Introduction

The preceding chapters have reviewed the extensive scientific evidence about the diverse diseases and other adverse effects caused by tobacco use. Policy actions to control tobacco use have long been motivated and informed by the knowledge that smoking causes multiple diseases and decreases life expectancy. To support policy actions and decision-making based on health evidence, quantitative estimates of the burden of disease associated with smoking in the population are made to characterize the size of the smoking epidemic and the potential benefit of tobacco control. These population-level estimates complement the epidemiologic studies that describe the risks to individuals associated with various smoking patterns.

For diseases attributable to a causal risk factor, such as smoking, epidemiologic methods can be used to estimate the disease burden associated with that risk factor for a particular population. Estimates can be based on several types of indicators, such as premature mortality, excess morbidity, disability-adjusted life years lost, changes in disability-adjusted life expectancy, quality-adjusted life years lost, years of potential life lost (YPLL), and economic costs of illness (U.S. Department of Health and Human Services [USDHHS] 2004).

The estimation of population-attributable fraction (PAF)—the percentage of the disease morbidity or mortality that is attributable to an exposure—is central to calculating burden. The calculation of PAF for a particular risk factor represents a form of quantitative risk assessment (National Research Council 1983). Risk assessment is a systematic approach that translates research findings for the purpose of guiding the implementation and evaluation of public health programs and policies (Samet et al. 2006). The elements of a risk assessment include hazard identification (does the exposure cause disease), exposure assessment (what is the population pattern of the exposure), dose-response assessment (how does risk vary with duration and amount of exposure), and risk characterization (what is the disease burden caused by the exposure). PAF was originally proposed in a classic paper by Levin (1953), and the application of this approach to smoking was described in the 1989 and 2004 Surgeon General's reports (USDHHS 1989, 2004).

Measuring changes in smoking-attributable mortality (SAM) periodically provides an ongoing indication of the burden of disease caused by tobacco use. This information can be used to reinforce the importance of comprehensive tobacco control programs at the national and state levels. For example, policymakers and decisionmakers can compare the impact of tobacco use with that of other risk factors when making decisions about resource allocation and needs (McGinnis and Foege 1993). These estimates are also useful for assessing the impact of changes in the prevalence of smoking or the risks associated with smoking over time.

This chapter first describes methods that are used to estimate the burden of disease attributable to smoking, particularly SAM, and then focuses on updates to the Smoking-Attributable Mortality, Morbidity, and Economic Costs (SAMMEC) system from the Centers for Disease Control and Prevention (CDC). These updates reflect the findings of this report which expands the list of diseases caused by smoking. SAMMEC has also been modified to incorporate recent risk estimates which are based on findings in cohort studies over the last decade (see Chapter 11, “General Morbidity and All-Cause Mortality”).

The chapter then reviews past estimates of smoking-attributable morbidity and discusses how previous estimates could be updated based on the new findings presented in other chapters in this report. Next, this chapter considers approaches to updating the estimated economic costs of smoking and uses the revised SAMMEC model to estimate the economic burden from active and passive smoking in the United States. Finally, the chapter summarizes international estimates of the global burden of smoking and exposure to secondhand smoke. This chapter is limited to considering the mortality risks from cigarette smoking and does not include those of other tobacco products, either singly or in combination with cigarettes.

Prior to the estimates in this report, CDC last published estimates of smoking-attributable morbidity and mortality in 2008. For the period 2000–2004, CDC estimated approximately 393,000 annual smoking-attributable premature deaths from 19 disease categories and 4 adverse health outcomes in infants that were causally associated with smoking (CDC 2008). An additional 740 deaths from residential fires caused by smoking were counted toward that total as were 49,400 deaths from lung cancer and coronary heart disease (CHD) attributed to exposure to secondhand smoke that were computed separately from deaths caused by active smoking. For the 5-year period, the resulting annual total of attributable mortality was 444,000. Other recent estimates are described in Appendix 12.1.

Methodology Used by CDC to Compute Smoking-Attributable Mortality in the United States

The overall approach to estimating SAM includes the following components:

• Identifying those diseases caused by (cigarette) smoking;
• Developing relative risk (RR) estimates for those diseases for current and former smokers in comparison to lifetime nonsmokers;
• Developing estimates of smoking prevalence for the populations and years of interest;
• Estimating disease- and gender-specific PAFs by age group; and
• Applying the PAFs to disease-specific mortality data for the population to estimate SAM.

Understanding the parameters of PAF allows researchers to describe any uncertainties associated with the resulting PAF estimates and acknowledges the cross-sectional nature of the SAM estimates for particular calendar time points. The estimates represent the SAM for a population with a defined smoking prevalence profile and a set of disease-specific RR estimates for a given year, based on the assumption that the RR estimates accurately represent those in the population of interest.

There are several methods that have been used for calculating SAM (see Appendix 12.1). The first approach historically, the PAF calculation, is used most commonly and can be calculated as:

where P is the prevalence of exposure in the population and RR is the relative risk for disease associated with exposure assumed for the population. The formula shows that PAF varies from 0 (if either P = 0 or RR = 1) to approximately 1 at very high values for P or RR (Figure 12.1).

Figure 12.1

The relationship of relative risk (RR) to the population-attributable fraction at different prevalence levels. Source: Michael Thun, unpublished data. Note: RR = relative risk.

This approach currently underlies the SAMMEC methodology (CDC 2011). The PAF and variants have also been referred to as assigned share, excess risk, etiologic fraction, attributable proportion, attributable risk, and incidence density fraction (Levin 1953; Walter 1976; incidence density fraction (Levin 1953; Walter 1976; Rothman 1986; Greenland and Robins 1988; USDHHS 1989; Greenland 1999). These measures estimate the burden (usually premature mortality) from all diseases and conditions combined or from the specific diseases attributed to smoking. When PAF is multiplied by the reported number of deaths in these disease categories, numbers of deaths for a given time period that are attributable to tobacco use can then be estimated. Based on this first application of the attributable risk calculation to available case-control data, Levin (1953) reported that 62–92% of all cases of lung cancer in study populations are caused by smoking.

Methodologic Issues in SAM Calculation

The methodology to estimate PAF for smoking—including sources of data and their limitations, potential confounding factors, sources of uncertainty, and approach to causal inference—has been reviewed and evaluated in detail in the past, including in the 2004 Surgeon General's report (USDHHS 2004), by the U.S. General Accounting Office ([USGAO] 2003) of the U.S. Congress, by an expert panel convened by CDC (SciMetrika 2010), and in the international epidemiologic literature. This section presents an overview of the general uncertainties of SAM estimates and the resulting limitations in their interpretation and application.

SAM estimates are based on assumptions that include some level of uncertainty. Not all uncertainties are encompassed by the confidence intervals (CIs), which reflect the statistical uncertainty (USDHHS 2004). Nonetheless, the estimates provide policymakers and the public with a general understanding of the magnitude of the burden imposed on the nation by cigarette smoking.

SAMMEC is not an annual surveillance system because single-year SAM calculations can be affected by small numbers of cause-specific deaths and random year-to-year variations. Also, as noted elsewhere in this chapter, when prevalence is declining, the smoking-attributable fraction (SAF) and hence the SAM will be underestimated. These effects are moderated by averaging the estimates over 5 years of prevalence data. When repeated periodically on this longer-term timeframe, SAMMEC estimates can provide insights into the consequences of the changing tobacco epidemic. The estimates are particularly useful if updated inputs are available.

Most estimates of SAM do not include mortality caused by cigar smoking, pipe smoking, or smokeless tobacco use. For example, an estimated 1,000 deaths in the United States were attributable to pipe smoking in 1991 (Nelson et al. 1996). This limitation reflects the lack of appropriate RRs related to tobacco products other than cigarettes. To date, these products cause many fewer deaths than cigarettes. However, given the dynamic nature of the tobacco-related environment, assessment of risk due to other tobacco products use is an emerging priority, particularly because of the introduction of tobacco products claiming to reduce exposure (Samet and Wipfli 2009) and increased dual use of tobacco products (Tomar et al. 2010). Dual use (i.e., use of cigarettes and another product) may complicate estimation of SAM, particularly if dual use extends to persons in age ranges where most smoking-caused deaths occur.

Confounding by such factors as alcohol consumption, education, income, blood pressure, diabetes, and other lifestyle factors has also been a concern. Generally, positive confounding has been postulated. Thun and colleagues (2000) examined the potential for confounding by lifestyle factors to bias SAM estimates and found minimal consequences of potential confounding. The representativeness of the Cancer Prevention Study II (CPS-II) cohort has also been reviewed with regard to SAMMEC estimates (Sterling and Weinkam 1987; Levy 2000). This issue has also been considered and set aside as a source of significant bias (Thun et al. 2000; USDHHS 2004). GAO (2003) reviewed CDC's assumptions, methods, and data sources for SAMMEC and concluded that SAMMEC estimates are sound, noting that appropriate attention had been paid by CDC to the issues of confounding and representativeness.

Further concerns relate to the estimation of the prevalence of smoking and the selection of RRs. The RR values used in past SAMMEC publications (Table 12.1) are based on the first 6 years of follow-up of CPS-II (1982–1988) (Thun et al. 1997b; CDC 2011). However, RR estimates associated with smoking can change over time for specific diseases (Thun and Heath 1997; Thun et al. 1997a). Patterns of smoking may change as might the toxicity and resulting risks of tobacco products. For example, compared to earlier cohorts, more recent cohorts of women began smoking at a younger age; consequently the age-specific RRs are higher (see Chapter 13, “Patterns of Tobacco Use Among U.S. Youth, Young Adults, and Adults”). Changes in tobacco product composition, which could affect risk, have occurred over time (USDHHS 2010). Changing rates in the comparison population of never smokers also affect RRs. As reviewed in Chapter 8, “Cardiovascular Diseases,” declining death rates from cardiovascular diseases in lifelong nonsmokers during the past half century indicate that the risk of cardiovascular death has decreased for the overall population.

Table 12.1

Relative risks for adult mortality from smoking-related diseases, adults 35 years of age and older, based on Cancer Prevention Study II, United States.

