Reconstructing and predicting the spatial evolution of Carbapenemase-producing Enterobacteriaceae outbreaks

Abstract

Intro Controlling the spread of infectious diseases requires correctly targeting preventative resources. When poorly deployed, these resources are misspent and have an inefficient impact. In a dynamic environment where sources of outbreaks are ever-changing, how to optimally deploy these resources is difficult to identify and can moreover change over time. In addressment, we outline and test network-driven framework that accounts for underlying patient mobility, and outbreak dynamics in hospitals to predict the temporal and spatial arrival of carbapenemase-producing Enterobacteriaceae (CPE) outbreaks. Methods We reconstructed CPE-outbreaks using a novel formulation based on transmission-dynamics, contact-interactions, and microbiology data. For each outbreak, we then examine their spatial evolution and, using background hospital population movement (entire patient population from Imperial College Healthcare NHS Trust between 2018-08-17 and 2022-02-03), we predict the arrival times of new CPE cases across wards. Findings The background mobility-network contained ward-transitions from 181,512 patients (178 wards). Based on the construction, the network comprises a single giant component and acts as a medium for disease transmission. For the results of a fitted regression predicting outbreak arrival, we included locational attributes, in addition to network distances. Overall, we found that effective distance (a graphbased path measure shown previously epidemiologically predictive) contained unique predictive power compared to edge weight. However, the effective distance could be complemented by information regarding patient demographics (Age and Sex) of patient transfers; their predictive power is suggestive of specific sub-population mobility as more important drivers of CPE. Conclusion We investigated a network-driven framework showing the potential to anticipate arrival-times of hospital CPE outbreaks. In including additional information, we also showed how specific hospital population movements were key drivers of CPE. In furthering our results, we next plan to investigate additional diseases, validate our findings beyond our current dataset, and explore further locational attributes.