New research on predicting healthcare-acquired infections has implications for COVID-19

Governing Body Fellow Professor Felix Reed-Tsochas co-authors the new research.

 Dsc6773 Jeffrey Lienert

Lead study author Dr. Jeffrey Rewley

New research by Dr. Jeffrey Rewley, a Green Templeton Social Epidemiology DPhil alumnus, and Governing Body Fellow Professor Felix Reed-Tsochas, in collaboration with Dr. Laura Koehly and Dr. Chris Marcum of the National Institutes of Health (NIH), develops and tests a novel method for predicting healthcare-acquired infections in patients using hospital data. The data allow the number of hours a patient spends in the same hospital ward as a patient suspected of infection to be captured, and this measure can then be used to predict the likely transmission of four different infections with high accuracy.

The new research, published in the Journal of Hospital Infection, is accompanied by a Healthcare Infection Society blog post by Dr. Rewley discussing the prediction method and how it may apply to contact tracing for COVID-19.

Professor Reed-Tsochas explains more about the paper’s analysis:

“The paper is based on the analysis of an extensive dataset from the Oxford University Hospitals NHS Foundation Trust, combining administrative information (e.g. when patients are admitted, which ward they are located on, or when they are discharged) and medical information (e.g. blood tests). We use the administrative information to construct a so-called ‘co-presence measure’, which records how long any two patients are in spatial proximity to each other. We then show that this measure can be used to identify patients at risk of being infected, based on how long they have been in the vicinity of other patients who are suspected of harbouring an infection.

Identifying patients at risk of infection in this manner provides an opportunity to prioritise tests for these patients more effectively than can be done with existing approaches. We also show that in principle, it could lead to detecting some infections up to a day earlier, on average, potentially resulting in better containment of an outbreak.

The focus of our research is on so-called nosocomial infections (e.g. the bugs like MRSA that are notorious for spreading within hospitals), and the data underpinning our research corresponds to the period before COVID-19 emerged. Hence, our research doesn’t study COVID-19 directly, but it is possible to consider how our methodological advance may be applied to the current global pandemic as well.”

With respect to the current COVID-19 pandemic, the research shows first that hospitals may be able to use the developed algorithm to predict nosocomial COVID-19 spread without large financial or labor overhead. Extending this further, the predictive power of the algorithm used here could be incorporated into mobile apps designed to detect community spread of COVID-19 via co-presence.

Because the data used here were not minutely-detailed in terms of location (the researchers only knew what bed a patient was assigned to, not their specific location at all times), the strong predictive power indicates that such mobile apps may function well without minutely-detailed data themselves, limiting the potential for ethical quandaries surrounding privacy issues. Read more about the research article and its implications for COVID-19.

Dr. Rewley is an Advanced Fellow for Health Services Research at the Philadelphia VA Medical Center and the Perelman School of Medicine. He completed his PhD in Social Epidemiology at the NIH and Oxford University, where the work discussed below was completed. Professor Felix Reed-Tsochas is Academic Tutor and a Governing Body Fellow of Green Templeton College. He is Professor of Complex Systems, Saïd Business School, University of Oxford.

Created: 2 September 2020