Background: Valuation of the economic cost of antimicrobial resistance (AMR) is important for decision making
and should be estimated accurately. Highly variable or erroneous estimates may alarm policy makers and hospital
administrators to act, but they also create confusion as to what the most reliable estimates are and how these
should be assessed. This study aimed to assess the quality of methods used in studies that quantify the costs of
AMR and to determine the best available evidence of the incremental cost of these infections.

Methods: In this systematic review, we searched PubMed, Embase, Cinahl, Cochrane databases and grey literature
sources published between January 2012 and October 2016. Articles reporting the additional burden of
Enterococcus spp., Escherichia coli (E. coli), Klebsiella pneumoniae (K. pneumoniae), Pseudomonas aeruginosa (P.
aeruginosa) and Staphylococcus aureus (S. aureus) resistant versus susceptible infections were sourced. The included
studies were broadly classified as reporting oncosts from the healthcare/hospital/hospital charges perspective or
societal perspective. Risk of bias was assessed based on three methodological components: (1) adjustment for
length of stay prior to infection onset and consideration of time-dependent bias, (2) adjustment for comorbidities
or severity of disease, and (3) adjustment for inappropriate antibiotic therapy.

Results: Of 1094 identified studies, we identified 12 peer-reviewed articles and two reports that quantified the
economic burden of clinically important resistant infections. Two studies used multi-state modelling to account for
the timing of infection minimising the risk of time dependent bias and these were considered to generate the best
available cost estimates. Studies report an additional CHF 9473 per extended-spectrum beta-lactamases -resistant
Enterobacteriaceae bloodstream infections (BSI); additional €3200 per third-generation cephalosporin resistant
Enterobacteriaceae BSI; and additional €1600 per methicillin-resistant S. aureus (MRSA) BSI. The remaining studies
either partially adjusted or did not consider the timing of infection in their analysis.

Conclusions: Implementation of AMR policy and decision-making should be guided only by reliable, unbiased
estimates of effect size. Generating these estimates requires a thorough understanding of important biases and
their impact on measured outcomes. This will ensure that researchers, clinicians, and other key decision makers
concerned with increasing public health threat of AMR are accurately guided by the best available evidence.