COMPARISONS OF INFANT-MORTALITY USING A PERCENTILE-BASED METHOD OF STANDARDIZATION FOR BIRTH-WEIGHT OR GESTATIONAL-AGE

Comparisons of infant, perinatal, or neonatal mortality across populat ions with different birthweight or gestational age distributions are p roblematic. Summary measures with adjustment for birthweight or gestat ional age frequently are invalid or lack interpretability. We propose a percentile-based method of standardization for comparing infant, per inatal, or neonatal mortality across populations that have different d istributions of birthweight and/or gestational age. The underlying con cept is a simple one: comparable health for two population groups will be expressed as equal rates of disease or mortality at equal quantile s in the two distributions of birthweight or gestational age. We descr ibe this method mathematically and present an example comparing mortal ity rates for African American vs European-American infants in North C arolina. When gestational age is transformed to its rank, the well-kno wn crossover in mortality rates, in which preterm African-American inf ants die at lower rates but term infants at higher rates, disappears: African-Americans show higher mortality rates at any percentile of ges tational age. With homogeneous mortality race ratios! a summary statis tic becomes meaningful. We also demonstrate adjustment for percentile transformed gestational age or birthweight in multiple logistic regres sion models. Percentile standardization is easily implemented, has adv antages over other methods of internal standardization such as that of Wilcox and Russell, and communicates an intuitive public health-based concept of equality of mortality across populations.