Other uncertainties may also influence SAM estimates, including potential differences in the strength of association between smoking and disease or death across different racial/ethnic groups, different socioeconomic strata, and different age groups—all of which are potential modifiers of risk. A particular aspect of heterogeneity is related to the age groups for which RRs and SAM are estimated. Most deaths from cardiovascular diseases occur at older ages, and the U.S. population is aging. In 2006, 36% of all deaths from coronary heart disease occurred in persons 85 years of age and older, and 29% occurred in persons 75–84 years of age (Heron et al. 2009). At the same time, most smokers quit smoking, or die because of it, by 75 years of age. With increasing age, particularly above 60–70 years, the RR for death from cardiovascular disease declines sharply. Crude age stratification in the estimation of PAF will not adequately reflect this age-related change, potentially leading to overestimation of the PAF; however, SAMMEC originally included only two age categories (35–64 years of age and 65 years of age and older).

Another issue is that SAMMEC estimates are based on the prevalence of current and former smokers at the present time. However, the deaths that occur during a given year are primarily among persons who began smoking 30–50 years earlier. Many people quit smoking during the later decades of the twentieth century (Malarcher et al. 1997). The RRs of former smokers are lower than those of current smokers for most diseases and for any cohort; the risks for former smokers reflect the distribution of times since quitting. Unless smoking behavior (including cessation) is stable over time, cross-sectional SAM estimates do not accurately reflect the risks of past cohorts of smokers. When the prevalence of smoking is declining, as in the United States (see Chapter 13), the SAMMEC methodology will tend to understate the number of deaths caused by smoking (USDHHS 2004). Table 12.2 shows the annual prevalence of current and former smoking among adults, 35 years of age and older, from 1965–2011. The ratio of former smokers to current smokers has greatly increased over the past half-century for both men and women.

Table 12.2

Annual prevalence of current smoking and former smoking among adults, 35 years of age and older, for selected years; National Health Interview Survey (NHIS), United States, 1965–2011.

Using survey data to derive estimates of exposure (e.g., prevalence of current and former cigarette smoking) is another source of uncertainty in SAM calculations. Although population-based surveys using self-reported data provide reasonably consistent estimates of adult smoking prevalence and are generally considered to be sufficiently accurate for tracking the general pattern of tobacco use in populations, a comparison of smoking self-reports and the biomarker cotinine in National Health and Nutrition Examination Survey (NHANES) data indicates that some underestimation is likely (Caraballo et al. 2001; Brener et al. 2003; USDHHS 2004, 2006, 2010) (see Chapter 13). The ongoing decrease in the percentage of U.S. households possessing landline phones and the decreasing participation rates in telephone surveys of households with landline phones (Steeh et al. 2001; Biener et al. 2004; Delnevo and Bauer 2009) can result in an underestimation of current smoking in landline telephone-based surveys (Blumberg et al. 2008; Delnevo and Bauer 2009) because smokers are more likely to live in wireless-only households than nonsmokers (Blumberg et al. 2006, 2008) and are less likely to participate in health-related surveys (Galea and Tracy 2007). Although this issue would not affect prevalence estimates from the National Health Interview Survey (NHIS), which is based on household sampling, it could have affected state smoking prevalence estimates from the Behavioral Risk Factor Surveillance System (BRFSS) before 2011 (the year cell phone sampling was instituted); states often use BRFSS data to calculate state-specific SAM. Finally, neither NHIS nor BRFSS include institutionalized populations and persons in the military, which prevents the generalization of the results to these groups. A probable net consequence of these survey issues is some underestimation of smoking in the U.S. population in widely cited state and national surveys of smoking behavior (see Chapter 13). Any downward bias in survey estimates of smoking will lead to underestimation of SAM.

2013 Update to SAMMEC Methodology

In 2013, CDC updated its SAMMEC methodology for adults to incorporate RRs based on more recent datasets. The update also refined the age ranges used in SAMMEC to more accurately capture the changes in risk with age. These changes reflect recommendations made by an expert panel convened by CDC in 2009–2010 to review its methodology for estimating SAM in the United States (SciMetrika 2010) and advise on whether updates were needed. The expert panel noted that the SAMMEC methodology had been evaluated repeatedly and was found to provide a credible indication of the mortality burden of the disease consequences of smoking. Thus, the panel did not find a need for substantive changes to the PAF methodology (SciMetrika 2010). However, the panel recommended that RRs be updated and calculated separately, to the extent possible, for individual racial/ethnic groups and older age strata. In addition, the panel noted that as sufficient evidence emerges to conclude that causal associations exist between smoking and new health conditions, data for these additional diseases should be included in future SAM estimates. The specific changes made are described below.

Adult RR Estimates

Subsequent to the previous SAMMEC estimates, Thun and colleagues (2013) pooled data from five large contemporary cohort studies: the National Institutes of Health-AARP Diet and Health Study, the American Cancer Society's CPS-II Nutrition Cohort (a subset of the original CPS-II mortality study), the Women's Health Initiative (WHI), the Nurses' Health Study, and the Health Professionals Follow-Up Study. Each had updated smoking and endpoint information for participants 55 years of age and older during 2000–2010. Cox proportional hazards regression was used to calculate RRs of current smokers and former smokers, with the latter group limited to those who had quit smoking at least 2 years before the start of the follow-up period. Models were adjusted for age or age plus cohort, race, and educational level. The multivariable-adjusted RR of death was similar for men and women: about 2.8 for all causes in current smokers and 1.5 for all causes in former smokers. RRs for men and women were also very similar for chronic obstructive pulmonary disease (COPD), CHD, and stroke.

Thun and colleagues (2013) compared death rates and RRs in the pooled contemporary cohort with those from previous cohorts, CPS-I (1959–1965) and CPS-II (1982–1988), when both smoking prevalence and the background death rate among never smokers were higher. Among men, RRs for current smokers increased from the 1960s to the 1980s and then plateaued, with the exception of a continuing increase in smoking-related mortality from COPD. Among women, RRs for current smokers increased across all the time periods so that they are now equal to those of men. Death rates by age and gender for lung cancer and COPD increased markedly over time from each of these cohorts to the next (Figure 12.2). Generalizability of the pooled cohort samples to the full U.S. population is a potential concern. The study populations included higher percentages of Whites and highly educated persons than are found in the general population. However, estimated RRs for participants with only a high school education or less were generally similar to, or larger than, the estimates for those with a college education; thus, the overall RRs calculated by Thun and colleagues (2013) are likely to lead to some underestimation of SAM.

Figure 12.2

Changes over time in annual death rates from lung cancer and chronic obstructive pulmonary disease (COPD). Source: Thun et al. 2013. Reprinted with permission from Massachusetts Medical Society, © 2013. Note: Data were obtained from the first (more...)

Another analysis of data from a large, contemporary cohort (Jha et al. 2013) similarly found that adjusting for race/ethnicity, educational level, alcohol consumption, and adiposity had little effect on risk estimates. Jha and colleagues (2013) matched data from the National Death Index (1986–2006) to records of participants, 25 years of age and older, in the NHIS from 1997–2004. In this study, the hazard ratio for death from any cause for current smokers versus nonsmokers was 3.0 for women and 2.8 for men. The lifespan of current smokers was 11–12 years shorter than that of nonsmokers.

The analysis by Jha and colleagues (2013) also has limitations. In particular, NHIS participants are interviewed once only; thus, smoking histories are not updated in the interval between initial interview and death and transitions from current smoker status to former smoker status are not captured. To avoid using outdated smoking history data, which would inevitably misclassify some former smokers as current smokers at the time of death, Jha and colleagues (2013) limited their analysis to NHIS participants from 1997–2004.

Comparable results on changes in RRs over time were found in a large cohort study carried out in the United Kingdom. Pirie and colleagues (2013) used national mortality records through 2010 to assess mortality among a cohort of women who were 52 years of age or older when recruited in 1996–2001. Participants were resurveyed 3 years after recruitment. Those who were current smokers at baseline had a 2.76 mortality rate ratio compared to nonsmokers. Those who remained current smokers at the 3-year resurvey had a mortality rate ratio of 2.97, translating to a lifespan reduction of 11 years.

These studies by Thun and colleagues (2013), Jha and colleagues (2013), and Pirie and colleagues (2013) provide compelling evidence that RRs for smoking have increased over the past decades, particularly for women. Therefore, in 2013, CDC began using RRs derived from a contemporary pooled cohort of adults 55 years of age and older in SAMMEC. This cohort is based on the one created by Thun and colleagues (2013). The original published report from this pooled cohort (Thun et al. 2013) did not include RRs for the age-specific categories needed for SAMMEC calculations (age groups 55–64, 65–74, and ≥75) or for all smoking-related causes of death now included in SAMMEC. Therefore, the investigators responsible for the datasets represented in the pooled cohort provided the RRs shown in Table 12.3 to CDC's Office on Smoking and Health. These estimates include additional data not included in the original report (Thun et al. 2013) from 2 years of follow-up (2009–2010) that became available from the CPS-II Nutrition Cohort after the original publication, as well as updated outcome information from the WHI.

Table 12.3

Relative risks by smoking status and age group, adults 35 years of age and older, United States.

Also, in women 55 years and older, the RRs for “other” vascular conditions (atherosclerosis, aortic aneurysm, other vascular conditions) were modified to exclude data from the National Heart, Lung, and Blood Institute's (NHLBI's) WHI because WHI does not ascertain these conditions. In addition, RRs for the category of smoking-attributable cancers other than lung cancer (a category which includes acute myeloid leukemia, but not other types of leukemia) were calculated excluding all leukemias from WHI because WHI did not distinguish acute myeloid leukemia from other forms of leukemia.

Comparable estimates are not available for adults younger than 55 years of age. The NHIS study by Jha and colleagues (2013) included younger adults and used a nationally representative sample. However, in constraining the dataset to only the most recent years, the study had only a limited dataset that was not sufficiently large to provide stable disease-specific RR estimates for those younger than 55 years of age. Therefore, CDC elected to continue using CPS-II as the RR source for younger adults. Since the RR estimates for populations 55 years of age and older have remained high or increased in recent years (Thun et al. 2013), it is assumed that the CPS-II RR estimates for younger adults are conservative.

Smoking-Attributable Mortality in Adults and Infants, United States, 2005–2009

This section provides the SAM estimates for the United States for the period 2005–2009. The general SAMMEC methodology has been modified as described above. The prevalence data for males and females 35 years of age and older came from NHIS for 1965–2011 (Table 12.2).

