- Email: [email protected]
Where there are no data
Public Health (2001) 115, 394–400 ß R.I.P.H.H. 2001 www.nature.com/ph
Where there are no data: what has happened to life expectancy in Georgia since 1990? I Badurashvili1, M McKee2*, G Tsuladze1, F Mesle´3, J Vallin3 and V Shkolnikov4 1 Georgian Centre of Population Research, Tbilisi, Georgia; 2European Centre on Health of Societies in Transition, London School of Hygiene and Tropical Medicine, London, UK; 3Institut National d’E´tudes De´mographiques, Paris, France; and 4 Max Planck Institute for Demographic Research, Rostock, Germany
In recent years there has been a considerable increase in understanding of changes in mortality in Russia and some other former Soviet republics. However, the situation in the republics of the Caucasus remains poorly understood. Information on Georgia is especially fragmentary as a fifth of the country remains outside government control, there has been large scale migration since 1991, and the introduction of fees for vital registration has compromised the quality of official statistics. The aim of the study is to produce plausible estimates for life expectancy in Georgia for the period 1990 – 1998 and thus to assess whether Georgia has undergone changes similar to other former Soviet republics in the post-independence period. Four models were used to construct life tables. Model 1 used officially published statistics on deaths and population. Model 2 applied new estimates of population derived from household surveys to the observed deaths. Model 3 adjusted model 2 for under-registration at extremes of life, with parameter estimates derived from a survey of infant mortality and comparison of observed rates with Coale-Demeny standard life tables. Model 4 arose following inspection of death rates by cause that revealed implausible discontinuities in cancer mortality rates and involved applying the estimates of underregistration that this finding implied to model 3. The four models produce quite different estimates of life expectancy, differing by 7.8 y for men and 6.8 y for women by 1998. In any of the models, however, Georgia does not appear to have experienced the marked deterioration in life expectancy seen in Russia following the transition to independence. Importantly, Georgia had also not experienced a marked improvement in life expectancy during the 1985 Soviet anti-alcohol campaign, again unlike other Soviet republics. Official statistics substantially over-estimate life expectancy at birth in Georgia. Despite undergoing a civil war, life expectancy in Georgia has been less affected by the transition than has Russia and the overall trends in mortality since the mid 1980s suggest that this may be because alcohol has played a smaller role in these changes than it did in Russia. Public Health (2001) 115, 394–400. Keywords: life expectancy; demography; Georgia; Russia
Introduction For most of the twentieth century detailed objective information on events in the territory occupied by the Soviet Union was essentially inaccessible to the international community. First, although the Soviet system collected vast amounts of data to support its system of central planning, the information was often suspect, reflecting a ‘teleological’ approach to economic planning in which it was legitimate to make data fit any desired objective.1 Second, information was shrouded in secrecy, with any *Correspondence: M McKee, European Centre on Health of Societies in Transition, London School of Hygiene and Tropical Medicine, Keppel Street, London WC1E 7HT, UK. E-mail: [email protected] Accepted 4 September 2001
disclosure likely to be considered espionage and punished accordingly.2 Finally, despite the obligation to supply certain information to international agencies, such as mortality data to the World Health Organization, some information was withheld, such as cases of homicide and some infectious diseases and, in the early 1970s, the supply of mortality data ceased entirely.3 This situation has changed enormously since the mid1980s due, first, to the policy of glasnost introduced under Gorbachev4 and, second, to the opening to the west following the collapse of the Soviet Union in 1991. Patterns of health and their determinants in the former Soviet Union are now much better understood than before. This increasing volume of research has, however, been distributed unevenly, with most focussing on Russia and the Baltic Republics and rather less on other former Soviet republics even though it is now clear that there is enormous diversity
Life expectancy in Georgia I Badurashvili et al
395
in patterns of health within what was the Soviet Union.5 There has been especially little research on health in the three former Soviet republics in the Caucasus region, Georgia, Armenia and Azerbaijan. This paper seeks to redress partially this balance by examining trends in life expectancy in one of these countries, Georgia, since it gained independence in 1991. Such studies are important because they shed additional light on the health consequences of social and economic transition in the less well known countries of the former Soviet Union and thus on the determinants of health in this region. Where data are more easily available and are believed to be generally of high quality, it has been striking how countries that appear otherwise to have pursued very different social and economic pathways, such as Russia and Estonia, have exhibited very similar trends in mortality. This raises a question of whether there has been a common post-Soviet mortality pattern. There are, however, reasons to suspect that countries such as Georgia may be different. In the Soviet period, Georgia was distinguished by a life expectancy that was somewhat greater than in the rest of the Soviet Union. In 1989, male life expectancy at birth in Georgia was 68.2 y, 4 y longer than in Russia.6 The corresponding figure for women was 75.8 y, 1.2 y longer than in Russia. Importantly, Georgia, unlike Russia and many other former Soviet republics was little affected by the 1985 anti-alcohol campaign and its sequelae. While male life expectancy at birth increased by 2 y in Russia between 1985 and 1986, the corresponding figure for Georgia was only 0.1 y. Since 1991, however, data on health in Georgia have been fragmentary. In January 2000, the World Health Organization Health for All database, which is the standard source for international mortality data in Europe, contained
data for independent Georgia only for the year 1994.6 Then, reported life expectancy at birth for men was 67.2 y, almost 10 y longer than in Russia, with the figure for females 76.8 y, 5.6 y greater than in Russia. There are several reasons for the lack of data. Like other parts of the Caucasus, Georgia has suffered a series of conflicts. Armed conflict between government and opposition forces broke out soon after independence, with the first president, Zviad Gamsakhurdia, being overthrown. Soon after, secessionist movements emerged in South Ossetia and Abkhazia, (Figure 1). The Abkhazian movement, with backing from Russia, defeated the Georgian forces and Abkhazia remains de facto independent, as does the area around Tkhinvali, the capital of South Ossetia. These conflicts have been associated with large-scale population movement, with an estimated 280 000 people displaced from Abkhazia and Ossetia now living in Georgia. There is therefore some uncertainty about the population denominator. It is also known that many of the deaths resulting from these conflicts, estimated to be about 12 000, were never recorded by the civil registration system. The most problematic issue was, however, emigration. Following independence, Georgian borders were opened. This important change affected not only the migration out of the ex-USSR but also exchanges between other Republics of the ex-USSR, especially Russia, with many Russian citizens returning to Russia. Migration out-flows also increased dramatically in the early 1990s due to the civil war and the political crisis, but the civil registration system failed to capture the scale of these flows. An additional obstacle to the collection of accurate data was the introduction, following independence, of a fee for issuance of death certificates. These were retained for civil certificates (containing basic information about the
