Introduction

Tobacco use has been considered to be the leading modifiable behavioral risk factor and the leading cause of premature death in the United States 1. Healthy People 2010 had listed smoking as a leading health indicator and included the objective of reducing the prevalence of smoking to 12% among US adults aged 18 and older 2. Yet, when the draft version of Healthy People 2010 was circulated, researchers questioned the attainability of the current goal based upon the inability to meet the previous goal (of 15%) 3. In response to the tobacco goal, investigators noted that, where feasible, data-driven analysis can and should be used in setting Healthy People objectives 3, 4.

When examining the progress for a given subgroup or geographical region, investigators tend to focus on the impact of smoking on mortality. Recently, the US Centers for Disease Control and Prevention (CDC) calculated the smoking-attributable mortality (SAM) and years of potential life lost both for the United States and for the individual states between 2000 and 2004 and compared these numbers with rates from 1996 to 1999 5, 6. Because of declines in the smoking prevalence in the majority of states, SAM rates decreased in 49 states and the District of Columbia. Nevertheless, the absolute SAM within states did not decline over time, most likely because of the long-term increases in the population and additional diseases known to be associated with smoking 7.

These studies do not measure the nonfatal impact of smoking, as assessed by health-related quality of life (HRQOL) and such an omission is important, given that smoking affects both morbidity and mortality 8, 9. As noted by the Secretary's Advisory Committee on Health Promotion and Disease Prevention, a single number, such as quality-adjusted life-years (QALYs), would be particularly useful in quantifying the overall health impact of risk factors such as smoking 10, 11. QALYs use preference-based measurements of HQROL to provide an assessment of the overall burden of diseases associated with both mortality and morbidity 12. Analyses of QALYs are especially useful for quantifying the impact of particular modifiable risk factors, analyzing health disparities at the national or local (community) levels and for small sociodemographic subgroups, and examining changes over time.

The main purpose of this study is to examine both the trend and the geographic variation of the overall health burden of smoking in the United States by calculating smoking-related QALYs lost from 1993 to 2008 for the entire nation and for each of the 50 states and District of Columbia. We also examined the impact of the state cigarette excise tax on the state smoking-related QALYs lost. Because tobacco has been the target of interventions at the national, state, and local levels, most of which were enacted over the past two decades, trend data would be particularly valuable 13-15.

Materials and Methods Data and Measurements

Population HRQOL scores were from the 1993 to 2008 Behavioral Risk Factor Surveillance System (BRFSS), the largest ongoing state-based health survey in the United States The BRFSS sampled noninstitutionalized civilian adult residents aged 18 years and older from each of the 50 states, the District of Columbia, and the three US territories 16. In this analysis, we excluded respondents from the three territories because of a lack of detailed mortality data in these areas, giving a total sample size of 3,590,540. The annual sample sizes ranged from 102,263 in 1993 to 406,749 in 2008. The calculations of mean scores and standard error of estimates were adjusted for the complex sampling design of combined state-level data as well as the unequal probability sampling weights and poststratification weights.

Since 1993, the BRFSS asked three questions about respondents' physically unhealthy days, mentally unhealthy days, and days with activity limitation during the past 30 days (with the exception of 2002 when only 22 states asked these questions) 17. To calculate QALYs, these unhealthy days measures need to be converted to preference-based scores 18. We applied the previously constructed formula for estimating EQ-5D index scores, a QALY compatible preference-based HRQOL measure, from the unhealthy days of the BRFSS respondents 18. The respondents' EQ-5D scores were calculated based on the three unhealthy days questions, self-rated general health, and age.

The National Center for Health Statistics produced the National Health Interview Survey (NHIS) Linked Mortality Files by linking eligible adults in the NHIS data to the National Death Index to obtain mortality follow-up through December 31, 2002 19. The NHIS contains sociodemographics and health behavior questions similar to the BRFSS. We used the 1997 to 2000 cohort of the NHIS Linked Mortality File and the annual mortality data (available at: http://wonder.cdc.gov) compiled by the CDC to estimate premature deaths due to smoking.

