Economic Benefits of Achieving Realistic Smoking Cessation Targets in Australia

Abstract

Objectives. We estimated the economic impact of reductions in the prevalence of tobacco smoking on health, production, and leisure in the 2008 Australian population.

Methods. We selected a prevalence target of 15%. Cohort lifetime health benefits were modeled as fewer incident cases of tobacco-related diseases, deaths, and disability-adjusted life-years. We estimated production gains by comparing surveyed participation and absenteeism rates of adult smokers and ex-smokers valued according to the human capital and friction cost approaches. We estimated household production and leisure gains from time use surveys and valued these gains with the appropriate proxy.

Results. In the 2008 Australian population, an absolute reduction in smoking prevalence of 8% would result in 158 000 fewer incident cases of disease, 5000 fewer deaths, 2.2 million fewer lost working days, and 3000 fewer early retirements and would reduce health sector costs by AUD 491 million. The gain in workforce production was AUD 415 million (friction cost) or AUD 863 million (human capital), along with gains of 373 000 days of household production and 23 000 days of leisure time.

Conclusions. Lowering smoking prevalence rates can lead to substantial economic savings and health benefits.

The deleterious impact of tobacco smoking on health is well understood,¹ and the health benefits of reducing the proportion of a country's citizens who smoke tobacco have been quantified numerous times.^2,3 For example, in the most recent Australian Burden of Disease Study, conducted in 2003, tobacco use was responsible for the greatest disease burden (7.8%) in the country among the 14 risk factors reviewed.⁴ In recent years, with comprehensive tobacco control programs, Australian tobacco smoking prevalence rates have decreased by about 1 percentage point per year.⁵

Australian tobacco control programs include media education campaigns, legislation (bans on smoking in public places, graphic warnings on packaging, point-of-sale advertising bans), and heavy tobacco product taxation. Although smoking-related cancers have declined after several decades of successful tobacco control efforts,⁶ further efforts to expand the range of effective prevention strategies are needed and should be similarly beneficial. In 2004 and 2005, 23% of adults (aged 18 years or older) in Australia were current tobacco smokers, with 21% smoking daily and 2% less than daily; 30% were ex-smokers; and 47% had never smoked.⁷

Quantifying the importance of investing in further prevention initiatives is crucial given that only 3% of Australian health expenditures are devoted to prevention and health promotion.^8,9 Such efforts are timely in that they provide added impetus to the important work of the National Preventative Health Taskforce (which is focused on tobacco, obesity, and alcohol use)¹⁰ and lend support to the comprehensive strategies and work plans of important Australian health promotion organizations with tobacco control objectives (the Victorian Health Promotion Foundation and Quit).

There is limited information about the potential impact of reducing the prevalence of tobacco smoking as opposed to attempting to completely eliminate it. The consequent impact of improved health on paid and unpaid production and on leisure time have similarly not been estimated for a scenario of realistic reduced smoking prevalence levels.

We estimated the economic impact of reductions in the prevalence of tobacco smoking on health, production, and leisure in the 2008 Australian adult population. This study was part of a project funded by the Victorian Health Promotion Foundation, completed in 2009, that sought to evaluate the health, economic, and financial benefits of reductions in the prevalence of 6 important risk factors (alcohol use, physical inactivity, high body mass index, tobacco smoking, inadequate consumption of fruit and vegetables, and intimate partner violence). The project is described in greater detail elsewhere¹¹ (details are also available at http://vichealth.vic.gov.au/Resource-Centre/Publications-and-Resources/Research/Health-and-economic-benefits-of-reducing-disease-risk-factors.aspx).

METHODS

We based our estimate of the prevalence of tobacco smoking among Australian adults (23%) on the 2004–2005 National Health Survey (NHS).⁷ We adopted the NHS definition of the exposure group (current smokers) as those who smoke once or more per day on a regular or irregular basis. We defined the nonexposed group (ex-smokers) as those who no longer smoke on a regular or irregular basis. Thus, we compared current smokers with ex-smokers as opposed to those who had never smoked.

