To assess quality of care, groups of care units that cared for the same patients at various stages of end-stage renal disease, might be more appropriate than the centre level. These groups constitute "communities" that need to be delineated to evaluate their practices and outcomes. In this article, we describe the use of an agglomerative (Fast Greedy) and a divisive (Edge Betweenness) method to describe dialysis activities in France. The validation was based on the opinion of the field actors at the regional level of the REIN registry. At the end of 2018, ESRD care in France took place in 1,166 dialysis units. During 2016-2018, 32 965 transfers occurred between dialysis units. With the Edge Betweenness method, the 1,114 French dialysis units in metropolitan France were classified into 156 networks and with the Fast Greedy algorithm, 167 networks. Among the 32 965 transfers, 23 168 (70%) were defined in the same cluster by the Edge Betweenness algorithm and 26 016 (79%) in the same cluster by the Fast Greedy method. According to the Fast Greedy method, during the study period, 95% of patients received treatment in only one network. According to the opinion of the actors in the field, the Fast Greedy algorithm seemed to be the best method in the context of dialysis activity modelling. The Edge Betweenness classification was not retained because it seemed too sensitive to the volume of links between dialysis units.