Characterizing antimicrobial selection pressure on resistance across four major causes of infections: Insights from the US Veterans Affairs Healthcare System, 2007-2021.

Background There is conflicting evidence of the role of antibiotic in driving antimicrobial resistance. Existing studies asses use-resistance relationships typically only for few pathogens, for narrow time window, or only in one geographic space, and/or for few pathogen-antibiotic combinations and often only one pathogen and antibiotic (class) at a time. No systematic evidence of whether some antibiotics are more efficient at selecting for resistance than others. Here, we tested the hypothesis that each antibiotic class exerts selection for the antimicrobial susceptibility patterns resistant to it, using data on four major pathogens in the US Veterans Affairs Healthcare System. Methods We analysed clinical microbiology data from electronic health records from patients admitted to 138 Veterans Affairs Medical Centers with acute care wards across the USA from Feb 1, 2007, to Dec 31, 2021. We used clinical isolates classified as hospital-onset Staphylococcus aureus, Escherichia coli, Klebsiella pneumoniae, and Pseudomonas aeruginosa. We quantified inpatient antibiotic prescribing as days of therapy (DOT) per 1000 patient-days and the antibiogram incidence as the number of incident isolates with that antibiogram per 1,000 admissions. We performed time trend analyses for antibiogram incidence and antibiotic prescribing using generalized estimating equations and reported average annual percentage changes (AAPC). We performed hierarchical multinomial logistic regressions to estimate the effect of inpatient facility-level prescribing on antimicrobial susceptibility patterns. We included calendar time, facility characteristics, and community prevalence of a respective antibiogram as covariates to account for ecological confounding. Results From 2007 to 2021, the incidence of hospital-onset isolates declined for all four pathogens except E. coli isolates resistant to third-generation cephalosporins (3GC) and co-resistant to either fluoroquinolones or beta-lactam/beta-lactamase inhibitors (BL/BLI), or both. Over the same period, antibiotic prescribing generally remained stable or decreased, with the notable exception of 3GC prescribing, which increased from 2007 to 2019 (AAPC: 2.4%, 95% CI: 1.3%-3.5%). In general, higher facility-level use of a given antibiotic was associated with increased resistance to that antibiotic, although exceptions were observed. Fluoroquinolones consistently selected for resistance across the four pathogens. In S aureus, each additional 14-day course of fluoroquinolones (per 1,000 patient-days) was linked to a 5.9% (95% CI: 0.8%-11.2%) increase in the odds of isolating a fluoroquinolone-resistant, macrolide-susceptible, methicillin-resistant phenotype. Anti-staphylococcal beta-lactams were not associated with methicillin-resistant S aureus. In Enterobacterales, each additional 14-day course of 3GC treatment increased the odds of isolating 3GC- and BL/BLI-resistant E coli by 7.4% (95% CI: 1.8%-13.4%) and K pneumoniae by 4.1% (95% CI: 0.0%-8.7%). For P aeruginosa, antipseudomonal carbapenem use selected for carbapenem-resistant isolates; the largest effect was observed in carbapenem-resistant phenotypes susceptible to fluoroquinolones and 3GC, where each additional 14-day course increased the odds by 22.6% (95% CI: 13.4%-32.7%). Interpretation Our findings demonstrate that fluoroquinolones play a key role in selecting for resistance across multiple pathogens, underscoring their significance as a driver of AMR. Although these results highlight the importance of cautious fluoroquinolone use and targeted stewardship strategies, we also observed that reductions in fluoroquinolone prescribing were offset by increased 3GC prescribing and co-resistance in E coli. Consequently, effective AMR mitigation requires comprehensive stewardship approaches that address multiple antibiotic classes in tandem rather than in isolation.