This article was downloaded by: [Harvard Library] On: 29 December 2014, At: 02:18 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Population Studies: A Journal of Demography Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/rpst20

Infant mortality in Kyrgyzstan before and after the break-up of the Soviet Union a

b

c

d

Michel Guillot , So-jung Lim , Liudmila Torgasheva & Mikhail Denisenko a

University of Pennsylvania

b

Utah State University

c

National Statistical Committee of the Kyrgyz Republic

d

National Research University Higher School of Economics Published online: 22 Oct 2013.

To cite this article: Michel Guillot, So-jung Lim, Liudmila Torgasheva & Mikhail Denisenko (2013) Infant mortality in Kyrgyzstan before and after the break-up of the Soviet Union, Population Studies: A Journal of Demography, 67:3, 335-352, DOI: 10.1080/00324728.2013.835859 To link to this article: http://dx.doi.org/10.1080/00324728.2013.835859

PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http:// www.tandfonline.com/page/terms-and-conditions

Population Studies, 2013 Vol. 67, No. 3, 335–352, http://dx.doi.org/10.1080/00324728.2013.835859

Infant mortality in Kyrgyzstan before and after the break-up of the Soviet Union Michel Guillot1, So-jung Lim2, Liudmila Torgasheva3 and Mikhail Denisenko4

Downloaded by [Harvard Library] at 02:18 29 December 2014

1

University of Pennsylvania; 2Utah State University; 3National Statistical Committee of the Kyrgyz Republic; 4 National Research University Higher School of Economics

There is a great deal of uncertainty over the levels of, and trends in, infant mortality in the former Soviet republics of Central Asia. As a result, the impact of the break-up of the Soviet Union on infant mortality in the region is not known, and proper monitoring of mortality levels is impaired. In this paper, a variety of data sources and methods are used to assess levels of infant mortality and their trend over time in one Central Asian republic, Kyrgyzstan, between 1980 and 2010. An abrupt halt to an already established decline in infant mortality was observed to occur during the decade following the break-up of the Soviet Union, contradicting the official statistics based on vital registration. Infants of Central Asian ethnicity and those born in rural areas were also considerably more at risk of mortality than suggested by the official sources. We discuss the implications of these findings, both for health policy in this seldom studied part of the former Soviet Union and for our understanding of the health crisis which it currently faces.

Keywords: infant mortality; Kyrgyzstan; Central Asia; former Soviet Union; indirect estimation techniques; data quality [Submitted December 2011; Final version accepted January 2013]

Introduction The countries of the former Soviet Union have experienced severe health crises in recent decades (Meslé et al. 1996; Brainerd 1998; Shkolnikov et al. 1998; Cornia and Paniccia 2000; Stuckler et al. 2011). The severity with which the crisis affected each country varied greatly, but all experienced substantial declines in life expectancy in the years following the break-up of the Soviet Union in 1991. Even today, a number of them have barely recovered their pre-1991 level of life expectancy. In Russia, for example, the life expectancy at birth for both sexes was 69.03 in 2010, that is, a value that is virtually the same as the value of 69.02 observed in 1991 (European Health for All Database). Trends in infant mortality reported in the region contrast greatly with the negative trends in overall mortality. In most former Soviet republics, official vital registration statistics (as reported in the European Health for All Database) indicate that infant mortality rates (IMRs), calculated as the ratio of infant deaths to the number of live births occurring in the same year, have not increased substantially © 2013 Population Investigation Committee

during the post-independence period, and today’s reported levels are substantially lower than they were in 1991. In Russia, for example, although the rate of decline stalled in the 1990s, the country’s official IMR subsequently shows an impressive trajectory of decline, as depicted in Figure 1. The country’s IMR for 2010 was 7.6 per 1,000 live births, less than half the 1991 value of 18.1 per 1,000 (European Health for All Database). The deteriorations in life expectancy, which began before the break-up of the Soviet Union (Shkolnikov et al. 2004), have been overwhelmingly due to an increase in adult mortality, especially amongst those in the 20–59 age range, and not to changes in infant mortality (Vallin and Meslé 2001). The contrast between the trajectories of infant mortality and adult mortality in the former Soviet republics, together with information on causes of death, has led researchers to conclude that, in Russia at least, the unfolding health crisis was unlikely to be linked primarily to any deterioration in either curative medical care or public health (Chen et al. 1996; Shkolnikov et al. 1998; Anderson 2002). Had that deterioration occurred, increases in

336

Michel Guillot et al. 50

IMR (per 1000)

40

30

20

Downloaded by [Harvard Library] at 02:18 29 December 2014

10 1980

1985

1990

1995

2000

2005

2010

Year Kyrgyzstan

Russia

Figure 1 Infant mortality rates (IMRs) derived from official vital registration data, Russia and Kyrgyzstan, 1980–2010 Source: European Health for All Database, available at: http://data.euro.who.int/hfadb/

communicable diseases, to which infants are particularly vulnerable, would have been expected along with substantial increases in the IMR. Vital registration (VR) statistics for the former Soviet republics of Central Asia indicate that trends in infant mortality in these countries were similar to those of Russia. The VR-based IMR figures for one such republic, Kyrgyzstan, over the 1980–2010 period are displayed next to those for Russia in Figure 1 (European Health for All Database). IMR levels were higher in Kyrgyzstan, but no sustained increase in the country’s IMR was reported in the wake of the events of 1991. There was a small, short-term increase between 1991 and 1993, but this was followed by over a decade of substantial decline. In fact, the VR data indicate a narrowing of the gap— absolute and relative—between the IMRs of Russia and Kyrgyzstan over the course of the 1990s. The sharp increase in Kyrgyzstan’s IMR between 2003 and 2005 will be discussed later in the paper. Although the overall pattern of decline in the VRbased IMR in Kyrgyzstan is not dissimilar, albeit at a different level, to that observed in Russia, it has been regarded with considerable doubt. Scholars have argued that the decreases in IMR reported for Kyrgyzstan and other Central Asian republics during the 1990s are likely to be artifactual, the result of

declines in the completeness of VR thought to have occurred in the years following the collapse of the Soviet Union (Becker et al. 1998; Hohmann and Garenne 2005). While there is debate over whether trends in IMR decreased, increased, or stagnated between 1980 and 2010 in the Central Asian republics, there is also considerable uncertainty over the level of infant mortality. There is a consensus that the rates based on VR data and reported by national statistical offices are generally too low (Anderson and Silver 1986, 1997; Velkoff and Miller 1995; Aleshina and Redmond 2005), but the degree of underestimation and how it may have changed over time remain in doubt. The latter, of course, has implications for the interpretation of the VR-based trends in IMR. If the magnitude of the underestimation remained constant over time, then the VR trends would still be informative. However, if the degree of underestimation changed over time then both the trends and levels of infant mortality reported in the VR data are misleading. This paper builds on a number of recent studies which have produced corrected trends in infant mortality for former Soviet republics, including Armenia (Hakobyan and Yepiskoposyan 2006; Duthé et al. 2010), Georgia (Duthé et al. 2010), Ukraine (Meslé and Vallin 2012), Uzbekistan (Hohmann and Garenne 2005), and Moldova (Penina et al. 2010). Central Asia remains an understudied region in this literature, a gap we sought to address with the study reported in this paper. We used a number of data sources and methods to assess the validity of Kyrgyzstan’s VR-based levels and trends in infant mortality between 1980 and 2010. The data used were drawn from detailed VR tables, censuses, and surveys. Many of these sources, including some VR tables and census tables, are not published and were obtained directly from the National Statistical Committee (NSC) of the Kyrgyz Republic. The period between 1980 and 2010 was taken as the focus of this study because these three decades spanned a number of years both before and after the break-up of the Soviet Union in December 1991, and therefore an ample timeframe for the assessment of whether the break-up coincided with a marked deviation from earlier mortality trends. It should be noted that before 1980 the VR-based IMR had been increasing in Kyrgyzstan and in other Central Asian republics (Guillot 2007). These increases, which were sufficient to affect national IMR trends in the Soviet Union, have been the topic of a lively debate among scholars (Davis and Feshbach 1980; Eberstadt 1981; Anderson and Silver

