The study team, led by Jianhong Yu and colleagues in the hospital’s Department of Clinical Laboratory, began with 18,581 laboratory records drawn from twelve quarterly microbiology datasets. Applying the Clinical and Laboratory Standards Institute’s recommended rule of retaining only the first isolate of each bacterial species per patient per calendar year, they removed 1,615 repeat records, leaving 16,966 isolate-level observations from 15,095 unique patients. Yearly volumes rose steadily, from 5,081 isolates in 2023 to 5,536 in 2024 and 6,349 in 2025. Gram-negative organisms dominated the cohort at 75.9 percent, with Gram-positive bacteria accounting for the remaining 24.1 percent. Respiratory specimens were the leading source at 43.5 percent of isolates, followed by urine at 21.3 percent, abdominal fluid at 13.1 percent, pus and secretions at 12.4 percent, and blood at 7.8 percent.

The demographic and care-setting structure of the cohort shaped everything that followed. General wards contributed 91.2 percent of isolates, intensive care units 7.8 percent, and outpatient departments a mere 0.9 percent. By age, 24.2 percent of isolates came from patients under 18 years, 31.3 percent from adults aged 18 to 64, and 44.5 percent from patients 65 and older. Escherichia coli was the most frequently recovered organism at 28.6 percent of the total, followed by Klebsiella pneumoniae at 10.4 percent, Pseudomonas aeruginosa at 9.8 percent, Haemophilus influenzae at 9.2 percent, and Staphylococcus aureus at 8.5 percent. Pediatric isolates were markedly enriched for H. influenzae, Moraxella catarrhalis, and Streptococcus pneumoniae, whereas P. aeruginosa, K. pneumoniae, Acinetobacter baumannii, and enterococci were more prominent among older-adult and intensive-care isolates, a distribution that reflects both the sampling frame and the underlying ecology of infection across age groups.

Methodologically, the study took deliberate steps to avoid artifacts that plague many local surveillance reports. All available minimum inhibitory concentrations and inhibition-zone diameters were reinterpreted against a single breakpoint framework, the CLSI M100 35th edition from 2025, eliminating bias from changing interpretive criteria across study years. Bacterial identification relied on the VITEK 2 Compact automated system and VITEK MS MALDI-TOF mass spectrometry, with susceptibility testing performed by VITEK 2 cards, Kirby-Bauer disk diffusion, gradient diffusion, and the ATB system for fastidious organisms. Resistance proportions used antimicrobial-specific tested denominators, so isolates lacking an interpretable result for a given drug were excluded from that drug’s denominator rather than silently counted as susceptible. Temporal associations were estimated with binomial logistic regression, modeling calendar year ordinally and using patient-clustered robust standard errors to account for the fact that the same patient could contribute multiple species or observations across years. Adjusted models included calendar year, age, sex, and care setting, and exploratory organism-drug comparisons were controlled with the Benjamini-Hochberg false discovery rate.

The headline resistance findings split cleanly along two axes. Carbapenem resistance in E. coli remained uncommon throughout, affecting 70 of 4,848 isolates, or 1.4 percent, and showed no significant temporal trend, moving only from 1.3 percent in 2023 to 1.7 percent in 2025. K. pneumoniae carbapenem resistance was similarly low at 63 of 1,769 isolates, or 3.6 percent, fluctuating between 4.4, 2.7, and 3.7 percent across the three years without a monotonic direction. These figures sit well below the CHINET 2025 national surveillance report, which recorded imipenem and meropenem resistance of 2.4 and 2.6 percent for E. coli but a striking 21.9 and 21.0 percent for K. pneumoniae, underscoring how sharply local epidemiology can diverge from national aggregates.

