level. These tables are the main source of mortality statistics used in the present study.
This system has been in effect with little change since the late 1950s under the control of the Health Care Ministry and the State Committee on Statistics (Goskomstat). During this period, four Soviet classifications were used in succession (Meslé et al., 1992). The oldest of these, in use from 1955 to 1964, was rather different from the International Classification of Diseases (ICD), but the other three, used in 1965-1969, 1970-1980, and since 1981, are more closely related to ICD-7, ICD-8, and ICD-9, respectively.
For further analysis, it is important to point out three peculiarities of the Soviet certification and registration system.
First, although the Soviet classification has become more closely related to the ICD definitions since 1965, important differences remain. The nosological categories are much more aggregated than those of the ICD (210 items instead of more than 2,000 in 1965-1969, and since 1970, 185 instead of almost 3,000 in ICD-8 and about 5,000 in ICD-9) (see also Kingkade and Arriaga, in this volume). Correspondences between items in the Soviet and ICD classifications are described in special Goskomstat documents (Goskomstat, 1981).
Second, until 1988, certain cause-of-death items in the Soviet classification were not used in the regular statistical tables; these causes (cholera, plague, suicide, homicide, and accidents at work) were hidden for political reasons. To maintain correct totals for all causes combined, the hidden causes were mixed with ill-defined ones. However, the hidden causes were tabulated separately in a top secret special table, which we finally were allowed to access for the years 1963 to 1987. The analysis of these data provided below helps clarify the reasons why these causes were kept so secret. In 1970, for instance, the standardized death rate by homicide was about eight times as high as the European average rate.
Finally, the system for coding of causes of death is decentralized, which means some regional differences in coding practice exist in spite of uniform instructions from the Health Care Ministry and Goskomstat. Goskomstat makes no attempt to verify or revise cause-of-death codification provided by regional statistical offices. When analyzing data by republics, it is obvious that the varying rules applied can produce different effects or introduce changes at different times. We can, however, hope for less heterogeneity within the republics.
Very few data on causes of death were published in the former Soviet Union. In the 1960s and early 1970s, only some aggregated data on cardiovascular and cancer mortality were published. The situation became even worse during the years 1974-1987, when the Soviet government decided to forbid any publication