference for women (OR 1.01 [95%CI 0.82-1.24]). The direction of the association of living in remote areas with being a non-smoker was reversed after adjustment for socio-economic position variables as well as age and gender (OR 1.25 [95% CI 1.06-1.48]). The Northern Territory (NT) had the highest proportion of smokers (55.7%), but the differences between jurisdictions were mostly not statistically significant. After adjusting for age and gender, Indigenous people in the remote NT who had ever smoked were less likely than other Indigenous people to have quit (OR 0.37 [95% CI 0.27-0.53]). Indigenous people in Victoria were less likely than others to have never smoked (OR 0.67 [95% CI 0.52-0.85]). The dataset only allowed the separate identification of Torres Strait Islanders (from Aboriginal people) in Queensland. The only statistically significant difference between these Indigenous groups was that Torres Strait Islanders from Queensland who had ever smoked were more likely to have quit (OR 1.60 [95% CI 1.062.40]) than other Indigenous people in Australia.
Socio-economic position Higher socio-economic position (SEP), as measured by each of the variables in Table 2, was strongly associated with being a non-smoker rather than a smoker, after controlling for age and gender. The association between education levels attained and being a non-smoker was more modest than for other SEP variables but was still statistically significant. Similar patterns were found when investigating the association between SEP and never-smoking versus ever-smoking and ex-smokers versus current smokers. However, in almost all categories of SEP and other variables, there was a greater absolute difference in the prevalence of never-smokers to ever-smokers than of ex-smokers to ever-smokers (results not shown). Those in the highest SEP category for all of the nine SEP variables were 10.5 times (95% CI 6.9-15.9) more likely to be non-smokers than those in the lowest SEP category for all the SEP variables. Figure 1 shows a clear income gradient in the likelihood of Indigenous people smoking: increasing household income was associated with increasing likelihood of being a non-smoker. However, smoking rates were still relatively high (37.1% [95% CI 28.2-46.9]) among the 6% of Indigenous people from households with incomes in the highest quintile of Australian household incomes. While there were 1,348 people with missing or not stated answers for income, the smoking status of these people was no different to those with a stated income.
Table 1: Proportion current smokers, ex-smokers and never-smokers by demographic characteristics (n=9,289). Variable
% current smokers (95% CI)
% ex-smokers (95% CI)
% never-smokers (95% CI)
Total (all 15 years and over)
51.2
(48.9-53.6)
15.4
(14.2-16.7)
33.4
(31.4-35.4)
Gender Male Female
53.1 49.5
(50.1-56.1) (46.7-52.3)
15.5 15.3
(13.6-17.5) (13.7-17.1)
31.5 35.2
(28.7-34.3) (32.7-37.7)
Age 15-24 years 25-34 years 35-44 years 45-54 years 55+ years
48.4 57.6 57.6 49.2 35.4
(43.7-53.1) (53.9-61.3) (53.5-61.6) (43.9-54.4) (30.3-40.9)
7.4 12.3 16.6 21.0 34.0
(5.8-9.4) (10.2-14.7) (13.7-20.1) (17.6-25.0) (29.2-39.2)
44.3 30.1 25.8 29.8 30.6
(39.8-48.9) (26.6-33.9) (22.4-29.4) (25.6-34.4) (26.3-35.2)
Region Non-remote Remote
50.0 54.7
(47.0-52.9) (51.9-57.5)
16.7 11.8
(15.2-18.4) (10.4-13.4)
33.4 33.5
(31.0-35.8) (30.9-36.2)
Jurisdiction New South Wales Victoria Queensland South Australia Western Australia Northern Terrritory Tasmania/Aust. Capital Territory
52.3 53.5 50.7 47.9 48.4 55.7 45.3
(47.1-57.5) (48.6-58.3) (46.6-54.9) (42.9-53.0) (43.7-53.2) (51.0-60.3) (40.9-49.8)
14.4 20.8 16.1 17.6 17.1 9.3 18.8
(12.0-17.1) (16.9-25.3) (13.5-19.1) (14.3-21.6) (13.5-21.3) (7.3-11.8) (15.9-22.1)
33.3 25.8 33.2 34.5 34.5 35.0 35.9
(29.2-37.7) (21.8-30.1) (29.8-36.8) (30.2-39.0) (29.9-39.4) (30.6-39.8) (31.9-40.0)
Torres Strait Islanders Torres Strait Islanders in Queensland
47.8
(41.3-54.4)
19.9
(15.2-25.5)
32.3
(27.8-37.2)
Source: Weighted data from National Aboriginal and Torres Strait Islander Social Survey 2002.