Figure 1 Map of Georgia. Source: CIA. The World Factbook, 1999. Public Health
Life expectancy in Georgia I Badurashvili et al
396
individual) until 1999 and are still in place for the more detailed medical certificates, which report the cause of death. These fees were considerable, amounting to more than the monthly pension. As burials can take place without a certificate, there was believed to be considerable avoidance of payment. Payments have also been retained for birth certificates but, in general, they must be paid before discharge from hospital is permitted. The scale of this problem has been explored in a series of surveys conducted by the Centre for Medical Statistics and Information of the Georgian Ministry of Health and the State Department for Statistics of Georgia. Although these surveys covered only some regions, they found that about 20% of deaths in Georgian hospitals were not captured by the civil registration system. Obviously, in circumstances such as those pertaining in Georgia, it is impossible to ascertain exact measures of life expectancy at birth. Instead, we argue, it is possible, on the basis of clearly stated assumptions, to derive a range within which the true values are likely to lie. This may be far from perfect but it does represent an improvement on the alternative of having no information on which to base policy decisions. On the basis of new estimations of population and deaths, this paper compares several estimates of Georgian mortality for the 1990s to provide some understanding of the health consequences of political transition in Georgia. In particular, it examines how Georgia compares with Russia, where these changes are much better understood. Methods Population estimates For the purposes of this paper, the population of Georgia is limited to that under de facto government control, thus excluding Abkhazia and Tkhinvali, which contain an estimated 250 000 people and from where even incomplete mortality data are unavailable. The last population census was conducted in Georgia in 1989. In subsequent years, the Georgian Statistical Office estimated populations from this census, on the basis of registered births and deaths by age in each calendar year and registered migrants, officially recorded by the Georgian Ministry of Interior Affairs. This method had been accurate as long as Georgian population movement was, as in all republics of the former Soviet Union, under tight administrative control. Then, migrations were at a very low level. Hardly anyone could leave the Soviet Union and although internal migration did occur between Republics, it was closely monitored. After 1991, when the USSR collapsed and, simultaneously, frontiers opened, migration flows increased dramatically and registration failed to capture many of them. The traditional method used by the Georgian Statistical Office to produce population estimates was thus no longer appropriPublic Health
ate. In 1997 and 1998, the Georgian Centre of Population Research (GCPR) made new estimates of population for the 1990s. These estimates were based on population projections from 1989, combining data from vital statistics with specific estimates of migration based on a series of small socio-demographic surveys conducted during the 1990s by the Institute of Demography and Sociological Research of the Georgian Academy of Sciences. These suggested that the net emigration flow could have reached up to 800 000 between 1990 and 1996, with the major loss taking place around 1993. Thus, the net emigration flow between 1990 and 1998 was finally estimated by GCPR as 1.1 million. Consequently, the 1999 total population was estimated to be 4.1 million instead of the official 5.2 million. In 1998, it was possible to refine these estimates when the Georgian Statistical Office undertook a multi-round Representative Household Survey. This broadly supported the GCPR population estimate for 1999, also estimating the total population to be 4.1 million in 1998. The process has been described in detail elsewhere but, in brief,8 the population was estimated from a regionally representative nation-wide household survey, including approximately 2800 households.9 The survey was undertaken in late 1998. Data on respondents, broken down by age and sex, were extrapolated from the sample to the estimated 1.2 million households in Georgia. This process yielded a population estimate for the survey year 1998. The population in the years from 1989 (the last year in which a census was undertaken) to 1997 were estimated by linear interpolation, supplemented by data from the various households surveys taking place in this period that had included questions on migration among household members. However, for the calculation of age-specific mortality rates these population estimates needed to be refined, in particular at the older ages. Adjustments were performed by the GCPR by comparing the observed results with that expected from the model life table. The results were published in the 1998 Georgian Demographic Yearbook.8 Mortality estimates Based on the new population data, a range of plausible age specific death rates were calculated and used to construct life tables and thus generate figures for life expectancy at birth for the years 1990 to 1998. Four models were developed. Model 1 simply used the death rates published by the Statistical Office. Model 2 used mortality rates derived from officially registered deaths in each age group divided by the newly estimated population numbers. Even though this takes into account more accurate estimates of migration flows, for the reasons already noted it is extremely likely that mortality rates are still underestimated and so life expectancy is probably overestimated. Model 3 considers that the results of Model 2 still underestimate mortality at young and old ages. For infant
Life expectancy in Georgia I Badurashvili et al
397
deaths, separate data were obtained from data collected by the Ministry of Health from its network of hospitals, polyclinics and health centres. These generated an infant mortality rate of 20.9 per thousand in 1998, compared with the official figure of 15.2 (1034 deaths identified compared with 710 reported to the Statistical Office). It was thus assumed that infant deaths were under-reported by about 33%. Unfortunately, the Ministry of Health data are not broken down by sex so we assumed that under-registration was the same for males and females. It was also assumed that there was comparable under-registration of deaths among older children and these were increased by 33% at ages 1 – 4 and 25% at ages 5 – 9. It is, however, important to note that childhood mortality in all parts of the former Soviet Union is relatively low nevertheless so overall life expectancy is not especially sensitive to these assumptions. Second, a comparison between mortality rates from Model 2 and those from Coale-Demeny standard life tables10 was made. In 1989, male death rates fit well with the standard life table at ages above 65. By 1998, however, observed death rates, which are also close to the standard between ages 65 and 75, are clearly below it above age 75. Assuming that the true rates should also fit the standard life table at older