We determined feasible reductions in the prevalence of tobacco smoking through a lengthy process comprising an extensive literature review, discussions with a study-specific advisory committee made up of a broad range of experts from the government and nongovernment sectors with a focus in health economics and health promotion or policy, and consultations with subject matter experts. We used an Arcadian mean to provide a valid prevalence target to model economic and health impacts in Australia. The term Arcadian mean has been used in several comparative economic studies^3,12,13 following its introduction by Armstrong in his study of age-standardized mortality rates between genetically similar countries.¹⁴ We adopted this strategy because data on the effectiveness of Australian population prevention initiatives were limited by many factors, including the use of cross-sectional rather than longitudinal study designs, heterogeneity of the interventions and target populations considered, and outcome measures that excluded reductions in smoking prevalence.

After a search for comparable cultural and demographic areas, California was selected (over Sweden and the United States) even though its tobacco prevalence survey captured cigarette smoking only (excluding pipe and cigar smoking). The reason was that the population of California is broadly comparable to that of Australia both demographically and culturally, and the California Tobacco Control Program, with its strong and sustained government involvement, has been very successful in reducing smoking prevalence rates.¹⁵ In California during 2004 and 2005, the cigarette smoking prevalence was 15% (10% daily smokers and 5% less than daily smokers).¹⁵ Although this figure may underestimate the prevalence of tobacco smoking in California, evidence from Australia suggests that the subgroup of those who smoke pipes and cigars are also likely to be cigarette smokers¹⁶; as a result, we believed that this target was realistic for Australia.

We modeled 2 realistic targets in terms of reductions in tobacco smoking prevalence: an ideal target (8% absolute reduction), representing what might be achieved in the medium and long term on the basis of current knowledge, and a progressive target, reflecting a shorter term goal and set as attainment of half of the ideal reduction target (4%). We report the net differences in health status and economic costs between the current (attributable) and avoidable burden caused by tobacco smoking as the benefits of reductions in prevalence.

Data Analyses

We developed population simulation models in Excel 2007 (Microsoft Corp, Redmond, WA) to determine the potential lifetime benefits of reducing the prevalence of tobacco smoking in the 2008 Australian adult cohort (those aged 15 years or older). We developed simulation models for workforce participants from previous work undertaken by Deakin Health Economics for the Victoria Department of Treasury and Finance.¹⁷ Additional models were developed to estimate the lost leisure time and household production associated with diseases attributable to tobacco smoking among both working and nonworking adults. We adjusted cost data (expressed in 2008 Australian dollars) from other years by applying health price inflators.⁸ The time horizon for economic benefits was the remaining lifetime of the 2008 Australian population cohort. We applied a 3% discount rate for lifetime benefits¹⁸ and varied this rate in our sensitivity analyses (substituting 0%, 5%, and 7%).

Data Sources

We derived age- and gender-specific prevalence estimates, population-attributable risk fractions, health status estimates (incident cases of disease and death related to tobacco smoking), and information on disability-adjusted life-years (DALYs) from the 2003 Australian Burden of Disease Study data files,⁴ which were made available for this study. Excel data files from the 2000–2001 Disease Costs and Impact Study¹⁹ were used to estimate changes in health sector costs.

We derived data on demographic characteristics, employment status, and health-related actions of tobacco smokers and ex-smokers from the 2004–2005 NHS Confidentialised Unit Record Files with the approval of the Australian Bureau of Statistics.^7,20 The NHS contains self-reported information and includes data on the health status of Australians, their use of health services, and other health-related lifestyle variables. We used data from the Confidentialised Unit Record Files to estimate characteristics of the Australian population, assigning weights (expansion factors) to individual records consistent with the representative sampling strategy.

We derived information on household production and leisure time from the 2006 Australian Bureau of Statistics Time Use Survey.²¹ Data on current average wages were obtained from the Australian Bureau of Statistics and published government pay scale summaries.^22,23

Workforce production gains model. Estimates of workforce production gains and losses reflected differences in workforce participation and absenteeism rates associated with smoking status (smoker or ex-smoker) along with changes in the health status of the population after achievement of smoking prevalence reductions. The 2 techniques used most frequently in economic evaluations designed to measure and value production gains and losses are the human capital approach and the friction cost approach.

