The problem of processing medical data contained in the information systems of various hospitals is very complex. An important issue in this respect is the dispersion of data due to their heterogeneity and location. Analysing and processing data from different sources, different hospitals, requires the use of advanced tools at different stages of data processing. One such stage, whose results often influence subsequent phases of data processing, is feature selection and attribute ranking. This process allows for the assessment of attribute importance and helps to understand which attributes have the greatest impact on the outcome. In the paper, the authors propose a research methodology to provide a hierarchical approach for global ranking construction, for the distributed data, using dataset from the medical domain. The main objective is to identify patterns that represent knowledge common to most hospitals which can be applied to individual hospitals as local data sources and at the global level. The analysis and comparison of the results obtained at various stages of the global ranking construction are provided, highlighting the complexity of the task and the validity of the proposed approach.