Research Question: How are acute hospitals in France connected to one another by the patients they share?

In this study, data from all inpatient presentations in France over a three-year period were combined to examine how often patients were admitted to pairs of hospitals on successive admissions. These data were converted into a undirected, unipartite network of hospitals connected by the frequency with which they shared patients.

This weighted adjcacency matric underwent Markov Multiscale Community detection - an unsupervised network clustering techniuque - to partition the French health system into patient-sharing clusters of hospitals.

The map below shows the results of a stable partition of the patient-sharing network with hospitals coloured according to their assigned patient-sharing cluster.

The evident spatial consistency of the map is as a result of the strong role spatial proximity between providers plays in determining the sharing of hospital patients.