Universal versus targeted chlorhexidine and mupirocin decolonisation and clinical and molecular epidemiology of Staphylococcus epidermidis bloodstream infections in patients in intensive care in Scotland, UK: a controlled time-series and longitudinal genotypic study

Abstract

Background

There are concerns that biocide skin and mucous membrane decolonisation, which is widely used to prevent health-care-associated infections in intensive care units (ICUs), might select for multidrug-resistant pathogens. We aimed to evaluate the effects of de-escalating from universal to targeted skin and nasal decolonisation on Staphylococcus epidermidis bloodstream infections (SE-BSI).

Methods

We did a retrospective, before-after-control-impact time-series analysis and longitudinal genotypic study in two ICUs with divergent decolonisation practice in tertiary care hospitals of adjacent health boards in Scotland, UK. Participants were aged at least 16 years and admitted between July 1, 2009, and Feb 28, 2022. There were no exclusion criteria for the study. In ICU one (intervention site) universal decolonisation in all admissions was de-escalated to targeted decolonisation of meticillin-resistant Staphylococcus aureus (MRSA) carriers on Feb 1, 2019, while in ICU two (control site) targeted decolonisation was applied throughout. We collected bloodstream infection data from all causes, including clinically significant SE-BSI. Antimicrobial susceptibility testing was used to define meticillin-resistant S epidermidis (MRSE) and chlorhexidine susceptibility. We used multilocus sequence typing to identify sequence types from archived SE-BSI isolates. Whole-genome sequencing was applied to a sample from ICU one. The primary outcomes were incidence densities of all bloodstream infections, SE-BSI, and meticillin-resistant S epidermidis bloodstream infections (MRSE-BSI), and the percentage probability that SE-BSI were MRSE-BSI. The effects of de-escalation on primary outcomes were estimated by differences between the intervention and control sites, before and after de-escalation, using a before-after-control-impact time-series design. Secondary outcomes included the proportion of multidrug resistant sequence types, carriage of mobile genetic elements and genes for multidrug resistance and biofilm production.

Findings

Between July 1, 2009, and Feb 28, 2022, S epidermidis was identified in 334 (45%) of 735 bloodstream infections in ICU one, of which 197 occurred before the de-escalation intervention in Feb 1, 2019, and S epidermidis was identified in 167 (60%) of 278 bloodstream infections in ICU two. There was no increase in all bloodstream infection incidence coinciding with de-escalation in ICU one, whereas MRSE-BSI incidence declined significantly from 10·4 cases per 1000 occupied bed days (OBDs; 95% credible interval [CrI] 7·2–15·4) to 4·3 cases per 1000 OBDs (2·5–6·7), as did the percentage probability of MRSE (from 89·2%, 95% CrI 77·8–96·5 to 56·7%, 34·3–77·5%). No significant changes in the primary outcomes were seen in ICU two. MRSE-BSI incidence density was positively associated with chlorhexidine use, but not mupirocin use. De-escalation was associated with a reduced proportion of SE-BSI due to multidrug-resistant sequence types and reduced carriage of mobile genetic elements and genes for multidrug resistance and biofilm production, as observed by multi-locus sequence typing and whole genome sequencing.

Interpretation

In ICU settings with low MRSA incidence, the benefits of universal decolonisation should be balanced against the risks of selecting MRSE sequence types adapted for invasive and device-associated infection.

Funding

National Health Service Grampian Charity.