Table 12.4 provides average annual SAM for the United States for 2005–2009. The results indicate that cigarette smoking and exposure to tobacco smoke led to at least 480,000 premature deaths annually in the United States. Among adults 35 years of age and older, 163,700 smoking-attributable deaths were caused by cancer, 160,600 by cardiovascular and metabolic diseases, and 113,100 by pulmonary diseases (see Tables 12.4 and 12.5 for detailed results). Smoking during pregnancy resulted in an estimated 1,015 infant deaths annually during 2005–2009. Based on previously published estimates (Max et al. 2012), an estimated 7,330 (4.63%) lung cancer and 33,950 (8.23%) CHD deaths annually were attributable to exposure to secondhand smoke. The average annual SAM estimates also include 620 deaths from smoking-attributable residential fires based on data from the National Fire Protection Association (NFPA) and National Fire Incident Reporting System (Figure 12.3) (Hall 2012).

Table 12.4

Annual deaths and estimates of smoking-attributable mortality (SAM) for adults 35 years of age and older, total and by gender, United States, 2005–2009.

Table 12.5

Average annual smoking-attributable mortality (SAM) for males 35 years of age and older, total and by age group, United States, 2005–2009.

Figure 12.3

Trend in civilian deaths in smoking-material home fires, United States, 1980–2010. Source: Hall 2012. Reprinted with permission from NFPA's report, “The Smoking-Material Fire Problem”, © 2013. Note: NFPA = National Fire (more...)

Smoking caused approximately 254,100 deaths in males (Table 12.5) and 183,300 deaths in females (Table 12.6), for a total of 437,400 deaths in the United States for each year from 2005–2009 from the new list of smoking-related diseases (i.e., the 19 diseases formerly used in the SAMMEC and the five diseases newly linked to smoking in this report). For men, 35 years of age and older, the counts of annual smoking-attributable deaths were 100,300 for cancers, 95,600 for cardiovascular and metabolic diseases, and 58,200 for pulmonary diseases (Table 12.5). For women, 35 years of age and older, the annual SAM was 63,400 for cancers, 65,000 for cardiovascular diseases, and 54,900 for pulmonary diseases (Table 12.6).

Table 12.6

Average annual smoking attributable mortality (SAM) for females 35 years of age and older, total and by age group, United States, 2005–2009.

These results differ from those obtained for 2005–2009 using only CPS-II RRs for the 19 diseases included in the past (Table 12.7). Compared with CPS-II alone, the new RRs produce a higher lung cancer SAM estimate for women (53,400 vs. 48,200) and a lower lung cancer SAM estimate for men (74,300 vs. 77,200). SAM for other cancers is similar across the two methods for both men and women (36,000 with the new RRs vs. 36,900 for CPS-II alone for both genders combined). SAM for pulmonary diseases with the new RRs is also somewhat higher for both genders (113,100 vs. 108,100). The biggest difference is for cardiovascular diseases; SAM for both CHD and other cardiovascular disease SAMs are greatly increased with the new RRs. For men, cardiovascular and metabolic diseases SAM is estimated at 95,600 (vs. 70,300 with CPS-II RRs); for women, cardiovascular SAM is estimated at 65,000 (vs. 41,300 with CPS-II RRs). In total, for cancers, cardiovascular diseases, and pulmonary diseases with both genders combined, the overall annual average SAM estimate for 2005–2009 is 437,400, about 15% higher than would have been calculated using RRs from only CPS-II (382,000).

Table 12.7

Average annual smoking-attributable mortality (SAM) for adults 35 years of age and older, total and by gender, United States, 2005-2009: Cancer Prevention Study II (CPS-II) relative risks (RR's) vs. CPS-II/contemporary cohorts RR's.

In previous infant SAMMEC calculations, the prevalence of prenatal smoking has been obtained from birth certificates. However, the 1989 version of the birth certificate and the 2003 revised birth certificate differ with respect to how smoking is ascertained. State uptake of the 2003 revised birth certificate has been gradual and it is expected that not all states will have implemented the revised birth certificate until 2014. Thus, birth certificate-based smoking data are not comparable across all states during the last decade. Therefore, for this report, the prevalence of prenatal smoking for 2005–2009 was calculated based on data from the Pregnancy Risk Assessment Monitoring System (PRAMS). PRAMS uses a self-administered questionnaire that is completed by women 2–6 months after delivering a live-born infant. Data from 34 states participating in PRAMS (Alaska, Arkansas, Colorado, Delaware, Florida, Georgia, Hawaii, Illinois, Maine, Maryland, Massachusetts, Michigan, Minnesota, Mississippi, Missouri, Nebraska, New Jersey, New Mexico, New York, North Carolina, Ohio, Oklahoma, Oregon, Pennsylvania, Rhode Island, South Carolina, Tennessee, Texas, Utah, Vermont, Washington, West Virginia, Wisconsin, and Wyoming) and New York City were included if the overall weighted response rate for a given state and year was at least 70% for data collected during 2005–2006 and at least 65% for data collected starting in 2007. These PRAMS states represented 70% of live births in the United States in 2009 (Martin et al. 2011). Data from these states were pooled to estimate the national prevalence of smoking during the last 3 months of pregnancy. The average prevalence of prenatal smoking during 2005–2009 was estimated at 12.3% for the 34 PRAMS states and New York City, and this estimate was used as an approximation for the national prevalence of prenatal smoking. Using this source of prevalence data and the new RRs described previously, maternal smoking was estimated to result in 1,015 infant deaths annually in 2005–2009, 582 in males and 431 in females (Table 12.8). Of these, 614 were attributed to prenatal conditions and 401 to SIDS, which is causally related to both maternal smoking during pregnancy and exposure to secondhand smoke as an infant (USDHHS 2006).

Table 12.8

Average annual perinatal deaths attributable to smoking, United States, 2005–2009.

Based on the results of Max and colleagues (2012), an estimated 33,951 CHD deaths and 7,333 lung cancer deaths were due to exposure to secondhand smoke among adults, 20 years of age and older, in 2006. Table 12.9 shows the estimated deaths attributable to exposure to secondhand smoke among U.S. adults, 20 years of age and older, for 2006. Although Max and colleagues (2012) also calculated the number of infant deaths attributable to exposure to secondhand smoke, those estimates are not included here because the relevant health conditions are encompassed in the SAMMEC calculations for infants.

Table 12.9

Deaths, years of productive life lost, and value of lost productivity from coronary heart disease and lung cancer attributable to exposure to secondhand smoke among nonsmoking adults, 20 years of age and older, United States, 2006.

CDC's national smoking-attributable burden totals have also typically included an estimate of smoking-attributable deaths from residential fires (CDC 2005, 2008). NFPA publishes estimates of the average annual number of civilian deaths attributed to smoking-material fires in the United States. These estimates are based on information reported to U.S. municipal fire departments and on information obtained from surveys from the NFPA and National Fire Incidence Reporting System. The average annual number of deaths attributed to smoking-material fires in homes between 2006–2010 was 620 (336 in males, 284 in females) (Hall 2012). These fires also caused an estimated 1,570 civilian injuries and $663 million in direct property damage. As the prevalence of smoking has decreased and requirements for fire-resistant mattresses and upholstery and for “fire-safe” cigarettes have been implemented, the number of smoking-material related fires and deaths has decreased. From 1980–2010, smoking-material fires decreased by 73% and civilian deaths in home structure fires decreased by 70% (Figure 12.3) (Hall 2012).

The average annual SAM for the United States for 2010–2014 (Table 12.15) is at least 480,000 premature deaths caused by cigarette smoking and exposure to secondhand smoke; however, this estimate does not include deaths caused by use of cigars, pipes, other forms of combusted tobacco (e.g., roll-your-own cigarettes, hookah pipes, bidis; see Chapter 13 for description of these various products), nor smokeless tobacco products. As discussed in Chapter 13, the use of these products has increased in recent years; while the methodology for estimating the current population burden from the use of these tobacco products remains under discussion, the number of deaths caused by these products is expected to be in the thousands per year (Shapiro et al. 2000). Also, the estimated burden due to the new causal conclusion that exposure to secondhand smoke causes stroke (see Chapter 8) has not been published. However, based upon the methodology used to compute the CHD deaths caused by exposure to secondhand smoke (Max et al. 2012), over 8,000 stroke deaths annually may be attributable to secondhand smoke. Hence, the average annual total SAM for the United States due to smoking any combusted tobacco product or exposure to secondhand smoke is likely approaching 500,000 per year.

Projected Smoking-Related Deaths Among Youth, United States, 2012

Due to the slow decline in the prevalence of current smoking and initiation among youth and young adults (see Chapter 13, Table 13.19 and Figure 13.26), the annual burden of smoking-attributable mortality can be expected to remain at high levels for decades into the future. Although there is a trend of increasing success of quit attempts (see Chapter 13, Figure 13.17), the risks of premature death from smoking-related illness among former smokers and continuing smokers were higher in cohort study data from 2000–2012 than from earlier cohort studies (see Chapters 11 and 12).

In 1996, CDC projected the future impact of smoking on the health of children and teenagers based on the assumption that current tobacco use patterns would persist across the lives of this cohort of youth (CDC 1996). The future probability that a young adult smoker would die prematurely of a smoking-related cause was estimated to be 32% (CDC 1996). The methodology used to compute this probability of smoking-attributable mortality (PSAM) among young adult smokers is described in Appendix 12.2. The same CDC methodology was used to calculate updated estimates on the number of youth in the United States who will become future smokers and will die prematurely of a smoking-related illness (Tables 12.2.1 and 12.2.2). Because the prevalence of smoking in a birth cohort peaks during early adulthood (see Chapter 13), the average prevalence of smoking among adults 18–30 years of age in each state during 2011–2012 was used to estimate the future prevalence of smoking during early adulthood for the 0–17-years-of-age birth cohort in 2012. The number of persons 0–17 years of age in each state in 2012 was multiplied by the state-specific prevalence of smoking among those 18–30 years of age to calculate the number of youth expected to become regular smokers in each state. Overall, the estimated number of future smokers from the 0–17-years-of-age birth cohort in 2012 in the United States was 17,371,000 (ranging from 22,300 in the District of Columbia to 1,557,800 in Texas) (Table 12.2.1).

Table 12.2.1

Prevalence of current smoking among adults, 18–30 years of age, and projected number of persons, 0–17 years of age, who will become smokers and die prematurely as adults because of a smoking-related illness, by state—United States, (more...)

Table 12.2.2

Proportion (%) of persons, 0–17 years of age, who are projected to become smokers and die prematurely as adults because of smoking-related illness, by state—United States, 2012.

