Antimicrobial selection and multidrug resistance in four major pathogens in the US Veterans Affairs health-care system: a multicentre, retrospective, cohort study

Résumé fourni par la source

BACKGROUND: Systematic evidence indicating the effect of antimicrobial selection pressure on resistance remains scarce. We aimed to estimate the effect of prescription behaviour of key antimicrobial classes on antimicrobial resistance (AMR) in patients across US Veterans Affairs medical centres (VAMCs). METHODS: In this multicentre, retrospective, cohort study, we identified all hospital-onset isolates (cultures that were collected on or after 3 days of admission) of Staphylococcus aureus, Escherichia coli, Klebsiella pneumoniae, and Pseudomonas aeruginosa from 138 US VAMCs between Feb 1, 2007, and Dec 31, 2021, using routine antimicrobial susceptibility and prescription data from Veterans Affairs administrative records. Facility-year periods with more than 30% missing susceptibility testing results were excluded. The study population comprised individuals admitted to acute care wards, and analyses were restricted to inpatient settings. The primary outcomes were the associations between facility-level prescription and resistance data and the temporal trends in pathogen-specific AMR profiles and prescription. AMR profiles were defined as pathogen-specific susceptibility patterns to clinically relevant antimicrobial panels. AMR profile incidence was quantified as the number of isolates per AMR profile category (susceptible, single-class resistant, or multidrug resistant) per 1000 admissions; facility-level prescription rate was defined as inpatient days of therapy per 1000 patient-days. The temporal trends were modelled using generalised estimating equations to estimate average annual percentage changes (AAPC). Multilevel multinomial logistic regression-related recent facility-level antimicrobial exposure (days of therapy per 100 patient-days in the preceding 14 days) to the relative odds of each AMR profile compared with the broadly susceptible profile. We adjusted models for calendar year, community-level AMR incidence, and facility-level characteristics (including census region, facility complexity, and facility rurality) to address potential confounding. To assess robustness, we performed sensitivity analyses by varying the exposure window, incorporating alternative model specifications, and using multiple imputation for missing antimicrobial susceptibility data under different assumptions. FINDINGS: We analysed 84 657 (51·6%) of 164 047 hospital-onset isolates across four pathogens-namely, S aureus (n=17 722), E coli (n=25 777), K pneumoniae (n=22 908), and P aeruginosa (n=19 299). Among 68 910 participants, 96·4% (66 430) were male and 69·8% (48 119) were classified as White, 23·4% (16 092) as Black or African American, and 8·7% (7414) as Hispanic or Latino. The median age was 70 years (IQR 62-79). Hospital-onset incidence declined for infections by all pathogens, except for those caused by third-generation or fourth-generation cephalosporin-resistant E coli. Antimicrobial prescription rates remained stable or decreased, except for third-generation or fourth-generation cephalosporin prescription rate, which increased during 2007-19 (AAPC 3·1% [95% CI 2·1 to -4·1]; p<0·0001). Fluoroquinolone exposure was consistently associated with resistance development in all pathogens. In S aureus, each unit increase in fluoroquinolone exposure was associated with 4·6% (1·5-7·7; p=0·0033) of increased relative odds of fluoroquinolone-resistant, macrolide-susceptible, meticillin-resistant S aureus. Anti-staphylococcal β-lactam prescription behaviour was not linked to the prevalence of meticillin-resistant S aureus. Third-generation or fourth-generation cephalosporin exposure increased the relative odds of third-generation or fourth-generation cephalosporin-resistant and β-lactam-β-lactamase-resistant E coli by 4·8% (2·1-7·5; p=0·00040) and K pneumoniae by 4·0% (1·9-6·2; p=0·00020. Carbapenem exposure increased the relative odds of carbapenem-resistant, fluoroquinolone-susceptible, and β-lactam-β-lactamase inhibitor-susceptible P aeruginosa by 15·7% (9·4-22·4%; p<0·0001). Community-level AMR prevalence was associated with increased relative odds of corresponding hospital-onset resistance profiles across most pathogen-profile combinations. Sensitivity analyses using alternative exposure windows, model specifications, and multiple imputation for missing susceptibility data yielded consistent results. INTERPRETATION: Consistent associations between fluoroquinolone prescription and resistance patterns across multiple pathogens reinforce stewardship rationale for restricting fluoroquinolone use, whereas emerging co-resistance of fluoroquinolones with that of third-generation or fourth-generation cephalosporins in E coli highlights the need to monitor multidrug resistance patterns as prescription behaviours shift. Effective stewardship requires coordinated strategies across antimicrobial classes, informed by pathogen-specific selection dynamics. By analysing AMR profiles across multiple pathogens and linking individual-level clinical isolates with facility-level antimicrobial prescription patterns, this study moves beyond single-pathogen, single-antimicrobial approaches to assess how antimicrobial use shapes resistance patterns. FUNDING: US Centers for Disease Control and Prevention, the Agency for Healthcare Research and Quality, and the Department of Veterans Affairs.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.