A Bayesian spatiotemporal model for prevalence estimation of a VRE outbreak in a tertiary care hospital

A Atkinson; B Ellenberger; V Piezzi; T Kaspar; O Endrich; A B Leichtle; M Zwahlen; J Marschall

doi:10.1016/j.jhin.2021.12.024

A Bayesian spatiotemporal model for prevalence estimation of a VRE outbreak in a tertiary care hospital

J Hosp Infect. 2022 Apr:122:108-114. doi: 10.1016/j.jhin.2021.12.024. Epub 2022 Jan 26.

Authors

A Atkinson¹, B Ellenberger², V Piezzi³, T Kaspar³, O Endrich⁴, A B Leichtle⁵, M Zwahlen⁶, J Marschall⁷

Affiliations

¹ Department of Infectious Diseases, Bern University Hospital, Inselspital, University of Bern, Bern, Switzerland. Electronic address: andrew.atkinson@insel.ch.
² Insel Data Science Centre, Bern University Hospital, Inselspital, University of Bern, Bern, Switzerland.
³ Department of Infectious Diseases, Bern University Hospital, Inselspital, University of Bern, Bern, Switzerland.
⁴ Medical Directorate, Bern University Hospital, Inselspital, University of Bern, Bern, Switzerland.
⁵ Insel Data Science Centre, Bern University Hospital, Inselspital, University of Bern, Bern, Switzerland; University Institute of Clinical Chemistry, Bern University Hospital, Inselspital, University of Bern, Bern, Switzerland.
⁶ Institute of Social and Preventive Medicine, University of Bern, Bern, Switzerland.
⁷ Department of Infectious Diseases, Bern University Hospital, Inselspital, University of Bern, Bern, Switzerland; Division of Infectious Diseases, Washington University School of Medicine, St. Louis, MO, USA.

PMID: 35090955
DOI: 10.1016/j.jhin.2021.12.024

Abstract

Background: There was a nosocomial outbreak of vancomycin-resistant enterococci (VRE) at the hospital between 1^st January 2018 and 31^st July 2020. The goals of this study were to describe weekly prevalence, and to identify possible effects of the introduction of selected infection control measures.

Methods: A room-centric analysis of 12 floors (243 rooms) of the main hospital building was undertaken, including data on 37,558 patients over 22,072 person-weeks for the first 2 years of the outbreak (2018-2019). Poisson Bayesian hierarchical models were fitted to estimate prevalence per room and per week, including both spatial and temporal random effects terms.

Results: Exploratory data analysis revealed significant variability in prevalence between departments and floors, along with sporadic spatial and temporal clustering during colonization 'flare-ups'. The oncology department experienced slightly higher prevalence over the 104-week study period [adjusted prevalence ratio (aPR) 4.8, 95% confidence interval (CI) 2.6-8.9; P<0.001; compared with general medicine], as did both the cardiac surgery (aPR 3.8, 95% CI 2.0-7.3; P<0.001) and abdominal surgery (aPR 3.7, 95% CI 1.8-7.6; P<0.001) departments. Estimated peak prevalence was reached in July 2018, at which point a number of new infection control measures (including the daily disinfection of rooms and room cleaning with ultraviolet light upon patient discharge) were introduced that resulted in decreasing prevalence (aPR 0.89 per week, 95% CI 0.87-0.91; P<0.001).

Conclusion: Relatively straightforward but personnel-intensive cleaning with disinfectants and ultraviolet light provided tangible benefits in getting the outbreak under control. Despite additional complexity, Bayesian hierarchical models provide a more flexible platform to study transmission dynamics.

Keywords: Bayesian modelling; Outbreak; Prevalence; Screening; VRE.

MeSH terms

Bayes Theorem
Cross Infection* / epidemiology
Cross Infection* / prevention & control
Disease Outbreaks
Gram-Positive Bacterial Infections* / epidemiology
Gram-Positive Bacterial Infections* / prevention & control
Humans
Prevalence
Tertiary Care Centers
Vancomycin-Resistant Enterococci*