Based on the application of PSAM (0.32) to the state-specific estimates of potential smokers, the overall number of potential future smoking-attributable deaths among youth 0–17 years of age during 2012 in the United States was 5,557,000 (ranging from 7,000 in the District of Columbia to 498,000 in Texas) (Table 12.2.1). Based on the estimated PSAM variance and the state-specific sampling errors on estimates of smoking prevalence from the BRFSS, the estimated number of overall smoking-related deaths in the United States was predicted to vary on a statistical basis by less than or equal to 115,000 deaths. The CIs did not account for other sources of uncertainty, such as future changes in risk of dying from smoking or in quitting rate patterns.

These state-specific estimates were also used to calculate the proportion of youth, 0–17 years of age, who are projected to die prematurely from a smoking-related illness (Table 12.2.2). At the state level, estimates varied almost threefold, from 4.4% in Utah to 12.3% in West Virginia. Overall, 7.5% of youth from the 0–17-years-of-age birth cohort in 2012 in the United States are projected to die prematurely from a smoking-related illness if current rates of smoking and risk of disease associated with smoking persist. Therefore, an estimated 5.6 million youth currently aged 0–17 years of age will die prematurely of a smoking-related illness (Table 12.2.2).

Smoking-Attributable Morbidity Estimates

The most recent previous national estimate of smoking-attributable morbidity for the United States was published for the year 2000 (CDC 2003). For that prior report, the estimates of the prevalence of smoking-related medical conditions were obtained from NHANES III (1988–1994) for current, former, and never smokers to compute the SAFs of morbid conditions. Using the smoking prevalence estimates from BRFSS for the combined years of 1999–2001, it was estimated that 8.6 million (95% CI, 6.9–10.5 million) persons in the United States had 12.7 million (95% CI, 10.8–15.0 million) smoking-attributable serious medical conditions (CDC 2003). These estimates represent the numbers of people living with a smoking-caused disease at the time of the survey (Table 12.10). For diseases with a high mortality rate, such as lung cancer, the prevalence is low because there are few long-term survivors. For many of the conditions, the numbers of self-reported diseases were higher among former smokers than among current smokers. This pattern reflects the quitting of smoking by those who develop a smoking-caused disease, particularly later in life (USDHHS 2004).

Table 12.10

Number and percentage of cigarette smoking-attributable conditions among current and former smokers, by condition, United States, 2000.

In making these estimates, CDC noted that the self-reported data on the prevalence of the medical conditions probably substantially underestimate the true disease burden, particularly for COPD (CDC 2003). Additionally, it was noted that the scope of the medical conditions was limited to the diseases for which the NHANES had survey questions and those that previous Surgeon General's reports had concluded were caused by smoking. Finally, as reviewed in Chapter 11, smoking affects various additional acute and chronic conditions related to the quality of life, health status, and general morbidity.

In the present report, smoking and exposure to secondhand smoke have been causally linked to additional adverse health outcomes that were not considered in the 2003 estimates from CDC. These new causal associations for specific diseases link active smoking with diabetes, colorectal cancer, liver cancer, and tuberculosis, and exposure to secondhand smoke and stroke. For each of these diseases, there is an excess burden attributable to smoking and is potentially avoidable through tobacco control. Additionally, Chapter 11 reviews the evidence regarding the excess morbidity attributable to smoking as reflected in overall health status and general morbidity that may come from still unidentified associations between smoking and disease and through indirect pathways, such as diminished immune function. For the cancers, respiratory and cardiovascular diseases, and other adverse health outcomes, issues that merit attention in future updates of the estimates of smoking-attributable morbidity are outlined below.

Smoking-Attributable Economic Costs

This section covers smoking-attributable economic costs resulting from lost productivity and health care expenditures.

Loss of Productivity

The productivity losses calculated here only represent the present value of future earnings (PVFE) from paid labor and of foregone future imputed earnings from unpaid household work that is unrealized as a consequence of early mortality. Past estimations of PVFE values to use for productivity loss calculations were based on values for the United States in 2000 (Haddix et al. 2003) and an assumed 1% productivity rate and 3% discount rate. Estimates of PVFE for later years were estimated by applying annual changes in the non-seasonally-adjusted employment cost index for total compensation for civilian workers (U.S. Bureau of Labor Statistics n.d.). Male-specific PVFEs were used for both males and females, to account for historical gender bias in compensation.

Productivity losses were estimated by multiplying age-specific YPLL by age-specific PVFE, with total productivity determined by adding subtotals across age and disease categories. Estimates are based on deaths in adults 35–79 years of age.

YPLLs are calculated by multiplying age-specific SAM for each disease category by age group-specific years of life remaining. Years of life remaining are based on U.S. life table data published by the National Center for Health Statistics at CDC. For this report, U.S. life tables were current through 2008; for these calculations, years of life remaining for 2009 were considered equal to those for 2008. Total YPLL is determined by adding subtotals across age and disease categories.

Grosse and colleagues (2009) published updated estimates of PVFE for the United States for 2007. PVFEs for total production for both genders combined were used for estimating productivity losses for 2005–2009, with 2007 values serving as the baseline and values for earlier and later years estimated by annual change in the employment cost index. Due to the use of different assumptions and parameters, the PVFE values published by Grosse and colleagues (2009) were conservative compared to the PVFE estimates published by Haddix and colleagues (2003).

Table 12.11 lists estimated average annual smoking-attributable productivity loss by disease category and gender for the United States from 2005–2009. For 2005–2009, the value of lost productivity attributable to premature death from smoking, based on the 19 diseases used in prior SAMMEC estimates (CDC 2008), was $107.6 billion—$69.6 billion in men and $38 billion in women. Cancers accounted for $44.5 billion of lost productivity costs, and cardiovascular and metabolic diseases accounted for $44.7 billion, and pulmonary diseases accounted for $18.4 billion. Using all-cause mortality, the value of lost SAM would be $150.7 billion—$105.6 billion in men and $45.1 billion in women. Additionally, the value of lost productivity due to premature deaths caused by exposure to secondhand smoke was estimated to be $5.7 billion (Table 12.9). Because these figures account only for lost productivity due to premature mortality and not for lost productivity due to morbidity they significantly underestimate the full value of lost productivity costs due to smoking.

Table 12.11

Average annual value of lost productivity attributable to death from cigarette smoking, adults 35–79 years of age, United States, 2005–2009.

Updated Estimates of Smoking-Attributable Health Care Expenditures

The smoking-attributable health care expenditure is one important component of smoking-attributable economic costs. Although the prevalence of smoking continues to decline in the United States, smoking-related health care expenditures still account for an estimated 5–14% of the total health care expenditures in the United States (Levy and Newhouse 2011; Congressional Budget Office [CBO] 2012). The analytical approaches used to estimate attributable expenditures vary depending on the methodology adopted and the time horizon considered in the analysis. In terms of the former, some studies use a disease-specific approach, in which the attributable expenditure for each major smoking-related disease is estimated as the product of the total health care expenditure for the disease and the share of attributable expenditure. The sum of these disease-specific attributable expenditures provides the total health care expenditure that is attributable to smoking. Other studies take a regression approach, in which health care expenditures are compared by smoking status while controlling for other factors that may differ among current, never, and former smokers. In terms of time horizons, most previous studies on smoking-attributable health care expenditures have taken a cross-sectional approach, in which attributable expenditures were calculated for a point in time. A few took a lifetime approach, in which the attributable expenditures were considered over an individual's life expectancy (Manning et al. 1991; Sloan et al. 2004).

In order to account for the methodologic differences and to provide a reasonable range of smoking-attributable health care expenditures, this section presents estimated attributable expenditures obtained from three different approaches: (1) updated estimates by type of medical services based on the expenditure SAFs in the SAMMEC, (2) updated estimates by age and gender based on an approach originated by Solberg and colleagues (2006), and (3) national estimates by source of fund from a regression analysis using the data from the Medical Expenditure Panel Survey (MEPS) (Xu et al. in press). Although these approaches are not the only ways to estimate the smoking-attributable health care expenditures, they are commonly used. To be consistent with SAM, all of these estimates are cross-sectional. Use of these approaches explores the sensitivity of estimates to the method selected.

Smoking-Attributable Health Care Expenditures by Type of Medical Services

The approach published by Miller and colleagues (1999) involves estimation of expenditure smoking-attributable fractions for five categories of personal health care expenditures—ambulatory care, hospital care, prescriptions, nursing home care, and other care (including home health care, durable and nondurable medical equipment, and other professional services) while accounting for potential confounders. The expenditure SAFs are calculated based on a 2-stage econometric model. Estimates of smoking-attributable health care expenditures are for adults 19 years of age and older and exclude dental expenditures. Values for category-specific health care expenditures are based on data from the Centers for Medicare & Medicaid Services ([CMS] 2012c).

Based on this approach, smoking-attributable medical expenditures were estimated to be $75.5 billion for 1998 (CDC 2002), and $96 billion for 2004 (CDC 2008). Updated estimates for the United States in 2009 were produced based on this approach, using expenditure SAFs for 2004 (CDC n.d.). Expenditures for persons 19 years of age and younger were excluded using 2004 age-specific expenditure data published by CMS (2012a). Updated overall and category-specific expenditure estimates are presented in Table 12.12. Overall, an estimated $132.5 billion of health care expenditures in adults 19 years of age and older were attributable to smoking in 2009, an approximate 38% increase over the 2004 figure. This accounts for 7.6% of all health care expenditures (excluding dental expenditures) in adults ages 19 years and older, consistent with the 6–8% range reported by Warner and colleagues (1999). This figure excludes costs attributable to exposure to secondhand smoke.

Table 12.12

Aggregate health care expenditures attributable to cigarette smoking by type of service among adults, 19 years of age and older, United States, 2009.

Smoking-Attributable Health Care Expenditures by Age and Gender

This approach, using the latest evidence-based RRs of smoking-related disease events, presents an option to distribute smoking-attributable health care expenditures by age, gender, and smoking status. The major advantage of this approach is that it can create estimates closely reflecting the known epidemiologic risks of smoking and benefits of cessation by age and gender. In this approach, per capita health care expenditures by smoking status are first estimated based on the projected 2012 per capita personal health care expenditures for 19 years of age and older from CMS (2012b) and relative cost ratios of health care expenditures for current and former smokers compared to never smokers. These cost ratios are calculated from per-person relative health care costs by smoking status reported by Musich and colleagues (2003). In that study, costs were calculated controlling for age, gender, and the presence of chronic diseases (Musich et al. 2003). With the estimated per capita health care expenditures by smoking status, the national smoking-attributable health care expenditures and expenditure SAFs can be obtained using national prevalence figures for current, former, and never smokers. The national smoking-attributable health care expenditures are then apportioned by age and gender group according to the distribution of hospitalization days for smoking-related diseases.

