Analysis of neonatal mortality:is standardizing for relative birth weight biased?
1 Departments of Pediatrics and of Epidemiology and Biostatistics, McGill University, Montreal, H3A 2T5, Canada
2 Section of Epidemiology and Biostatistics, Department of Obstetrics, Gynecology and Reproductive Sciences, University of Medicine and Dentistry of New Jersey, Robert Wood Johnson Medical School, New Brunswick, New Jersey, 08901-1977 USA
BMC Pregnancy and Childbirth 2004, 4:9 doi:10.1186/1471-2393-4-9Published: 4 June 2004
Infant mortality has traditionally been analyzed as a function of birth weight and birth weight-specific mortality. Often, however, when comparing two populations, the population with higher overall mortality has lower mortality at low birth weights and a reversed pattern at higher birth weights. Methods standardizing birth weight, such as the "relative birth weight", have been proposed to eliminate these crossover effects, but such methods do not account for the separate contributions to birth weight of gestational age and fetal "growth."
Using data for singleton U.S. Blacks (n = 3,683,572) and Whites (n = 18,409,287), we compared neonatal mortality, gestational age, and the difference between the observed birth weight and the optimal birth weight (the weight at which neonatal mortality was lowest) among Black and White infants at the same relative birth weight.
At relative birth weights below zero, gestational ages were, on average, 2.4 ± 1.5 (mean ± standard deviation) weeks shorter for Blacks than for Whites for the same relative birth weight. At relative birth weights above zero, no differences were observed in gestational age, but the optimal birth weight occurred at a much higher relative birth weight in Whites than in Blacks (4150 vs. 3550 g).
Our results suggest that comparisons of neonatal mortality between groups using "relative" birth weight may be potentially biased by differences in gestational age at low birth weights, and by the distance from the optimal birth weight at higher birth weights.