SMOKING AS A FACTOR OF REDUCED LIFE EXPECTANCY IN RUSSIA

Smoking is an important factor of preventable mortality in Russia. According to the author’s calculations, based on international estimates of the relative risks of mortality for smokers and Russian data on smoking prevalence, self-assessment of health and mortality by causes of death, in 2017 the life expectancy of smokers was 5.2-5.3 years lower than that of non-smokers, while healthy life expectancy was 2.6-3.2 years lower.


INTRODUCTION
Smoking is an important factor in preventable mortality. Beginning, as a rule, already in adolescence 1 , by the age of 30-35 it becomes the cause of significant differences in mortality between smokers and non-smokers.
Due to the higher prevalence of smoking among men (in 2017, according to the HSE RLMS, 43% of men and 14% of women in Russia smoked 2 ), smoking is especially noticeable for male mortality, including at working ages.

LITERATURE REVIEW
Since 2003, life expectancy in Russia has been growing, making this the longest period of improvement in mortality rates since the mid-1960s . But a comparison with the values of half a century ago does not show an improvement in the mortality structure; the age distribution of mortality from large classes of causes has not changed, which allows us to conclude that there are no signs of a second epidemiological transition in Russia (Vishnevsky 2015).
The work (Shkolnikov et al. 2013) notes that the improvement in mortality dynamics in the 2000s was mainly due to the decrease in mortality from cardiovascular diseases and external causes among adults. The authors note a larger effect on the reduction in mortality from circulatory diseases and especially from cerebrovascular diseases, which makes the current situation in Russia partly similar to the changes both in the countries of Central Europe in the 1990-2000s and in Western European countries in the 1970s, what is called the "cardiovascular revolution." But the question of how sustainable this trend is trend remains open.
The same authors in the study ) note that excess mortality in Russia can be reduced, including by further reducing tobacco consumption (along with a decrease in alcohol consumption and an increase in the effectiveness of the national healthcare system).
The work (Grigoriev et al. 2014) explores the components and factors of the steady decline in mortality in Russia observed since 2003. Using data on mortality by cause of death, the main elements of recent improvements were determined -a combination of factors due to changes in behavior (reduced alcohol consumption, improved diet) and the implementation of healthcare policies (using drugs to lower blood pressure, expanding access and using high-tech medical and surgical interventions), as well as an improved economic situation of the population. The authors conclude that, despite a certain similarity between the Russian trend and the initial stages of mortality reduction in a number of European countries (France, Poland and Estonia), one cannot exclude the possibility of a return to stagnation of mortality due to the insufficient willingness of the authorities and society to pay adequate attention to public health care.
In addition, Russia's significant lag in life expectancy behind not only developed countries, but also behind countries with similar levels of economic development persists. Thus, life expectancy in Russia is significantly lower than the level predicted by the Preston curve, which establishes a relationship between life expectancy and GDP per capita. According to the results presented in the work (Andreev, Shkolnikov 2018), in 2010 the difference between model and actual life expectancy for Russia was 8.7 years and was the highest among the 57 countries included in the calculations.
A significant effect of smoking on mortality has been repeatedly proven, including for Russian data. For example, according to the results of a prospective study of mortality among residents of St. Petersburg over the course of 30 years, mortality from all causes among men was for smokers more than double the mortality of never-smokers (Plavinsky, Plavinskaya 2012). A retrospective study of male mortality in Tomsk in 1990Tomsk in -2001 showed that between the ages 15-74, smoking increases the risk of death by 1.5 times, and at working age -more than 2 times (Efimova et al. 2017). Female smoking is no less dangerous. In a prospective study of mortality among residents of Moscow aged 20 and older, the relative risks of mortality from all causes for moderate smokers and ex-smokers were assessed as 1.86 and 1.25, respectively (Konstantinov et al. 2007).
A study carried out on the panel data of a national representative survey of the Russian Monitoring of the Economic Situation and Health of the National Research University -Higher School of Economics (hereinafter RLMS) showed that smoking is comparable in terms of its influence on mortality to excessive consumption of alcohol: relative mortality risks for smokers and respondents with a dangerous type of alcohol consumption were 1.64 and 1.56, respectively (Denisova 2010).
The tobacco epidemic in Russia began in the Soviet era, when it was widespread mainly among men. In the 1990s, it continued to gain momentum: smoking prevalence among men reached its maximum and then stopped for a long time, while the proportion of women smokers almost doubled over this period (Perlman et al. 2007).
By the end of the 2000s, when the negative consequences of mass smoking became apparent not only to experts, but to society as a whole, the tobacco control policy was significantly strengthened: a ban on smoking in public places was introduced, advertising was prohibited, excise taxes on cigarettes and other tobacco products began to increase noticeably 3 . Over the past decade, WWW.DEMREVIEW.HSE.RU noticeable changes have occurred in smoking in Russia: for the first time in the post-Soviet period, the prevalence of smoking has steadily declined.
In a recent work (Quirmbach, Gerry 2016), it was concluded that the dynamics of smoking in Russia as a whole repeat the main stages of the development of the tobacco epidemic previously observed in other countries. In particular, the authors note a change in the behavior of different generations of smokers, as well as a steady decrease in gender differences in the prevalence of smoking.