Specifically, for each smoking-related disease h, age group i, and gender j, hospitalization days are distributed by smoking status using algebraic manipulation of the formula:

$${\mathit{\text{DAYST}}}_{h,i,j}=\sum _{k\text{ε}S}{\mathit{\text{DAYSN}}}_{h,i,j,k}*R{R}_{h,i,j,k}$$

where DAYST and DAYSN are the number of hospitalization days for all smokers, including both current and former smokers, and never smokers, respectively, and the smoking status s is defined by three values for k representing never, current, and former smokers. RR_h,i,j,k is the RR of hospitalization for smoking status k, age group i, and gender j relative to never smokers from the same demographic group (RR = 1.0 for never smokers). After solving for DAYSN, RRs are used to calculate days of hospitalization for current and former smokers. The portion of days across all smoking-related diseases for each age, gender, and smoking status group (PD_i,j,k) is then:

$$P{D}_{i,j,k}=\frac{{\Sigma }_h{\mathit{\text{DAYSN}}}_{h,i,j,k}*R{R}_{h,i,j,k}}{{\Sigma }_h\Sigma k\epsilon s{\mathit{\text{DAYSN}}}_{h,i,j,k}*R{R}_{h,i,j,k}}$$

In practice, the RR for hospitalizations is replaced by the proxy, the RR of mortality. Finally, the resulting proportions can be applied along with estimates of smoking prevalence and national smoking-attributable health care expenditures to estimate the distribution of expenditures by age, gender, and smoking status that reflect relative disease risk.

The number of hospitalization days comes from the 2010 National Hospital Discharge Survey. Hospitalization days are weighted to a nationally representative sample and standard errors on counts and percentages are tabulated using the generalized variance curves provided by the National Hospital Discharge Survey (NHDS) in the public use data set. The prevalence figures for current smokers, former smoker and never smokers are estimated from the 2012 NHIS.

Results from this approach suggest that smoking accounted for 8.66% of total annual health care expenditures in the United States in 2012. The projected personal health care expenditure is $2,031.2 billion in 2013, after excluding dental expenditures and expenditures for persons 19 years of age or younger based on 2004 age specific expenditure data published by CMS (2012a,b). Consequently, the smoking-attributable health care expenditure in 2013 is estimated around $175.9 billion. Of the total, $94.2 billion was contributed by current smokers and $81.7 billion was contributed by former smokers. Table 12.13 presents smoking-attributable health care expenditures by age and gender.

Table 12.13

Aggregate health care expenditures ($ in billions) attributable to cigarette smoking among adults, 35 years of age and older, by age group and gender, United States, 2012.

Smoking-Attributable Health Care Expenditures by Source of Funds

This regression approach is based on a two-part model to calibrate the impact of smoking independently from the impact of other factors that are correlated with smoking and may affect health care expenditures. The data used in the analysis come from the MEPS. The MEPS is a nationally representative survey of the civilian noninstitutionalized U.S. population that provides detailed information on health care use and medical expenditures. MEPS respondents can be directly linked to the NHIS, as they are drawn from the NHIS household samples within the preceding 2 years. The NHIS, designed to be a major source of information on the health of the civilian noninstitutionalized U.S. population, collects detailed smoking history information from respondents. Therefore, respondents' smoking status in the analysis comes from the NHIS. The final data set contains approximately 41,000 observations from the 2006–2010 MEPS that are linked to the 2004–2009 NHIS.

The two-part model, a standard statistical technique for analyzing health care spending data (Finkelstein et al. 2009; CBO 2012), separately estimates the probability of having any medical expenditure in the first part and then estimates annual medical expenditure conditional on having positive expenditures in the second part. The estimates from each part are then combined to estimate annual smoking-attributable health care expenditures.

In each component of the model, health care expenditures excluding dental are considered as a function of the respondent's smoking status and individual sociodemographic and health characteristics. Each respondent in the analysis was categorized as a current smoker, a recent quitter (quit smoking within the last 5 years), a long-term quitter (quit smoking for more than 5 years), or a never smoker. Studies have found that former smokers, and particularly recent quitters, might have higher expenditures than continuing smokers (Fishman et al. 2003, 2006; Hockenberry et al. 2012), as quits are often prompted by symptoms of disease that continue to cause care utilization in the years following quitting. This issue has been addressed in different ways in the existing literature, by excluding recent quitters (Solberg et al. 2006), categorizing recent quitters as current smokers (Sloan et al. 2004; Jha et al. 2013), or mathematically smoothing over expenditure increases in the period after successfully quitting in a way that effectively excludes recent quitters from the analysis of spending (CBO 2012). This analysis includes an independent group of recent quitters to address this issue.

In order to account for differences in sociodemographic and health characteristics by smoking status, all of the regressions included controls for age (18–24 years of age, 25–44 years of age, 45–64 years of age, 65–74 years of age, and 75 years of age and older), gender, race/ethnicity (Whites, Blacks, Hispanics, Others), education (less than high school, high school, some college, college and above), marital status (married or cohabitating, never married, not cohabitating, divorced/separated/widowed), family income as percent of federal poverty level (<100%, 100–124%, 125–200%, 200–399%, 400% and above), indicators of alcohol consumption (excessive drinkers, nonexcessive drinkers, and nonusers), indicators of body weight (underweight, normal weight, overweight, and obese), an indicator of health insurance coverage (when applicable), the receipt of flu shots, use of seatbelt, taking more risks than average person, and belief in own ability to overcome illness without medical help (yes vs. no), and geographic location (four census regions), and the year fixed effect.

A linear probability model was used for the first part. Based on the specification tests, a generalized linear model with a log link and gamma distribution was used in the second part (Manning and Mullahy 2001). In addition to the total health care expenditure, separate two-part models were also performed by source of fund (out-of-pocket, private, Medicaid, and Medicare, other federal, and others). In all analyses, bootstrapped standard errors were obtained.

Smoking-attributable fractions of health care expenditures are estimated using the following formula:

$${\mathit{\text{SAFE}}}_{ij}=\frac{{\mathit{\text{EXPS}}}_{ij}-{\mathit{\text{EXPNS}}}_i}{{\mathit{\text{EXPNS}}}_i}$$

where EXPS_ij is the predicted level of expenditures given individual i's smoking status j (current smokers, recent quitters, or long term quitters), whereas EXPNS_i is the predicted level of expenditures if individual i had never been a smoker and SAFE_ij represents the various attributable fractions. Specifically, using the estimated model, each individual's EXPS_ij is obtained as the product of the predicted probability of having an expenditure given the smoking status and the predicted expenditure conditional on the expenditure being positive. The calculation for EXPNS_i is the same except that the smoking indicator is set to never smoking. The SAFE_ij of the population is obtained by averaging over the population for each current or former smoker.

One limitation, however, is that the MEPS expenditure estimates have been shown to be 38% lower than comparable estimates from the Personal Health Care Expenditures reported by CMS, since the MEPS sample is limited to noninstitutionalized civilians and it does not include costs for some services such as long-term care stays longer than 45 days (Sing et al. 2006).

Therefore, the annual expenditures presented are estimated based on the 2010 National Health Expenditures by health insurance enrollment from CMS (2012b) and SAFs by source of fund estimated from the MEPS. Specifically, these expenditures are calculated as the product of SAFs estimated via MEPS by total health care expenditures for the corresponding category reported.

The estimates from the NHIS-linked MEPS data suggest that 8.7% of total health care expenditure was attributable to cigarette smoking between 2006–2010. Based on the 2010 Personal Health Care Expenditures report by CMS, smoking contributed to more than $170 billion in health care expenditures in total (Table 12.14). In particular, roughly 3.4% of out-of-pocket health care expenditure (approximately $8.5 billion), 4.4% of private health insurance expenditure (approximately $33.7 billion), 15.2% of Medicaid expenditure (approximately $40.1) billion, or 9.6% of Medicare expenditure (approximately $45 billion) was smoking attributable. In other words, more than 60% of annual health care expenditures associated with smoking in the United States were reimbursed by public funds, either Medicaid, Medicare, or other federal funds.

Table 12.14

Smoking-attributable fraction (SAF) and aggregate health care expenditures attributable to cigarette smoking by source of fund, National Health Expenditure Accounts (NHEA), United States, 2010.

Synthesis of Findings

In the preceding sections, smoking-attributable health care expenditures have been estimated based on three different approaches providing a range of figures. None of the total smoking-attributable health care expenditures estimated from these approaches is dependent on a list of specific smoking-attributable conditions. Annual smoking-attributable estimated health care expenditures are between $132.5 billion in 2009 to $175.9 billion in 2013. These estimates are far higher than the $95.9 billion estimate for 2004 by CDC (2008). In comparison, if the CDC estimate for 2004 had been simply adjusted using the Consumer Price Index (all items and medical care-specific) to 2012, the resulting estimates would have been $116.56 billion (all items) or $128.32 billion (medical care-specific).

These estimated attributable expenditures also suggest that smoking has accounted for approximately 7–9% of total annual health care spending in the United States during recent years. In addition to total smoking-attributable health care expenditures, each approach provides specific estimates for different subpopulations in the United States. Results in Table 12.13 suggest that annual attributable health care expenditures may vary by age, from $6.2 billion for those between 35–44 years of age to approximately $50 billion for both those between 65–74 years of age and those 75 years of age and above, while estimates in Table 12.14 imply that more than 60% of the attributable health care expenditures are likely paid by public funds. Based on expenditure SAFs in the SAMMEC, Table 12.12 indicates that smoking contributes $67 billion in expenditures for hospitals, $21 billion in ambulatory care, $10.6 billion in nursing home care, $25.5 billion in prescription drugs, and $8.2 billion in other services.

These three approaches each have their own limitations. The updated SAMMEC approach depends on SAFs for expenditures estimated in 2004 and the data used for the estimation comes from the 2000–2004 MEPS. These attributable fractions are likely outdated and thus may cause an underestimation of smoking-attributable health care expenditures.