Downloaded by [Harvard Library] at 02:18 29 December 2014

Infant mortality in Kyrgyzstan 1986; Velkoff and Miller 1995), but they were beyond the scope of our study. We chose to study infant mortality, rather than mortality amongst those aged less than 5, for both substantive and practical reasons. The substantive reason was that infant mortality is a more responsive measure of changes in the conditions which affect mortality, because it involves a shorter period of exposure to the risk of mortality after birth. The practical reason was that the NSC of the Kyrgyz Republic collects more detailed information on infant mortality than on under-5 mortality. Although the use of infant mortality as a measure of health conditions has a long tradition in demography, the international health community favours the use of mortality under the age of 5 years (under-5 mortality). This is partly because, in retrospective surveys, measures of under-5 mortality are less subject to age misreporting error. Surveys were only one of several data sources used in our study, and therefore the advantages of using under-5 mortality were less clear. Assessing levels and trends in infant mortality in Central Asia has obvious policy relevance. The precise monitoring of infant mortality is paramount for purposes of health policy, especially since the overall nature of the health crisis in former Soviet Central Asia, by comparison with Russia, remains largely unknown. Because levels of infant mortality were observed to decline in Russia after the mid-1990s, it has been argued, as mentioned previously, that deterioration in the level of health services was not a major factor in the country’s health crisis. The extent to which such a conclusion was also valid in the case of the Central Asian republics remains to be determined, largely because of the considerable amount of uncertainty surrounding their trends in infant mortality. Using data from Kyrgyzstan, we first review the errors and biases relevant to the measurement of infant mortality that can be identified in the VR statistics of the former Soviet context in general, and in Central Asia in particular. Uncorrected trends in infant mortality, calculated from official VR data, are presented, followed by a discussion of the ways in which these patterns seem implausible, given comparisons within the data. We then present results based on a variety of data sources, including censuses, vital registration, and surveys, which suggest ways of addressing the deficiencies identified in the VR-based measures of infant mortality. Finally, we propose ‘adjusted’ levels and trends in infant mortality, and then discuss the implications of the patterns revealed for our understanding of the health crisis in this seldom studied part of the former Soviet Union.

337

Sources of error in VR-based infant mortality rates In the demographic literature on infant mortality in the former Soviet republics, there is a consensus that the VR-based levels are too low, and that underestimation is particularly severe in the Central Asian republics (Anderson and Silver 1997; Aleshina and Redmond 2005). When independent estimates of infant and child mortality in the Central Asian region were produced by Demographic and Health Surveys (DHS), it became clear that rates of infant mortality in the region had been underestimated since at least the late Soviet period. In Kyrgyzstan, for example, the 1997 DHS estimated an average IMR of 66.2 per 1,000 live births for the period 1988–97 (Macro International 1998, p. 97). In contrast, the VR-based mortality rate for the same period was 30.8 per 1,000. There are several reasons why VR-based infant mortality in the former Soviet Union may have been underestimated. The Soviet definitions of a ‘live birth’ and a ‘stillbirth’ differed markedly from those used by the World Health Organization (WHO) (Anderson and Silver 1986). Some births counted as live births under the WHO definition were counted as stillbirths or miscarriages by the Soviets, and were not the subject of a birth certificate or a death certificate. The WHO counted every birth showing any sign of life as a live birth, whereas under the Soviet definition a baby had to be seen to breathe before a live birth was recorded. Moreover, of those Soviet babies who did draw breath, any born before the end of the 28th week of gestation, or who weighed under 1,000 grams, or were less than 35 centimetres in length, and who subsequently died in the first 7 days of life, were considered to be ‘miscarriages’ rather than ‘live births’ by the Soviets. These differences in definition meant that IMRs calculated in Soviet republics were considerably lower than they would have been had the WHO international standard definition been used. Using information on the 1960 US birth cohort, Anderson and Silver (1986) estimated that IMRs calculated using the Soviet definition would need to be adjusted upwards by 22–25 per cent if the WHO definition was used. After the break-up of the Soviet Union, the newly independent republics progressively adopted the international standard definition of a live birth. In Kyrgyzstan, this change occurred in 2004 and, as shown in Figure 1, the change generated a 43 per cent increase in the VR-based IMR between 2003 and 2005, from 20.7 to 29.7

Downloaded by [Harvard Library] at 02:18 29 December 2014

338

Michel Guillot et al.

deaths per 1,000 live births (European Health for All Database). Anderson and Silver (1999) and Aleshina and Redmond (2005) discuss additional factors which lowered the VR-based rates of IMR in the former Soviet Union even further. These include the following: the misreporting of live births as stillbirths, even for cases which would have met the Soviet definition of a live birth; the misreporting of deaths occurring late in the first year of life as deaths occurring after the child’s first birthday, which generated an unusually high number of reported deaths at the age of 12 completed months or beyond; and the under-registration of infant deaths. These problems, which occur to some degree in all former Soviet Republics, are believed to have been particularly severe in Central Asian republics, perhaps owing to the region’s lower level of socio-economic development (Anderson et al. 1994).

Patterns of infant mortality in the uncorrected vital registration data As a first step in our analysis, we examined the uncorrected VR data on which the official statistics were based. These data were collected from the archives of the NSC of the Kyrgyz Republic and consisted of unpublished aggregate tables of births and deaths in Kyrgyzstan over the 1980–2010 period. The distribution of annual births and infant deaths separately by urban and rural residence and by ethnicity were given in Form 3 for the years 1980–87 and Form A03 for the years 1988–2010. The distribution of deaths below age 2 by age at death in months was given, again by place of residence and ethnicity, on Forms 4a and 4z for the years 1980–87, and S40 and S45 for the years 1988–2010. These tables were calculated for internal purposes as part of the government’s routine tabulation plan put in place during the Soviet period. Following the breakup of the Soviet Union, the NSC continued to produce these tables using a similar tabulation plan. Only a small portion of the tabulated VR data is published in Kyrgyzstan’s annual demographic yearbook. The official IMR reported in the NSC’s annual demographic yearbook and in the European Health for All database is produced from these tabulations. The numbers of births and infant deaths by place of residence and ethnicity for each year of the study period culled from these tables are presented in the Appendix. Using these figures, annual IMRs were calculated in the conventional manner, dividing the annual

number of deaths under 1 year by the number of births occurring in that year. IMRs calculated in this way provide an approximation of 1q0, the probability that children born in a particular year will die before the age of 1. This approach was used because information on infant deaths by year of birth (which is necessary for calculating a correct 1q0) was available at the national level for Kyrgyzstan but not for all the subgroups studied in our research. In the absence of large fluctuations in the annual number of births, the conventional IMR provides a close approximation of 1q0 (Preston et al. 2001, p. 36), and a comparison of the two figures carried out at the national level in Kyrgyzstan showed that discrepancies between the two were negligible.

Infant mortality by place of residence An analysis of the detailed VR tables obtained from the NSC largely confirmed the problems with registration in Kyrgyzstan outlined above. The left-hand panel of Figure 2 shows the VR-based IMRs for the country’s urban and rural areas. Around the world in the late twentieth century, IMRs tended to be higher in rural areas (United Nations 1982; Preston and Haines 1991; Bocquier et al. 2011), owing in part to rural residents’ lower levels of socio-economic status and their greater difficulty in accessing health facilities. This was certainly true of most former Soviet republics (Kingkade and Arriaga 1997), including Kyrgyzstan in the early 1980s. However, as Figure 2 shows, the differential in IMR between the urban and rural areas decreased during the 1980s until, in 1992, there was a crossover. By the late 1990s, the VR-based IMR for rural areas was substantially lower than that for urban areas. In our view, this finding was highly implausible. The problems affecting registration of infant deaths mentioned earlier were likely to have been particularly acute in rural areas, and had the potential to create a rural advantage in infant survival that was entirely spurious. A further possible source of error that might have affected the urban/rural differential, in addition to the sources listed above, was the misreporting of place of residence on both birth and death certificates. The literature on Soviet mortality suggests that many deaths to rural residents may have been reported as ‘urban’ because the deceased died in an urban hospital and information was lacking on their actual place of residence (Anderson and Silver 1997). This problem would not have affected trends observed at the national level, but would have affected trends by place of

Infant mortality in Kyrgyzstan by Ethnicity

50

50

40

40

IMR (per 1000)

IMR (per 1000)

by Residence

30

20

Downloaded by [Harvard Library] at 02:18 29 December 2014

339

30

20

10

10 1980

1985

1990

1995

2000

2005

2010

1980

1985

1990

Year Urban

1995

2000

2005

2010

Year Rural

Slavs

Central Asians

Figure 2 Infant mortality rates (IMRs) by place of residence and ethnicity derived from official vital registration data, Kyrgyzstan, 1980–2010 Source: Unpublished vital registration tables (Forms 3 [1980–87] and A03 [1988–2010]).

residence, potentially contributing to the implausible urban/rural differential in IMR observed in Kyrgyzstan after 1992.