ages, under-registration among the elderly was estimated as follows: 10% at 75 – 79, 32% at 80 – 84 and 35% at 85 þ . The situation for women is less clear as there may be greater misreporting of age but overall the pattern does not seem very different. Assuming that the under-registration of female deaths must be at least as high as that of males, we have chosen to apply male coefficients of under-registration to the female deaths. The use of life tables in this way is only possible by assuming that the revised mortality rates at ages 10 – 64 are correct. Evidence from other countries does suggest that under-registration is often greatest at the extremes of age. This is likely also to be the case in Georgia, in particular because the crude death rate (using official data) in those aged over 65 fell by almost 15% between 1992 and 1994, there was no decrease in the rate among those aged 0 – 64. Nonetheless, in the circumstances pertaining in Georgia, there must be some concern about data at all ages. For this reason a fourth model was developed. This employed an innovative approach involving estimation of deaths by major categories of causes (neoplasms, cardiovascular diseases, external causes, others). In the absence of any evidence to the contrary it was assumed that there was no systematic differential under-registration of particular causes of death. Thus the proportion of deaths by major cause in each age group, as recorded in official statistics, was applied to the estimated total death rate from model two. However inspection of trends showed a decline in the death rate from neoplasms of 23% between 1990 and 1994 and 13% between 1990 and 1998. This was considered implausible. Although deaths from cancer have declined slightly in countries such as Russia during this period,11 the reduction observed in these data is considerably higher than seen elsewhere and, unlike other causes of death, nowhere in
the former Soviet Union have cancer deaths undergone large year-to-year fluctuations. Of course it is possible that the observed decline in cancer deaths could be due to a change in coding, with deaths due to cancer being incorrectly coded as, for example, pneumonia. Given the crises facing the Georgian health care system this is certainly plausible. However the decline has been equally great in both cancers where the scope for misdiagnosis is considerable, such as lung cancer, and those where misdiagnosis is less likely, such as breast cancer. Furthermore there was no corresponding increase in deaths from possible substitute labels, such as those of the respiratory system. The estimates of deaths in each age group and for both genders were thus recalculated assuming that deaths from cancer had remained constant. The exception was for deaths in age groups at the extremes of life, where cancer is a relatively minor cause of death. These had already been adjusted upwards in model 2 and so were not adjusted for a second time. In all other age groups, the total number of deaths was multiplied by a factor representing the estimated under-registration of cancer deaths in that age group. As in the other models, life expectancies at birth were then recalculated for males and females for the years 1991 to 1998. This, if anything, will overestimate deaths slightly if it is assumed that, like Russia, there has been a small decline in cancer mortality. These models yield four sets of life expectancies at birth which can be considered to represent reasonable estimates of the upper and lower boundaries of the true value. Although life expectancy at birth is a useful summary measure, it is also important to know about changes in age specific death rates. Changes in age specific mortality were examined by first comparing the ratio of deaths in each age group in 1994 with those in 1990. The years chosen are the last complete year of the Soviet Union, and 1994. Although the initial estimate of male life expectancy at birth, based on the earlier analyses, indicated that male life expectancy at birth was actually lowest in 1993, the war with Abkhazia did not end until September of that year. Thus, given the inevitable uncertainty about data from 1993, the following year was chosen. This provides information on changes during the period when overall life expectancy was falling. Second, age specific rates in 1998 were compared with those from 1994, covering the period when life expectancy was improving. In the interests of brevity, only those values derived from model 4 are shown. These data were compared with published data on Georgia available from the World Health Organization Health for All database, the most widely used source of international comparative health data. They were also compared with data from Russia, derived from the same source. Results Figure 2 shows population pyramids for 1989 (the year of the last Soviet census) and 1998. This shows an ageing of Public Health
Life expectancy in Georgia I Badurashvili et al
398
Figure 2 Population pyramids for Georgia.
the population, with those over 65 accounting for 13.9% of the population in 1998, compared with 9.7% in 1991. This is consistent with the known dramatic fall in birth rate as well as considerable emigration among young adults. Figure 3 shows trends in Georgian life expectancy at birth according to the four models, as well as a comparison with Russia. Even if the estimates from model 4 are not necessarily the most probable they suggest that by 1998 the true life expectancy at birth could be substantially lower than that shown in official statistics, by 7.8 y for men and 6.8 y for women. In none of the models did the level of life expectancy decline to that seen in Russia in the early 1990s. Figure 4 shows changes in age specific death rates. In view of space constraints, this and subsequent analyses are confined to males, who in all parts of the former Soviet Union have been most affected by transition. Only the estimates from model 4 are shown. One line shows the ratio of death rates in 1994 compared with 1991, when mortality was increasing. The other shows the ratio of death rates in 1998 compared with 1991, including the recent fall in mortality. This shows a pattern that has similarities but also some differences from that seen in Russia. As in Russia, mortality among children continued to fall throughout the period and mortality in early middle aged people rose and then fell, although not recovering fully to 1991 levels. Again, as in Russia, there was a steep rise in mortality of young people between 1991 and 1994 but, unlike Russia where it has remained at the 1994 level, in Georgia it appears to have continued to increase. Again, unlike Russia, deaths among the late middle aged increased considerably to 1994 but then fell. Clearly, given the many caveats about the data, these figures must be treated with great caution but, in general, they provide a degree of reassurance that the changes observed are broadly consistent with those seen elsewhere. As noted earlier, the WHO Health for All database contains figures for life expectancy at birth in Georgia for only one year since independence, 1994. The figure for males, at 67.2 y is somewhat higher than that suggested by model 4 (64.8 y), as is the figure for females, at 76.7 Public Health
compared with 74.6. Interestingly, the figure supplied to WHO is 1.5 y higher than the last recorded official value in 1990, an increase that seems implausible. A more detailed analysis by cause is inappropriate, given the many limitations of the data. From official statistics, however, it appears that the increase in deaths among young men was primarily due to external causes and cardiovascular diseases, with the former increasing by 22% among those aged 20 – 29 and the latter by 87% among those aged 20 – 39.