Although the second approach originated by Solberg and colleagues (2006) can closely reflect the known epidemiologic risks of smoking and the benefits of cessation by age and gender, the estimated attributable expenditures depend on the relative cost ratios of health care expenditures for current and former smokers compared to never smokers. In addition, the use of mortality RRs as a proxy for the RR of hospitalizations implicitly assumes that the event-fatality rate is constant across smoking status. If instead, for example, current smokers have a lower event-fatality rate than former smokers for a particular smoking-related disease, then the RRs of death for former smokers compared to current smokers would be higher than the RR of events. Consequently, the calculations might overestimate the economic benefits of quitting. Second, if there are differences in case-events by smoking status and those differences are not constant with respect to age, then the distribution of expenditures by age group could also be impacted by using mortality RRs as a proxy. Differences in disease cases by smoking status have not been systematically studied in detail and, therefore, it is difficult to predict the direction and extent of any biases introduced to those estimates.

Finally, although a two-part model is commonly used to model health expenditures, the robustness of the estimates depends on the extent to which all of the factors of health care spending and mortality are accounted for. For example, in an analysis using a similar approach, the CBO (2012) concluded that differences in demographic characteristics account for $130 (12%) of the gap in annual expenditures between current or former smokers and nonsmokers who otherwise resemble smokers in the 45–64 age group, $380 (26%) of the gap in the 65–74 age group, and $460 (26%) of the gap in the 75-and-over age group. In the regression approach, an extensive set of factors are included, in addition to the regional fixed effects. After controlling for similar factors, Jha and colleagues (2013) concluded that additional adjustments did not materially affect their estimated smoking impacts on hazard ratios of mortality. This finding provides indirect evidence to support the specification used in the analysis.

Summary

The estimated SAFs of health care expenditures from all three approaches (7.6%, 8.7%, and 8.7%) are within the range of findings from existing cross-sectional studies which extend from 6.54% in Miller and colleagues (1999) to 14% in Warner and colleagues (1999). Particularly, these estimates are very close to the most recent estimated SAFs reported in the CBO's report (2012), which concluded that 7% of total annual spending on health care in the United States between 2002–2008 was attributable to cigarette smoking. Thus, the various estimates, coming from different data sets and methodologies, are consistent in showing that smoking has continued to cause a significant portion of health care expenditures in the United States.

Total Smoking-Attributable Mortality, 1965–2014

Table 12.15 provides the estimated smoking-attributable mortality for the period 1965–2014. The 2004 Surgeon General's report provided an estimated cumulative total for SAM for 1965–1999 (USDHHS 2004, Chapter 7 and Appendix). The total SAM estimates for 1965–1999 were derived from annual population-attributable risk estimates for the period beginning with the publication of the first Surgeon General's report on the health consequences of smoking in 1964.

Table 12.15

Smoking-attributable mortality, total and by gender, United States, 1965–2014.

Specific details on the methodology used to compute the population-attributable risk estimates for 1965–1999 were provided in the 2004 Surgeon General's report (USDHHS 2004, Chapter 7 and Appendix). Deaths from cigar smoking, pipe smoking, and smokeless tobacco use were not included in these estimates, nor were deaths from fires and exposure to secondhand smoke for the period of 1965–1999. The mortality RR estimates for the 19 disease categories among adults causally associated with smoking were obtained from data from CPS-I and CPS-II: CPS-I data (1959–1965) were used for 1965–1971; CPS-II data (1982–1988) for 1982–1999; and the midpoint RRs between CPS-I and CPS-II were used for 1972–1981.

The average annual SAM estimates for the United States from 2000–2004 have been published (CDC 2008) and briefly described above. For the period 2000–2004, CDC estimated the annual SAM for 19 disease categories based on the mortality RR estimates from CPS-II (1982–1988). Annual estimates of smoking-attributable premature deaths for four health outcomes in infants, deaths from residential fires caused by smoking, and deaths from lung cancer and CHD attributed to exposure to secondhand smoke also were published (CDC 2008).

The 2013 update to SAMMEC methodology was discussed earlier in this chapter. Estimates for 2005–2009 for men and women are shown in Tables 12.5 and 12.6. Updated estimates of infant deaths and deaths attributable to exposure to secondhand smoke for 2005–2009 are shown in Tables 12.8 and 12.9. The average annual SAM for the United States for 2005–2009 are provided in Table 12.7.

From 1965–2009, smoking caused an estimated 5.8 million cancer deaths, 7.0 million cardiovascular and metabolic disease deaths, 3.2 million respiratory disease deaths, and 103,355 infant deaths (Table 12.15). Since 1980, there have been an estimated 32,530 smoking-attributable residential fire-related deaths (Figure 12.3). Since Hall (2012) reports that smoking-material fires dropped by more than one-half when smoke detectors became more widely used in the late 1970s, it can be conservatively estimated that the number of residential fire deaths from 1965–1979 was at least 50,000 (e.g., 2,000 per year). Thus, for the period 1965–2009, the total number of smoking-attributable residential fire-related deaths can be estimated to be about 82,530 (50,000 + 32,530). Deaths attributable to exposure to secondhand smoke have not been estimated for 1965–1990; however, since 1990 the total number of deaths attributable to exposure to secondhand smoke is estimated to be about 3,400 annually for lung cancer and 35,000 for CHD (Steenland 1992; USEPA 1992; CDC 2005, 2008). Since exposures to secondhand smoke were much higher in nonsmokers in earlier decades (USDHHS 2006), it can be estimated that the deaths attributable to exposure to secondhand smoke back to 1965 could be estimated to be about at similar rates for the 35-year period (1965–1999). Thus, for the period 1965–2009, the total number of premature deaths attributable to exposure to secondhand smoke can be estimated to be about 1.8 million. Hence, for 1965–2009, the total estimated deaths attributable to smoking and exposure to secondhand smoke was about 18.0 million.

From 2010–2014, the number of deaths caused by smoking and exposure to secondhand smoke is estimated to continue at levels similar to that from 2005–2009—namely, about 480,000 per year. Therefore, the estimated total for the years 2010–2014 would be approximately 2.4 million additional deaths caused by smoking and exposure to secondhand smoke, and the total estimate for the 50-year period, from 1965–2014, would be approximately 20.4 million deaths caused by smoking and exposure to secondhand smoke.

Summary

Cigarette smoking remains the single leading cause of preventable mortality in the United States and causes a high morbidity burden. The costs of health care are substantial. This chapter reviewed various methods for assessing the disease burden of smoking-related illnesses, including epidemiologic calculations, indirect estimates, and model-based approaches for assessing SAM. These estimates are not strongly biased by potential confounding factors, even though smokers and nonsmokers tend to have different profiles for a number of lifestyle-related risk factors for disease. Economic disease burden estimates assess the costs of smoking to governments and society in general. Both types of assessments provide compelling evidence that programs and policies are needed to continue the progress toward ending the tobacco epidemic.

Conclusions

1. Since the first Surgeon General's report on smoking and health in 1964, there have been more than 20 million premature deaths attributable to smoking and exposure to secondhand smoke. Smoking remains the leading preventable cause of premature death in the United States.
2. Despite declines in the prevalence of current smoking, the annual burden of smoking-attributable mortality in the United States has remained above 400,000 for more than a decade and currently is estimated to be about 480,000, with millions more living with smoking-related diseases.
3. Due to the slow decline in the prevalence of current smoking, the annual burden of smoking-attributable mortality can be expected to remain at high levels for decades into the future, with 5.6 million youth currently 0 to 17 years of age projected to die prematurely from a smoking-related illness.
4. Annual smoking-attributable economic costs in the United States estimated for the years 2009–2012 were between $289–332.5 billion, including $132.5–175.9 billion for direct medical care of adults, $151 billion for lost productivity due to premature death estimated from 2005–2009, and $5.6 billion (in 2006) for lost productivity due to exposure to secondhand smoke.

Implications

Estimates of the attributable burden of disease from smoking have value for policy formulation and decision-making. They may be useful for motivating action and for assigning priorities. For smoking, the enormity of the estimates is a powerful impetus for action. In addition, economic cost-of-illness studies on tobacco-related diseases can help inform policymakers about the economic benefits of supporting comprehensive tobacco use prevention and control programs, and implementing effective policies and regulations to reduce tobacco use in the United States. The estimates of expenditures are essential for cost-effectiveness and cost-benefit analyses. Uniformly, such analyses provide a strong basis for implementing effective tobacco control strategies. Additionally, these estimates may be useful in estimating SAM in other countries (Appendix 12.1).

Acknowledgment

CDC's Office on Smoking and Health acknowledges the contribution of the results shown in Table 12.6 from the following studies: the National Institutes of Health-AARP Diet and Health Study, the American Cancer Society Cancer Prevention Study-II Nutrition Cohort, the Women's Health Initiative, the Nurses' Health Study, and the Health Professionals Follow-Up Study.

