Russia, and other countries of the former Soviet Union and central and eastern Europe, have undergone dramatic fluctuations in mortality rates.1 One problem in conducting studies on the reasons for these changes is precisely that countries in transition might have limited data systems. There are questions about the validity of official data, and epidemiologic studies are scarce.
We present a quick and effective approach to assess predictors of mortality in a population. It borrows from demographers’ indirect methodology to estimate mortality in countries without vital statistics. A number of indirect demographic methods using survey or census data, often called “Brass techniques,”2,3 have been developed to estimate mortality from information on the survival of close kin (such as spouses and parents) when conventional data are unavailable. These methods use simple information on the number of close kin and on how many of them have died. We modified this method for literate and numerate populations, and showed that the method, based on spouses, is a useful tool to study mortality in Russia.4 In this study, we extended the data collection to form a population-based cohort study with data on individual characteristics of the siblings (eg, education, smoking, and alcohol consumption) to assess the effect of these characteristics on mortality.
Death rates in Russia are among the highest in any industrialized country. The rise in mortality between 1991 and 1994 is unparalleled in contemporary countries with vital statistics. This increase comprised more than 2 million extra deaths above the long-term trends.5 Analyses of routine data have suggested that educational inequalities existed before the transition and increased during the 1990s.6 If the new method was a reliable way of gathering data, we would expect smoking and low education to predict mortality. If the data perform as expected, we can use them to test hypotheses about other factors that might contribute to the high mortality rate in Russia, such as alcohol consumption and marital breakdown.7-9
A cross-sectional survey of a national sample of the Russian population was conducted in the spring of 2001 by the Russian Centre for Public Opinion Research (VCIOM) agency in collaboration with the New Russian Barometer survey program; the program primarily focuses on social and economic impact of the societal transformation. The population sample was selected as follows. The whole Russian Federation was first stratified into 22 regions, and each region was further stratified into urban and rural areas. Within this framework, towns and settlements were randomly selected proportionately to population size. Within these locations, primary sampling units (locations) were randomly drawn. In each primary sampling unit, an address was randomly selected, and interviewers were instructed to seek a face-to-face interview at every nth eligible household. At each address, the interviewer asked for a respondent matching an age–sex–education grid, and if more than 1 respondent was eligible, the person with the most recent birthday was selected. In total, 3254 households containing an eligible respondent were identified. Of these, 1060 declined to be interviewed, 160 were unable to answer as a result of poor health or other reasons, and 34 interviews were started but not completed. Exactly 2000 face-to-face interviews were successfully completed, yielding an overall response rate of 62%.
In addition to questions on age, sex, socioeconomic characteristics, and social and political attitudes, the participants were asked whether they have or had siblings; if so, they were asked when their eldest sibling was born, whether he or she was still alive at the time of the survey, and if not, how old was he or she when he or she died and the cause of death if known. We collected information on 1 sibling only and chose the oldest 1 to maximize the number of deaths. The study participants reported their eldest sibling’s education, marital status, and smoking and drinking habits, either at the time of the survey or just before death. Two indicators of alcohol consumption were available. First, the frequency of drinking vodka or other spirits, and second, the frequency of drinking more than half a bottle of vodka (0.5 L) at one time (binge drinking). In addition, participants answered a question about any lack of food and medicines in their family during their childhood; a positive answer to any of these was taken as an indicator of childhood deprivation of their siblings.