In brief, the human capital approach accounts for all future income lost from an individual who leaves the workforce as a result of death or disability, whereas the friction cost approach assumes that individuals will be replaced after a specified period (the friction period), and thus production losses to society are reduced. There is debate in the literature about which method is preferable.^24,25 Despite its limitations, the human capital approach is still the dominant methodology used to measure production costs in cost of illness studies. However, we prefer the friction cost approach because it has a stronger logical connection to cost effects on industry and has been used frequently in recent studies. In our study, the friction period was assumed to be 3 months^24,26 (we substituted 6 months in our sensitivity analyses). We also used the human capital approach to calculate workforce production gains as an additional sensitivity analysis.

The workforce production gains model involved following a theoretical cohort of Australians (aged 15–65 years) and estimating the production gains and losses and taxation effects of the expected health benefits of tobacco smoking cessation from the cohort's working years until retirement. We defined workforce participation as part-time work, full-time work, or an active search for work. “Presenteeism,” a measure and valuation of the extent to which individuals are less productive at work because of ill health, was not included because of data limitations. We captured uncertainty in wages, participation rates, and absenteeism in the reported survey standard errors.^7,22,23

Household production and leisure time model. We considered it important to capture aspects of production that extend beyond those participating in the paid workforce. We defined household production as hours of time spent performing nonpaid household duties such as cooking, shopping, cleaning, child care, and maintenance. We used the replacement cost method to estimate the value of these duties, assuming that such services would be purchased commercially when a person in the household was ill. Unit prices for household production were based on 2008 average wage rates for commercially available domestic services and child care. Leisure time comprised social and community interactions and recreation and leisure activities. We used a conventional opportunity cost method to estimate the value of changes in leisure time, applying one third of average weekly earnings for men and women during 2008.

We multiplied estimates of leisure time and household production hours per day, derived from the Australian Bureau of Statistics 2006 Time Use Survey, by estimates of days of absenteeism associated with illness and days out of role (i.e., days on which people are unable to carry out normal activities such as attending educational courses or performing household duties) as a result of illness, derived from the 2004–2005 NHS for smokers and ex-smokers and categorized according to gender and workforce status (working, not in the labor force, retired). We then used activity-specific unit prices to assign a monetary value to the net difference in days of household production and leisure time between smokers and ex-smokers (i.e., the gain).

Health sector cost estimation. Although the Disease Costs and Impact Study¹⁹ is a descriptive investigation apportioning total health sector costs across diseases via a systematic costing methodology and the disease categorizations used in the Australian Burden of Disease Study, it does not provide costs apportioned to risk factors. To estimate the health sector costs of tobacco smoking, we calculated the portion of total health sector costs attributable to diseases related to smoking using population-attributable risk fractions.⁴

Moreover, the Disease Costs and Impact Study provides annual estimates and includes all health sector costs associated with the treatment of both incident and prevalent cases of disease. We did not attempt to model lifetime health expenditure costs from these data but maintained a conservative approach, assuming that annual health sector costs attributed to the diseases associated with smoking would approximate the costs of treating incident cases of disease.

Uncertainty analyses. We conducted multivariable probabilistic uncertainty analyses using @RISK version 4.5 for Excel.²⁷ We modeled input variables as known distributions rather than single values in instances in which uncertainty existed (e.g., life-years remaining). We used a minimum of 4000 simulations with Monte Carlo sampling to estimate a mean and 95% uncertainty interval around each outcome parameter generated.

RESULTS

Table 1 presents demographic data and days of reduced activity resulting from ill health among tobacco smokers and ex-smokers in the 2008 adult Australian population by gender, age group, and workforce status. In comparison with smokers, ex-smokers tended to be older and were more likely to be out of the workforce; also, after retirement age they had more days out of role as a result of illness. Tobacco smokers in the workforce had more days off than ex-smokers.

TABLE 1.

Demographic Characteristics of Smokers and Ex-Smokers, by Gender and Workforce Status: Australia, 2004–2005