References

• American Cancer Society. Cancer Facts & Figures 2013. Atlanta (GA): American Cancer Society; 2013.
• American Lung Association. Trends in COPD (Chronic Bronchitis and Emphysema): Morbidity and Mortality. 2013. [April 23, 2013]. < http://www​.lung.org/finding-cures​/our-research​/trend-reports/copd-trend-report​.pdf>.
• Anderson RN, Minino AM, Hoyert DL, Rosenberg HM. Comparability of cause of death between ICD-9 and ICD-10: preliminary estimates. National Vital Statistics Reports. 2001;49(2):1–32. [PubMed: 11381674]
• Biener L, Garrett CA, Gilpin EA, Roman AM, Currivan DB. Consequences of declining survey response rates for smoking prevalence estimates. American Journal of Preventive Medicine. 2004;27(3):254–7. [PubMed: 15450639]
• Blumberg S, Luke J, Cynamon M. Telephone coverage and health survey estimates: evaluating the need for concern about wireless substitution. American Journal of Public Health. 2006;96(5):926–31. [PMC free article: PMC1470567] [PubMed: 16571707]
• Blumberg S, Luke J, Cynamon M, Frankel M. Advances in Telephone Survey Methodology. Hoboken (NJ): John Wiley & Sons; 2008. Recent trends in household telephone coverage in the United States; pp. 56–86.
• Brener ND, Billy JO, Grady WR. Assessment of factors affecting the validity of self-reported health-risk behavior among adolescents: evidence from the scientific literature. Journal of Adolescent Health. 2003;33(6):436–57. [PubMed: 14642706]
• California Environmental Protection Agency. Health effects of exposure to environmental tobacco smoke. Final report. Sacramento (CA): Office of Environmental Health Hazard Assessment, California Environmental Protection Agency; 1997. [May 14, 2013]. < http://www​.oehha.org​/air/environmental_tobacco/finalets.html>.
• California Environmental Protection Agency. Proposed Identification of Environmental Tobacco Smoke as a Toxic Air Contaminant Part B: Health Effects, Chapter 7 Carcinogenic Effects. Sacramento (CA): California Environmental Protection Agency, Office of Environmental Health Hazard Assessment; 2005. 7.4.1. Breast cancer; pp. 7-76–7-132.
• Callaghan WM, MacDorman MF, Rasmussen SA, Qin C, Lackritz EM. The contribution of preterm birth to infant mortality rates in the United States. Pediatrics. 2006;118(4):1566–73. [PubMed: 17015548]
• Caraballo RS, Giovino GA, Pechacek TF, Mowery PD. Factors associated with discrepancies between self-reports on cigarette smoking and measured serum cotinine levels among persons aged 17 years or older: Third National Health and Nutrition Examination Survey, 1988–1994. American Journal of Epidemiology. 2001;153(8):807–14. [PubMed: 11296155]
• Centers for Disease Control and Prevention. Projected smoking-related deaths among youth—United States. Morbidity and Mortality Weekly Report. 1996;45(44):971–4. [PubMed: 8965796]
• Centers for Disease Control and Prevention. Annual smoking-attributable mortality, years of potential life lost, and economic costs—United States, 1995–1999. Morbidity and Mortality Weekly Report. 2002;51(14):300–3. [PubMed: 12002168]
• Centers for Disease Control and Prevention. Cigarette smoking-attributable morbidity—United States, 2000. Morbidity and Mortality Weekly Report. 2003;52(35):842–4. [PubMed: 12966360]
• Centers for Disease Control and Prevention. Annual smoking-attributable mortality, years of potential life lost, and productivity losses—United States, 1997–2001. Morbidity and Mortality Weekly Report. 2005;54(25):625–8. [PubMed: 15988406]
• Centers for Disease Control and Prevention. Smoking-attributable mortality, years of potential life lost, and productivity losses—United States, 2000–2004. Morbidity and Mortality Weekly Report. 2008;57(45):1226–8. [PubMed: 19008791]
• Centers for Disease Control and Prevention. Smoking-Attributable Mortality, Morbidity, and Economic Costs (SAMMEC). 2011. [March 1, 2012]. < http://apps​.nccd.cdc​.gov/sammec/whats_new.asp>.
• Centers for Disease Control and Prevention. Inpatient surgery. 2013. [April 23, 2013]. < http://www​.cdc.gov/nchs/fastats/insurg​.htm>.
• Centers for Disease Control and Prevention. Smoking-Attributable Mortality, Morbidity, and Economic Costs (SAMMEC). [May 10, 2013]. n.d.; < https://apps​.nccd.cdc​.gov/sammec/show_esaf_data.asp>.
• Centers for Medicare & Medicaid Services. Age and gender. 2012. [May 10, 2013]. < http://www​.cms.gov/Research-Statistics-Data-and-Systems​/Statistics-Trends-and-Reports​/NationalHealthExpendData​/Age-and-Gender.html>.
• Centers for Medicare & Medicaid Services. National health expenditure data. 2012. [May 10, 2013]. < http://www​.cms.gov/Research-Statistics-Data-and-Systems​/Statistics-Trends-and-Reports​/NationalHealthExpendData/>.
• Centers for Medicare & Medicaid Services. Projected. 2012. [May 10, 2013]. < http://www​.cms.gov/Research-Statistics-Data-and-Systems​/Statistics-Trends-and-Reports​/NationalHealthExpendData​/NationalHealthAccountsProjected.html>.
• Centers for Medicare & Medicaid Services. State (residence). 2012. [May 10, 2013]. < http://www​.cms.gov/Research-Statistics-Data-and-Systems​/Statistics-Trends-and-Reports​/NationalHealthExpendData​/NationalHealthAccountsStateHealthAccountsResidence.html>.
• Ciruzzi M, Pramparo P, Esteban O, Rozlosnik J, Tartaglione J, Abecasis B, Cesar J, De Rosa J, Paterno C, Schargrodsky H. Case-control study of passive smoking at home and risk of acute myocardial infarction. Argentine FRICAS Investigators. Factores de Riesgo Coronario en America del Sur. Journal of the American College of Cardiology. 1998;31(4):797–803. [PubMed: 9525549]
• Conen D, Everett BM, Kurth T, Creager MA, Buring JE, Ridker PM, Pradhan AD. Smoking, smoking cessation, [corrected] and risk for symptomatic peripheral artery disease in women: a cohort study. Annals of Internal Medicine. 2011;154(11):719–26. [PMC free article: PMC3111942] [PubMed: 21646555]
• Congressional Budget Office. Raising the Excise Tax on Cigarettes: Effects on Health and the Federal Budget. Washington: Congressional Budget Office; 2012. [July 25, 2013]. < http://www​.cbo.gov/sites​/default/files/cbofiles​/attachments/06-13-Smoking​_Reduction.pdf>.
• Delnevo CD, Bauer UE. Monitoring the tobacco use epidemic III: The host: data sources and methodological challenges. Preventive Medicine. 2009;48(1 Suppl):S16–S23. [PubMed: 18851990]
• Delnevo CD, Gundersen DA, Hagman BT. Declining estimated prevalence of alcohol drinking and smoking among young adults nationally: artifacts of sample undercoverage? American Journal of Epidemiology. 2008;167(1):15–9. [PubMed: 17977896]
• Dietz PM, England LJ, Shapiro-Mendoza CK, Tong VT, Farr SL, Callaghan WM. Infant morbidity and mortality attributable to prenatal smoking in the U.S. American Journal of Preventive Medicine. 2010;39(1):45–52. [PubMed: 20547278]
• Finkelstein EA, Trogdon JG, Cohen JW, Dietz W. Annual medical spending attributable to obesity: payer-and service-specific estimates. Health Affairs. 2009;28(5):w822–w31. [PubMed: 19635784]
• Fishman PA, Khan ZM, Thompson EE, Curry SJ. Health care costs among smokers, former smokers, and never smokers in an HMO. Health Services Research. 2003;38(2):733–49. [PMC free article: PMC1360912] [PubMed: 12785570]
• Fishman PA, Thompson EE, Merikle E, Curry SJ. Changes in health care costs before and after smoking cessation. Nicotine & Tobacco Research. 2006;8(3):393–401. [PubMed: 16801297]
• Fontham ET, Correa P, Reynolds P, Wu-Williams A, Buffler PA, Greenberg RS, Chen VW, Alterman T, Boyd P, Austin DF, et al. Environmental tobacco smoke and lung cancer in nonsmoking women. A multicenter study. JAMA: the Journal of the American Medical Association. 1994;271(22):1752–9. [PubMed: 8196118]
• Galea S, Tracy M. Participation rates in epidemiologic studies. Annals of Epidemiology. 2007;17(9):643–53. [PubMed: 17553702]
• Gavin NI, Wiesen C, Layton C. Review and Meta-Analysis of the Evidence on the Impact of Smoking on Perinatal Conditions Built into SAMMEC II. Atlanta (GA): Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion; 2001. Final Report to the National Center for Chronic Disease Prevention and Health Promotion. Research Triangle Institute, RTI Project No. 7171-010.
• Greenland S. Relation of probability of causation to relative risk and doubling dose: a methodologic error that has become a social problem. American Journal of Public Health. 1999;89(8):1166–9. [PMC free article: PMC1508676] [PubMed: 10432900]
• Greenland S, Robins JM. Conceptual problems in the definition and interpretation of attributable fractions. American Journal of Epidemiology. 1988;128(6):1185–97. [PubMed: 3057878]
• Grosse SD, Krueger KV, Mvundura M. Economic productivity by age and sex: 2007 estimates for the United States. Medical Care. 2009;47(7 Suppl 1):S94–S103. [PubMed: 19536021]
• Haddix AC, Teutsch SM, Corso PS, editors. Prevention Effectiveness: A Guide to Decision Analysis and Economic Evaluation. 2nd ed. New York: Oxford University Press; 2003.
• Hall JR Jr. The Smoking-Material Fire Problem. Quincy (MA): National Fire Protection Association, Fire Analysis and Research Division; 2012.
• He J, Ogden LG, Bazzano LA, Vupputuri S, Loria C, Whelton PK. Risk factors for congestive heart failure in U.S. men and women: NHANES I epidemiologic follow-up study. Archives of Internal Medicine. 2001;161(7):996–1002. [PubMed: 11295963]
• Heron M, Hoyert DL, Murphy SL, Xu J, Kochanek KD, Tejada-Vera B. Deaths: final data for 2006. National Vital Statistics Reports. 2009;57(14):1–134. [PubMed: 19788058]
• Hockenberry JM, Curry SJ, Fishman PA, Baker TB, Fraser DL, Cisler RA, Jackson TC, Fiore MC. Healthcare costs around the time of smoking cessation. American Journal of Preventive Medicine. 2012;42(6):596–601. [PMC free article: PMC3358703] [PubMed: 22608375]
• Jha P, Ramasundarahettige C, Landsman V, Rostron B, Thun M, Anderson RN, McAfee T, Peto R. 21st-century hazards of smoking and benefits of cessation in the United States. New England Journal of Medicine. 2013;368(4):341–50. [PubMed: 23343063]
• Levin M. The occurrence of lung cancer in man. Acta Unio Internationalis Contra Cancrum. 1953;9(3):531–41. [PubMed: 13124110]
• Levy RA. Estimating the numbers of smoking-related deaths. JAMA: the Journal of the American Medical Association. 2000;284(18):2319–20. [PubMed: 11066174]
