Figure 1.
The structure of the hospital referral network in England, based on the NHS Hospital Episode Statistics for the year 2006–07.
In this period, 7,420,219 patients were admitted to 146 acute NHS hospital trusts, for a total of 12,929,171 health-care episodes (corresponding to 143 inpatients and 249 admissions per 1000 inhabitants). Markers indicate hospitals; squares, diamonds, large dots and small dots denote respectively the acute teaching, large, medium and small acute hospitals. The thickness of the lines between nodes indicates the number of patients that are referred between hospitals. Different colours indicate regional hospital clusters as identified by community detection algorithm and defined as hospitals that share more patients among themselves than with other hospitals. Typically, regional hospital clusters are centred around acute teaching hospitals, and have a total number of hospitals ranging from 5 (in Sheffield) to 25 hospitals (London South & West). Hospital clusters are numbered according to size.
Figure 2.
Differences between hospital clusters in the English hospital referral network.
A) The size distribution and composition of hospital clusters, categories are acute teaching (gray), large (blue), medium (red) and small acute hospitals (black). B) The MRSA bacteraemia incidence rate per hospital cluster, between 2001 and 2009, although the overall MRSA bacteraemia incidence rate declines, the ordering of clusters remains largely the same; London S&W (cluster 1) presents highest rates in all years, whereas Sheffield (cluster 12) shows lowest rates in all years except 2001–02 C) The correlation between the number of hospitals within a cluster and the mean incidence rate is significant in all years. Larger clusters show higher rates in all years.
Figure 3.
Correlation between proportion of potentially infectious patients among all admitted patients (infectious relative indegree, IRI) and the MRSA bacteraemia incidence rate at hospital level, in England between 2001 and 2009.
A) The proportion of potentially infectious patients among all admitted patients (log IRI) by hospital category. This proportion increases with hospital category, from small acute care hospitals to teaching hospitals. B) The MRSA bacteraemia incidence rate per hospital, between 2001 and 2009 (thin lines), and the mean per hospital category (Thick Lines), the MRSA incidence rate is highest in acute teaching hospitals. C) Correlation between the hospital log IRI and MRSA bacteraemia incidence rate for all regional hospital clusters. Over the 8 years, 20 times a cluster showed a significant positive correlation, while none showed a significant negative correlation. D) Partial correlation coefficient between the hospital log IRI and MRSA bacteraemia incidence rate for all hospitals, adjusted for incidence differences of regional clusters. Hospitals with a high degree of connectedness show higher MRSA rates than their lesser connected counterparts.
Figure 4.
The simulated spread of MRSA at national level (assuming no interventions and equal effective case reproduction numbers) for England (red) and The Netherlands (blue).
A) Proportion of hospitals with MRSA positive patients, arrows show the number of hospitals in both countries, showing faster dispersal in England as compared to the Netherlands. B) Mean MRSA prevalence among hospitals. C) The distribution of time to 50 hospitals infected. D) The percentage of simulated introductions of MRSA resulting in an epidemic.