Data
The calculations used estimates of the relative risks of mortality from a number of diseases caused by smoking on the data of the second round of the American Cancer Prevention Study CPS II (Oza et al. 2011). The relative risk values used are shown in Table 1. The CPS II survey 4 was conducted by the American Cancer Society over a 24-year period and included observations of 1.2 million people in all 50 states, the District of Columbia and Puerto Rico. The purpose of this large-scale prospective survey was to examine the relationship between individual factors and lifestyle characteristics, including nutrition and bad habits, and the risk of developing cancer and other diseases.
The survey questionnaire included questions about the physical and demographic characteristics of respondents, their personal and family history of cancer and other diseases, the use of drugs and vitamins, the reproductive behavior of women, work in harmful industries, eating habits, smoking and drinking, physical education and sports and other aspects of personal behavior. Over 24 years of observation (from 1982 to 2006), 491,188 deaths were recorded as part of the survey, with the cause of death being recorded for 99.3% of them. Estimates of the relative mortality risks obtained using the survey data have a significantly higher level of detail both by gender and age groups and by the list of diseases compared to similar Russian prospective mortality surveys (Plavinsky, Plavinskaya 2012; Konstantinov et al. 2007;Efimova et al. 2017).
To estimate the prevalence of smoking and health status, data from the RLMS of the Higher School of Economics 5 from 1994 to 2017 were used.
The main source of information for estimating mortality from diseases caused by smoking is the Russian NES 6 database on fertility and mortality (mortality rates by cause of death and average annual population for five-year age and gender groups, 1994-2017). 4 See (Calle et al. 2002), and the American Cancer Society site survey URL: https://www.cancer.org/research/weconduct-cancer-research/epidemiology/cancer-prevention-study -2.html 5 Russian monitoring of the economic situation and public health of the HSE (RLMS HSE), conducted by the Higher School of Economics National Research University and Demoscope LLC with the participation of the

Estimated smoking prevalence and poor / good health by gender and age
To assess the prevalence of smoking, RLMS data for 1994-2007 were used. Smokers were considered those respondents who gave a positive answer to the question "Do you currently smoke?", and ex-smokersthose who gave a negative answer to the question about smoking at present and a positive answer to the question "Have you ever smoked?" In the research literature, the problem of accounting for the delayed effect of smoking on health and mortality is addressed in different ways. One possible solution is the Peto-Lopez approach (Peto et al. 1992), which involves the use of the smoking impact ratio (SIR) on public health. The value of the impact ratio of smoking, SIR, is determined by comparing mortality from lung cancer in the study population and in the reference group, usually borrowed from the previously mentioned American Cancer Prevention Survey CPS II. The fact is that lung cancer in the vast majority of cases is caused precisely by active or passive smoking. The impact level of smoking, SIR, is defined as: where CLC and NLC are the mortality rates for lung cancer for the studied population as a whole and for those who never smoked, respectively, and S * LC and N * LC are the mortality rates for lung cancer for smokers and non-smokers in the control group. The N * LC/NLC correction factor is necessary in cases where, in addition to smoking, there are other factors that significantly affect mortality from lung cancer, for example, the use of coal for heating and cooking, a high level of incarceration, etc. In the course of our study, an attempt was made to apply the SIR method to Russian data starting from the 2000s. However, due to the very high mortality rate from lung cancer in the first half of the 2000s, the accumulated negative effect of smoking for some groups, in particular for men 45-49 years old, turns out to be higher than 100% (Figure 1). Due to the impossibility of using the SIR coefficient, the proportion of the population affected by the negative effects of smoking was estimated using data on smoking prevalence (National Center ... 2014;Oza et al. 2011). Many authors (Kong et al. 2016;Liutkute et al. 2017) use data on smoking prevalence with a lag of 8, 10, or 20 years to take into account the delayed effect of smoking on health and mortality. The period of observation of smoking in the RLMS began in 1994, for which reason it was decided to use a lag of 10 years, making it possible to track the dynamics of the indicator from 2004 through 2017.
To estimate the share of the population with poor health, RLMS data for 2004-2017 were used. Respondents were considered to have poor health if they answered the question "How do you assess your health?" with the answer "bad" or "completely bad". A similar method of assessment is used in the methodology for calculating the indicator "Healthy life expectancy (years)" 7 .

