Nomenclature, definitions, and methodological approaches to estimate the association between antimicrobial treatment and clinical outcomes of drug-resistant bloodstream infections

Background Antimicrobial resistance increases the risk of misaligned initial antibiotic treatment (IAT), as susceptibility data are typically delayed. The causal effect on patient outcomes, however, is unclear due to reliance on observational studies with methodological heterogeneity. Objectives To describe the terminology and definitions for IAT misalignment and evaluate methods used to analyse its association with mortality and hospital length of stay (LOS) for patients with drug-resistant bloodstream infections (BSIs). Methods A systematic review. Data sources PubMed and EMBASE: January 1990 to August 2024. Study eligibility criteria We included studies on drug-resistant BSIs caused by ESKAPE pathogens (Enterococcus faecium, Staphylococcus aureus, Klebsiella pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa, Enterobacter species, and other Enterobacterales). Eligible studies defined IAT misalignment and assessed its effect on mortality/LOS. Participants Patients with drug-resistant BSIs. Exposure (mis)aligned IAT. Assessment of risk of bias Revised versions of the Joanna Briggs Institute tools. Methods of data synthesis Qualitative synthesis. Results From 3627 screened publications, 187 studies were included, predominantly cohort studies (n = 183). The most common terminology for IAT misalignment was “(in)appropriate” (n = 139, 74.3%), followed by “(in)adequate” (n = 34, 18.2%). Definitions primarily considered in vitro susceptibility to prescribed antibiotic(s) (n = 184, 98.4%), with up to nine additional criteria. Impact of (in)appropriate IAT on mortality (n = 186) was mostly evaluated using logistic or Cox regression, including various confounder selection methods, showing an association in 122 of 186 studies (65.6%). Admission-to-infection time and infection-to-treatment time were rarely considered. Impact of (in)appropriate IAT on LOS was shown in two of nine studies. Only four studies explicitly analysed postinfection LOS. No study scored a low risk of bias, due to residual confounding and time-dependent bias. Discussion Wide variability of IAT definitions and impact analysis was observed, with a high risk of bias, hindering data aggregation and limiting understanding of the causal effect of inappropriate IAT on clinical outcomes. Guidelines are required to improve study quality and harmonize future research.