Discussion Faced with a country beset by war, with large scale migration and the added complication of an uncertain reduction in the quality of civil registration, it would be tempting to abandon any hope of saying anything about trends in mortality. Information on life expectancy is, however, important to inform policies in many areas of social policy in Georgia. It is also important for other reasons. Georgia exhibits some of the features seen in other postSoviet states but the fluctuations in life expectancy, both before and after the collapse of the Soviet Union, were less marked. This observation merits further, more detailed examination but this may now be impossible because of the absence of valid, contemporaneous data. Notwithstanding the many caveats, one clear finding emerges, regardless of which model is used. Georgia was unusual among Soviet republics in the 1980s in that it did not exhibit large changes in life expectancy in response to the 1985 anti-alcohol campaign. It is known that the campaign was implemented enthusiastically in Georgia, with the destruction of about a quarter of the republic’s vineyards, but commentators have also noted that Soviet era Georgia had significantly fewer social problems attributed to alcohol than had most other Soviet republics.12 This would suggest that the underlying contribution of alcohol to overall mortality had been less in Georgia than in Russia,
Life expectancy in Georgia I Badurashvili et al
399
Figure 3
Life expectancy at birth under different assumptions.
Figure 4 Changes in age specific death rates (males) in Georgia 1989 – 1998.
which could plausibly be explained by Georgia being a predominantly wine drinking region, unlike Russia where vodka is drunk, commonly in binges.13 In terms of life expectancy, Georgia, although predominantly Christian, thus behaved more like the republics where adherence to a traditional Islamic way of life, and thus a limited use of alcohol, has persisted, such as Uzbekistan. This observation is of considerable importance because the relatively small decline in life expectancy in Uzbekistan post-independence, a country that has undergone only limited liberalisation, has been used by conservative politicians in Russia to argue against the rapid introduction of market principles in Russia. These data also say something about Russia. It must be a matter of great concern that, even though Georgia has Public Health
Life expectancy in Georgia I Badurashvili et al
400
undergone almost total economic collapse and has been beset by civil war, the decline in life expectancy at birth was still somewhat less than in Russia where the threats, though substantial, have been much less. In short, for whatever reason, the people of Russia seem to have been much less able to face the challenges of transition, whether because of a lack of social networks, the role that alcohol plays in society, or some other reason. This study has many limitations. There is, however, a danger of the perfect becoming the enemy of the good and it does provide a model that can be used in the neighbouring republics of the Caucasus that face similar problems. It illustrates what can be done by bringing together data that already exist, rather than simply consigning to the bin headline figures that are known to be problematic, as are published figures for Georgian life expectancy. The techniques used are well known but one, the use of trends in cause specific mortality, in this case cancer, to validate overall mortality trends is innovative, if simple. Even so, we are still unable to provide a precise figure for life expectancy in Georgia but we can indicate the potential scale of the error in official data, which may over-estimate life expectancy at birth by up to 8 y. These findings at least place some boundaries around what is likely to be happening to life expectancy in Georgia and thus can contribute in a small part to the ongoing policy debate in that country.
Acknowledgements IB undertook some of this work at the London School of Hygiene and Tropical Medicine while in receipt of a fellowship from the Open Society Institute in New York. MM’s work in the former Soviet Union is supported by the UK Department for International Development. FM and JV’s work in the former Soviet Union is supported by the Institut National d’E´tudes Demographique and VS’s work
Public Health
by the Centre for Human Demography and Ecology. We are grateful to Remis Prokhoskas for advice on an earlier draft.
References 1 Service R. A History of Twentieth Century Russia. Allen Lane: London, 1997, p 171. 2 Andrew C, Mitrokhin V. The Mitrokhin Archive: The KGB in Europe and the West. Allen Lane: London, 1999. 3 Shkolnikov VM, Mesle´ F, Vallin J. Health crisis in Russia I. Recent trends in life expectancy and causes of death from 1970 to 1993. Population 1996; 8: 123 – 154. 4 Gorbachev M. Memoirs. Doubleday: London, 1996. 5 McKee M. The health effects of the collapse of the Soviet Union. In: Poverty and Inequality in Health. Leon D, Walt G (eds). Oxford University Press: Oxford, 2001, pp 17 – 36. 6 World Health Organization. Health for All database. WHO: Copenhagen, 2000. 7 Goldenberg S. Pride of Small Nations: The Caucasus and Post-Soviet Disorder. Zed: London, 1994. 8 Tsuadze G, Badurashvili I. The Demographic Yearbook of Georgia. National Centre of Population Research: Tbilisi, 1999. 9 State Department for Statistics of Georgia. Georgian Households, Economical-statistical Book. State Department for Statistics of Georgia: Tbilisi, 1998. 10 Coale AJ, Demeny P. Regional Model Life Tables and Stable Populations (2nd ed). Academic Press: New York, 1983. 11 Shkolnikov VM, McKee M, Vallin J, Aksel E, Leon D, Chenet L, Mesle´ F. Cancer mortality in Russia and Ukraine: validity, competing risks, and cohort effects. Int J Epidemiol 1999; 28: 19 – 29. 12 White S. Russia Goes Dry. Cambridge University Press: Cambridge, 1995. 13 Bobak M, McKee M, Rose R, Marmot M. Alcohol consumption in a national sample of the Russian population. Addiction 1999; 94: 857 – 866.