Because we were interested in determinants of adult mortality, only eldest siblings who had reached 25 years of age were included in these analyses. Age- and sex-specific mortality risks for specified calendar periods could be approximated by using deceased eldest siblings as the numerator and the number of eldest siblings as the denominator.4,10,11 In this article, we did not estimate such overall death risks because our primary interest was in the effect of individual-level covariates on mortality within the cohort of eldest siblings. We therefore calculated the Cox proportional hazard ratios (relative risks) to assess the effect of eldest siblings’ characteristics on their survival. Because we relied on survey responders’ reports about their siblings, siblings with unknown vital status were excluded from the analyses. For other variables, we included the response “unknown” as a separate category. For drinking frequency in men, we used those who drank a “few times a year” as the reference group. (The use of abstainers as the baseline category in studies of the effects of alcohol has been criticized because this group can contain exdrinkers in poor health). However, Russian women have low rates of drinking,12,13 and abstainers were the more appropriate reference group. For binge drinking, we used “never” as the reference category in both sexes (because only a minority of drinkers binge and nonbinge drinkers are not abstainers).
We present results separately for men and women. In addition to age-adjusted relative risks (RR), we also report relative risks adjusted for education and smoking. We found that women were more likely to report deaths of eldest siblings than men; the relative risk was 1.23 (95% confidence interval [CI] = 0.93–1.64) for reports by women as compared with men after controlling for year of birth of the sibling. We attribute this to the recognized tendency for men to be poorer at recalling vital events of relatives.14 In subsequent analyses, we therefore treat reports from men and women as separate strata in our Cox models. We also stratified for the calendar period during which the eldest siblings were born. We examined the associations of covariates and mortality by groups of causes of death, but only the group of all cardiovascular deaths among men was sufficiently large enough for analyses.
Of 2000 participants, 1674 (84%) had at least 1 sibling; the mean number of siblings was 1.19. Of these, information on date of birth or date of death was missing for 57 siblings, and in a further 237 cases, the sibling had not reached 25 years, either before death or at the time of the survey. Therefore, we have full information on the survival status of 682 male and 698 female siblings who lived beyond age 25. These siblings were included in the subsequent analyses, and their characteristics are shown in Table 1. The mean age of eldest siblings was 49.5 (standard deviation [SD], 13.9) for brothers and 50.2 (SD, 15.5) for sisters. The mean difference between the year of birth of the male and female eldest siblings and the study respondents was 3.9 (SD, 6.8) and 3.5 (SD, 6.8) years, respectively; quartiles of the age gap between eldest sibling and the respondent are shown in Table 1. The age gap between respondents and their eldest siblings had an approximately normal distribution for both men and women. Among the 203 deaths included in the analyses, 10% occurred before 1950 and 50% occurred after 1989.
Tables 2 and 3 show the association between mortality and personal characteristics of the siblings. We found strong associations between mortality and education. Men and women with primary education had over 4 times higher mortality than those with a completed university degree, although the number of deaths in the highly-educated group was small and the standard error for this group was high. Additional adjustment for alcohol did not change these estimates. There were no substantial effects of reported childhood deprivation or of number of siblings, taken as proxies for childhood circumstances.
For both men and women, divorced people had somewhat higher mortality; never-married women had lower risks and never-married men had higher risks; widowed subjects had the lowest risks. Frequent contact, including coresidence, with a sibling was associated with higher mortality; this probably reflects low geographic mobility and possible dependence of those in poor health. The group with missing values on the contact variable had particularly low relative risks; this is probably because the responders were unaware of whether their sibling had died. Additional adjustment for frequency of contacts with siblings did not materially change the results on other variables; if anything, the rate ratios became somewhat stronger.
Regular smoking was associated with approximately doubled mortality in men and more than 3-fold mortality risk in women, but the proportions of heavy smokers among women was much lower than among men. The effect of smoking was largely independent of other characteristics.
Mortality risk was elevated among both men and women who drank vodka more than once a month. For both sexes, however, the relative risk was reduced by approximately one third after controlling for smoking and education. Men consuming at least half a bottle of vodka per occasion at least once a week had 2.4 times higher risk of death than men binging infrequently or never, after adjustment for smoking and education. Among women, the risk of death was increased in all categories of binging frequency; however, the number of binging women was small and so the confidence intervals were wide and the population attributable risk was small. When women in all categories of binge frequency were combined (never vs. ever), the age-adjusted relative risk of binge drinking was 3.0 (CI = 1.6-5.8), and the relative risk adjusted for education and smoking was 2.8 (CI = 1.4-5.5). We were unable to separate the 2 indices of drinking; the 2 measures are correlated and stratified or mutually adjusted estimates were unstable.