Statistical Analyses

The smoking associated excess deaths (Δi) in age interval i were estimated based on: 1) hazard ratios (Hi) of current smokers versus nonsmokers estimated from the NHIS Linked Mortality File; 2) smoking prevalence (pi) estimated from the BRFSS; and 3) the mortality rate (mi) for the United States and the individual states from the detailed annual mortality statistics:

QALYs are defined using preference-based HRQOL scores (e.g., EQ-5D index) which provide an assessment of the burden of disease 20. The EQ-5D index uses summary scores with anchors at 0 (death) to 1 (perfect health) that represent population preferences for different health states. Thus, 1 year of life lived at an EQ-5D index value of 0.8 is equal to 0.8 QALYs.

The total QALYs lost contributed by a risk factor is the sum of the QALYs lost in the current year because of the decrease in HRQOL score (morbidity) and the future QALYs lost in the expected life years due to premature deaths (mortality) 20. To compare QALYs lost between states and in different years, we used the proportion of the population to calculate QALYs, which can be converted to absolute total QALYs by multiplying the adult population in each state/year. The QALYs lost due to morbidity contributed by a risk factor is defined as the potential annual QALYs that would be gained if those at risk (current smokers) had a mean EQ-5D index score that was equal to the score of the reference group (nonsmokers). Let pi be the percent of adults who smoked for age interval i, estimated from the BRFSS data. The QALYs lost due to morbidity contributed by smoking is:

where and are the mean EQ-5D scores for the age interval i of nonsmokers and smokers, respectively.

The future QALYs lost in the expected life years due to excess deaths (Δi) is the product of smoking-related excess deaths and the potential quality-adjusted life expectancy that would be gained if those at risk had a mean EQ-5D index score that was equal to the score of the reference group. The quality-adjusted life expectancy is defined as the mean EQ-5D index score multiplied by each corresponding expected life year 20:

where Lexp is life expectancy at 18 years old for the general population of each of the US states.

Results

Figure 1 depicts both the smoking prevalence over time and the EQ-5D index scores for nonsmokers and smokers. From 1993 to 2008, the percent of US adults who smoked decreased from 22.7% to 18.5%, or an 18.4% decrease. During this period, the population mean HRQOL score (EQ-5D index) decreased from 0.886 to 0.865. Although EQ-5D index scores decreased for both nonsmokers and smokers, the rate of decline was much faster for smokers, as illustrated by the difference in EQ-5D scores between nonsmokers and smokers increasing from 0.016 (= 0.889–0.873) in 1993 to 0.047 (= 0.874–0.827) in 2008, or a 194% increase. This trend of widening EQ-5D index score gaps between nonsmokers and smokers was more pronounced for women than for men. Specifically, the difference of EQ-5D index scores between nonsmokers and smokers increased 126% for men and 266% for women. Women also had worse mean HRQOL scores than men did, about 0.024 points lower, and, despite EQ-5D index scores declining for both men and women, the difference of EQ-5D index scores between men and women was relatively stable over time. Of note, despite worsening HRQOL, quality-adjusted life expectancy increased slightly over this time period for both smokers and nonsmokers because of an increase in life expectancy (figure not shown).

Smoking prevalence rate and EQ-5D index score by smoking status among US adults, 1993 to 2008.

Although the percent of adults who smoked declined significantly from 1993 to 2008 (i.e., decreased 14.6% for men and 22.8% for women) (Fig. 2), smoking-related QALYs lost remained relatively stable at approximately 0.0438 QALYs lost per population overall. With regard to gender, smoking contributed 0.0535 QALYs lost per population for men in 2008, much higher than the number for women (0.0339 QALYs in 2008) (Fig. 2). Nevertheless, the trend analysis showed that the smoking-related QALYs lost decreased by 2.5% for men but increased by 12.6% for women.

Smoking rate and smoking-related quality-adjusted life-years (QALYs) lost by gender, 1993 to 2008.

Disaggregating the QALYs lost into morbidity and mortality, the majority of QALYs lost was from mortality, particularly for men. For men, 81% of smoking-related QALYs lost was due to mortality whereas, for women, 68% of QALYs lost was due to mortality (Fig. 3). This difference in the mortality component of QALYs lost between men and women accounted for most of the gender difference in total smoking-related QALYs lost. By contrast, the morbidity component of QALYs lost was approximately the same for men and women. Also, although the smoking-related QALYs lost due to mortality decreased by 11% for men and 4% for women, the QALYs lost due to morbidity increased by 59% for men and 83% for women.

