A Modified Louvain Approach for Medical Community Detection using Geographic Data

Physicians create a network of associations and trust among themselves which can be observed to have an impact on the patient referral process. However, patients may sometimes disregard these referrals due to the far distances between them and the physicians. In this work, we modified the Louvain community detection algorithm to discover partitions in the intra-medical relationships network that are consistent with the locations of medical offices and may guide better referrals. We collected data from nearly 10 million appointments between 2017 and 2023, performed by 1 448310 patients in the Belo Horizonte metropolitan area. Our approach resulted in the induction of concise communities of high modularity, geographically well- partitioned, and consistent with time, highlighting the robustness of our modified Louvain algorithm.