The only mortality subgroup with sufficiently large numbers of deaths was all cardiovascular deaths in men (n = 48). For these deaths, the associations of education and smoking were similar to those observed for all causes of death. Drinking vodka more than once a month was associated with a relative risk of 2.0 (CI = 0.95-4.4), and binge drinking once a week or more often with a relative risk of 3.2 (CI = 1.2-8.3) (not shown in Table).
In this study, using a convenience cohort based on a Russian population sample, we found that all-cause mortality was elevated in men and women with low education, among smokers, and among subjects who frequently drank spirits or who consumed large amounts of vodka at a single session.
In its original form, the indirect methods simply collected information on numbers of total live-born and surviving siblings for use in countries without vital registration systems and where levels of numeracy are low.2,3,15 In these circumstances, the methods require assumptions such as simple patterns of mortality change, and they are subject to considerable bias. In a population with high literacy and numeracy, such as Russia, however, the method should be more reliable and can be extended to an indirect cohort study.
The indirect methodology has been previously piloted in a study of mortality of spouses in Russia.4 The method produced mortality estimates consistent with the official statistics, and mortality risks of husbands were associated with education and material circumstances of their wives.4 In the present study, we applied the same method to siblings of the study participants. Data on siblings have several attractive features, including the fact that information can be obtained over all adult age ranges and on childhood circumstances of the sibling.
The records on eldest siblings’ vital status and their probabilities of death by certain ages are not influenced by the survival of respondents. The information collected for a given eldest sibling by this method would be exactly the same as that obtained from any other valid data collection system. On the other hand, selection of a particular subject (eldest sibling) into the sample can be affected by several mechanisms. First, the probability of a person being included in the sample is proportional to the number of living siblings so that those with a large number of siblings are overrepresented (and no information is collected on only-children). However, we found that sibling group size effects are very small and therefore we report unweighted data; the lack of relation with sibling group size leads us to infer that the exclusion of only-children does not bias the estimates, although this is not certain. The second issue is whether there are biases resulting from the fact that the eldest sibling is usually either the first or second birth. There are birth-order effects on infant and child mortality. However, we have no evidence that there are substantial birth-order effects among adults. Finally, because reports are available only from surviving respondents, correlated mortality between siblings would lead to a downward bias. We did simulations, which suggested that the effects on the estimated level of mortality are small, and in any case, our main interest is in differentials among risk categories as measured by relative risks.
Although we excluded eldest siblings with unknown vital status, the overall mortality risk was somewhat lower than the official statistics suggest. This was true especially for male siblings and for estimates based on reports by male respondents. This could be the result of respondents’ lack of knowledge of a sibling’s death, not counting dead siblings, and possible inflation of age of the sibling. It is likely that there was some under-ascertainment of mortality of eldest siblings who had infrequent contact with the respondent. However, the results within the cohort were similar when we controlled for the frequency of contacts with the respondent and when we excluded siblings who had limited contact with respondents. As long as the underestimation of deaths is not related to the exposures of interest, it does not affect the estimation of socioeconomic differentials and effects of smoking and drinking within the sample of siblings. In addition, it is likely that reporting of variables such as smoking and drinking is more accurate for siblings who are still alive or who died only recently than for those who died a long time ago. This would lead to underestimation of the underlying associations between behaviors and mortality. Reporting of some behaviors, such as drinking, might differ for living and deceased siblings. However, our Russian colleagues did not identify such a problem in an ongoing case-control study of fatal heart disease in Novosibirsk (Malyutina, oral communication, March 2003).