Smoking-related quality-adjusted life-years lost due to mortality and morbidity by gender, 1993 to 2008.

As noted in Table 1, smoking prevalence decreased for nearly all of the states (with the exception of Alabama, Mississippi, Oklahoma, and the District of Columbia) from the 1993 to 1997 to 2004 to 2008 time periods. Utah had the greatest percentage improvement over time (27.7% decline) whereas the District of Columbia had the greatest (or worst) percentage increase in smoking prevalence by 5.0%. Utah also had the lowest smoking prevalence in the nation at 10.5% in 2004 to 2008, whereas Kentucky had the highest smoking prevalence at 27.6%. By contrast, smoking-related QALYs lost increased slightly (by 3.2%) between these two time periods for the entire nation and increased for 37 states. States with the worst smoking-related QALYs lost are located in the southeast and southwest. During the 2004 to 2008 period, the top five states with the most smoking-related QALYs lost were Kentucky, Oklahoma, Mississippi, West Virginia, and Tennessee. The five states with the least smoking-related QALYs lost were Utah, California, Connecticut, Minnesota, and Hawaii. This pattern resembled the ranking of smoking prevalence, with the correlation coefficient between the smoking prevalence and smoking-related QALYs lost being 0.87. Regarding percentage changes from 1993 to 1997 to 2004 to 2008, smoking-related QALYs lost decreased more than 5% in Connecticut, Utah, California, District of Columbia, New York, Colorado, and New Jersey but increased more than 15% in Mississippi, Arkansas, Oklahoma, Kentucky, Alabama, Tennessee, and New Mexico. The percentage change in smoking-related QALYs lost between the 1993 to 1997 period and the 2004 to 2008 period also was positively correlated to the percentage change in smoking prevalence (r = 0.63).

