Introduction and objective:
This study examines the geographical variation of amenable mortality in Poland, focusing primarily on the role of health care resources at the level of administrative districts and regions, and selected area socioeconomic characteristics as explanatory factors. The concept was used of amenable mortality, based on the assumption that deaths from certain causes should not occur in the presence of timely and effective health care.

Material and Methods:
Standardized death rates (SDR) from causes considered amenable to health care and, separately, for ischaemic heart disease (IHD), were calculated for each of 379 districts (NUTS 4 level) in Poland in 1991–1995 and 2006–2010, using unit mortality data from the National Causes of Death Register. The analytical procedure involved spatial analysis of the distribution of amenable mortality rates, selection of explanatory variables and fitting multilevel regression models using area-level and regional-level characteristics.

Results:
The results indicate that mortality from conditions which have become amenable to medical intervention has generally decreased in all districts of Poland in the past two decades. Considerable territorial variation in mortality can be observed. Since the 1990s, these differences have been reduced for IHD-related mortality and have increased for amenable mortality.

Conclusions:
The presented analysis only partly confirms the correlation between variables reflecting the infrastructure of health care resources and the territorial variation in mortality from these two categories of causes of death. Significant correlations with variation in mortality are revealed for the number of primary care physicians (at district level) and the number of specialist practitioners (at regional level). However, after controlling for socioeconomic variables, such as education and low income, the effect of the health care infrastructure-related variables was considerably reduced. The multi-level models also revealed a substantial variation at the regional level, which implies that there are other unobserved contextual influences on amenable mortality at this level.