Cookies on this website

We use cookies to ensure that we give you the best experience on our website. If you click 'Accept all cookies' we'll assume that you are happy to receive all cookies and you won't see this message again. If you click 'Reject all non-essential cookies' only necessary cookies providing core functionality such as security, network management, and accessibility will be enabled. Click 'Find out more' for information on how to change your cookie settings.

Oxford University co-lead: Prof Sarah Walker

UKHSA co-lead: Prof Susan Hopkins

Our strategy is to exploit large-scale linked EHR data from multiple sources to answer the following key questions:

  • How can routine surveillance be automated optimally?
  • What populations are at greatest risk of different AMR&HAI?

Our approach will take advantage of the in-depth linked regional data from the Infections in Oxfordshire Research Database (IORD) for initial analyses, before testing generalisability in larger national datasets (Hospital Episode Statistics (HES) linked to the PHE Second Generation Surveillance System (SGSS)) which have less granularity.


Mandatory surveillance of healthcare associated infections and pathogens using linked electronic health record data  

We are investigating how we can extend data linkage approaches developed for COVID-19 between UKHSA’s Second Generation Surveillance System (SGSS) and NHS Digital’s emergency care datasets (ECDS), secondary use statistics (SUS) and Hospital Episodes Statistics (HES) to other pathogens subject to mandatory surveillance (including E. coli bacteraemia), validate these against traditionally captured data in the mandatory surveillance database HCAI-DCS (particularly in terms of the Trust that each episode is assigned to), and compare availability, completeness, and agreement of additional fields derived from HES/SUS, as well as timeliness. 

A component of this project has been the development, implementation and optimisation of methods to automate the monitor of quality of big data feeds.   


Implementation of whole genome sequencing (WGS) for surveillance and control of Clostridium difficile – in conjunction with Interventions them 

To generate recommendations for the optimal role of WGS in different kinds of outbreaks and AMR&HAI surveillance, we are using an implementation study, first updating the “network” modelling of patient movements between NHS Trusts to identify sentinel sites for C. difficile surveillance, then characterising contemporaneous background diversity circulating C. difficile strains in order to interpret new sequences which will be generated as the Clostridium difficile Ribotying Network (CDRN) moves to WGS.  


Mandatory surveillance of healthcare associated infections and pathogens using linked electronic health record data  

This project builds directly on learning from COVID-19, utilising methods developed to identify changepoints in infection surveillance datastreams and to identify “at risk” populations for COVID-19.  


Amoxicillin prescribing and E. coli bloodstream and urine infections  

We are taking advantage of the “natural experiment” of reduced community prescribing due to COVID-19, and new linked individual level community prescription records with SGSS and HES to investigate the association between wide-scale amoxicillin and related co-selection antibiotic prescribing in the community and the incidence of E. coli bloodstream infection (BSI) or urinary tract infection (UTI) at an individual-level in England. 


Machine learning to predict antimicrobial usage, resistance and HAI at a Trust level  

We are testing a range of machine learning methods as potential tools to predict the future use of different antimicrobials, individually, by class and using WHO Access/Reserve/Watch groups, resistance and HAI, from historical use, resistance, and HAI at a Trust level in order to assess the impact of current use on subsequent AMR&HAI and ideally test counterfactual scenarios around different future strategies.  


Characterisation of “normal” response to (effectively) treated sepsis using physiological measurements to avoid antibiotic escalation and encourage early antibiotic stopping    

This project aims to characterise “normal” response to culture-positive and culture-negative sepsis in patients with blood cultures receiving (effective) antibiotics using physiological responses (e.g., temperature, heart rate, respiratory rate), laboratory test results (e.g., C-reactive protein). Growth curve models have been used to explicitly estimate percentiles of expected responses day by day, which will then be compared with duration of antibiotic treatment, and escalation/switching to estimate response percentiles at which cessation and switching/escalation generally occur. We will then investigate algorithms that could be used to help stop or escalate antibiotics appropriately.