Where there are no data: what has happened to life expectancy in Georgia since 1990? I Badurashvili1, M McKee2*, G Tsuladze1, F Mesle´3, J Vallin3 and V Shkolnikov4 1 Georgian Centre of Population Research, Tbilisi, Georgia; 2European Centre on Health of Societies in Transition, London School of Hygiene and Tropical Medicine, London, UK; 3Institut National d’E´tudes De´mographiques, Paris, France; and 4 Max Planck Institute for Demographic Research, Rostock, Germany
In recent years there has been a considerable increase in understanding of changes in mortality in Russia and some other former Soviet republics. However, the situation in the republics of the Caucasus remains poorly understood. Information on Georgia is especially fragmentary as a fifth of the country remains outside government control, there has been large scale migration since 1991, and the introduction of fees for vital registration has compromised the quality of official statistics. The aim of the study is to produce plausible estimates for life expectancy in Georgia for the period 1990 – 1998 and thus to assess whether Georgia has undergone changes similar to other former Soviet republics in the post-independence period. Four models were used to construct life tables. Model 1 used officially published statistics on deaths and population. Model 2 applied new estimates of population derived from household surveys to the observed deaths. Model 3 adjusted model 2 for under-registration at extremes of life, with parameter estimates derived from a survey of infant mortality and comparison of observed rates with Coale-Demeny standard life tables. Model 4 arose following inspection of death rates by cause that revealed implausible discontinuities in cancer mortality rates and involved applying the estimates of underregistration that this finding implied to model 3. The four models produce quite different estimates of life expectancy, differing by 7.8 y for men and 6.8 y for women by 1998. In any of the models, however, Georgia does not appear to have experienced the marked deterioration in life expectancy seen in Russia following the transition to independence. Importantly, Georgia had also not experienced a marked improvement in life expectancy during the 1985 Soviet anti-alcohol campaign, again unlike other Soviet republics. Official statistics substantially over-estimate life expectancy at birth in Georgia. Despite undergoing a civil war, life expectancy in Georgia has been less affected by the transition than has Russia and the overall trends in mortality since the mid 1980s suggest that this may be because alcohol has played a smaller role in these changes than it did in Russia. Public Health (2001) 115, 394–400. Keywords: life expectancy; demography; Georgia; Russia
Introduction For most of the twentieth century detailed objective information on events in the territory occupied by the Soviet Union was essentially inaccessible to the international community. First, although the Soviet system collected vast amounts of data to support its system of central planning, the information was often suspect, reflecting a ‘teleological’ approach to economic planning in which it was legitimate to make data fit any desired objective.1 Second, information was shrouded in secrecy, with any *Correspondence: M McKee, European Centre on Health of Societies in Transition, London School of Hygiene and Tropical Medicine, Keppel Street, London WC1E 7HT, UK. E-mail: [email protected] Accepted 4 September 2001
disclosure likely to be considered espionage and punished accordingly.2 Finally, despite the obligation to supply certain information to international agencies, such as mortality data to the World Health Organization, some information was withheld, such as cases of homicide and some infectious diseases and, in the early 1970s, the supply of mortality data ceased entirely.3 This situation has changed enormously since the mid1980s due, first, to the policy of glasnost introduced under Gorbachev4 and, second, to the opening to the west following the collapse of the Soviet Union in 1991. Patterns of health and their determinants in the former Soviet Union are now much better understood than before. This increasing volume of research has, however, been distributed unevenly, with most focussing on Russia and the Baltic Republics and rather less on other former Soviet republics even though it is now clear that there is enormous diversity
Life expectancy in Georgia I Badurashvili et al
395
in patterns of health within what was the Soviet Union.5 There has been especially little research on health in the three former Soviet republics in the Caucasus region, Georgia, Armenia and Azerbaijan. This paper seeks to redress partially this balance by examining trends in life expectancy in one of these countries, Georgia, since it gained independence in 1991. Such studies are important because they shed additional light on the health consequences of social and economic transition in the less well known countries of the former Soviet Union and thus on the determinants of health in this region. Where data are more easily available and are believed to be generally of high quality, it has been striking how countries that appear otherwise to have pursued very different social and economic pathways, such as Russia and Estonia, have exhibited very similar trends in mortality. This raises a question of whether there has been a common post-Soviet mortality pattern. There are, however, reasons to suspect that countries such as Georgia may be different. In the Soviet period, Georgia was distinguished by a life expectancy that was somewhat greater than in the rest of the Soviet Union. In 1989, male life expectancy at birth in Georgia was 68.2 y, 4 y longer than in Russia.6 The corresponding figure for women was 75.8 y, 1.2 y longer than in Russia. Importantly, Georgia, unlike Russia and many other former Soviet republics was little affected by the 1985 anti-alcohol campaign and its sequelae. While male life expectancy at birth increased by 2 y in Russia between 1985 and 1986, the corresponding figure for Georgia was only 0.1 y. Since 1991, however, data on health in Georgia have been fragmentary. In January 2000, the World Health Organization Health for All database, which is the standard source for international mortality data in Europe, contained
data for independent Georgia only for the year 1994.6 Then, reported life expectancy at birth for men was 67.2 y, almost 10 y longer than in Russia, with the figure for females 76.8 y, 5.6 y greater than in Russia. There are several reasons for the lack of data. Like other parts of the Caucasus, Georgia has suffered a series of conflicts. Armed conflict between government and opposition forces broke out soon after independence, with the first president, Zviad Gamsakhurdia, being overthrown. Soon after, secessionist movements emerged in South Ossetia and Abkhazia, (Figure 1). The Abkhazian movement, with backing from Russia, defeated the Georgian forces and Abkhazia remains de facto independent, as does the area around Tkhinvali, the capital of South Ossetia. These conflicts have been associated with large-scale population movement, with an estimated 280 000 people displaced from Abkhazia and Ossetia now living in Georgia. There is therefore some uncertainty about the population denominator. It is also known that many of the deaths resulting from these conflicts, estimated to be about 12 000, were never recorded by the civil registration system. The most problematic issue was, however, emigration. Following independence, Georgian borders were opened. This important change affected not only the migration out of the ex-USSR but also exchanges between other Republics of the ex-USSR, especially Russia, with many Russian citizens returning to Russia. Migration out-flows also increased dramatically in the early 1990s due to the civil war and the political crisis, but the civil registration system failed to capture the scale of these flows. An additional obstacle to the collection of accurate data was the introduction, following independence, of a fee for issuance of death certificates. These were retained for civil certificates (containing basic information about the