	Male Current Smokers	Male Ex-Smokers	Female Current Smokers	Female Ex-Smokers
Age group, y
15–64, no. (95% CI)	1 829 391 (1 747 967, 1 910 816)	1 548 199 (1 466 488, 1 629 910)	1 440 908 (1 362 169, 1 519 647)	1 367 902 (1 289 423, 1 446 381)
≥65, no. (95% CI)	109 022 (88 760, 129 283)	667 528 (633 132, 701 924)	83 396 (65 460, 101 333)	349 654 (317 660, 381 648)
≥15, mean (95% CI)	39.8 (39.1, 40.4)	54.3 (53.6, 55.0)	39.8 (39.1, 40.4)	49.1 (48.3, 49.9)
Workforce membera
Overall % (95% CI)	81 (79, 83)	62 (60, 64)	65 (62, 68)	56 (54, 59)
Days off work, mean (95% CI)	0.38 (0.26, 0.50)	0.33 (0.22, 0.44)	0.47 (0.36, 0.58)	0.33 (0.23, 0.43)
Non–workforce member
Overall % (95% CI)	19 (17, 21)	38 (36, 40)	35 (32, 38)	44 (41, 46)
Days of reduced activity among those aged 15–64 y, mean (95% CI)	2.04 (1.43, 2.65)	2.52 (1.91, 3.12)	1.95 (1.54, 2.36)	1.27 (0.88, 1.66)
Aged ≥65 y
Overall % (95% CI)	5.6 (4.7, 6.7)	30.1 (28.6, 31.7)	5.5 (4.4, 6.8)	20.4 (18.6, 22.2)
Days of reduced activity, mean (95% CI)	0.81 (0.22, 1.40)	1.44 (1.10, 1.78)	1.20 (0.40, 2.00)	2.03 (1.56, 2.50)

Note. CI = confidence interval. Mean numbers of days were measured over a 2-week period.

Source. Data were derived from the 2004–2005 National Health Survey.⁷

^aAmong those aged 15 years or older; includes unemployed individuals seeking work and those aged 65 years or older.

At a population level, we found that if the prevalence of tobacco smoking in the 2008 Australian adult cohort decreased from 23% to 15%, important opportunity cost savings from the reduction of diseases associated with this risk factor could be achieved. According to the friction cost approach, the largest component of the total potential opportunity cost savings modeled would occur in the health sector, followed by workforce and household–leisure (Figure 1).

FIGURE 1.

Total potential opportunity cost savings from reductions in tobacco smoking: Australia, 2008.

Note. FCA = friction cost approach (preferred conservative estimate).

We found that if the prevalence of tobacco smoking could be reduced to the ideal target, potential opportunity cost savings of AUD 491 million to the health sector and AUD 415 million in production (according to the friction cost approach) and leisure could be realized (or AUD 2942 million according to the human capital approach) over time. The 455 000 annual new cases of smoking-related disease could be reduced by 158 000, the 16 000 annual deaths attributed to smoking could be reduced by 5000, and the 205 000 DALYs could be reduced by 71 000.

If the prevalence of tobacco smoking could be reduced from 23% to 19% (the progressive target), potential opportunity cost savings of AUD 246 million to the health sector and AUD 207 million in production (according to the friction cost approach) and leisure (or AUD 1478 million according to the human capital approach) could be realized over time from the reduction by 79 000 in annual new cases of smoking-related disease, the reduction by 3000 in deaths attributed to smoking, and the reduction by 36 000 in DALYs. Tables 2 and 3 present the full benefits modeled, including 95% uncertainty intervals around the point estimates for each target in health status, economic, and financial terms.

TABLE 2.

Health Status and Economic Outcomes of Reductions in Smoking Prevalence: Australia, 2008

Health Status and Economic Outcomes	Thousands, Mean (95% CI)
Ideal target reduction
Yearly
DALYs	71
Incidence of disease	158
Mortality	5
Lifetime
Leisure, d	23 (−825,a 891)
Absenteeism, d	2203
Out of home-based production role, d	373 (−164,a 888)
Early retirement, no.	3
Progressive target reduction
Yearly
DALYs	36
Incidence of disease	79
Mortality	3
Lifetime
Leisure, d	12 (−413,a 445)
Absenteeism, d	887
Out of home-based production role, d	186 (−82,a 444)
Early retirement, no.	2

Notes. CI = confidence interval; DALY = disability-adjusted life-year. DALYs were calculated for all age groups. Leisure and home-based production were calculated for individuals aged 15 years or older. Absenteeism and early retirement were calculated for individuals aged 15 to 64 years. Values are in thousands (rounded).

^aLosses rather than gains possible from achieving the target.

TABLE 3.