Infant mortality by ethnicity Kyrgyzstan is a multi-ethnic country in which the various ethnic groups have different patterns of urban and rural residence, different educational levels, and different standards of living. It is also likely that the quality of data on vital events differs by ethnic group. With this in mind, a comparison of the IMRs derived from the VR data for the different ethnic groups can be useful in helping to distinguish real from spurious trends. The right-hand panel of Figure 2 shows patterns of infant mortality for the two broad ethnic groups in Kyrgyzstan: ‘Central Asians’ (comprising Kyrgyz, Uzbeks, Kazakhs, Tajiks, and Turkmens), who were native to the region, and ‘Slavs’ (Russians, Ukrainians, and Byelorussians). These two groups accounted for between 92 and 96 per cent of annual births in the republic during the period covered by our analysis. The numbers are provided in the Appendix.

The IMRs estimated for the two ethnic groups reveal some further implausible patterns. As shown in Figure 2, the Central Asians apparently had substantially higher infant mortality than the Slavs in the 1980s. This is consistent with the established pattern of social stratification between the better-off Slavs and the less-well-off native Central Asian groups (Kahn 1993). During the 1990s, however, rates of infant mortality among Central Asians declined continuously, while among Slavs they remained more or less constant. As a result, by the late 1990s, the excess infant mortality of Central Asians had disappeared. This seems implausible, given the persistent economic disadvantage of Central Asians relative to Slavs during both the 1990s and 2000s, which has been documented by several surveys (Ackland and Falkingham 1997; World Bank 1999, 2007). The pattern of infant mortality by ethnicity is related to the patterns based on place of residence, because Central Asians live disproportionately in Kyrgyzstan’s rural areas. The patterns shown in Figure 2 suggest that the degree of underestimation of infant mortality is particularly large among Central Asians residing in rural areas. Given the trends

340

Michel Guillot et al.

shown in Figure 2, it seems likely that the underestimation worsened during the 1990s amongst this particular group.

Further specific deficiencies in Kyrgyzstan’s VR data are revealed by age patterns of infant mortality. Figure 3 shows an approximation of the probability that a baby aged exactly x months will die before age x + 1 months (1qx in life table notation, but with the subscripts expressing months rather than years) in Kyrgyzstan in 1983, which was the earliest year for which data on deaths below the age of 2 years by age in months were available. Since it was not possible to calculate either a period or a cohort life table by age in months with the available data, we approximated period 1qx by treating the number of births occurring in 1983 and the number of deaths by age in months below age 2 in 1983 as if they applied to the same synthetic birth cohort. This approximation is acceptable as long as the number of births does not vary greatly over three consecutive years. Calculations for those years for which data were available (which allowed deaths by age in months to be related to the

0.020

0.015

1qx

Downloaded by [Harvard Library] at 02:18 29 December 2014

Infant mortality by month of age

number of survivors in the corresponding birth cohorts) show that the probabilities estimated using this form of approximation were virtually the same as the ‘true’ probabilities. Applying this approach to the VR data for 1983, it would appear that mortality was highest during the first month of life, increased between ages 1 and 3 months, decreased again, and then increased once more between ages 11 and 12 months. Once again, such a pattern is highly implausible. Typically, mortality decreases with age from birth until it reaches a minimum point at around 10 years of age. Some instances in which mortality increases with age following an initial decline during the first few months of life have been documented in the literature, but these instances seem to be limited to certain high-mortality sub-Saharan African populations (Garenne 1982; Pison and Langaney 1985; Delaunay et al. 2001; Sankoh et al. 2006), particularly those with a high prevalence of malaria (Abdullah et al. 2007) or HIV (Marinda et al. 2007), or in some historical European populations (Knodel and Kintner 1977). The epidemiological contexts of these populations have little in common with that of contemporary Kyrgyzstan. A more likely explanation for the age pattern shown in Figure 3 is errors in

0.010

0.005

0 0

3

6

9

12

15

18

21

24

Age x (months) 1983

2003

2005

Figure 3 The probability of death within 1 month, by age in exact months, (1qx),1 derived from official vital registration data, Kyrgyzstan, 1983, 2003, and 2005 1

Subscripts express months, rather than years. Source: Unpublished vital registration tables (Forms 3, 4a, 4z [1980–87], and A03, S40, and S45 [1988–2010]).

Downloaded by [Harvard Library] at 02:18 29 December 2014

Infant mortality in Kyrgyzstan the data. If early infant deaths were misclassified as stillbirths and deaths under 3 months of age were under-registered to a large extent, there would appear to be an increase in mortality between the second and fourth months, and mortality at ages 0–2 months would be underestimated relative to mortality at ages 3–11 months. Similarly, the elevated values of 1qx around 12 months are probably because deaths occurring late in the first year of life were misreported as deaths occurring in the early months of the second year. It has been suggested that the misreporting was practised by medical professionals as a means of reducing the apparent IMR (Ksenofontova 1994; Aleshina and Redmond 2005). Figure 3 also shows the probability of death by age in months for 2003 and 2005, indicating that by 2005, the peaks around ages 3 and 12 months had disappeared, and a more classic age profile had been achieved. The VR data thus show clear patterns of error. While some of the adjustments necessary to correct for these errors were made using the detailed registration data, this work was complemented by the use of additional sources of information, and these are discussed in the next section.

Data and methods In the course of our study, we calculated a variety of estimates of IMR in Kyrgyzstan, which are shown in Figure 5 and discussed below. These estimates were based on three main sources of information: (1) two sample surveys—a DHS conducted in 1997 and a Multiple Indicator Cluster Survey (MICS) undertaken in 2006; (2) two censuses taken in 1989 and 1999; and (3) the detailed VR information, discussed above. These data sources were analysed using various estimation methods, as detailed below.

Sample surveys The 1997 DHS of Kyrgyzstan is a nationally representative survey of about 3,800 women aged 15–49, conducted as part of the DHS programme (Macro International 1998). This survey collected full birth histories, from which direct estimates of infant and child mortality could be derived. We are aware, however, that direct estimates of infant mortality from the DHS are subject to a greater degree of error than estimates of under-5 mortality, partly because of the misreporting of child’s age at death, in particular age heaping around age 1, the critical age which determines whether or not a death should be counted as that of an infant (Guillot et al. 2012).

341

Rather than estimate the IMR directly, therefore, we used direct estimation to calculate under-5 mortality from the DHS and then converted it into a measure of infant mortality using model life tables. The Coale and Demeny West family of life tables (Coale et al. 1983) was used for this procedure, because for the most recent, and reliable, years in Kyrgyzstan’s VR data the age pattern of under-5 mortality—in terms of 1q0 relative to 4q1 (the subscripts representing years in this instance)— conformed best to this family. To obtain estimates of IMR at the national level, we followed DHS practice by calculating mortality estimates for the three 5-year periods that preceded the survey date (1982–87, 1987–92, and 1992–97) plotted at the mid-points in Figure 5 (Macro International 1998). We calculated these from the original DHS microdata. However, for estimates of the IMRs for urban and rural areas (shown in Figure 6), we undertook calculations for only two 5-year periods (1987–92 and 1992–97) because residence status in the DHS was determined by the mother’s residence at the time of the survey, and we wanted to ensure that this variable was reasonably applicable to the retrospective information on child survival. We were unable to calculate a reliable estimate of infant mortality amongst Slavs, because the number of child deaths for this group reported in the DHS was too small. For each of the three 5-year periods (1982–87, 1987–92, and 1992–97) just eight, four, and three such deaths, respectively, were reported. This was not a problem for the Central Asian group, for which estimates of IMR could be calculated for each of the three 5-year periods (shown in Figure 7). Rather than full birth histories, the 2006 MICS collected only summary birth histories, noting just the number of children ever born and the number surviving to the date of the survey (MICS 2007). This information was converted into IMR estimates using Brass’s method and Coale and Demeny’s West model (United Nations 1983). Information on child loss contained in the responses of women in the 15–19 and 20–24 age groups was ignored in our calculations because, as is usually the case when using Brass’s method, estimates of child mortality amongst such women were implausibly high and out of line with the estimates derived from the responses of mothers in older age groups. As with the data from the DHS, and for the same reason, the MICSbased IMR estimates for urban and rural areas were calculated only for the 10 years preceding the survey (see Figure 6), and again, because of sample-size

342

Michel Guillot et al.

issues, the MICS data could not be used to estimate infant mortality amongst the Slavs.

