Fig. 6

Visualisation of the multilayer network. Nodes represent clusters, and directed edges between the nodes are weighted according to the rate at which patients move between them. The direction of the edges is always from the node with lower to the one with higher label; in most cases, this is from left to right. The edges are coloured depending on the average age of the transitioning patients: light (dark) green edges indicate their low (high) average age. Node size is proportional to the size of the corresponding cluster. The node colour indicates the cluster mortality, with white standing for low and red standing for high mortality. For clarity, in the visualisation, we removed cluster 0 (the “healthy cluster”) which is connected to all other nodes. The layout has been generated by a force-directed algorithm after which nodes have been displaced from left to right by a distance proportional to the average age associated with its links, i.e. patient age tends to increase from left to right. To improve readability, we only display a filtered version of the network with the links representing statistically robust transition rates greater than 0.5%. Nodes which are isolated under this threshold have been aligned in the upper left corner. For elderly patients, we observe network regions of low, medium and high mortality. Patients access the high-mortality regions via the high-risk clusters 112 and 120 (blue circles)