Appraise infant mortality evolution at different territorial scales, identify disparities in terms of trend evolution and explore the contribution of some socioeconomic factors (educational attainment, unemployment rate and ethnicity) in the existing regional differences of infant mortality levels. As the aim of our study is to identify geographic disparities based on different spatial tendencies in infant mortality evolution, trend analysis was the most suited method. Cluster analysis was used to organize data into meaningful structures and allowed the identification of relatively homogenous groups of counties and municipalities. With the help of GIS techniques we have created cartographic material for a better visualization of the results. Linear regression was used to analyze the contribution of some socioeconomic factors on regional differences in infant mortality level. The results have demonstrated that, although the general downward trend is characteristic for infant mortality evolution in Romania, important variations between different geographic areas and between different social classes remain. Determined by the level of analysis, the results will vary. The model that summarizes the combined influence of all variables on infant mortality at national level explained 25% of the infant mortality variability and only 7% at district level. Of all independent variables, percentage of population with no education was the only indicator identified as having a significant influence on infant mortality. Multilevel analysis of infant mortality can give to scientists a better perspective and a better understanding of its relations with socioeconomic and biological determinants.