Mortality from smoking-related diseases
To estimate the mortality from diseases caused by smoking, we calculated the coefficient of additional risk due to smoking, the PAF (population attributable fraction), separately for the types of diseases and age and gender groups: is the prevalence of smoking in the age and gender group j among neversmokers, ex-smokers and smokers, respectively, and (1) and (2) are the relative death risks for ex-smokers and smokers from disease i in the age and gender group j compared to neversmokers. The prevalence of smoking depending on the status of the smoker for various age and gender groups was calculated on the basis of RLMS data with a lag of 10 years. The relative risks of death were taken from studies using the CPS II survey data ( (Oza et al. 2011: table 1) and others) and are presented in table 1.
The number of additional deaths AM (here, mortality attributed to smoking) for a given disease is calculated as AM = OM × PAF, where OM is the total number of deaths from a given cause of death, taken from official mortality data.

Age-specific mortality rates for smokers, ex-smokers and non-smokers
Knowing the relative risks of death (RR), smoking prevalence (p), and mortality rates by cause of death for five-year age groups ( ), we can calculate the mortality rates from disease i for age group j for smokers ( (2) ), ex-smokers ( (1) ) and never-smokers ( (0) ): (2) = * (0) + (1) * (1) + (2) * (2) ; (1) = * (1) Then, using the standard technique for constructing mortality tables 8 , we can calculate the life expectancy for smokers, ex-smokers and never-smokers. Similarly, dividing the totality of people in each age group into two groups, those with poor health and the rest, the values of the healthy life expectancy for these three groups of the population are calculated.

The contribution of changes in mortality by age groups and causes of death to the dynamics of life expectancy (decomposition method)
When analyzing the contribution to the increase in life expectancy of changes in mortality of certain age groups and mortality from various causes, the decomposition method was used (Andreev, Shkolnikov, Begun 2002); see also recent examples of the application of the method in (Timonin et al. 2017;Papanova, Shkolnikov, Timonin 2019). The calculations used a numerical decomposition program from a technical report (Andreev, Shkolnikov 2012

The dynamics of mortality due to smoking
Since 2004, Russia has witnessed a long period of decline in mortality and an increase in life expectancy, against which, according to the calculations, the mortality from diseases caused by smoking was reduced. According to the results, in 2017 smoking was the direct cause of 211.5 thousand deaths among men and 23.4 thousand deaths among women (table 2). The absolute values of mortality due to smoking have been steadily decreasing since 2005, generally repeating the changes that occurred during this period in all-cause mortality. However, data across the entire population mask significant gender differences: the main decrease in tobacco mortality occurred among men.
One possible explanation of what is happening is offered by the theory of the tobacco epidemic (Lopez, Collishaw, Piha 1994). The tobacco epidemic among men began earlier and is now at a more mature stage compared to the female one. For women, smoking in many respects continues to be one of the manifestations of emancipation or, on the contrary, an attribute of more successful male behavior, as well as a possible struggle with stress, overweight, etc. Use of lagged data on smoking prevalence (to take into account the delayed effect of smoking on health, when calculations looked at data with a 10-year shift) only enhances this effect. And if for men smoking in 1994-2007 stabilized (albeit at very high values), the proportion of women smokers during this period increased markedly. As a result, even the favorable dynamics of mortality from cardiovascular diseases in the second half of the 2000s and 2010s did not allow a reduction in the relative rates of tobacco mortality for women ( Figure 2); see also (Kalabikhina, Kuznetsova 2019). Note: Data on smoking prevalence is taken with a 10-year lag in order to take into account the effect of the delayed effects on health.