The time period covered by these data do not correspond to a well-defined time period or cohort, but includes information back to the earliest years of births of the siblings of the survey participants. The questionnaire did not ask about any specific time period, and the experience covered is weighted toward the period just before survey date because nearly all respondents provide information on that period (only a relatively small proportion of older people can provide information about earlier periods) and because most deaths occurred over the last decade. Therefore, the risk factors for mortality identified in this study would be most relevant for the 1980s and 1990s.
The validity of the study is further supported by the fact that the results, with respect to the effects of education and smoking, are consistent with the literature. Shkolnikov et al.6 found that in Russia in the early 1990s the ratio of mortality from all causes for those with lower than secondary education compared with higher education was 1.7 in men and 1.5 in women. In our data, the age-adjusted relative risk for these 2 groups were 2.2 (CI = 1.4-3.5) for men and 2.1 (CI = 1.1-4.0) for women. In the Russian part of the US-USSR Lipid Research Clinic cohort,16 mortality from all causes among women with less than high school education was 1.6 times higher than in women with more education. The relative risks in our study were larger than in these studies. This might be in large part because of the use of a finer classification of education or to an increase in educational gradient in mortality during the transformation period. A study in population sample in the Russian city of Novosibirsk in 2001 found a more than 5-fold difference in prevalence and progression of ultrasound diagnosed carotid artery atherosclerosis between men with primary and university education (Ryabikov, written communication, September 2002). On the other hand, the confidence intervals around the relative risks in our study were relatively wide. A selective underreporting of death in siblings with university education is unlikely.
Differentials in mortality by marital status in Russia have not been published. In this study, divorced men showed the increased risk found in other eastern European populations,17,18 but among the never-married, men had higher mortality rates and women had lower mortality rates. The low reported rates for widowed people could reflect a tendency for some respondents to describe the marital status of their deceased bereaved siblings as “married” rather than “widowed” because that might reflect how they remembered their sibling’s marital status. Calendar period of birth or other factors did not account for the observed pattern.
The effects of smoking observed in this study are within the range reported in the literature. For example, in the study of British doctors, mortality from all causes in smokers was approximately double of that in nonsmokers.19 In the U.S. Cancer Prevention Study, the mortality excess in smokers was almost 3-fold higher.20 The effects of smoking in women are similar.21 The somewhat higher estimates of relative risk in women in our study, compared with men, might again be the result of random error. It is also possible that Russian women who smoke (still relatively uncommon22) are a particular group with multiple disadvantages and unmeasured risk factors that account for the apparently large relative risk.
The results on alcohol are intriguing. Circumstantial evidence suggests a link between alcohol consumption and increase in general mortality in Russia,7,8,23 but the role of alcohol in increasing mortality in Russia has been disputed.24,25 Several studies reported high prevalence of heavy or binge drinking in Russia,12,13,26-28 but the results of 2 cohort studies of alcohol and mortality in Russia published so far are inconsistent. One study did not find any association at all,29 and in the second study, mortality from all causes and cardiovascular diseases was approximately doubled in a small group of frequent heavy drinkers (approximately 5% of men in the cohort).30
In Western populations, the association between alcohol intake and mortality from all causes is approximately J-shaped,31 and there is a growing evidence that the elevated mortality among heavy drinkers (the upper part of the J-shaped curve) is at least partly the result of increased cardiovascular risk.32,33 Several recent studies in Western populations suggest that heavy or binge drinking might indeed increase all-cause and cardiovascular mortality.33-38
The present study is consistent with these findings and with the proposition that alcohol might be an important risk factor for mortality in Russia. We found that mortality was increased among men who drank vodka several times a month and among women who drank vodka more than once a month, as well as in men and women who were frequent binge drinkers. The effect of drinking was seen for all causes and for cardiovascular deaths in men. The fact that mortality in men who drank more often than once a week was not higher than among those drinking several times a month could be the result of the fact that frequent drinkers often drink smaller amounts per session, and smaller doses might be less harmful than large ones.