Downloaded by [Harvard Library] at 02:18 29 December 2014

Census-based estimates of infant mortality In the 1989 Soviet census, the long form of the census questionnaire (administered to a random sample of 25 per cent of the total population) asked women to report the number of children ever born and number surviving (Goskomstat Kirgizkoy SSR 1990, p. 3). The same questions were asked of all women aged 15 and above when the first post-Soviet census of Kyrgyzstan was taken in 1999. As with the MICS data, this information could be used to estimate levels and trends in infant mortality using Brass’s method. Databases exist that hold the original individual-level data returned during the two censuses. These are held by the NSC, from which we were able to obtain detailed tabulations of the information on children ever born and children surviving, including tabulations by place of residence and ethnicity. The two census databases contain the responses of many more subjects than the DHS or the MICS, and allowed us to estimate the mortality of Slavic infants. In addition, the 1989 census provided the only data on summary birth histories collected before the break-up of the Soviet Union. The fact that the same question was asked in both 1989 and 1999 allowed us to make pre- and post-independence comparisons, something which was not possible with the data from the sample surveys. When estimating rates of infant mortality from the census data, we again used Coale and Demeny’s West model, ignored mortality information based on younger women’s age groups, and considered the information by place of residence for no more than 10 years before the time of the census. The rates estimated from the census data are shown in Figures 5–7.

Registration information on death by month of age As discussed earlier, Figure 3 illustrates two major problems detected in the figures reported in the VR data: (i) under-reporting of mortality below age 3 months and (ii) age heaping of deaths around age 12 months. It is of interest that neither of these errors affects the probability that a child aged exactly 3 months will die before they achieve the exact age of 24 months. Of course, if neither a death below age 3 months nor the corresponding birth was registered, the reported proportion of survivors at

exact age 3 months will not be affected. Similarly, if a death occurring between the exact ages of 3 and 12 months was misreported as occurring between the exact ages of 12 and 24 months, the numerator of the probability of dying between ages 3 and 24 months would not be affected. In the following discussion the probability that a baby aged exactly 3 months will die before the exact age of 24 months is represented as ‘21q3’, using the traditional demographic notation but with the subscripts referring to age in months rather than years. The measure 21q3 is not totally free of errors. In particular, an undercount of the deaths occurring in the 3–24 months age range will lower the estimate of 21q3. However, the pattern of errors shown in the data from 1983 in Figure 3 suggests that the most serious errors took place in the recording of deaths that occurred during the first 3 months of life and around the first birthday, which affected the IMR but not 21q3. (An undercount of deaths below the age of 3 months, if the corresponding births were actually recorded, would lower estimates of 21q3 because such an undercount would affect the denominator, rather than the numerator, of the probability, but the impact on 21q3 would be small relative to that of an undercount of deaths in the age range 3–24 months.) As in the case of 1qx in Figure 3, 21q3 could be estimated for each year from Kyrgyzstan’s VR data by treating deaths below 2 years of age in that year and births in the same year as if they applied to the same synthetic birth cohort. Thus the denominator of 21q3 for a given year, that is, the number of survivors at exact age 3 months, was estimated by subtracting the deaths below 3 months during that year from the number of births in the same year. The numerator of 21q3 was estimated using the number of deaths between exact ages 3 and 24 months during that same year. Levels and trends in 21q3 are useful because a strong relationship exists between 21q3 and the IMR, as illustrated in Figure 4 which, in the absence of more general models of the relationship, presents data from Sweden (1904–48) (Statistika Centralbyrån 1904–48) and England & Wales (1922–55) (General Register Office 1922–55), together with the fitted log–log linear relationship between the two indicators. Sweden and England & Wales were chosen for their high data quality and their long time series. For the two countries, the 21q3 for each year was approximated by relating deaths at the appropriate ages in that year to the number of births in the same year, as had been done for Kyrgyzstan. The time periods for which figures are shown in Figure 4 were so chosen that the observed levels of

Infant mortality in Kyrgyzstan

343

0.10

0.08

IMR

0.06

0.04

0.02

Downloaded by [Harvard Library] at 02:18 29 December 2014

0 0

0.01

0.02

0.03

0.03

0.03

0.06

0.07

0.08

21q 3

Sweden

England & Wales

Kyrgyzstan

Figure 4 The relationship between vital registration-based 21q31 and the infant mortality rate (IMR): Sweden 1904–48, England & Wales 1922–55, and Kyrgyzstan 1981–20102 1

The probability that a child aged exactly 3 months will die before reaching the age of exactly 24 months. The curve fitted to the data is based on a log–log linear model estimated using the data from Sweden and England & Wales. Source: Sweden, Statistika Centralbyrån (1904–48); England & Wales, General Register Office (1922–55); Kyrgyzstan, unpublished vital registration tables (Forms 3, 4a, 4z [1980–87], and A03, S40, and S45 [1988–2010]). 2

21q3 in Sweden and England & Wales (shown along the x-axis) were within the same range as those in Kyrgyzstan during the study period. Figure 4 also presents data points for Kyrgyzstan, derived from the VR data. As Figure 4 shows, the VR-based IMRs for Kyrgyzstan were much lower than might have been expected given the reported levels of 21q3 and the patterns seen in Sweden and England & Wales. Because 21q3 is robust to the two types of errors discussed earlier, this measure was used as the basis for the calculation of adjusted estimates of IMR in Kyrgyzstan. The adjusted IMR values were obtained by using VR-based 21q3 values for Kyrgyzstan as an explanatory variable and calculating corresponding IMR values predicted on the basis of the log– log relationship between 21q3 and IMR observed in Sweden and England & Wales. This method of correction appears reasonable, because in Figure 4 the data points for Kyrgyzstan for the years 2006–10, which have the lowest levels of 21q3 and lie at the left of the graph, conform well to the pattern observed in Sweden and England & Wales—a marked change from the data points relating to previous dates. The timing of this change towards the model pattern

coincides with Kyrgyzstan’s adoption of WHO’s definition of a live birth, and indicates an important improvement in the way information on infant mortality was collected in the country. It should be noted that the adjustment based on q 21 3 proposed here is similar to that developed by Kingkade and Arriaga (1997) to estimate levels of infant mortality in the Soviet republics in 1990, except that the latter authors did not take into account mortality observed during the second year of life. Instead, they used 6q4 (with subscripts expressing age in months) as the basis for their correction. Therefore, their correction procedure did not take into account errors which may have arisen from the misreporting of infant deaths as deaths occurring above age 1.

Results Infant mortality Kyrgyzstan

at

the

national

level

in

Figure 5 compares the trend of IMR based on Kyrgyzstan’s uncorrected VR data with the various

344

Michel Guillot et al.

IMR (per 1000)

80

60

40

20 1980

1985

1990

1995

2000

2005

2010

Downloaded by [Harvard Library] at 02:18 29 December 2014

Year VR Census 1999 MICS 2006 Final

Census 1989 DHS 1997 21q 3 adjustment

Figure 5 Comparison of estimates of the infant mortality rate (IMR), Kyrgyzstan, 1980–2010 Notes: VR = vital registration; DHS = Demographic and Health Survey; MICS = Multiple Indicator Cluster Survey. Source: Vital registration: unpublished vital registration tables (Tables 3, 4a, 4z [1980–87], and A03, S40, and S45 [1988–2010]). Census 1989 and 1999: unpublished tabulations produced from Kyrgyzstan’s 1989 and 1999 census microdata. DHS 1997: DHS 1997 microdata. Macro International (1998). MICS 2006: MICS 2006 microdata. United Nations Children’s Fund (2007).

estimates of IMR derived from the data and methods described above. The estimates all lie within a rather narrow range, and, for the most part, at a considerably higher level than the VRbased trend. In the years where they overlap, the census-based and the 21q3-based IMR estimates are somewhat lower than the estimates from the DHS, and the MICS estimates are lower still, but on the whole the different series provide a similar picture of the evolution of infant mortality in Kyrgyzstan, increasing our confidence that they bring us closer to the true picture of the pattern of infant mortality in the country over the study period. Of the different series, the one that appears most informative is that calculated using the ‘21q3 adjustment’, which provides annual estimates and is based on a consistent data source and method for the entire period of analysis. In contrast, the other approaches provide only a broad indication of levels and trends for rather limited time periods. Since the 21q3-adjusted series agrees overall with the series produced by the other

approaches for those years in which they overlap, we think it reasonable to privilege it in the discussion below. In this section we focus on the period before 2004, the year in which the standard definition of a live birth changed in Kyrgyzstan. The period 2004–10 will be discussed in a later section. The adjusted estimates of IMR confirm that the VR-based rates in the 1980s and 1990s were largely underestimated and that the amount of underestimation changed over time. If the rates calculated using our 21q3 adjustment are compared with the VR-based figures, a progressive diminution of the gap between the two can be seen during the 1980s, suggesting a decline in the degree to which the IMR calculated from VR data was underestimated. During the 1990s, however, the gap between the VRbased and the adjusted rates widens again, suggesting an increase in the amount of underestimation within the registration data. As a consequence of the change in the extent of under-registration the trend in the adjusted IMR is quite different from that shown by the VR-based IMR. While the VR-based IMR does not depart substantially from its earlier trend and continues to decline during the 1990s, the 21q3-adjusted IMR abruptly stops declining in 1990, only resuming its downward trend after 1999. In none of the series of adjusted estimates of IMR do we find evidence for a large surge in infant mortality during the 1990s, however; overall, the levels of IMR in the 1990s were lower than those observed in the 1980s. It would therefore appear that the decline observed in the VR-based IMR across the 1990s was spurious, the result of a deterioration in data quality following the break-up of the Soviet Union.