• Levy DE, Newhouse JP. Assessing the effects of tobacco policy changes on smoking-related health expenditures. In: Bearman PS, Neckerman KM, W L, editors. After Tobacco: What Would Happen If Americans Stopped Smoking? New York: Columbia University Press; 2011. pp. 256–89.
• Malarcher AM, Schulman J, Epstein LA, Thun MJ, Mowery P, Pierce B, Escobedo L, Giovino GA. Methodological issues in estimating smoking-attributable mortality in the United States. American Journal of Epidemiology. 2000;152(6):573–84. [PubMed: 10997548]
• Manning W, Keeler E, Newhouse J, Sloss E, Wasserman J. The Costs of Poor Health Habits. Cambridge (MA): Harvard University Press; 1991.
• Manning WG, Mullahy J. Estimating log models: to transform or not to transform? Journal of Health Economics. 2001;20(4):461–94. [PubMed: 11469231]
• Martin JA, Hamilton BE, Ventura SJ, Osterman MJ, Kirmeyer S, Mathews TJ, Wilson EC. Births: final data for 2009. National Vital Statistics Reports. 2011;60(1):1–70. [PubMed: 22670489]
• Max W, Sung HY, Shi Y. Deaths from secondhand smoke exposure in the United States: economic implications. American Journal of Public Health. 2012;102(11):2173–80. [PMC free article: PMC3477960] [PubMed: 22994180]
• McGinnis JM, Foege WH. Actual causes of death in the United States. JAMA: the Journal of the American Medical Association. 1993;270(18):2207–12. [PubMed: 8411605]
• Miller LS, Zhang X, Rice DP, Max W. State estimates of total medical expenditures attributable to cigarette smoking, 1993. Public Health Reports. 1998;113(5):447–58. [PMC free article: PMC1308416] [PubMed: 9769770]
• Miller VP, Ernst C, Collin F. Smoking-attributable medical care costs in the USA. Social Science and Medicine. 1999;48(3):375–91. [PubMed: 10077285]
• Musich S, Faruzzi SD, Lu C, McDonald T, Hirschland D, Edington DW. Pattern of medical charges after quitting smoking among those with and without arthritis, allergies, or back pain. American Journal of Health Promotion. 2003;18(2):133–42. [PubMed: 14621408]
• National Heart, Lung, and Blood Institute. Morbidity and Mortality: 2012 Chartbook on Cardiovascular, Lung, and Blood Diseases. 2012. [May 14, 2013]. < http://www​.nhlbi.nih​.gov/resources/docs/2012_ChartBook.pdf>.
• National Research Council. Risk Assessment in the Federal Government: Managing the Process. Washington (DC): National Academy Press; 1983. [PubMed: 25032414]
• Nelson DE, Davis RM, Chrismon JH, Giovino GA. Pipe smoking in the United States, 1965-1991: prevalence and attributable mortality. Preventive Medicine. 1996;25(2):91–9. [PubMed: 8860273]
• Pierce JP, Messer K, White MM, Kealey S, Cowling DW. Forty years of faster decline in cigarette smoking in California explains current lower lung cancer rates. Cancer Epidemiology, Biomarkers and Prevention. 2010;19(11):2801–10. [PubMed: 20852009]
• Pirie K, Peto R, Reeves GK, Green J, Beral V, Million Women Study C. The 21st century hazards of smoking and benefits of stopping: a prospective study of one million women in the UK. Lancet. 2013;381(9861):133–41. [PMC free article: PMC3547248] [PubMed: 23107252]
• Pirkle JL, Flegal KM, Bernert JT, Brody DJ, Etzel RA, Maurer KR. Exposure of the U.S. population to environmental tobacco smoke: the Third National Health and Nutrition Examination Survey, 1988 to 1991. JAMA: the Journal of the American Medical Association. 1996;275(16):1233–40. [PubMed: 8601954]
• Samet JM, White RH, Burke TA. Epidemiology and risk assessment. In: Brownson RC, Petitti DB, editors. Applied Epidemiology: Theory to Practice. 2nd ed. New York: Oxford University Press; 2006. pp. 125–63.
• Samet JM, Wipfli H. Unfinished business in tobacco control. JAMA: the Journal of the American Medical Association. 2009;302(6):681–2. [PubMed: 19671910]
• SciMetrika. Review of the Methodology to Assess Smoking-Attributable Mortality in the United States, 2010. Research Triangle Park (NC): Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, Office on Smoking and Health; 2010.
• Shapiro JA, Jacobs EJ, Thun MJ. Cigar smoking in men and risk of death from tobacco-related cancers. Journal of the National Cancer Institute. 2000;92(4):333–7. [PubMed: 10675383]
• Sing M, Banthin JS, Selden TM, Cowan CA, Keehan SP. Reconciling medical expenditure estimates from the MEPS and NHEA, 2002. Health Care Financing Review. 2006;28(1):25–40. [PMC free article: PMC4194973] [PubMed: 17290666]
• Sloan FA, Ostermann J, Conover C, Taylor DH, P G. The Price of Smoking. Cambridge (MA): MIT Press; 2004.
• Solberg LI, Maciosek MV, Edwards NM, Khanchandani HS, Goodman MJ. Repeated tobacco-use screening and intervention in clinical practice: health impact and cost effectiveness. American Journal of Preventive Medicine. 2006;31(1):62–71. [PubMed: 16777544]
• Steeh C, Kirgis N, Cannon B, DeWitt J. Are they really as bad as they seem? Nonresponse rates at the end of the 20th century. Journal of Official Statistics. 2001;17(2):227–47.
• Steenland K. Passive smoking and the risk of heart disease. JAMA: the Journal of the American Medical Association. 1992;267(1):94–9. [PubMed: 1727204]
• Sterling TD, Weinkam JJ. Errors in estimates of smoking-related deaths derived from nonsmoker mortality. Risk Analysis. 1987;7(4):463–75. [PubMed: 3444933]
• Teo KK, Ounpuu S, Hawken S, Pandey MR, Valentin V, Hunt D, Diaz R, Rashed W, Freeman R, Jiang L, et al. Tobacco use and risk of myocardial infarction in 52 countries in the INTERHEART study: a case-control study. Lancet. 2006;368(9536):647–58. [PubMed: 16920470]
• Thun MJ, Apicella LF, Henley SJ. Estimating the numbers of smoking-related deaths (response to RA Levy). JAMA: the Journal of the American Medical Association. 2000;284(18):2319–20. [PubMed: 11066174]
• Thun MJ, Apicella LF, Henley SJ. Smoking vs other risk factors as the cause of smoking-attributable deaths: confounding in the courtroom. JAMA: the Journal of the American Medical Association. 2000;284(6):706–12. [PubMed: 10927778]
• Thun MJ, Carter BD, Feskanich D, Freedman ND, Prentice R, Lopez AD, Hartge P, Gapstur SM. 50-year trends in smoking-related mortality in the United States. New England Journal of Medicine. 2013;368(4):351–64. [PMC free article: PMC3632080] [PubMed: 23343064]
• Thun MJ, Heath CW Jr. Changes in mortality from smoking in two American Cancer Society prospective studies since 1959. Preventive Medicine. 1997;26(4):422–6. [PubMed: 9245660]
• Thun MJ, Lally CA, Flannery JT, Calle EE, Flanders WD, Heath CW Jr. Cigarette smoking and changes in the histopathology of lung cancer. Journal of the National Cancer Institute. 1997;89(21):1580–6. [PubMed: 9362155]
• Thun MJ, Myers DG, Day-Lally C, Calle EE, Flanders WD, Adams SL, Heath CW Jr. Changes in Cigarette-Related Disease Risks and Their Implications for Prevention and Control. Bethesda (MD): U.S. Department of Health and Human Services, Public Health Service, National Institutes of Health, National Cancer Institute; 1997. Age and the exposure-response relationships between cigarette smoking and premature death in Cancer Prevention Study II; pp. 383–413. Smoking and Tobacco Control Monograph No. 8. NIH Publication NIH Publication No. 97-4213.
• Tomar SL, Alpert HR, Connolly GN. Patterns of dual use of cigarettes and smokeless tobacco among U.S. males: findings from national surveys. Tobacco Control. 2010;19(2):104–9. [PMC free article: PMC2989167] [PubMed: 20008157]
• U.S. Bureau of Labor Statistics. ECT frequently asked questions. [May 10, 2013]. n.d.; < http://www​.bls.gov/ncs/ect/#tables>.
• U.S. Department of Health and Human Services. Reducing the Health Consequences of Smoking: 25 Years of Progress. A Report of the Surgeon General. Rockville (MD): U.S. Department of Health and Human Services, Public Health Service, Centers for Disease Control, National Center for Chronic Disease Prevention and Health Promotion, Office on Smoking and Health; 1989. DHHS Publication No. (CDC) 89-8411.
• U.S. Department of Health and Human Services. The Health Consequences of Smoking: A Report of the Surgeon General. Atlanta (GA): U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, Office on Smoking and Health; 2004.
• U.S. Department of Health and Human Services. The Health Consequences of Involuntary Exposure to Tobacco Smoke: A Report of the Surgeon General. Atlanta (GA): U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, Coordinating Center for Health Promotion, National Center for Chronic Disease Prevention and Health Promotion, Office on Smoking and Health; 2006. [PubMed: 20669524]
• U.S. Department of Health and Human Services. How Tobacco Smoke Causes Disease—The Biology and Behavioral Basis for Smoking-Attributable Disease: A Report of the Surgeon General. Atlanta (GA): U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, Office on Smoking and Health; 2010. [PubMed: 21452462]
• U.S. Department of Health and Human Services. Preventing Tobacco Use Among Youth and Young Adults: A Report of the Surgeon General. Atlanta (GA): U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, Office on Smoking and Health; 2012. [PubMed: 22876391]
• U.S. Environmental Protection Agency. Respiratory Health Effects of Passive Smoking: Lung Cancer and Other Disorders. Washington: U.S. Environmental Protection Agency, Office of Research and Development, Office of Air Radiation; 1992. Report No. EPA/600/6-90/0006F.
• U.S. General Accounting Office. CDC's April 2002 Report on Smoking: Estimates of Selected Health Consequences of Cigarette Smoking Were Reasonable. Washington: U.S. General Accounting Office; 2003.
• Walter SD. The estimation and interpretation of attributable risk in health research. Biometrics. 1976;32(4):829–49. [PubMed: 1009228]
• Warner KE, Hodgson TA, Carroll CE. Medical costs of smoking in the United States: estimates, their validity, and their implications. Tobacco Control. 1999;8(3):290–300. [PMC free article: PMC1763946] [PubMed: 10599574]
• Whincup PH, Gilg JA, Emberson JR, Jarvis MJ, Feyerabend C, Bryant A, Walker M, Cook DG. Passive smoking and risk of coronary heart disease and stroke: prospective study with cotinine measurement. British Medical Journal. 2004;329(7459):200–5. [PMC free article: PMC487731] [PubMed: 15229131]
• Xu X, Bishop E, Kennedy SM, Pechacek TF. Annual health care spending attributable to smoking: An update Health Affairs. in press.

Chapter 12 Appendices