Over the last 20 years in Austria infant mortality has declined from 11.2/1,000 life births (1985) to 4.7/100,000 (1997) but has remained constant since then. This stagnation is in contrast to the trend in Finland, where the infant mortality rate is both lower than in Austria and continues to decline. In attempting to understand this difference we concentrated on the spatial distribution of infant mortality in Austria in addition to the trend over time. We describe the regional distribution of infant mortality adjusted by risk factors over the period from 1984 to 2002 based on data from 1.6 million birth certificates. All variables we examined were significant due to the large number of observations. We calculated an R-squared measure to assess the ability of our regression model to predict the survival status of newborns. Only the variables birth weight, gestational age, infant's length at birth and to a lesser extent year of birth had relevant impacts in terms of predictive ability. All remaining variables did not notably contribute to the prediction of survival status of the newborn despite their significance. In the greater area of Styria, infant mortality is significantly lower than in the rest of Austria even when the mortality rates are adjusted for variables such as birth weight, gestational age, sex of the newborn and sociodemographic status of the mother. In the period from 1984 to 2002 about 1500 more infants would have survived the first year of life if the mortality rate in the rest of Austria had been the same as in this area. In our regression model many important risk factors were included. Nevertheless, we can not explain the observed spatial pattern in infant morality. Further analytic studies are needed to explore the impact of variables other than those contained in the birth certificates.