This study cannot answer the important question of whether drinking is causally associated with increased mortality, particularly from cardiovascular diseases. At this stage, it is also impossible to separate the effects of episodic binge drinking and regular heavy drinking. Nevertheless, these data and the literature are consistent with a possibility that alcohol, smoking, and unfavorable social conditions are involved in high mortality in Russia.
The plausibility of the results indicates that the cohort study based on the indirect demographic approach is reasonably reliable and sufficiently sensitive to detect moderately strong effects. It is much quicker and cheaper than a prospective cohort study, and the bias related to selection of subjects is probably smaller than in case-control studies. With a sufficient number of events, the method could also be extended to study trends in mortality by specific subgroup. For exposures that can be estimated from proxy informant recall, and in populations with high literacy and numeracy, this method provides a time- and cost-effective alternative to other study designs.
1.United Nations Children’s Fund. A decade of transition. Regional Monitoring Report No. 8. Florence: UNICEF; 2001.
2.Brass W, Coale AJ, Demeny P, et al. The Demography of Tropical Africa. Princeton, NJ: Princeton University Press; 1968.
3.Hill K, Trussel J. Further developments in indirect mortality estimation. Population Studies. 1977;31:313–334.
4.Bobak M, Murphy M, Pikhart H, et al. Mortality patterns in the Russian Federation: indirect technique using widowhood data. Bull WHO. 2002;80:876–881.
5.Shkolnikov VM, Cornia GA. The population crisis and rising mortality in transitional Russia. In: Cornia GA, Paniccia R. The Mortality Crisis in Transitional Economies. New York: Oxford University Press; 2000:253–279.
6.Shkolnikov V, Leon DA, Adamets S, et al. Educational level and adult mortality in Russia: an analysis of routine data 1979 to 1994. Soc Sci Med. 1998;47:357–369.
7.Leon DA, Chenet L, Shkolnikov V, et al. Huge variation in Russian mortality rates 1984-94: artefact, alcohol, or what? Lancet. 1997;350:383–388.
8.Shkolnikov VM, Nemtsov A. The anti-alcohol campaign and variations in Russian mortality. In: Bobadilla JL, Costello CA, Mitchell F, eds. Premature Death in the New Independent States. Washington, DC: National Academy Press; 1997:239–261.
9.Cornia GA. Labour market shocks, psychosocial stress and the transition’s mortality crisis. Research in Progress 4 Working Paper. United Nations University World Institute for Development Economics Research. Helsinki: UNU/WIDER; 1997.
10.United Nations. Manual X: indirect Techniques for Demographic Estimation. New York: United Nations; 1983.
11.United Nations. Estimation of Adult Survivorship Probabilities From Information on Orphanhood and Widowhood. Readings in population research methodology, vol 2. New York: United Nations; 1983:8;122–153.
12.Bobak M, McKee M, Rose R, et al. Alcohol consumption in a national sample of the Russian population. Addiction. 1999;94:857–866.
13.Malyutina S, Bobak M, Kurilovitch S, et al. Alcohol consumption and binge drinking in Novosibirsk, Russia, 1985-95. Addiction. 2001;96:987–995.
14.Poulain M, Riandey B, Firdion JM. Une experimentation franco-belgique sur la fiabilite des enquetes retrospectives: l’enquete 3B BIS [in French]. Population. 1991;46:65–88.
15.Hill K. The measurement of adult mortality: an assessment of data availability, data quality and estimation methods. In: Chamie J, Cliquet RL, eds. Health and Mortality: Issues of Global Concern. New York: United Nations Population Division; 1999;72–83.
16.Davis CE, Deev AD, Shestov DB, et al. Correlates of mortality in Russian and US women. The Lipid Research Clinics Program. Am J Epidemiol. 1994;139:369–379.
17.Hajdu P, McKee M, Bojan F. Changes in premature mortality differentials by marital status in Hungary and in England and Wales. Eur J Public Health. 1995;5:259–264.