Infant mortality by place of residence Figure 6 presents our results by place of residence. Here also, there is broad agreement between the figures derived using different data sources and methods, the 21q3 series once again providing the longest times series. According to the adjusted estimates, infant mortality in urban areas remained systematically lower than in rural areas, both before and after 1991; the crossover in the VR-based rates seen in Figure 2 does not appear in Figure 6. This is consistent with an expectation of differentials in infant mortality between urban and rural areas, and supports the hypothesis that the urban/rural crossover in the VR data was artifactual. Similarly, the figures shown in Figure 6 indicate that the underestimation of infant mortality in the VR data

Infant mortality in Kyrgyzstan Rural

80

80

60

60

IMR (per 1000)

IMR (per 1000)

Urban

40

20

40

20

1980

1985

1990

1995

2000

2005

2010

1980

1985

1990

Year

Downloaded by [Harvard Library] at 02:18 29 December 2014

345

VR Census 1999 MICS 2006

Figure 6 2010

1995

2000

2005

2010

Year Census 1989 DHS 1997 21q 3 adjustment

Comparison of estimates of the infant mortality rate (IMR) by place of residence, Kyrgyzstan, 1980–

Notes: VR = vital registration; DHS = Demographic and Health Survey; MICS = Multiple Indicator Cluster Survey. Source: Vital registration: unpublished vital registration tables (Tables 3, 4a, 4z [1980–87], and A03, S40, and S45 [1988– 2010]). Census 1989 and 1999: unpublished tabulations produced from Kyrgyzstan’s 1989 and 1999 census microdata. DHS 1997: DHS 1997 microdata. Macro International (1998). MICS 2006: MICS 2006 microdata. United Nations Children’s Fund (2007).

was much larger in rural areas, which is also consistent with expectation. The adjusted estimates further indicate that the stalling of mortality decline in the 1990s, seen at the national level in Figure 5, occurred in both urban and rural areas.

Infant mortality by ethnicity Adjusted IMR estimates for the two broad ethnic groups are shown in Figure 7. Ethnicity-specific deaths by month of age were available for the years 1996–2000 only, so, unfortunately, the 21q3 adjustment can be presented for these years only. For the most recent years, no survey data were available to fill this gap. Nonetheless, for those years for which adjusted estimates were available, it can be seen that infant mortality remained consistently higher among Central Asians and that there is little evidence that the ethnic differential between this group and the Slavs narrowed over time. Furthermore, the gap between the VR-based and the 21q3-adjusted estimates was much larger for the Central Asian population, indicating that the amount of under‐

estimation in their VR data was considerably greater than that in the data for the Slavs.

The impact of the change in the standard definition of a live birth We assumed that our use of the 21q3 adjustment would largely correct for any underestimation of infant mortality arising from the use in Kyrgyzstan of a definition of a live birth that differed from the international standard. If the increase in the country’s VR-based IMR between 2003 and 2005, seen in Figure 5, was due to the international standard definition being adopted, and if the 21q3 adjustment worked as expected, then no increase in the 21q3based IMR estimates should be observed. The difference between the VR-based and the adjusted IMRs should also be seen to decrease in this 3-year period. This is indeed what is seen in Figure 5. From 2005 onward, there is virtually no difference between the VR-based and 21q3-adjusted IMRs. The impact of the adoption of the WHO standard definition of a

346

Michel Guillot et al. Central Asians

80

80

60

60

IMR (per 1000)

IMR (per 1000)

Slavs

40

20

20 1980

Downloaded by [Harvard Library] at 02:18 29 December 2014

40

1985

1990

1995

2000

2005

2010

1980

1985

1990

Year

Figure 7

1995

2000

2005

2010

Year VR

Census 1989

Census 1999 MICS 2006

21q 3

DHS 1997 adjustment

Comparison of estimates of the infant mortality rate (IMR) by ethnicity, Kyrgyzstan, 1980–2010

Notes: VR = vital registration; DHS = Demographic and Health Survey; MICS = Multiple Indicator Cluster Survey. Source: Vital registration: unpublished vital registration tables (Tables 3, 4a, 4z [1980–87], and A03, S40, and S45 [1988– 2010]). Census 1989 and 1999: unpublished tabulations produced from Kyrgyzstan’s 1989 and 1999 census microdata. DHS 1997: DHS 1997 microdata. Macro International (1998). MICS 2006: MICS 2006 microdata. United Nations Children’s Fund (2007).

live and stillbirth can also be seen in Figure 3, which shows, as expected, that between 2003 and 2005 most of the increase in VR-based mortality in the first year of life takes place below 1 month of age. Changes in age heaping around the age of 12 months played a negligible role in the increase, because the scale of age heaping had greatly decreased since the early 1980s, and also because mortality around the age of 12 months had substantially declined over the same period. In the years after the adoption of the international definition of a live birth, a puzzling pattern is seen in the left-hand panel of Figure 2. Between 2003 and 2005, there was a much larger increase in the VRbased IMR in urban areas than in rural areas. This was unexpected because, in theory, the impact of the change in definition should not vary widely amongst the various subgroups of the population. Interestingly, in the right-hand panel of the figure, the Slavs and Central Asians both appear to have been equally affected by the definitional change. This makes the different experience in urban and rural areas even more puzzling since the Slavs live predominantly in the urban areas and Central

Asians live predominantly in the rural areas and there should be some parallel between the patterns by place of residence and the patterns by ethnicity. In this case, however, the patterns by ethnicity were consistent with expectations, while the patterns by residence were not. As mentioned above, urban vs. rural comparisons are likely to have been confounded, in part, by the misreporting of place of residence on birth and death certificates. The extent of this phenomenon, however, has never been fully explored. Comparison of VR-based and 21q3-adjusted IMRs in urban and rural areas for the period 2003–05 shown in Figure 6 confirms the existence of this problem in the case of Kyrgyzstan. Following the change to the internationally accepted definition of a live birth, the VRbased IMR for the country’s urban areas exhibits a large increase, to the extent that it reaches a level above that predicted on the basis of 21q3. However, in rural areas, the VR-based IMR increases only slightly over the 2003–05 period and remains significantly below those which are 21q3-adjusted. At the national level, as Figure 5 shows, the increase in the VR-based IMR post-2004 is large but does not

Downloaded by [Harvard Library] at 02:18 29 December 2014

Infant mortality in Kyrgyzstan rise beyond the 21q3-adjusted level, which is to be expected given the change in definition. A plausible explanation for the apparent over-reporting of infant mortality in urban areas—at least relative to the levels predicted by 21q3—while under-reporting remains high in rural areas, is that many rural deaths were misclassified as urban deaths. It is likely that such misreporting of place of residence during the registration process happened before 2004 as well. It does not appear so obvious during these years, because, with the Soviet definition of ‘live births’ and ‘stillbirths’ still in place, the VR-based IMR in urban areas was underestimated to such an extent that even with the additional, misclassified rural deaths included, it appeared lower than the 21q3-adjusted IMR. In fact, given the amount of underestimation in the IMR arising from the role played by the alteration in the definitions used, the gap between the VR-based and adjusted rates shown in Figure 6 might have been expected to be even larger in urban areas before 2004, had the misreporting of place of residence been absent. Interestingly, as shown in Figure 6, the gap between the urban areas’ VRbased and adjusted IMRs suddenly narrowed during the early 1990s, a pattern that was not evident in the figures for Kyrgyzstan as a whole shown in Figure 5. It seems likely that the misclassification of rural deaths as urban deaths worsened in the early 1990s, as part of the deterioration in the quality of the vital registration noted earlier, and that this contributed to the crossover in the VR-based urban and rural rates seen in the left-hand panel of Figure 2. Assuming that, in the absence of the misclassification of the place of residence, the ratio of the rural IMR to the urban IMR in the VR data would be equal to the same ratio in the 21q3-adjusted data, we estimated that in the 2006–10 period about 36 per cent of infant deaths occurring among rural residents were misclassified as ‘urban’ deaths. This proportion appears to have gradually increased over time, from about 7 per cent during the late 1980s. The increase was particularly large between 2003 and 2004, perhaps because of the increased registration of early neonatal deaths, which are more likely to occur in a hospital and therefore more subject to misclassification of residence. It is noteworthy that the 21q3-adjusted estimates present a surprising stagnation in IMR between the years 2004 and 2007, both at the national level (Figure 5) and by place of residence (Figure 6). It is possible that the adoption of the WHO standard definitions had an impact on the quality of VR data

347

beyond the mere issue of clarifying whether a woman had had a live birth or a stillbirth, and that following the change a greater number of deaths in the first year were registered, thus generating a possibly spurious stagnation in the rates of infant mortality.