Financial Outcomes of Reductions in Smoking Prevalence: Australia, 2008

Financial Outcomes	Millions AUD, Mean (95% CI)
Ideal target reduction
Health sector costs	491
Production costs HCA	2812 (943, 4667)
FCA	285 (101, 576)
Recruitment and training costs	84
Leisure-based production	−18a (−235,a 200)
Home-based production	147 (−65,a 350)
Total production HCA	2942 (1016, 4863)
FCA	415 (−57,a 914)
Total benefits HCA	3433 (1507, 5354)
FCA	906 (434, 1405)
Progressive target reduction
Health sector costs	246
Production costs HCA	1413 (544, 2297)
FCA	143 (53, 288)
Recruitment and training costs	42
Leisure-based production	−9a (−117,a 100)
Home-based production	74 (−33,a 175)
Total production HCA	1478 (568, 2389)
FCA	207 (−30,a 458)
Total benefits HCA	1724 (814, 2635)
FCA	453 (216, 704)

Note. CI = confidence interval; FCA = friction cost approach (preferred conservative estimate); HCA = human capital approach. The total opportunity cost savings (benefits) are the sum of the health sector offsets and the combined workforce, household, and leisure production effects. The mean estimates can be added together in this way, but not the uncertainty intervals, because both the components and the total were run as independent simulations. Recruitment and training costs are included in production gains and losses with the FCA only. Health sector, leisure, and home-based production estimates are based on individuals aged 15 years or older. Production gains and losses and taxation effects are based on individuals aged 15–64 years. Values are net present value with a 3% discount rate.

^aLosses rather than gains possible from achieving the target.

In the sensitivity analyses, the human capital approach and the friction cost approach estimates became lower as the discount rate increased, whereas the friction cost approach estimates increased as the length of the friction period expanded. The results of the complete sensitivity analyses have been reported elsewhere.¹⁶

DISCUSSION

This study contributes important new knowledge about the major impact of reductions in smoking-related deaths and disease on health sector expenditures, workforce production, household production, and leisure time. Overall opportunity cost savings were large with respect to both health sector cost savings and workforce production gains. The selection of reduction targets for a risk factor is the most important policy decision and starting point for analyses. Our modeling of Arcadian means, together with our reliance on expert opinion, ensured that our estimates were realistic and relevant to future disease prevention strategies in the Australian context. Although the California survey on smoking prevalence pertained to cigarette smoking rather than tobacco smoking, we believed that it was realistic and applicable to Australia for the reasons outlined earlier. Similar approaches could be applied in other countries where data are available.

Our estimates provide a fuller picture of what might be achieved in terms of production gains because they were applied to all members of society, not only those in the paid workforce. The ability and time to engage in household and leisure activities is increasingly recognized as essential to maintaining a healthy work–life balance, yet such activities are rarely accorded an economic worth.

The negative health impacts of tobacco use are so large that further smoking reduction programs are warranted to complement the existing Australian legislative and taxation framework. When coupled with comprehensive economic evaluations of potential interventions to reduce tobacco use, the aims of tobacco control organizations can be converted to evidence-based action plans. The roles of mass media education, cheaper pharmacotherapies, and tobacco product price increases are important given that reductions in tobacco use have such substantial benefits to society.

Limitations

Past levels of smoking, rather than current levels, are most influential in determining current health burdens, health sector costs, and production and leisure costs. In the short term, past changes in health status and production arising from reductions in smoking prevalence will be experienced before the effects of future reductions. Because reductions in risk of harm to health after cessation of exposure to tobacco occur over a period of years, the latter benefits will be observed over a more extended interval.²⁸ We did not incorporate time lags to the realization of health benefits in our analyses.

The main limitation of this study was our reliance on cross-sectional data from the NHS to identify the association between tobacco smoking and reduced productivity caused by ill health. The absence of robust longitudinal data means that our results must be regarded as broadly indicative. Other cross-sectional data issues existed in our comparisons of ex-smokers and current smokers. Ex-smokers might have been in poorer health than were current smokers because, as a group, they were older and less likely to be in the workforce (Table 1). This poor health may or may not have been related to past smoking. Ex-smokers of retirement age were more often out of role as a result of ill health than were current smokers. This led to some of the estimated benefits being negative, as reflected in the 95% uncertainty intervals of the household and leisure time estimates. Nonetheless, such a finding might be plausible in that it is not clear from our cross-sectional data whether ill health or the change in risk behaviors occurred first.

