observation per country. Table 6: Estimates of adjustment factors for infant mortality rates in CEE/CIS countries, 1987-2000 Average official infant mortality rate 1987-2000 Adjustment factor (per cent) Adjusted infant mortality rate 1987-2000 20.0 +65.0 33.07 Russia (10) 17.9 +26.5 22.61 Ukraine (7) 13.9 +54.5 21.52 Average +46.6 Source: Kingkade and Sawyer (2001), Table 2. Note: The adjustment factors represent average adjustment for years between 1987 and 2000 to make the pattern of infant death before the fourth month, and at the fourth and tenth months, consistent with historical patterns observed in US and German life tables. The number of years’ data from which the adjustment factors are calculated are in parentheses. 14 The mismatches outlined on Tables 5 and 6 for several countries in South Eastern Europe and the CIS do not constitute proof that infant deaths are being hidden or recorded as stillbirths or miscarriages. However, they do show that the pattern of infant death does not conform to expected patterns in many countries. Moreover, they suggest that the problem of undercounting of infant mortality may not just be confined to countries of the Caucasus and Central Asia, where surveys confirm this to be the case, but also to other countries in the CIS and South Eastern Europe, where survey data are not available. Mis-reporting early neonatal deaths as stillbirths It is also possible, using official statistics, to address the more specific issue of whether early neonatal deaths are being mis-reported as stillbirths. This is done by examining the relationship between stillbirths and early neonatal deaths in the official data. Given the mortality rate that prevailed (according to official counts) in several transition countries during the 1990s, one might expect to find a ratio of about one stillbirth for every early neonatal death. A ‘stillbirths ratio’ far in excess of 1:1 suggests that some early neonatal deaths were being misreported as stillbirths (Velkoff and Miller, 1995). Figure 1 shows stillbirths ratios for some countries and years where the official neonatal mortality rate ranged from 6.5 to 11 per 1,000 live births. Data are available for most CIS countries, but not for any countries in South Eastern Europe except Serbia and Montenegro. The expected 1:1 stillbirths ratio prevailed in Belgium and Greece in the 1970s, Estonia in the 1980s, and in Serbia and Montenegro in the 1990s. The 1:1 stillbirths ratio was also evident in Russia in the late 1980s, but it began to rise steeply from 1990. In Georgia, the ratio increased from 1992 onwards. In Turkmenistan, Ukraine and Uzbekistan, the stillbirths ratio climbed during the 1990s to about 2:1 or even 3:1. 15 Figure 1: Ratio of stillbirths to early neonatal deaths, 1970-2000 Source: WHO Health for All database. Note: Only those countries and years are selected for which the early neonatal mortality rate was between 6.5 and 11 per 1,000 live births. Depending on the country and the year, the definition of live birth may vary. The case of Georgia is interesting because this country officially adopted the WHO definition of live birth in 1994 (see Table 3), and because a study of birth outcomes in the year 2000 in hospitals in Tbilisi provides at least a partial independent check on what the expected stillbirths ratio should be. Asatiani (2001) examined medical records (primary medical history journals rather than medical death certificates) for 15,000 births at obstetrics clinics in Tbilisi, Georgia, in 2000 – a third of all births in Georgia in that year. In this sample, there were 22.3 stillbirths per 1,000 births, and 15.6 early neonatal deaths per 1,000 live births, giving a stillbirths ratio 1.43:1 – a far lower ratio than what the official data show, suggesting perhaps a degree of massaging of official infant death figures. Trends in the stillbirths ratio for Armenia, Kazakhstan and Kyrgyzstan, on the other hand, are difficult to explain. In these countries, the ratio remained at about 1:1 or declined towards this level during the 1990s. At face value, this would imply accurate counting of early neonatal deaths. Given the gap between survey based estimates and official data, and the evidence of underestimation of neonatal deaths in Table 5, this idea is difficult to sustain. Moreover, small scale analyses of hospitals in Armenia and Kazakhstan tend to confirm the practice of misreporting (Ministry of Health of and UNICEF Armenia, 2002; Wuhib, 1998; Ministry of Health and ZAO, 2002). Another possibility is that in these countries, a significant number of infant deaths are simply not reported at all, for example, because they follow home births1996 1998 2000 Ukraine Turkmenistan Russia Georgia Uzbekistan Armenia Estonia Belgium Greece Kazakhstan Kyrgyzstan Serbia & Montenegro 16 Mis-reporting post neonatal deaths as child deaths Another specific type of misreporting that can be investigated with an ‘internal consistency’ check is the possibility that deaths of older infants are recorded as occurring after their first birthday, and are thus not included in the infant mortality count. Velkoff and Miller (1995) cite a study by Ksefontenova (1990) which suggests that in the Muslim republics in particular, misreporting of the infant’s age was used to reduce the infant mortality count. We do not have official data for infant deaths by age in months, or by age in single years past their first birthday. However, it possible to compare official and