Problems of Interpreting Early Mortality Data

As already mentioned, relatively few sources of early mortality statistics can satisfy the Preston- Keyfitz-Schoen criteria. The deficiencies of such data are easily enumerated (Shryock and Siegal 1980):

1.

2. Misallocation of deaths by place of occurrence. In cases where large hospitals serve an extended catchment area, it is common for death notices to be listed incorrectly, indicating the residence of the deceased to be the metropolitan area in which the hospital is located rather than the outlying areas.

3. Misallocation of deaths by time of occurrence. Deaths are commonly tabulated on a year-of- Occurrence basis, and mistakes in reporting are inevitable.

4. Age misreporting. This is a problem mostly at the older end of the age spectrum, where there is a tendency for reported ages to exceed actual ages significantly. It is also a problem in the case of children.

5. Completeness of registration. This is a problem involving both the extent and the accuracy of regis­tration. In many countries, national registration systems were not in fact national, as in the United States with its original 10-state Death Registra­tion Area. In other countries, certain populations were explicitly excluded from early registration systems, such as Maoris of New Zealand and per­sons residing in the Yukon and the Northwest Territories of Canada. Even assuming that the entire population of a nation was included in the registration process, some degree of under­counting was inevitable. This tends to be the most serious problem in the case of infants, children under 5, and persons over 65.

6. Cause misreporting. Preston, Keyfitz, and Schoen (1972) have categorized problems in this area into errors in diagnosing cause of death and errors in translating an accurate diagnosis into an entry on the death registration form. The latter include errors in categorizing deaths according to interna­tional vital statistics convention (e.g., brain tu­mors were once classified under diseases of the nervous system, but are now grouped with neo­plasms) and, more important, errors arising from complex issues in the case of multiple causes of death. To these two areas of difficulty should be added biases arising from the extent and accuracy Ofregistration. Even an incomplete death registra­tion system can, in theory, reflect accurately the population’s cause-of-death structure. Unfortu­nately, differential completeness of registration can skew the cause-of-death structure. For exam­ple, if deaths among the elderly are systemati­cally underreported, the resulting cause-of-death structure will understate the role of chronic and degenerative diseases, among them neoplasms and cardiovascular disease.

Among these problems, the fifth and sixth typi­cally are of greatest concern, and it is to these that the remaining pages of this essay are devoted.