Figure 2. Standardized mortality rates for all causes and for diseases caused by smoking, 2004-2017, per 100,000
Source: Author's calculations.

Dynamics of life expectancy indicators of smokers and non-smokers
By calculating the mortality tables separately for smokers, ex-smokers and never-smokers (see Appendix), we can compare the values of life expectancy for these three groups, which together give the entire adult population. The largest differences in life expectancy, as expected, are observed for smokers and never-smokers, amounting to 5.3 years for men in 2017 and 5.2 years for women. The fact of quitting smoking significantly (by 2.6 years for men and 3.2 years for women) prolongs the life of quitters (table 3). The dynamics of healthy life expectancy indicators for smokers and non-smokers are presented in table 4. We used the methodology used in Russian official statistics, according to which people are considered healthy who rate their health as average, good or very good. A healthy life expectancy for both smokers and non-smokers is markedly lower than similar indicators of life expectancy, and differences for non-smokers are especially noticeable, since their additional life years compared to smokers are in old age, when self-evaluation of health decreases sharply.
As a result, the gap in life expectancy caused by smoking is reduced: if for life expectancy in 2017 it was more than 5 years, then for healthy life expectancy it is only 2.5-3 years. The healthy life expectancy of smokers depending on smoking status in 2017 was, for men, 59.8 years for smokers and 62.8 years for non-smokers, and for women -63 years for smokers and 65.4 years for non-smokers. Of particular note is the significant gap between life expectancy and healthy life expectancy of the population: even for non-smoking women, the group with the lowest mortality risks, the healthy life expectancy is barely more than 65 years. This is due, at least in part, to the characteristics of the indicator used for the proportion of healthy people based on respondents' self-evaluation. Estimates of mortality from diseases caused by smoking have a number of limitations. Thus, the calculations did not take into account passive smoking, as well as the effect that smoking of pregnant women has on the health of their unborn children. Another important limitation is the use of relative mortality risk values for smokers compared to non-smokers obtained for another population (US population 1982(US population -2006. The accuracy of the estimates directly depends on the quality of the RLMS data used to calculate the proportion of the healthy population and smoking prevalence.

THE CONTRIBUTION OF SMOKING TO CHANGES IN LIFE EXPECTANCY IN 2004-2017
The increase in life expectancy of men in 2017 compared to 2004 was mainly due to a decrease in mortality in older working age from 40 to 60 years, and a significant part of this improvement was associated with a decrease in mortality from diseases caused by smoking (Figure 3). Decomposition of the increase in the life expectancy of women, on the contrary, shows that the change in mortality from diseases caused by smoking practically did not affect the improvement of the situation, and for some age groups (50-54 and 55-59 years) it even contributed to a certain decrease in life expectancy.

Figure 4. Contribution of mortality from various causes to differences in life expectancy of smokers and non-smokers, 2004 and 2017
Source: Author's calculations.
We emphasize that the life expectancy of women smokers from 2004 to 2017 increased more than for non-smokers: the growth for smokers and non-smokers was 7.1 and 5.5 years, respectively (table 3). Similar dynamics were observed for healthy life expectancy: the advantage of non-smokers over smokers decreased by 0.5 years (table 4). Why did this happen? One possible explanation arises when studying the contribution of changes in mortality by age groups and causes of death to the dynamics of life expectancy (Figure 4) using the decomposition method. As can be seen from the data presented, for smokers (both men and women) in 2017, compared with 2004, the advantage of non-smokers over smokers was reduced at younger ages. In particular, the advantage of non-smoking women is especially noticeable in the age range from 50 to 69 years, which is mainly due to a change in mortality from cardiovascular diseases. For men, a similar reduction in the advantage of non-smokers over smokers occurs in younger ages (from 40 to 64 years). At older ages, non-smokers compensate for their lag in the increase in life expectancy, however, since for women this effect manifests itself later, its influence is insufficient and the overall increase in life expectancy of women smokers in the study period is higher.

INTERNATIONAL COMPARISONS
In order to compare the life expectancy of smokers and non-smokers in Russia with other countries, we used data from (Jha, Peto 2014: 62), which summarizes information on four countries where nationally representative studies of the dependence of mortality on the smoking status of the deceased were conducted. Thus, according to the results of such studies, in the UK for 35-yearold men the probability of surviving to 80 is 60% for those who have never smoked, and 26% for smokers, and the differences in life expectancy for these categories are 10 years. In Japan, the probability of survival is 68 and 41%, respectively, for non-smokers and smokers, and the difference in life expectancy is 9 years.