18.Broda G, Rywik S, Piotrowski W. 8-year Mortality From all Causes and Cardiovascular Disease in Relation to Education Level and Marital Status on Polish MONICA Warsaw Population. Warsaw: National Institute of Cardiology; 1994.
19.Doll R, Peto R, Wheatley K, et al. Mortality in relation to smoking: 40 years’ observations on male British doctors. BMJ. 1994;309:901–911.
20.Peto R, Lopez AD, Boreham J, et al. Mortality From Smoking in Developed Countries 1950-2000: Indirect Estimates From National Vital Statistics. Oxford: Oxford University Press; 1994.
21.Women and Smoking. A Report of the Surgeon General. Rockville, MD: US Department of Health and Human Services; 2001.
22.McKee M, Bobak M, Rose R, et al. Patterns of smoking in Russia. Tobacco Control. 1998;7:22–26.
23.Chenet L, McKee M, Leon D, et al. Alcohol and cardiovascular mortality in Moscow: new evidence of a causal association. J Epidemiol Commun Health. 1998;52:772–774.
24.Bobak M, Marmot M. Alcohol and mortality in Russia: is it different than elsewhere? Ann Epidemiol. 1999;9:335–338.
25.Balkau B. All the evidence points to alcohol being implicated in the recent changes in mortality in Russia. Ann Epidemiol. 1999;9:339–340.
26.Ryan M. Alcoholism and rising mortality in the Russian Federation. BMJ. 1995;310:646–648.
27.Carlson P, Vagero D. The social pattern of heavy drinking in Russia during transition. Evidence from Taganrog 1993. Eur J Public Health. 1998;8:280–285.
28.Nemtsov A. Alcohol Related Mortality in Russia, 1980-1990s. Moscow: Publisher not stated; 2001.
29.Deev A, Shestov D, Abernathy J, et al. Association of alcohol consumption to mortality of middle aged US and Russian men and women. Ann Epidemiol. 1998;8:147–153.
30.Malyutina S, Bobak M, Kurilovitch S, et al. Heavy and binge drinking and cardiovascular mortality: a prospective cohort study in Novosibirsk, Russia. Lancet. 2002;360:1448–1454.
31.Royal College of Physicians, Royal College of Psychiatrists, Royal College of General Practitioners. Alcohol and the Heart in Perspective. Sensible Limits Reaffirmed. Report of a working group. London: Royal College of Physicians; 1995.
32.Corrao G, Rubiatti L, Bagnardi V, et al. Alcohol and coronary heart disease: a meta-analysis. Addiction. 2000;95:1505–1523.
33.Puddey IB, Rakic V, Dimmitt SB, et al. Influence of pattern of drinking on cardiovascular disease and cardiovascular risk factors: a review. Addiction. 1999;94:649–663.
34.Kauhanen J, Kaplan GA, Goldberg DE, et al. Beer binging and mortality: results from the Kuopio ischaemic heart disease risk factors study, a prospective population based study. BMJ. 1997;315:846–851.
35.Kauhanen J, Kaplan GA, Goldberg DD, et al. Frequent hangovers and cardiovascular mortality in middle aged men. Epidemiology. 1997;8:310–314.
36.Rehm J, Greenfield TK, Rogers JD. Average volume of alcohol consumption, patterns of drinking, and all-cause mortality: results from the US National Alcohol Study. Am J Epidemiol. 2001;153:64–71.
37.Murray RP, Connett JE, Tyas SL, et al. Alcohol volume, drinking pattern, and cardiovascular disease morbidity and mortality: is there a U-shaped function? Am J Epidemiol. 2002;155:242–248.
38.Britton A, McKee M. The relation between alcohol and cardiovascular disease in Eastern Europe: explaining the paradox. J Epidemiol Commun Health. 2000;54:328–332.
© 2003 Lippincott Williams & Wilkins, Inc.