Viii. Use of Direct and Indirect Techniques for Estimating the Completeness of Death Registration Systems

This chapter reviews briefly the direct and indirect means of estimating the completeness with which deaths are recorded in a civil registration system. In accordance with the instructions of the organizers of the Meeting, it was designed as an adjunct to the paper presented by Moriyama,' on continuous registration systems. It should be noted that the methods can be used with little modification for systems other than civil registration, for example, survey reports on household deaths in the past year. The methods are important not simply because they provide performance measures for statistical systems. More important are the opportunities they offer for adjusting or "calibrating" those systems to provide better estimates of mortality conditions. The fact that a registration system is "incomplete" has often been used to justify ignoring its products, but if the degree of incompleteness can be accurately assessed, the system becomes virtually as useful as one that is "complete". It may be mentioned, however, that the less complete the underlying registration of deaths, the less tenable become the assumptions used in assessing the completeness level. As a rough rule of thumb, a registration system that records 60 per cent or more of deaths represents a very useful source of mortality information; if completeness is much below this level, however, problems of non-representativeness sharply limit the value of the data. It should be stated at the outset that any estimate of "true" crude or age-specific death rates implicitly provides an estimate of the completeness of death registration. Many methods of estimating the true mortality conditions have been devised, and the United Nations2 recently published a review.of these methods. These alternative estimates are usually based on surveys containing retrospective questions on the survival of children or of other kin. The surveys may include event histories complete with dating of events, as in the World Fertility Survey; or they may be limited to reports of cumulative numbers of events, in which case the proper dating may be assigned through indirect procedures. In either case, it is clear that comparisons can be made between the frequency of events during a particular period that were reported in the survey and those reported in civil registration. This chapter does not consider these independent methods for estimating mortality, which are the subject of several other papers submitted to the Meeting. Instead, it is confined to methods for evaluating completeness that use the data from the registration system itself as input to the evaluation procedure. These data are always used in conjunction with other information. If the other information consists of an independently constructed listing of deaths, then the evaluation is considered to be "direct". If it consists of other data on the population, particularly its age distribution or growth rate, then it is considered to be "indirect". Indirect procedures are reviewed first.