Final adjustments to IMR estimates at the national level While the 21q3 adjustment provided a most informative time series, the figures calculated using this procedure may still be underestimates since it is possible that deaths in the 3–24 months age range were undercounted in the VR data. The near perfect agreement between the VR-based and the 21q3adjusted IMRs for the years 2005–10 shown in Figure 5 should not, therefore, be interpreted as indicating that the VR-based statistics of IMR in Kyrgyzstan are now perfectly accurate. There is little information available from which a correction for such an undercount could be devised. However, it is likely that the DHS estimates are somewhat higher than the 21q3-adjusted rates for those years in which the two measures overlap because the VR-based levels of 21q3, which form the basis for the 21q3 adjustment, were probably underestimated to some degree. Therefore, we made a final adjustment to the national-level figure, to make the average adjusted IMR for the period 1982–97 match the DHS-based average for that period, that is, the 15-year period before the survey. This produced an adjustment factor of 1.129, which was applied to the entire series of 21q3-adjusted IMR estimates because, although the amount of underestimation of 21q3 may have changed over recent years, there was no information from which the variation could be evaluated. The ‘final’ adjusted figures therefore represent a combination of the useful information contained in the DHS and the VR-based levels of 21q3. With this final adjustment, our preferred IMR estimates for the period 1980– 2010 were produced. These ‘final’ estimates, which are shown in Figure 5, are also presented in Table 1, along with the original VR-based estimates of IMR (shown in Figure 5 and Figure 1), and the estimates which incorporate the 21q3 adjustment (also shown in Figure 5).

Discussion Our findings may be summarized as follows. First, we confirmed that the IMRs calculated from

348

Michel Guillot et al.

Table 1 Vital registration-based and adjusted infant mortality rates (IMRs), Kyrgyzstan, 1980–2010 Infant mortality rate (IMR): deaths per 1,000 live births

Downloaded by [Harvard Library] at 02:18 29 December 2014

Year 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

Vital registration 43.3 40.3 38.6 39.7 40.3 41.6 37.6 37.5 37.0 32.4 30.2 29.7 31.6 32.9 29.6 27.7 26.6 28.6 26.0 22.7 23.0 21.6 21.1 20.7 25.6 29.7 29.2 30.6 27.1 25.0 22.8

21q3

adjustment 1

n/a 71.5 63.0 67.7 66.5 66.8 61.9 62.7 60.6 50.5 47.2 43.7 47.3 51.5 49.0 47.1 44.5 49.7 47.3 42.7 40.2 35.6 32.7 31.0 28.8 29.6 29.2 29.6 26.2 24.8 23.3

Final estimate n/a1 80.7 71.1 76.4 75.0 75.4 69.9 70.8 68.4 57.1 53.3 49.3 53.4 58.1 55.4 53.2 50.2 56.2 53.4 48.2 45.4 40.2 36.9 35.0 32.6 33.4 33.0 33.4 29.5 28.0 26.3

1 The data needed for calculating 21q3 were not available in 1980. Therefore, the 21q3 adjustment series starts in 1981. Source: Unpublished vital registration tables (Forms 3, 4a, 4z [1980–87], and A03, S40, and S45 [1988–2010]).

registration data in Kyrgyzstan were subject to considerable underestimation owing to a combination of errors. Comparing the adjusted ‘final’ estimates with the VR-based estimates for the period 1980–2010, shown in Table 1, it can be seen that the VR-based IMR was 41 per cent lower, on average, than our adjusted IMR. However, the amount of mortality underestimation varied greatly over time. While improvements in the registration process were detected in the data from the 1980s, deterioration was observed in the 1990s, following the break-up of the Soviet Union. The decrease in VR-based infant mortality over the 1990s is therefore spurious. In reality, improvements in infant mortality stalled abruptly immediately after the break-up of the Soviet Union, and the IMR did not resume a

downward trajectory until 1999. We found no evidence to suggest that the IMR surged upward in the 1990s; in fact, it seems to have stagnated. Second, we found that the amount of underestimation in the VR data was disproportionately large in rural areas and among Central Asian ethnic groups. Consistent with these patterns of error, the crossover in the VR data between the rates of infant mortality in urban and rural areas was seen to be spurious, the result of a greater undercount of deaths in rural areas, as well as the misreporting of place of residence. We found that in fact the IMR remained consistently higher in rural areas throughout the study period. We also found that the decrease in the gap between the infant mortality of Slavs and that of Central Asians seen in the VR data was spurious; our estimates indicated that the true levels of IMR among Central Asians remained consistently higher. Third, we found that the quality of the VR data in Kyrgyzstan improved in recent years, following the adoption, in 2004, of the international standard definitions of live births and stillbirths. Nonetheless, it is probable that some underestimation of infant mortality is still taking place. Comparing the various estimates of IMR presented in Table 1, we suggest that the average officially reported IMR for the 2006–10 period was still lower than the ‘true’ value by about 11 per cent. In spite of the improvements, the misclassification of place of residence continues to generate spurious urban/rural comparisons. We estimate that in 2006–10 about 36 per cent of infant deaths occurring among rural residents were misclassified in the VR data as ‘urban’ deaths. These findings imply that it would be misguided to formulate health policies for Kyrgyzstan based on the uncorrected VR-based levels and trends in infant mortality. Infants born to Central Asian parents or born in rural areas remained considerably more at risk at the end of our study period than the VRbased statistics indicate. Resources in the republic need to be allocated, or re-allocated, to address these differentials. In light of our findings, it is useful to contrast mortality patterns in Kyrgyzstan with those observed in Russia. As noted in the introduction, the recent health crisis in Russia has been characterized by a surge in mortality at adult ages, while infant mortality remained relatively low. These patterns, together with information on causes of deaths, have led researchers to emphasize the role of adult behaviour, including alcohol consumption and violence, in the increase of mortality, and to relate this behaviour to the stark decline in the country’s

Downloaded by [Harvard Library] at 02:18 29 December 2014

Infant mortality in Kyrgyzstan economic and social conditions (Shkolnikov et al. 1998). In the case of Kyrgyzstan, a much poorer country where the economic crisis has been even more severe, adult mortality has, paradoxically, exhibited greater resilience, in part because of cultural differences influencing patterns of alcohol consumption (Guillot et al. 2011). However, the same cannot be said for mortality amongst Kyrgyzstan’s children. Our adjusted estimates show that the relative gap in infant mortality between Russia and Kyrgyzstan remained more or less constant during the 1990s, with both countries experiencing a stalling of progress. Thus, the narrowing of the gap between the two countries appearing in the uncorrected VR data and depicted in Figure 1 appears to be artifactual. Over this decade access to health care, which had been virtually free during the Soviet period, became more expensive in Kyrgyzstan and in the other Central Asian republics, as both official and under-the-table costs increased. Poverty levels also increased dramatically during this period, leaving many of the population unable to afford health care (Falkingham 2002), while at the same time the quality of the care that was provided declined substantially (McKee et al. 2002). In view of these changes, it is not surprising that the infant mortality decline stalled in the 1990s, but there is no evidence of an upward surge in infant mortality in Kyrgyzstan during the 1990s. As in Russia, declines in life expectancy in Kyrgyzstan were mostly due to mortality increases at adult ages. While this paper has focused primarily on Kyrgyzstan, our findings also have implications for other Central Asian republics. Many of the patterns found in Kyrgyzstan’s VR data are also present in the VR data reported for its neighbours. These registration statistics show, for example, that infant mortality declined in all Central Asian republics during the 1990s (European Health for All Database). An analysis of the DHS data for Uzbekistan questioned the veracity of these declines and even suggested the possibility that infant mortality in the region had increased during the 1990s (Hohmann and Garenne 2005). In view of such results, declines in infant mortality reported in the VR data for Kazakhstan, Tajikistan, and Turkmenistan in the 1990s should also be viewed with considerable scepticism. Finally, our study has some implications for choice of methods. Despite the deficiencies in the VR sources, the most informative findings come from a procedure that used VR data. Indirect methods and survey data have proved useful and provide broad indications of levels of infant mortality, but they often fail to provide precise information on short-

349

term fluctuations and differentials. For example, had only survey data been used, it would not have been possible to detect a stagnation in Kyrgyzstan’s IMR during the 1990s. Detailed VR data of the type used in this study remain underused in many less developed countries, in part because they are more difficult to access than DHS or other survey data, but also because it is assumed that they are too unreliable to be used as a basis for mortality estimation. This assumption, while true in many instances, does not apply in all less developed areas. Our study of the different data sources in Kyrgyzstan shows that, while the registration data had, and still have, a number of deficiencies, the extent of errors varied substantially by age and population subgroup. With this more nuanced evaluation, we believe that much can be learned about the true patterns of mortality from official VR data despite their deficiencies.