Figure 1 Map of Georgia. Source: CIA. The World Factbook, 1999. Public Health
Life expectancy in Georgia I Badurashvili et al
396
individual) until 1999 and are still in place for the more detailed medical certificates, which report the cause of death. These fees were considerable, amounting to more than the monthly pension. As burials can take place without a certificate, there was believed to be considerable avoidance of payment. Payments have also been retained for birth certificates but, in general, they must be paid before discharge from hospital is permitted. The scale of this problem has been explored in a series of surveys conducted by the Centre for Medical Statistics and Information of the Georgian Ministry of Health and the State Department for Statistics of Georgia. Although these surveys covered only some regions, they found that about 20% of deaths in Georgian hospitals were not captured by the civil registration system. Obviously, in circumstances such as those pertaining in Georgia, it is impossible to ascertain exact measures of life expectancy at birth. Instead, we argue, it is possible, on the basis of clearly stated assumptions, to derive a range within which the true values are likely to lie. This may be far from perfect but it does represent an improvement on the alternative of having no information on which to base policy decisions. On the basis of new estimations of population and deaths, this paper compares several estimates of Georgian mortality for the 1990s to provide some understanding of the health consequences of political transition in Georgia. In particular, it examines how Georgia compares with Russia, where these changes are much better understood. Methods Population estimates For the purposes of this paper, the population of Georgia is limited to that under de facto government control, thus excluding Abkhazia and Tkhinvali, which contain an estimated 250 000 people and from where even incomplete mortality data are unavailable. The last population census was conducted in Georgia in 1989. In subsequent years, the Georgian Statistical Office estimated populations from this census, on the basis of registered births and deaths by age in each calendar year and registered migrants, officially recorded by the Georgian Ministry of Interior Affairs. This method had been accurate as long as Georgian population movement was, as in all republics of the former Soviet Union, under tight administrative control. Then, migrations were at a very low level. Hardly anyone could leave the Soviet Union and although internal migration did occur between Republics, it was closely monitored. After 1991, when the USSR collapsed and, simultaneously, frontiers opened, migration flows increased dramatically and registration failed to capture many of them. The traditional method used by the Georgian Statistical Office to produce population estimates was thus no longer appropriPublic Health
ate. In 1997 and 1998, the Georgian Centre of Population Research (GCPR) made new estimates of population for the 1990s. These estimates were based on population projections from 1989, combining data from vital statistics with specific estimates of migration based on a series of small socio-demographic surveys conducted during the 1990s by the Institute of Demography and Sociological Research of the Georgian Academy of Sciences. These suggested that the net emigration flow could have reached up to 800 000 between 1990 and 1996, with the major loss taking place around 1993. Thus, the net emigration flow between 1990 and 1998 was finally estimated by GCPR as 1.1 million. Consequently, the 1999 total population was estimated to be 4.1 million instead of the official 5.2 million. In 1998, it was possible to refine these estimates when the Georgian Statistical Office undertook a multi-round Representative Household Survey. This broadly supported the GCPR population estimate for 1999, also estimating the total population to be 4.1 million in 1998. The process has been described in detail elsewhere but, in brief,8 the population was estimated from a regionally representative nation-wide household survey, including approximately 2800 households.9 The survey was undertaken in late 1998. Data on respondents, broken down by age and sex, were extrapolated from the sample to the estimated 1.2 million households in Georgia. This process yielded a population estimate for the survey year 1998. The population in the years from 1989 (the last year in which a census was undertaken) to 1997 were estimated by linear interpolation, supplemented by data from the various households surveys taking place in this period that had included questions on migration among household members. However, for the calculation of age-specific mortality rates these population estimates needed to be refined, in particular at the older ages. Adjustments were performed by the GCPR by comparing the observed results with that expected from the model life table. The results were published in the 1998 Georgian Demographic Yearbook.8 Mortality estimates Based on the new population data, a range of plausible age specific death rates were calculated and used to construct life tables and thus generate figures for life expectancy at birth for the years 1990 to 1998. Four models were developed. Model 1 simply used the death rates published by the Statistical Office. Model 2 used mortality rates derived from officially registered deaths in each age group divided by the newly estimated population numbers. Even though this takes into account more accurate estimates of migration flows, for the reasons already noted it is extremely likely that mortality rates are still underestimated and so life expectancy is probably overestimated. Model 3 considers that the results of Model 2 still underestimate mortality at young and old ages. For infant
Life expectancy in Georgia I Badurashvili et al
397
deaths, separate data were obtained from data collected by the Ministry of Health from its network of hospitals, polyclinics and health centres. These generated an infant mortality rate of 20.9 per thousand in 1998, compared with the official figure of 15.2 (1034 deaths identified compared with 710 reported to the Statistical Office). It was thus assumed that infant deaths were under-reported by about 33%. Unfortunately, the Ministry of Health data are not broken down by sex so we assumed that under-registration was the same for males and females. It was also assumed that there was comparable under-registration of deaths among older children and these were increased by 33% at ages 1 – 4 and 25% at ages 5 – 9. It is, however, important to note that childhood mortality in all parts of the former Soviet Union is relatively low nevertheless so overall life expectancy is not especially sensitive to these assumptions. Second, a comparison between mortality rates from Model 2 and those from Coale-Demeny standard life tables10 was made. In 1989, male death rates fit well with the standard life table at ages above 65. By 1998, however, observed death rates, which are also close to the standard between ages 65 and 75, are clearly below it above age 75. Assuming that the true rates should also fit the standard life table at older ages, under-registration among the elderly