Spread of Pathogens on Healthcare Institutions Networks: a modeling study – SPHINx

The methodological approach of the SPHINx project combines mathematical modelling,

analysis of large databases, and epidemiology.

At each scale explored (intra-hospital, inter-hospital, between hospitals and the community),

we developed mathematical models of HAI transmission. Depending on the size of the

population concerned and the context, these could be network, individual-centric or

compartmental, deterministic or stochastic models.

These models were informed using data coming from several sources, notably:

- The PMSI database on patient transfers between hospitals in France; and

- data from the iBird study on contact networks between patients and caregivers

collected at Berck maritime hospital in 2009.

At the hospital ward level, the project allowed to better understand the role played by

individual microbiota and antibiotics on the population-level selection of resistant bacteria.

At the healthcare institution level, the project led to the development of an individual-based

modelling framework to simulate pathogen spread over a network of individuals connected

through proximity interactions. This model was applied to both resistant bacteria and SARSCoV-

2 and helped inform control strategies in both contexts.

At the national level, the project led to the first reconstruction of the French healthcare

network connecting hospitals through patient transfers. Links between this network and the

spread of resistant bacteria were explored.

Finally, the project also helped model SARS-CoV-2 propagation at the community-hospital

interface, allowing to predict potential hospital saturation under various community spread

scenarios.

The obtained results are rich and have made it possible to explore a wide range of issues and

scales around the theme of healthcare networks, both within hospitals and between

hospitals. Their relevance to public health was highlighted during the COVID-19 crisis, and

their scientific contributions will continue to be explored in the coming years through

new projects.

* Duval A., Smith D., Guillemot D., Opatowski L., Temime L. (2019) CTCmodeler: An Agent-Based Framework to Simulate Pathogen Transmission Along an Inter-individual Contact Network in a Hospital. In: Rodrigues J. et al. (eds) Computational Science – ICCS 2019. Lecture Notes in Computer Science, vol 11537. Springer, Cham
* Duval A, Obadia T, Boëlle PY, Fleury E, Herrmann JL, Guillemot D, Temime L, Opatowski L, i-Bird Study group. Close proximity interactions support transmission of ESBL-K. pneumoniae but not ESBL-E. coli in healthcare settings. PLoS Comput Biol. 2019 May 30;15(5):e1006496
* Darbon A, Colombi D, Valdano E, Savini L, Giovannini A, Colizza V. Disease persistence on temporal contact networks accounting for heterogeneous infectious periods. bioRxiv 2018.
* Duval A, Obadia T, Martinet L, Boëlle PY, Fleury E, Guillemot D, Opatowski L, Temime L ; I-Bird study group.Measuring dynamic social contacts in a rehabilitation hospital : effect of wards, patient and staff characteristics. Scientific Reports. 2018 Jan 26 ;8(1):1686.
* Nekkab N, Astagneau P, Temime L, Crépey P. Spread of Hospital-Acquired Infections : A Comparison of Healthcare Networks. PLoS Comput Biol. 2017 Aug ;13(8):e1005666.
* Assab R, Nekkab N, Crépey P, Astagneau P, Guillemot D, Opatowski L, Temime L. Mathematical models of infection transmission in healthcare settings : recent advances from the use of network structured data. Current Opinion in Infectious Diseases. 2017 Aug ;30(4):410-418.

Despite advances in modern biology and medicine, healthcare-associated infections (HAI) have been occurring with increased incidence and severity over the last decades, becoming a major public health issue. A wide range of control strategies is already available, including hygiene measures, barrier precautions, antimicrobial stewardship, vaccination, patient isolation, cohorting, etc., based on screening programmes and surveillance systems. However, faced with the continued spread of HAI, the implementation of these strategies needs to be optimized and new control strategies must be considered and evaluated. Mathematical modelling and computer simulations are powerful tools which can help public-health practitioners examine possible courses of dissemination of HAI and assess the efficacy of control strategies.
The successful diffusion of a pathogen in a healthcare system results from a combination of processes operating at different scales. The way human populations are structured within and outside health-care institutions and the networks of patient transfers between wards and between institutions are layers that intrinsically interact with the biological layer of pathogen transmission and with the microbiological population dynamics layer of within-host pathogen selection. The processes operating at those different scales may have opposite effects on HAI dynamics which make HAI spread prediction and control strategies assessment difficult when considering only one scale. Yet, previously published modelling studies were typically limited to describing one of these scales, separately from the others. This separation challenges our ability to get a global understanding of HAI spread.
In this context, the objective of the SPHINx project is to propose a global approach to better understand and control the spread of HAI integrating various scales (hospital wards, hospitals and healthcare facilities networks, community) by developing a multi-scale computational framework.
First, we will build one model for each scale to understand how HAI can emerge, be selected and spread within and between healthcare facilities. Then, we will integrate these models in a multi-scale model to reproduce potential interferences between the different scales. The models will be informed using both published data and original data including data on patient transfers between healthcare facilities and the community at the French national level, as well as a detailed dataset on inter-individual contacts, microbial carriage status of patients and healthcare workers and antibiotic exposure in a single hospital over a 4-month period. Finally, we will use the developed multi-scale framework to assess and compare the potential impact of various control strategies on HAI spread and incidence, and assess their cost effectiveness from healthcare facility, payer and societal perspectives.
By increasing our knowledge of the healthcare-associated infection dynamics at various scales, the outputs of this project should contribute to design efficient and cost-effective integrated infection control measures at an individual, local, regional and national level.