Project Description

Study Title: The Burden of Antimicrobial Resistance and Antibiotic Treatment Failure in Low-, Lower-Middle, Upper-middle and High-Income Countries

Chief Investigator:Timothy R. Walsh

Sponsor: University of Oxford

The study will perform a prospective cohort study to evaluate the impact of BSIs caused by Enterobacteriaceae, Pseudomonas aeruginosa, Acinetobacter species, Staphylococcus aureus and Enterococcus species (EPASE-R) (see ‘Exposures’ below) on clinical outcomes. All patients with suspected cases of BSIs admitted in six selected wards such as the intensive care unit (ICU) (including paediatric ICU), high dependency unit (HDU) (including paediatric HDU), general surgery, general medicine, paediatric surgery and paediatric medicine with transit through emergency/admission department will be initially included. Participants with confirmed EPASE BSIs at study sites will be eligible for detailed data collection.

The study  will analyse patients with BSI present on admission (community-onset BSI [CO-BSI]) separately from those with hospital-onset BSI (HO-BSI) because these groups come from different ‘at-risk’ populations. 

  • CO-BSI (detected ≤48 hours after admission) are time-fixed and the ‘study base’ (i.e. at-risk population) are members of the community. The entire admission is considered attributable to the infection. 
  • HO-BSI (detected >48 hours after admission) are time-varying and the ‘study base’ (i.e. at-risk population) are inpatients at the same hospital. Only a fraction of the admission is considered attributable to the infection. 

The study  will consider the following two types of comparisons:  

EPASE-R BSI versus EPASE-S BSI: For all BSI (both CO-BSI and HO-BSI), the study will compare patients with EPASE-R BSI to patients with EPASE-S BSI (table 1). The study  will therefore measure the impact of replacing EPASE-R BSI with EPASE-S BSI. 

EPASE-R BSI versus uninfected and EPASE-S BSI versus uninfected: For HO-BSI, The study will compare patients with BSI to patients without BSI, stratified by susceptibilities of respective antibiotics. The stud will therefore measure the avoidable burden by reducing the number of EPASE-R BSI or EPASE-S BSI among hospitalised patients without replacement by other infections. For each patient with an HO-BSI that we include (i.e. exposed patients), The study will randomly select one patient who has been in the hospital for the same number of days (+/- 1 day) but who has not yet experienced a BSI (i.e. unexposed patients). This exposure density sampling approach is an efficient way to select a sub-cohort of the entire hospital population while still accounting for time dependency (figure 3).18 The study will identify the unexposed patient within 30 days of the corresponding exposed patient. In combination with data about the distribution of length of stay in each hospital, these controls will be upweighted to represent the full hospital population eligible as control on each day, thereby avoiding bias in the statistical models.