was estimated as follows: 10% at 75 – 79, 32% at 80 – 84 and 35% at 85 þ . The situation for women is less clear as there may be greater misreporting of age but overall the pattern does not seem very different. Assuming that the under-registration of female deaths must be at least as high as that of males, we have chosen to apply male coefficients of under-registration to the female deaths. The use of life tables in this way is only possible by assuming that the revised mortality rates at ages 10 – 64 are correct. Evidence from other countries does suggest that under-registration is often greatest at the extremes of age. This is likely also to be the case in Georgia, in particular because the crude death rate (using official data) in those aged over 65 fell by almost 15% between 1992 and 1994, there was no decrease in the rate among those aged 0 – 64. Nonetheless, in the circumstances pertaining in Georgia, there must be some concern about data at all ages. For this reason a fourth model was developed. This employed an innovative approach involving estimation of deaths by major categories of causes (neoplasms, cardiovascular diseases, external causes, others). In the absence of any evidence to the contrary it was assumed that there was no systematic differential under-registration of particular causes of death. Thus the proportion of deaths by major cause in each age group, as recorded in official statistics, was applied to the estimated total death rate from model two. However inspection of trends showed a decline in the death rate from neoplasms of 23% between 1990 and 1994 and 13% between 1990 and 1998. This was considered implausible. Although deaths from cancer have declined slightly in countries such as Russia during this period,11 the reduction observed in these data is considerably higher than seen elsewhere and, unlike other causes of death, nowhere in
the former Soviet Union have cancer deaths undergone large year-to-year fluctuations. Of course it is possible that the observed decline in cancer deaths could be due to a change in coding, with deaths due to cancer being incorrectly coded as, for example, pneumonia. Given the crises facing the Georgian health care system this is certainly plausible. However the decline has been equally great in both cancers where the scope for misdiagnosis is considerable, such as lung cancer, and those where misdiagnosis is less likely, such as breast cancer. Furthermore there was no corresponding increase in deaths from possible substitute labels, such as those of the respiratory system. The estimates of deaths in each age group and for both genders were thus recalculated assuming that deaths from cancer had remained constant. The exception was for deaths in age groups at the extremes of life, where cancer is a relatively minor cause of death. These had already been adjusted upwards in model 2 and so were not adjusted for a second time. In all other age groups, the total number of deaths was multiplied by a factor representing the estimated under-registration of cancer deaths in that age group. As in the other models, life expectancies at birth were then recalculated for males and females for the years 1991 to 1998. This, if anything, will overestimate deaths slightly if it is assumed that, like Russia, there has been a small decline in cancer mortality. These models yield four sets of life expectancies at birth which can be considered to represent reasonable estimates of the upper and lower boundaries of the true value. Although life expectancy at birth is a useful summary measure, it is also important to know about changes in age specific death rates. Changes in age specific mortality were examined by first comparing the ratio of deaths in each age group in 1994 with those in 1990. The years chosen are the last complete year of the Soviet Union, and 1994. Although the initial estimate of male life expectancy at birth, based on the earlier analyses, indicated that male life expectancy at birth was actually lowest in 1993, the war with Abkhazia did not end until September of that year. Thus, given the inevitable uncertainty about data from 1993, the following year was chosen. This provides information on changes during the period when overall life expectancy was falling. Second, age specific rates in 1998 were compared with those from 1994, covering the period when life expectancy was improving. In the interests of brevity, only those values derived from model 4 are shown. These data were compared with published data on Georgia available from the World Health Organization Health for All database, the most widely used source of international comparative health data. They were also compared with data from Russia, derived from the same source. Results Figure 2 shows population pyramids for 1989 (the year of the last Soviet census) and 1998. This shows an ageing of Public Health
Life expectancy in Georgia I Badurashvili et al
398
Figure 2 Population pyramids for Georgia.
the population, with those over 65 accounting for 13.9% of the population in 1998, compared with 9.7% in 1991. This is consistent with the known dramatic fall in birth rate as well as considerable emigration among young adults. Figure 3 shows trends in Georgian life expectancy at birth according to the four models, as well as a comparison with Russia. Even if the estimates from model 4 are not necessarily the most probable they suggest that by 1998 the true life expectancy at birth could be substantially lower than that shown in official statistics, by 7.8 y for men and 6.8 y for women. In none of the models did the level of life expectancy decline to that seen in Russia in the early 1990s. Figure 4 shows changes in age specific death rates. In view of space constraints, this and subsequent analyses are confined to males, who in all parts of the former Soviet Union have been most affected by transition. Only the estimates from model 4 are shown. One line shows the ratio of death rates in 1994 compared with 1991, when mortality was increasing. The other shows the ratio of death rates in 1998 compared with 1991, including the recent fall in mortality. This shows a pattern that has similarities but also some differences from that seen in Russia. As in Russia, mortality among children continued to fall throughout the period and mortality in early middle aged people rose and then fell, although not recovering fully to 1991 levels. Again, as in Russia, there was a steep rise in mortality of young people between 1991 and 1994 but, unlike Russia where it has remained at the 1994 level, in Georgia it appears to have continued to increase. Again, unlike Russia, deaths among the late middle aged increased considerably to 1994 but then fell. Clearly, given the many caveats about the data, these figures must be treated with great caution but, in general, they provide a degree of reassurance that the changes observed are broadly consistent with those seen elsewhere. As noted earlier, the WHO Health for All database contains figures for life expectancy at birth in Georgia for only one year since independence, 1994. The figure for males, at 67.2 y is somewhat higher than that suggested by model 4 (64.8 y), as is the figure for females, at 76.7 Public Health
compared with 74.6. Interestingly, the figure supplied to WHO is 1.5 y higher than the last recorded official value in 1990, an increase that seems implausible. A more detailed analysis by cause is inappropriate, given the many limitations of the data. From official statistics, however, it appears that the increase in deaths among young men was primarily due to external causes and cardiovascular diseases, with the former increasing by 22% among those aged 20 – 29 and the latter by 87% among those aged 20 – 39.
