Does gestational age in combination with birthweight provide better statistical adjustment of neonatal mortality rates than birthweight alone?

Between-area comparisons of neonatal mortality rates should be adjusted for differences in the underlying mortality risk. The traditional approach to this problem is to adjust neonatal mortality rates statistically for between-area differences in the birthweight distributions. However, in other types of perinatal research, birthweight is usually considered in combination with gestational age. For between-area comparisons of neonatal mortality rates, some researchers have argued that adjustment by gestational age in addition to birthweight might not be necessary. This present study used graphical methods based on a non-parametric version of Poisson regression to underline the importance of examining neonatal mortality rates by both gestational age and birthweight. Six years of data from a whole-population database (Queensland Perinatal Data Collection) were used. The analysis also illustrates the value of non-parametric modelling in perinatal epidemiology.