Notes 1 Michel Guillot is at the Population Studies Center, University of Pennsylvania, 3718 Locust Walk, Philadelphia, PA 19104, USA. E-mail: [email protected]. So-jung Lim is at Utah State University; Liudmila Torgasheva is at the National Statistical Committee of the Kyrgyz Republic; Mikhail Denisenko is at the National Research University Higher School of Economics, Moscow. 2 The authors would like to thank Zarylbek Kudabaev, former chairman of the National Statistical Committee of the Kyrgyz Republic, and Orozmat Abdykalykov, the current chairman, for providing them with access to much of the evidence used in this paper. They are also grateful to Larissa Mimbaeva and Elena Komandirova for supervising the data collection in Kyrgyzstan. Jane Falkingham provided useful comments on an earlier draft. This project was supported by grants from NICHD, R03 HD38752, and R01 HD045531.

References Abdullah, S., K. Adazu, H. Masanja, D. Diallo, A. Hodgson, E. Ilboudo-Sanogo, A. Nhacolo, S. OwusuAgyei, R. Thompson, T. Smith, and F. N. Binka. 2007. Patterns of age-specific mortality in children in endemic areas of sub-Saharan Africa, American Journal of Tropical Medicine and Hygiene 77(6): 99–105. Ackland, R. and J. Falkingham. 1997. A profile of poverty in Kygyzstan, in J. Falkingham, J. Klugman, S. Marnie, and J. Micklewright (eds.), Household Welfare in Central Asia. Houndmills, Basingstoke [England]: Macmillan Press; New York: St. Martin’s Press, pp. 81–99.

Downloaded by [Harvard Library] at 02:18 29 December 2014

350

Michel Guillot et al.

Aleshina, N. and G. Redmond. 2005. How high is infant mortality in central and Eastern Europe and the commonwealth of independent states? Population Studies 59(1): 39–54. Anderson, B. A. 2002. Russia faces depopulation? Dynamics of population decline, Population and Environment 23(5): 437–464. Anderson, B. A., K. Katus, and Brian D. Silver. 1994. Developments and prospects for population statistics in countries of the former Soviet Union, Population Index 60(1): 4–20. Anderson, B. A. and Brian D. Silver. 1986. Infant mortality in the Soviet Union: Regional differences and measurement issues, Population and Development Review 12(4): 705–738. Anderson, B. A. and Brian D. Silver. 1997. Issues of data quality in assessing mortality trends and levels in the New Independent States, in José Luis Bobadilla, Christine A. Costello, and Faith Mitchell (eds.), Premature Death in the New Independent States. Washington: National Academy Press, pp. 120–155. Anderson, B. A. and Brian D. Silver. 1999. The geodemography of infant mortality in the Soviet Union, 1950– 1990, in G.J. Demko, G. Ioffe, and Zh. Zayonchkovskaya (eds.), Population Under Duress: The Geodemography of Post-Soviet Russia. Boulder: Westview Press, pp. 73–103. Becker, Charles M., Damira I. Bibosunova, Grace E. Holmes, and Margarita M. Ibragimova. 1998. Maternal care vs. economic wealth and the health of newborns: Bishkek, Kyrgyz Republic and Kansas City, USA, World Development 26(11): 2057–2072. Bocquier, P., N. J. Madise, and E. M. Zulu. 2011. Is there an urban advantage in child survival in sub-Saharan Africa? Evidence from 18 countries in the 1990s, Demography 48 (2): 531–558. Brainerd, E. 1998. Market reform and mortality in transition economies, World Development 26(11): 2013–2027. Chen, L. C., F. Wittgenstein, and E. McKeon. 1996. The upsurge of mortality in Russia: Causes and policy implications, Population and Development Review 22 (3): 517–530. Coale A. J., P. G. Demeny, and B. Vaughan. 1983. Regional Model Life Tables and Stable Populations. New York, NY: Academic Press. 496 p. Cornia G. A. and R. Paniccia. 2000. The Mortality Crisis in Transitional Economies. Oxford: Oxford University Press. Davis, C. and M. Feshbach. 1980. Rising Infant Mortality in the U.S.S.R. in the 1970’s. US Bureau of the Census, Series P-95, No 74. Washington, DC: US Bureau of the Census.

Delaunay, V., J. F. Etard, M. P. Préziosi, A. Marra, and F. Simondon. 2001. Decline of infant and child mortality rates in rural Senegal over a 37-year period (1963– 1999), International Journal of Epidemiology 30(6): 1286–1293. Duthé, G., I. Badurashvili, K. Kuyumjyan, F. Meslé, and J. Vallin. 2010. Mortality in the Caucasus: An attempt to re-estimate recent mortality trends in Armenia and Georgia, Demographic Research 22(23): 691–732. Eberstadt, Nick. 1981. The health crisis in the USSR, New York Review of Books 28(2): 23–31. Falkingham, J. 2002. Poverty, affordability and access to health care, in M. McKee, J. Healy, and J. Falkingham (eds.), Health Care in Central Asia. Buckingham, Philadelphia, PA: Open University Press, pp. 42–56. Garenne, Michel. 1982. Variations in the age pattern of infant and child mortality with special reference to a case study in Ngayokheme (Rural Senegal). PhD. thesis, University of Pennsylvania, Philadelphia, PA. General Register Office. 1922–1955. Registrar General’s Statistical Review of England and Wales. London: Her Majesty’s Stationery Office. Goskomstat Kirgizkoy SSR. 1990. Itogi vsesoyuznoy perepisi naseleniya 1989 goda po Kirgizkoy SSR [Results of the allSoviet population census of 1989 for Kyrgyzstan] Tom 1. Frunze: Gosudarstvenny Komitet Kirgizkoy SSR po Statistike, 255 p. Guillot, Michel. 2007. Mortality in Kyrgyzstan since 1959: Real patterns and data artifacts, Espace-PopulationsSociétés 2007(1): 113–126. Guillot, M., N. Gavrilova, and T. Pudrovska. 2011. Understanding the ‘Russian mortality paradox’ in Central Asia: Evidence from Kyrgyzstan, Demography 48(3): 1081–1104. Guillot, M., P. Gerland, F. Pelletier, and A. Saabneh. 2012. Child mortality estimation: A global overview of infant and child mortality age patterns in light of new empirical data, PLoS Medicine 9(8): e1001299. Hakobyan, M. and L. Yepiskoposyan. 2006. Infant mortality in Armenia, 1992–2003, Economics & Human Biology 4(3): 351–358. Hohmann S. and M. Garenne. 2005. Enjeu politique des indicateurs de santé en Ouzbékistan soviétique et postsoviétique: l’exemple de la mortalité infantile. Paper presented at the XV IUSSP International Population Conference, July 18–23, 2005, Tours, France. Kahn, M. 1993. Les russes dans les ex-républiques soviétiques [Russians in former Soviet republics], Le Courrier des Pays de l’Est 376: 3–20. Kingkade, W. Ward, and Eduardo E. Arriaga. 1997. Mortality in the New Independent Sates: patterns and impacts, in José Luis Bobadilla, Christine A. Costello, and Faith Mitchell (eds.), Premature Death in the New

