Infant mortality rate is a very often used parameter to measure health status and socioeconomic progress of a community or population. The progress of the health and environmental condition can be measured spatially or temporally. The bias introduced by using the infant mortality rate without considering the differences in data collection of the health administrations in various countries can be corrected partially with a simple method. Especially when using the infant mortality rate for international studies, this bias should be corrected. Whenever the health condition of a country is to be analyzed, several parameters are basically used. Among them, the crude death rate was the first to be criticized for its lack of definition when used without adjusting to sex and age at least, if specific age rates cannot be obtained. Life expectancy also used widely, seems to be an incomplete parameter, if not accompanied by other attributes, like sex and age. That is to say, the wonder that the statistical analysis of health parameters offer to us, appear when the differences blossom, and classifications can be made. But the differences, especially if they are too glamorous, must be analyzed for the reasons at every step of the elaboration of the final data. This is especially important when comparing parameters from different countries, because what is meant at official level, is hard to be kept in normal practice. Much more difficult for the health administrations is to point out the origin of the difference of criteria in order to correct it, mainly due to the routine that generally is associated with data collection work. Considering the domestic comparison of data of a country through different periods as an isolated fact, the analysis of the differences is not strongly affected by external errors, because the existing bias, if any, is always constant, and of identic origin. When comparing cross-sectional data from an international point of view, these bias constitute the biggest handicap, and unless we are able to identify the origin, and perform a correction method, erroneous conclusions can be suspected. It has already been reported by several authors (Fedrick & Butler, 1972, and Anderson, 1981) the underregistration and miss-classification of mortality statistics derived from the "first day", "first week", "first month", and "first year" period considerations, that is, bias introduced in the origin of elaboration, which can be called primary step bias. A secondary step bias remains in the handling of already published information, derived from errors in textbooks or software packages (Bland & Altman, 1988, and Clarke & Whitfield, 1978) or from the different quality control established in health administrations (for example, in the WHO Statstics Annual 1987, still countries present data of 1980, while nearly 50% report from 1984). Still a terciary step bias is introduced by the consideration of the published data without correction for further analysis. This fact also has been recently reported (hurray, 1988).
View full abstract