Data Quality

The autopsy rate and the number of deaths attributed to "malignant neoplasms of other and unspecified sites" (ICD 195-199) frequently serve as indirect indicators of the quality of cause-of-death statistics. A direct quality criterion is the comparison of the statements made on death certificates with the clinical diagnosis of the fatal illness or with pathologic findings.

The autopsy rate in western Germany, at 8%, is substantially lower than the 25% autopsy rate in the former GDR. Other European countries also report higher frequencies of autopsies. The autopsy rates in Central Europe range from 9% in Poland to 50% in Hungary (Lee 1994).

Cancers of "other and unspecified sites" (ICD 195-199) were such a frequent cause of death in the mid-1970s that they represented the fifth leading reported cause of cancer deaths in males and the fourth leading cause in females. While these rates have fallen markedly since the early 1980s, ICD codes 195-199 are still among the ten most frequent causes of male and female cancer deaths in Germany (see Fig. 1, p. 24, and the table on cancer mortality, p. 46).

High rates in this category signify and underestimation of the "true" causes of cancer deaths. A decline means that the statements on death certificates have become more accurate over the decades, leading to an artifactual increase in the mortality rates for the diagnosed cancers. This must be considered when the secular mortality trends for specific types of cancer are interpreted.

The maps displaying these data use an absolute scale and cover the entire decade from 1981 to 1990. Compared with the maps for specific cancer sites, these maps illustrate the relatively high frequency of poorly specified causes of cancer deaths in the western part of Germany. As the maps indicate, it was uncommon in the former GDR to declare unspecified causes of cancer deaths. The maps also show marked regional differences in the tendency to cite primary tumors at unspecified sites as the cause of death.

Numerous authors have made direct comparisons of the statements on death certificates with clinical diagnoses and pathologic findings (surveys in Boyle 1989 and Lee 1994). Generally it has been found that a detection rate of 70-80% implies that malignant tumors are underreported as a cause of death. The accuracy of the statements varies for different cancer sites and is poorer in the older population than in younger patients (Modelmog et al. 1992, Hoel et al. 1993, Lee 1994, Ron et al. 1994). Accordingly, there appears to be an urgent need for improvement in the quality of the information placed on death certificates.

A study conducted in Germany found that 25% of the physicians surveyed did not know that cause-of-death statistics were kept. Another one-third of physicians were unaware that every death certificate is incorporated into these statistics (Müller and Bocter 1990). Thus, an initial step in improving the quality of cause-of-death statistics wold be to correct this information deficit and emphasize the importance of accurate death-certificate statements for public health policy and science.

The validity of mortality data from the former GDR constitutes a specific problem that is not addressed by indirect quality indicators and, except for the study cited above (Modelmog et al. 1992), has not been adequately investigated by direct comparisons. As mentioned, deaths in the former GDR were coded at once by the physician who filled out the death certificate, and the cause of death was checked only by the district physician. No quality control measures were implemented.

After German reunification, the standard coding practices used in western Germany were adopted in the eastern states, and a marked structural change was noted in the cause-of-death statistics in eastern Germany beween 1990 and 1991 (Brückner 1993): The number of deaths attributed to diseases of the circulatory system (ICD 390-459) decreased by 4.5%, and deaths attributed to diseases of the respiratory system (ICD 460-519) decreased by 12.6%. Meanwhile, the number of deaths in which a malignant neoplasm (ICD 140-208) was indicated as the cause of death rose by 13.3.%. The detailed data presented in the specific sections of "Cancer Mortality by Site" will show that this abrupt rise is based on a deficit of cancer deaths in 1990 and previous years, and that the deficit occurred mainly in the population age 65 and over (see also Modelmog et al. 1992).

In contrast to other general shifts in cause-of-death statistics, this underreporting cannot be explained in terms of noncompliance with WHO rules for the identification of underlying diseases. Incidence data from the cancer atlas of the former GDR show that the age-related cancer incidence is rising more slowly than in the west for the 65-and-over population (see Overview of Cancer Mortality below) and that, contrary to internationally published data, there is actually a downturn in the higher age range.

This phenomenon has not yet been fully explained. One interpretation may be that fewer diagnostic and therapeutic options were available to the elderly in the former GDR than to younger age ranges and in relation to western Germany (Ziegler and Stegmaier 1996). It is clear that the inconsistencies that have appeared in the cause-of-death statistics of the former GDR, especially those relating to cancer mortality, are in need of further analysis.

to top