Downloaded by [Harvard Library] at 02:18 29 December 2014

Infant mortality in Kyrgyzstan Independent States. Washington: National Academy Press, pp. 156–183. Knodel, John and Hallie Kintner. 1977. The impact of breast feeding patterns on the biometric analysis of infant mortality, Demography 14(4): 391–409. Ksenofontova, N. 1994. Trends in infant mortality in the USSR, in Wolfgang Lutz, Sergei Sherbov, and Andrei Volkov (eds.), Demographic Trends and Patterns in the Soviet Union Before 1991. London: Routledge, pp. 359–378. Macro International. 1998. Kyrgyz Republic Demographic and Health Survey. Calverton, MD: Macro International. Marinda, E., J. H. Humphrey, P. J. Iliff, K. Mutasa, K. Nathoo, E. G. Piwoz, L. Moulton, P. Salama, B. Ward, and the ZVITAMBO Study Group. 2007. Child mortality according to maternal and infant HIV status in Zimbabwe, The Pediatric Infectious Disease Journal 26(6): 519–526. McKee, M., J. Healy, and J. Falkingham. 2002. Poverty Health care systems in the Central Asian republics: An introduction, in M. McKee, J. Healy, and J. Falkingham (eds.), Health Care in Central Asia. Buckinghan: Open University Press, pp. 3–11. Meslé, F., V. M. Shkolnikov, V. Hertrich, and J. Vallin. 1996. Tendances récentes de la mortalité en Russie par cause [Recent trends in mortality by cause in Russia]. Paris: INED. Meslé, France and JacquesVallin. 2012. Mortality and Causes of Death in 20th-Century Ukraine. New York, NY: Springer. Demographic Research Monographs. 279 p. Penina, O., F. Meslé, and J. Vallin. 2010. Correcting for under-estimation of infant mortality in Moldova, Population-E 65(3): 499–514. Pison, G. and A. Langaney. 1985. The level and age pattern of mortality in Bandafassi (Eastern Senegal): Results from a small-scale and intensive multi-round survey, Population Studies 39(3): 387–405. Preston, S. H. and M. Haines. 1991. Fatal Years: Child Mortality in Late Nineteenth-Century America. Princeton: Princeton University Press. Preston, S. H., P. Heuveline, and M. Guillot. 2001. Demography: Measuring and Modeling Population Processes. London: Blackwell. Sankoh, Osman A., Pierre Ngom, Samuel J. Clark, Don de Savigny, and Fred Binka. 2006. Levels and patterns of

351

mortality at INDEPTH demographic surveillance sites, in R. Feachem, D. Jamison, W. M. Makgoba, K. Rogo, K. Hofman. and E. Bos (eds.), Disease and Mortality in sub-Saharan Africa, 2nd Edition. Oxford: Oxford University Press, for the World Bank. Shkolnikov, V. M., E. M. Andreev, D. A. Leon, M. McKee, F. Meslé, and J. Vallin. 2004. Mortality reversal in Russia: The story so far, Hygiea Internationalis 4(1): 29–80. Shkolnikov, V. M., G. A. Cornia, D. A. Leon, and F. Meslé. 1998. Causes of the Russian mortality crisis, World Development 26: 1995–2011. Statistika Centralbyrån. 1904–1948. Befolkningsrörelsen [Population movement]. Stockholm: Statistika Centralbyrån. Stuckler D., L. King, and M. McKee. 2011. Mass privatization and the post-communist mortality crisis: A crossnational analysis, Lancet 373(9661): 399–407. United Nations. 1982. Levels and Trends of Mortality Since 1950. New York, NY: United Nations. United Nations. 1983. Manual X: Indirect Techniques for Demographic Estimation. New York, NY: United Nations. United Nations Children’s Fund. 2007. Multiple Indicator Cluster Survey 2006, Krygyz Republic. Final Report. National Statistical Committee of the Kyrgyz Republic: Bishek, Kyrgyzstan. Vallin, J. and F. Meslé. 2001. Trends in mortality in Europe since 1950: Age-, sex-, and cause-specific mortality, in J. Vallin, F. Meslé, and T. Valkonen (eds.), Trends in Mortality and Differential Mortality. Strasbourg: Council of Europe Publishing, pp. 31–186. Velkoff, V. A. and J. E. Miller. 1995. Trends and differentials in infant mortality in the Soviet Union, 1970–90: How much is due to misreporting? Population Studies 49(2): 241–258. World Bank. 1999. Kyrgyz Republic: Update on Poverty in the Kyrgyz Republic. Report No. 19425-KG. Washington, DC: The World Bank. World Bank. 2007. Kyrgyz republic Poverty Assessment. Report No. 40864-KG. Washington, DC: The World Bank.

352

Appendix: registered births and deaths below age 1, Kyrgyzstan, 1980–2010, by place of residence and ethnicity

Downloaded by [Harvard Library] at 02:18 29 December 2014

By place of residence

Number of deaths below age 1 By ethnicity

By place of residence

By ethnicity

Year

Total

Urban

Rural

% Urban

Central Asian

Slavic

Other

% Slavic

Total

Urban

Rural

Central Asian

Slavic

Other

1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

107,278 113,434 117,235 120,708 126,075 128,460 133,728 136,588 133,710 131,508 128,810 129,536 128,352 116,795 110,113 117,340 108,007 102,050 104,183 104,068 96,770 98,138 101,012 105,490 109,939 109,839 120,737 123,251 127,332 135,494 146,123

34,450 36,046 35,670 36,597 39,289 39,352 40,796 41,732 40,803 38,943 37,404 37,616 35,599 31,070 29,327 30,081 28,700 25,919 28,494 28,328 28,193 28,491 30,195 31,866 37,381 35,600 39,414 41,402 43,974 46,924 49,905

72,828 77,388 81,565 84,111 86,786 89,108 92,932 94,856 92,907 92,565 91,406 91,920 92,753 85,725 80,786 87,259 79,307 76,131 75,689 75,740 68,577 69,647 70,817 73,624 72,558 74,239 81,323 81,849 83,358 88,570 96,218

32.1 31.8 30.4 30.3 31.2 30.6 30.5 30.6 30.5 29.6 29.0 29.0 27.7 26.6 26.6 25.6 26.6 25.4 27.3 27.2 29.1 29.0 29.9 30.2 34.0 32.4 32.6 33.6 34.5 34.6 34.2

79,690 n/a 89,147 93,034 97,147 100,477 105,006 107,757 106,773 107,321 106,846 109,558 110,828 102,542 97,711 104,410 95,990 90,967 92,921 92,490 85,778 87,505 90,301 94,578 98,586 99,020 108,858 111,683 115,594 123,795 134,203

18,985 n/a 18,544 17,975 18,676 17,901 18,241 18,371 16,983 15,238 13,845 12,598 10,744 8,285 6,976 7,143 6,571 5,984 6,232 6,217 5,943 5,684 5,535 5,609 5,829 5,691 5,892 5,752 5,924 5,584 5,473

8,603 n/a 9,544 9,699 10,252 10,082 10,481 10,460 9,954 8,949 8,119 7,380 6,780 5,968 5,426 5,787 5,446 5,099 5,030 5,361 5,049 4,949 5,176 5,303 5,524 5,128 5,987 5,816 5,814 6,115 6,447

17.7 n/a 15.8 14.9 14.8 13.9 13.6 13.4 12.7 11.6 10.7 9.7 8.4 7.1 6.3 6.1 6.1 5.9 6.0 6.0 6.1 5.8 5.5 5.3 5.3 5.2 4.9 4.7 4.7 4.1 3.7

4,643 4,577 4,523 4,788 5,086 5,346 5,034 5,122 4,952 4,258 3,889 3,848 4,058 3,839 3,262 3,250 2,871 2,920 2,708 2,360 2,225 2,123 2,128 2,186 2,812 3,258 3,526 3,771 3,453 3,393 3,337

1,107 1,022 1,036 1,062 1,118 1,181 1,154 1,300 1,252 1,114 1,033 1,007 1,097 1,127 956 980 866 847 821 756 836 785 870 880 1,427 1,617 1,802 1,963 1,852 1,893 1,888

3,536 3,555 3,487 3,726 3,968 4,165 3,880 3,822 3,700 3,144 2,856 2,841 2,961 2,712 2,306 2,270 2,005 2,073 1,887 1,604 1,389 1,338 1,258 1,306 1,385 1,641 1,724 1,808 1,601 1,500 1,449

3,839 n/a 3,711 3,982 4,233 4,522 4,343 4,359 4,249 3,631 3,367 3,369 3,583 3,416 2,948 2,933 2,620 2,644 2,445 2,118 1,973 1,902 1,908 1,960 2,552 2,977 3,219 3,442 3,152 3,150 3,080

515 n/a 495 501 477 491 399 464 401 381 305 291 267 226 186 177 124 166 136 127 143 112 130 115 140 163 158 173 167 135 143

289 n/a 317 305 376 333 292 299 302 246 217 188 208 197 128 140 127 110 127 115 109 109 90 111 120 118 149 156 134 108 114

Source: Unpublished vital registration tables (Forms 3 [1980–87] and A03 [1988–2010]).

Michel Guillot et al.

Number of births