Association between carbapenem resistance in invasive Klebsiella pneumoniae and antibacterial consumption in Türkiye: an ecological analysis using national surveillance data

Abstract

Antimicrobial resistance (AMR) is driven not only by antimicrobial consumption but also by healthcare-associated transmission, the spread of resistant clones, and infection prevention and control practices. In this ecological study, we jointly evaluated consumption of antibacterials for systemic use (ATC group J01) in Türkiye from 2014 to 2023 and carbapenem resistance in invasive Klebsiella pneumoniae isolates from 2016 to 2023, with Acinetobacter spp., Pseudomonas aeruginosa, and Escherichia coli described as secondary context. Consumption data were obtained from the WHO Regional Office for Europe Antimicrobial Medicines Consumption (AMC) Network, and resistance data from the Central Asian and European Surveillance of Antimicrobial Resistance (CAESAR) network and CAESAR-linked national publications, using harmonized definitions. Given the small number of annual observations, trends were assessed using Theil–Sen median slopes with exact Kendall tests, and consumption–resistance associations using exact permutation-based Spearman tests. A potential 2020 breakpoint (onset of the COVID-19 pandemic) was explored for the consumption series only, using segmented regression with HC3 heteroskedasticity-consistent standard errors. Consumption was higher in 2023 than in 2014 (42.7 vs. 34.7 defined daily doses per 1,000 inhabitants per day [DID]), but no monotonic trend was detected over the full period (Theil–Sen slope −0.10 DID/year; 95% CI −1.42 to 1.10; exact Kendall p = 1.000). Carbapenem resistance in invasive K. pneumoniae rose from 29.5% to 55.0% (Theil–Sen slope +3.83 percentage points/year; 95% CI 2.90 to 4.85; exact Kendall p < 0.001). No statistically significant association was found between J01 consumption and K. pneumoniae carbapenem resistance in concurrent (ρ = 0.262; exact p = 0.536), one-year lagged, or first-difference analyses. In the segmented model, the immediate 2020 level change was not significant (−6.47 DID; HC3 p = 0.184), whereas the post-2020 slope change was significant (+6.54 DID/year; p = 0.010). These findings are consistent with a resistance epidemiology that aggregate J01 consumption alone does not explain, although the short series limits statistical power. Class-specific, sector-specific, and One Health-linked surveillance is needed.