Some features of the distribution of birthweight of human infants

摘要

This paper is concerned with the distribution of birthweight of some 18,700 singleton live and still births taking place in Cornwall, Devon, Plymouth, and Exeter in 1965. Information collected as part of the statutory birth registration process was linked to the corresponding information about the birth recorded by the local authority health department. The birthweights were originally recorded in ounces, but the irregularities in the pattern of results suggest that the observations are subject to substantial observer error and are accurate at best to the nearest quarter pound.The general form of the birthweight distribution is examined and it is shown that the overall distribution can be effectively summarized by a mixture of two normal components. Because of the inaccuracies of the basic data concerning birthweight, it is not possible to obtain reliable estimates of all the five parameters necessary to describe a mixture of two normal components. However, the distribution (which we convert to the metric scale) can be described adequately in terms of the mean and standard deviation of the primary or major component and the proportion of births weighing less than 2,000 g, the corresponding figures for all single births being 3,380 g, 500 g, and 2·0% respectively.The population of single births was subdivided in turn on the basis of sex, social class, parity, and maternal age. The mean of the primary component was about 120 g higher in males than in females, about 130 g higher in parity 1 births than in parity 0 births, about 110 g higher in births to mothers of 30-34 years of age than in births to mothers of less than 20 years, and about 150 g higher in social classes 1 and 2 than in social class 5. When two or more attributes are considered in combination, the social class gradient is present in most of the various subpopulations formed by subdivision of the population in terms of the other attributes. However, there seem to be complex interactions in terms of sex, parity, and maternal age. For example, the estimates of the mean of the primary component for males of parity 0 increase with increasing maternal age, whereas the corresponding estimates for females of parity 0 decrease. The variability of the primary distribution is greater for males than for females and increases with decreasing social class, increasing parity, and increasing maternal age. The proportion of births weighing less than 2,000 g was similar in males and females but decreased with increasing social class. In terms of parity, the proportion was higher for parity 0 than for parity 1 but subsequently increased with increasing parity. A similar pattern exists for maternal age in male infants, the proportion decreasing with age to reach a minimum in the 25-29 years age group and subsequently increasing as maternal age increases further. Female infants, on the other hand, show a totally different pattern in the lowest maternal age groups.