Table 1. Smoking prevalence and smoking-related QALYs lost by US states for 1993 to 1997, 1998 to 2003, and 2004 to 2008 Smoking prevalence QALYs lost State 1993–1997 (%) 1998–2003 (%) 2003–2008 (%) % change* 1993–1997 1998–2003 2004–2008 % change* United States 22.8 22.6 19.8 −13.4 0.0417 0.0411 0.0431 3.2 Alabama 22.4 24.5 23.5 4.7 0.0512 0.0516 0.0606 18.2 Alaska 27.0 26.7 23.5 −13.0 0.0406 0.0433 0.0438 7.7 Arizona 22.3 21.1 18.5 −17.2 0.0407 0.0395 0.0415 2.0 Arkansas 26.4 25.8 23.5 −11.1 0.0495 0.0558 0.0603 21.7 California 17.9 17.6 14.6 −18.0 0.0344 0.0295 0.0292 −15.3 Colorado 23.0 21.0 18.8 −18.3 0.0361 0.0322 0.0335 −7.2 Connecticut 21.0 20.4 16.6 −20.9 0.0367 0.0333 0.0305 −16.9 Delaware 25.6 24.0 20.7 −19.3 0.0459 0.0423 0.0467 1.7 District of Columbia 17.6 21.1 18.4 5.0 0.0625 0.0573 0.0567 −9.2 Florida 22.9 22.4 19.9 −13.0 0.0448 0.0459 0.0472 5.5 Georgia 21.9 23.4 20.2 −8.1 0.0430 0.0469 0.0478 11.2 Hawaii 19.6 19.4 16.7 −14.6 0.0310 0.0296 0.0321 3.4 Idaho 20.1 20.5 17.6 −12.4 0.0329 0.0345 0.0359 9.2 Illinois 23.8 23.3 20.8 −12.5 0.0420 0.0402 0.0412 −1.9 Indiana 26.8 26.8 25.3 −5.7 0.0447 0.0472 0.0511 14.3 Iowa 22.6 22.8 20.3 −10.3 0.0362 0.0336 0.0378 4.4 Kansas 21.8 21.3 18.7 −14.2 0.0356 0.0383 0.0406 14.0 Kentucky 29.9 30.9 27.6 −7.5 0.0566 0.0624 0.0671 18.6 Louisiana 24.7 24.7 22.5 −9.0 0.0515 0.0505 0.0563 9.3 Maine 24.3 23.4 20.2 −16.8 0.0377 0.0390 0.0422 11.7 Maryland 20.5 21.1 17.6 −14.3 0.0407 0.0408 0.0396 −2.7 Massachusetts 21.6 19.6 17.4 −19.7 0.0378 0.0337 0.0366 −3.1 Michigan 25.5 25.4 21.8 −14.4 0.0425 0.0454 0.0451 6.1 Minnesota 21.4 20.4 18.6 −13.1 0.0314 0.0294 0.0317 0.9 Mississippi 23.4 24.8 24.0 2.6 0.0520 0.0554 0.0636 22.4 Missouri 26.5 26.7 24.0 −9.4 0.0473 0.0499 0.0527 11.4 Montana 21.1 20.6 19.3 −8.4 0.0412 0.0419 0.0450 9.2 Nebraska 21.5 21.8 19.7 −8.5 0.0348 0.0376 0.0359 2.9 Nevada 28.2 28.0 22.4 −20.7 0.0578 0.0551 0.0559 −3.3 New Hampshire 22.9 23.2 19.4 −15.0 0.0371 0.0365 0.0380 2.4 New Jersey 21.0 20.0 17.3 −17.5 0.0375 0.0359 0.0355 −5.4 New Mexico 21.9 22.6 20.4 −6.9 0.0415 0.0440 0.0486 16.9 New York 22.5 22.4 18.9 −16.2 0.0415 0.0358 0.0380 −8.4 North Carolina 26.2 25.5 22.3 −14.9 0.0448 0.0482 0.0499 11.4 North Dakota 21.6 21.6 19.7 −8.9 0.0318 0.0304 0.0348 9.6 Ohio 26.2 26.6 22.8 −13.2 0.0447 0.0481 0.0512 14.6 Oklahoma 24.2 25.5 25.4 4.9 0.0529 0.0542 0.0641 21.2 Oregon 21.8 21.2 18.0 −17.2 0.0394 0.0396 0.0402 2.1 Pennsylvania 23.9 24.3 22.0 −8.0 0.0450 0.0433 0.0477 6.1 Rhode Island 23.7 22.8 18.9 −20.2 0.0402 0.0395 0.0397 −1.3 South Carolina 24.0 25.2 22.2 −7.4 0.0471 0.0505 0.0530 12.7 South Dakota 21.8 23.2 19.6 −10.2 0.0365 0.0392 0.0392 7.4 Tennessee 26.9 25.7 24.5 −8.7 0.0521 0.0532 0.0609 17.0 Texas 22.8 22.3 19.2 −15.8 0.0399 0.0391 0.0430 7.6 Utah 14.6 13.1 10.5 −27.7 0.0229 0.0215 0.0192 −15.9 Vermont 22.7 21.4 18.3 −19.2 0.0365 0.0347 0.0351 −3.8 Virginia 24.0 22.5 19.1 −20.3 0.0373 0.0393 0.0385 3.5 Washington 22.4 21.3 17.2 −23.0 0.0367 0.0372 0.0362 −1.3 West Virginia 26.7 27.5 26.5 −0.7 0.0572 0.0564 0.0634 10.9 Wisconsin 23.1 23.3 20.6 −11.1 0.0325 0.0347 0.0358 10.1 Wyoming 23.0 23.5 21.2 −7.8 0.0385 0.0437 0.0434 12.8 * Percent change of QALYs lost from 1993–1997 to 2004–2008. QALYs, quality-adjusted life-years.

Finally, we examined the association between the state tobacco tax rate (per pack of 20 cigarettes) and the smoking prevalence, smoking-related QALYs lost, and percentage change over time (between 1993–1997 and 2004–2008) for both (Table 2). The analysis demonstrated that states with the lowest state tobacco tax (less than $0.60 in 2009), including South Carolina, Missouri, Mississippi, Virginia, Florida, North Carolina, Louisiana, Georgia, Alabama, and North Dakota, had the highest smoking prevalence, the smallest decline in smoking prevalence, the most smoking-related QALYs lost, and the largest increase in smoking-related QALYs lost.

Table 2. Relationship of state excise tobacco tax rate and state smoking prevalence and smoking related QALYs lost State tobacco tax* Smoking prevalence Smoking related QALYs lost 1993–1997 2004–2008 % change† 1993–1997 2004–2008 % change† Tax rate in 2009  ≥$2.00 22.4% 18.6% −17.0% 0.0401