Also, self-reported data are less reliable than actual measurement data because people exaggerate or understate, fail to remember accurately, misunderstand questions and diseases, or simply misreport information, and thus the direction of bias in our results is unclear. We did not control for other risk factors or socioeconomic status because we were not interested in subgroup analyses and most of the individuals surveyed had more than 1 risk factor.

Comparisons With Other Studies

It is difficult to compare our findings with previous Australian and international studies for 2 main reasons. First, the significant methodological differences between our study and previous research limit a direct comparison of findings. Previous researchers have variously chosen to model incident or prevalent cases, used different definitions of risk factors and disease, and used different unit costs and inclusion criteria; also, some have focused on disease-specific health sector costs associated with different risk factors. Second, previous analyses have generally excluded household production and leisure time activities. Therefore, our findings provide important new information, especially for policy-making regarding the value of investment in health promotion initiatives from a broader societal perspective.

The cost estimates associated with tobacco smoking reported by Collins and Lapsley¹² were generally higher than those of our study. For example, the net productivity cost of AUD 8 billion estimated by Collins and Lapsley is almost 10 times higher than the AUD 906 million we estimated with the friction cost approach (and 3 times the AUD 3300 million with the human capital approach). A key methodological difference is that Collins and Lapsley quantified costs on the basis of past and present tobacco smoking by comparing a counterfactual scenario in which no tobacco smoking had occurred with the prevalence in the current population. In contrast, our estimates were based on incident cases of tobacco-related disease averted in 2008 and the associated lifetime costs based on these incident cases. Although the study estimates are therefore not directly comparable, they provide a greater understanding of the magnitude of the avoidable costs associated with tobacco smoking.

Conservatism of Estimates

Our estimates are conservative because we assumed that only incident cases of diseases related to smoking would be reduced. Thus, we did not attempt to assess reductions in ongoing disease risk among those who were already ill; those with existing cases of cardiovascular disease might also benefit from quitting smoking. Also, we did not measure the likely benefits in future cohorts of further reductions in smoking prevalence. Such reductions could be expected to provide important lifetime benefits for each population targeted. These benefits would be marginally less pronounced than for the 2008 cohort given that fewer individuals in each population would be exposed to the risk factor assuming that no relapse to smoking occurred.

In addition, the study advisory committee's choice of the ideal Arcadian prevalence reduction of 8% could be argued to represent a rather conservative target because of the acknowledged and continuing effectiveness of Australia's strong health-promoting legislation and taxation framework for tobacco control. However, if larger reductions are observed through introduction or enhancement of existing strategies, then our results may be an underestimate of the potential economic and health status benefits. Interestingly, the Australian National Preventative Health Taskforce set the same target reduction independently of this project, providing further evidence that our target of an absolute reduction in smoking prevalence of 8% from current levels was realistic.¹⁰

Our use of the friction cost approach can also be considered a conservative preference, but we believe that this approach was appropriate for our research focus. As a sensitivity analysis, we also used the human capital approach to calculate results. The friction cost approach estimates of workforce production gains were only 13% of the human capital approach estimates, a considerable difference. Using the friction cost approach reduced the relative contribution of workforce production gains as a proportion of the total production benefits to 69%, as compared with 96% according to the human capital approach.

The most appropriate economic methods for quantifying and valuing household production and leisure time remain an area of continued debate.²⁹ Potential limitations include the fact that individuals may perform overlapping activities because of time constraints. Floro and Miles³⁰ reported that about one third of every activity involves at least 1 other simultaneous activity. Therefore, it can be difficult to quantitatively estimate changes in household production and leisure time in a precise manner.

Conclusions

We recommend caution in the interpretation of the opportunity cost savings reported here because these benefits will be achieved only through the adoption of effective interventions that will involve implementation costs. We did not include intervention costs in our analyses, and we assumed that effective interventions exist to achieve the target reductions in smoking prevalence. Furthermore, opportunity cost savings are not estimates of immediately realizable financial savings but, rather, estimates of resources used in current practice that could be available for other purposes.

Substantial health and economic benefits can be gained from sustained reductions in tobacco smoking prevalence in Australia. Such benefits are particularly relevant at a time when current health policy focuses on chronic disease prevention.