Discussion Faced with a country beset by war, with large scale migration and the added complication of an uncertain reduction in the quality of civil registration, it would be tempting to abandon any hope of saying anything about trends in mortality. Information on life expectancy is, however, important to inform policies in many areas of social policy in Georgia. It is also important for other reasons. Georgia exhibits some of the features seen in other postSoviet states but the fluctuations in life expectancy, both before and after the collapse of the Soviet Union, were less marked. This observation merits further, more detailed examination but this may now be impossible because of the absence of valid, contemporaneous data. Notwithstanding the many caveats, one clear finding emerges, regardless of which model is used. Georgia was unusual among Soviet republics in the 1980s in that it did not exhibit large changes in life expectancy in response to the 1985 anti-alcohol campaign. It is known that the campaign was implemented enthusiastically in Georgia, with the destruction of about a quarter of the republic’s vineyards, but commentators have also noted that Soviet era Georgia had significantly fewer social problems attributed to alcohol than had most other Soviet republics.12 This would suggest that the underlying contribution of alcohol to overall mortality had been less in Georgia than in Russia,
Life expectancy in Georgia I Badurashvili et al
399
Figure 3
Life expectancy at birth under different assumptions.
Figure 4 Changes in age specific death rates (males) in Georgia 1989 – 1998.
which could plausibly be explained by Georgia being a predominantly wine drinking region, unlike Russia where vodka is drunk, commonly in binges.13 In terms of life expectancy, Georgia, although predominantly Christian, thus behaved more like the republics where adherence to a traditional Islamic way of life, and thus a limited use of alcohol, has persisted, such as Uzbekistan. This observation is of considerable importance because the relatively small decline in life expectancy in Uzbekistan post-independence, a country that has undergone only limited liberalisation, has been used by conservative politicians in Russia to argue against the rapid introduction of market principles in Russia. These data also say something about Russia. It must be a matter of great concern that, even though Georgia has Public Health
Life expectancy in Georgia I Badurashvili et al
400
undergone almost total economic collapse and has been beset by civil war, the decline in life expectancy at birth was still somewhat less than in Russia where the threats, though substantial, have been much less. In short, for whatever reason, the people of Russia seem to have been much less able to face the challenges of transition, whether because of a lack of social networks, the role that alcohol plays in society, or some other reason. This study has many limitations. There is, however, a danger of the perfect becoming the enemy of the good and it does provide a model that can be used in the neighbouring republics of the Caucasus that face similar problems. It illustrates what can be done by bringing together data that already exist, rather than simply consigning to the bin headline figures that are known to be problematic, as are published figures for Georgian life expectancy. The techniques used are well known but one, the use of trends in cause specific mortality, in this case cancer, to validate overall mortality trends is innovative, if simple. Even so, we are still unable to provide a precise figure for life expectancy in Georgia but we can indicate the potential scale of the error in official data, which may over-estimate life expectancy at birth by up to 8 y. These findings at least place some boundaries around what is likely to be happening to life expectancy in Georgia and thus can contribute in a small part to the ongoing policy debate in that country.
Acknowledgements IB undertook some of this work at the London School of Hygiene and Tropical Medicine while in receipt of a fellowship from the Open Society Institute in New York. MM’s work in the former Soviet Union is supported by the UK Department for International Development. FM and JV’s work in the former Soviet Union is supported by the Institut National d’E´tudes Demographique and VS’s work
Public Health
by the Centre for Human Demography and Ecology. We are grateful to Remis Prokhoskas for advice on an earlier draft.
References 1 Service R. A History of Twentieth Century Russia. Allen Lane: London, 1997, p 171. 2 Andrew C, Mitrokhin V. The Mitrokhin Archive: The KGB in Europe and the West. Allen Lane: London, 1999. 3 Shkolnikov VM, Mesle´ F, Vallin J. Health crisis in Russia I. Recent trends in life expectancy and causes of death from 1970 to 1993. Population 1996; 8: 123 – 154. 4 Gorbachev M. Memoirs. Doubleday: London, 1996. 5 McKee M. The health effects of the collapse of the Soviet Union. In: Poverty and Inequality in Health. Leon D, Walt G (eds). Oxford University Press: Oxford, 2001, pp 17 – 36. 6 World Health Organization. Health for All database. WHO: Copenhagen, 2000. 7 Goldenberg S. Pride of Small Nations: The Caucasus and Post-Soviet Disorder. Zed: London, 1994. 8 Tsuadze G, Badurashvili I. The Demographic Yearbook of Georgia. National Centre of Population Research: Tbilisi, 1999. 9 State Department for Statistics of Georgia. Georgian Households, Economical-statistical Book. State Department for Statistics of Georgia: Tbilisi, 1998. 10 Coale AJ, Demeny P. Regional Model Life Tables and Stable Populations (2nd ed). Academic Press: New York, 1983. 11 Shkolnikov VM, McKee M, Vallin J, Aksel E, Leon D, Chenet L, Mesle´ F. Cancer mortality in Russia and Ukraine: validity, competing risks, and cohort effects. Int J Epidemiol 1999; 28: 19 – 29. 12 White S. Russia Goes Dry. Cambridge University Press: Cambridge, 1995. 13 Bobak M, McKee M, Rose R, Marmot M. Alcohol consumption in a national sample of the Russian population. Addiction 1999; 94: 857 – 866.