112
AUSTRALIAN AND NEW ZEALAND JOURNAL OF PUBLIC HEALTH © 2008 The Authors. Journal Compilation © 2008 Public Health Association of Australia
2008 vol. 32 no. 2
Living Well
Social determinants of being an Indigenous non-smoker
Crime and justice
Alcohol
The strongest associations with being a non-smoker were for those not arrested or not incarcerated in the last five years (see Table 3). Those arrested were 4.5 times less likely and those incarcerated were 4.0 times less likely to be non-smokers, adjusted for age and gender. The magnitude of these associations was reduced after further adjustment for the SEP variables, but both those arrested or incarcerated were still less than half as likely to be non-smokers. Those who had been neither arrested, jailed, victims of violence nor used legal services were 9.9 (95% CI 6.4-15.2) times more likely to be non-smokers than those who answered yes to all four questions, adjusted for age and gender. The magnitude of the association of these variables was much greater with neversmoking versus ever-smoking (combined OR 13.4 [95% CI 8.022.4]) than for ex-smokers versus smokers (combined OR 3.9 [95% CI 2.2-6.8]).
Not drinking alcohol was strongly associated with being a nonsmoker, with having quit, and with having never smoked (rather than ever smoked). The association with being a non-smoker decreased with increasing levels of short and long-term alcohol consumption (results not shown).
Removal from family Indigenous people who had not been removed from their natural family were twice as likely to be a non-smoker, to never have smoked or to have quit. This association was the same when adjusted for age and gender and when also adjusted for the SEP variables. Indigenous people who had a relative removed from their family were statistically significantly less likely (OR 0.78 [95% CI 0.64-0.95]) to have never smoked (versus ever smoked). A more modest association with being a non-smoker was not statistically significant; however, there were 1,775 missing values for this variable, leading to less statistical power.
Table 2: Socio-economic position: odds ratios of being a non-smoker versus being a smoker, adjusted for age and gender.
n
Incomea Lowest quintile Second quintile Third, fourth or highest quintile
7,954
Education Year 10 or below or never attended Year 11 or 12 or non-school qualifications
9,169
Employment Not in labour force Employed in public or private sector Unemployed or CDEP
9,188
Principal source of income Government cash pensions (including CDEP) Not Government pensions
9,200
Financial stress Could not raise $2,000 within a week Could raise $2,000 within a week
8,872
Housing Renter or other Owner or purchaser
9,289
Transport access Access to motor vehicles No access to motor vehicles/no licence
9,262
Information technology Used computer last 12 months Not used computer Not used Internet last 12 months Used Internet
9,289
%
Odds ratio
(95% CI)
42.5 28.2 29.3
1.00 1.49c 2.50c
(1.20-1.85) (1.96-3.18)
60.5 39.5
1.00 1.60c
(1.36-1.89)
40.4 33.5 26.1
1.00 2.03c 0.76b
(1.64-2.52) (0.61-0.94)
60.8 39.2
1.00 2.52c
(2.09-3.03)
57.0 43.0
1.00 2.33c
(1.94-2.79)
72.5 27.5
1.00 2.87c
(2.36-3.48)
55.0 45.0
1.00 0.50c
(0.42-0.59)
55.7 44.3 58.9 41.1
1.00 0.45c 1.00 2.35c
(0.38-0.53) (1.98-2.78)
Notes: (a) Equivalised household gross weekly income using national quintiles from 2002 General Social Survey. (b) p
