Fig. 1

Spreader selection for hospital ward contact network (\(\tau = 1\)). CDI determined communities, defined for a are highlighted in b–f to visualise the iterative process of spreader selection when \(s=1\) (a–c) and \(s=0.1\) (d–f). In a, d the original network is embedded in a Euclidean space defined by the dominant eigenvectors (\(\text {v}_{1}\) and \(\text {v}_{2}\)) of \(L_e\). The node with the largest \(\text {v}_{1}\) value is the chosen spreader. In b, e, \(L_e\) is updated by removing connections from the chosen spreader. In c, f \(L_e\) is updated again by removing connections from the chosen spreader in b, e respectively. Markers denote hospital role and dot colour denotes community where community influence is ranked in a according to largest \(\text {v}_{1}\) entry in each community. Node size is mostly proportional to weighted degree, with a minimum size limit used to aid visibility