A methodology for evaluating size at birth

Data from the Swedish Medical Birth Registration, 1977-78, were utilized to develop methods of calculating reference standards for evaluating size at birth. Using the clinical information available, a 'healthy' sub-group was extracted. The individual distributions of birthweight (BW), birth length (BL) and birth headcircumference (BHc) at each week of gestational age (GA) were modelled following some truncation of their ranges. Application of the Box-Cox power family of transformations was generally found to improve the normality of the data. Certain percentages (such as 2.28, 50 and 97.72) were linked through a smoothing device. Features of the results include a positively skewed BW distribution in most GA weeks, a normal positively skewed BW distribution in most GA weeks, a normal distribution of BL and BHc at term, a larger relative variation in BW at lower GA than at term, and a distinct sigmoid shape of the median and other BW curves. Compared with five commonly quoted standards, the present data show the least deceleration in the fullterm region. Disregarding the clinical information, 'primary' sub-groups were also extracted using decomposing techniques. These gave rise to similar BW/GA relationships. Using data from the WHO comparative study, the method just mentioned was applied to standardize BW and BW-specific early neonatal mortality rates. This resulted in an improved description of biological events.