Figure 5. Probability of surviving to a certain age for people over 30 in Russia, the USA and India
Source: Russia -author's calculations; USA - (Jha et al. 2013), India - (Jha et al. 2008). Figure 5 presents comparisons of the situation in Russia with two other countries mentioned in the article (Jha, Peto 2014): the USA (Jha et al. 2013) and India (Jha et al. 2008).
The United States was chosen as a large economically developed country, India as a country with a value of life expectancy closer to Russia's. It should be noted that for both countries we are talking about estimates of mortality caused by smoking at the beginning of the 2000s. For India, estimates were made of the probabilities of surviving to age 70, which is more relevant when comparing with Russian data. For men, they came to 59 and 39%, respectively, for non-smokers and smokers (similar values for Russia were 64 and 47%, respectively). A feature of India is the almost complete absence of gender differences in the effect of smoking on life expectancy: the proportion of women surviving to 70 years of age is 62 and 38% for non-smokers and smokers, respectively.
Comparison of the proportion of those living up to 80 years old with similar American indicators shows a significant lag in Russia in the expected life expectancy of non-smokers. If in the USA 61% of thirty-year-old men and 70% of women live to the age of 80, in Russia only 37 and 57%, respectively, do so. Similar values for smokers are much closer: in the USA -26% of men and 38% of women, in Russia -20% of men and 37% of women. This suggests that, in addition to smoking, in Russia there are other factors that significantly reduce life expectancy, including excessive consumption of alcohol and the low quality of public healthcare (How to overcome ... 2016).

CONCLUSION
Smoking is an important factor in preventable mortality. Long-term studies conducted both in Russia and abroad show that smokers face significantly higher mortality risks from many diseases of the cardiovascular system, respiratory system and neoplasms. This paper provides a quantitative assessment of the effect of smoking on the mortality rate of the Russian population in 2004-2017, using indicators such as standardized mortality rates, life expectancy and healthy life expectancy for smokers and non-smokers. For the estimates, we used international data on the relative risks of mortality from various diseases and Russian data on the prevalence of smoking, ill health and mortality by cause of death.
Mortality from diseases caused by smoking, according to estimates, in 2017 exceeded 220 thousand people, having significantly decreased compared to the beginning of the 2000s, following the decrease in overall mortality from cardiovascular diseases observed during this period. Among women, mortality from smoking has not declined, due to the less mature tobacco epidemic among women: having started later than among men, it has not yet reached peak smoking prevalence rates and the mortality caused by it, and smoking remains attractive to many teenage girls and young women.
According to calculations, smoking significantly reduces life expectancy. In 2017, the differences in life expectancy of smokers compared to never-smokers were 5.3 years for men and 5.2 years for women. The fact of quitting smoking noticeably (by 2.6 years for men and 3.2 years for women) prolongs the life of quitters.
Compared to 2004, the life expectancy of smokers and non-smokers increased significantly, and for smokers, the increase in life expectancy was greater, resulting in a decrease in the advantage of non-smokers in life expectancy in 2004-2017 of more than 1.5 years (in healthy life expectancy -0.5 years). The decomposition of differences in life expectancy between smokers and non-smokers shows that the advantage of women non-smokers mainly decreased in the age range from 50 to 69 years, the result primarily of a change in mortality from cardiovascular diseases. In men, a similar reduction in the benefits of non-smokers compared to smokers occurs in younger ages (from 40 to 64 years). At older ages, however, non-smokers compensate for their lag in growth of life expectancy, since for women this effect appears later, its influence is insufficient and the overall increase in the life expectancy of women smokers in the study period is higher.
Compared with developed countries (USA, UK, Japan), the life expectancy of smokers and non-smokers in Russia shows more noticeable gender differences. Another important feature of Russia is a significant lag in the life expectancy of non-smokers. If in the USA 61% of thirty-yearold men and 70% of women live to the age of 80, in Russia only 37 and 57%, respectively, do so. Similar values for smokers are much closer, which additionally confirms the existence in Russia, in addition to smoking, of other factors that significantly reduce life expectancy, including lifestyle-related factors (alcohol abuse, unbalanced diet, untimely medical treatment, etc.). WWW.DEMREVIEW.HSE.RU