Validating a prediction tool to determine the risk of nosocomial multidrug-resistant Gram-negative bacilli infection in critically ill patients: A retrospective case–control study. (September 2020)

• Record Type: Journal Article
• Authors: Boyd, Sara E.; Vasudevan, Anupama; Moore, Luke S.P.; Brewer, Christopher; Gilchrist, Mark; Costelloe, Ceire; Gordon, Anthony C.; Holmes, Alison H.
• Is Part Of: Journal of global antimicrobial resistance. Volume 22(2020)
• Page Start: 826
• Page End: 831
• Publication Date: 2020-09
• Subjects: Nosocomial infection -- Gram-negative bacilli -- Antimicrobial resistance -- Intensive care unit -- Bedside prediction tool -- Critical care; Drug resistance -- Periodicals; Drug resistance -- Periodicals; Drug resistance; Periodicals; 616.9041
• DOI: 10.1016/j.jgar.2020.07.010
• Languages: English
• ISSNs: 2213-7165

Abstract

Highlights: Validation of GSDCS score for prediction of nosocomial RGNB infection. Geotemporally matched RGNB and SGNB patients were included. AUC was 0.75 in the validation cohort comparable with 0.77 in the derivation cohort. Efficient bedside prediction tool to identify risk of RGNB in ICU patients. Effective tool to differentiate between low, medium and high risk of RGNB. Abstract: Background: The Singapore GSDCS score was developed to enable clinicians predict the risk of nosocomial multidrug-resistant Gram-negative bacilli (RGNB) infection in critically ill patients. We aimed to validate this score in a UK setting. Method: A retrospective case–control study was conducted including patients who stayed for more than 24 h in intensive care units (ICUs) across two tertiary National Health Service hospitals in London, UK (April 2011–April 2016). Cases with RGNB and controls with sensitive Gram-negative bacilli (SGNB) infection were identified. Results: The derived GSDCS score was calculated from when there was a step change in antimicrobial therapy in response to clinical suspicion of infection as follows: prior Gram-negative organism, Surgery, Dialysis with end-stage renal disease, prior Carbapenem use and intensive care Stay of more than 5 days. A total of 110 patients with RGNB infection (cases) were matched 1:1 to 110 geotemporally chosen patients with SGNB infection (controls). The discriminatory ability of the prediction tool by receiver operating characteristic curve analysis in our validation cohort was 0.75 (95% confidence interval 0.65–0.81), which is comparable with the area under the curve of the derivation cohort (0.77). The GSDCS score differentiated between low- (0–1.3), medium- (1.4–2.3) and high-risk (2.4–4.3) patients for RGNB infection ( P < 0.001) in a UK setting. Conclusion: A simple bedside clinical prediction tool may be used to identify and differentiate patients at low, medium and high risk of RGNB infection prior to initiation of prompt empirical antimicrobial therapy in the intensive care setting.