Among the non-fermenting Gram-negative bacteria, the picture was far grimmer and tightly linked to care intensity. Carbapenem-resistant P. aeruginosa accounted for 190 of 1,659 isolates, or 11.5 percent overall, but reached 27.7 percent among ICU isolates compared with 9.9 percent in general wards. Carbapenem-resistant A. baumannii was even more concentrated, at 128 of 380 isolates, or 33.7 percent overall, rising to 68.8 percent in intensive care versus 21.8 percent on general wards. In multivariable models adjusted for age, sex, and calendar year, ICU or critical-care status remained independently associated with both phenotypes, with adjusted odds ratios of 2.80 for CRPA and 4.29 for CRAB, both highly statistically significant. Notably, the unadjusted upward trend in CRPA, which rose from 8.7 percent in 2023 to 13.5 percent in 2025, was attenuated and no longer significant after case-mix adjustment, suggesting that shifting patient composition rather than a pure secular drift may have driven much of the apparent increase. CRAB, by contrast, declined numerically from 38.6 to 29.4 percent without reaching statistical significance.

Two Gram-positive findings deserve particular attention. Methicillin-resistant S. aureus accounted for roughly a quarter of S. aureus isolates, at 24.1 percent in general wards and 24.4 percent in intensive care, slightly below the 28.4 percent reported nationally by CHINET in 2025, and no S. aureus isolate showed resistance to vancomycin or linezolid. More unexpected was the phenotypic linezolid non-susceptibility in Enterococcus faecalis, observed in 60 of 325 isolates, or 18.5 percent, with annual rates holding steady between 17.3 and 19.0 percent and no temporal trend. That figure is substantially higher than the 5.9 percent linezolid resistance rate CHINET reported for E. faecalis in 2025. The authors are careful with the terminology: all enterococcal isolates initially flagged as non-susceptible to vancomycin or linezolid underwent gradient-diffusion retesting, but reference broth microdilution and molecular confirmation were not performed, so the finding is reported as phenotypic non-susceptibility rather than confirmed resistance. Genome-level research has shown that enterococcal linezolid resistance typically arises through 23S ribosomal RNA mutations or transferable genes such as optrA, poxtA, and cfr, with optrA being the most common element in E. faecalis strains of Chinese origin that have spread globally, but confirming any such mechanism here would require the molecular work the study did not perform.

The clearest adjusted temporal signal in the entire dataset came from adult H. influenzae. Ampicillin resistance among isolates from patients aged 18 and older climbed from 35 of 50 isolates, or 70.0 percent, in 2023 to 214 of 243 isolates, or 88.1 percent, in 2025, an absolute increase of 18.1 percentage points with an adjusted odds ratio per year of 1.73 that persisted after case-mix adjustment. Amoxicillin-clavulanate resistance rose even more dramatically, from 8 of 50 isolates, or 16.0 percent, to 145 of 243 isolates, or 59.7 percent, a 43.7 percentage-point jump with an adjusted odds ratio per year of 3.06. Beta-lactamase positivity increased in parallel, from 70.0 to 84.8 percent, and the descriptive MIC distributions shifted accordingly: the ampicillin MIC50 for adult isolates rose from 16 mg/L in 2023 to 64 mg/L in 2024 and 2025, while the amoxicillin-clavulanate MIC50 moved from 4 to 8 mg/L by 2025 with observed values reaching 64 mg/L. The authors note that the concurrent rise in beta-lactamase production, MICs, and amoxicillin-clavulanate resistance is compatible with additional non-beta-lactamase mechanisms such as altered penicillin-binding protein 3, the target of ftsI mutations that drive beta-lactamase-negative ampicillin-resistant and beta-lactamase-positive amoxicillin-clavulanate-resistant phenotypes, but without molecular testing this mechanistic explanation remains hypothetical. They also caution that the adult H. influenzae denominator grew substantially across the study years, so changes in culture ordering, testing utilization, or case mix could have contributed to the observed pattern.

For clinical practice, the authors frame their results as local susceptibility evidence rather than direct evidence of treatment effect. The low carbapenem resistance rates in E. coli and K. pneumoniae point to strong in vitro carbapenem activity against most isolates in this dataset but do not justify indiscriminate empirical carbapenem use. The concentration of CRAB and CRPA in intensive care argues for setting-specific cumulative antibiograms, review of prior cultures and antimicrobial exposure when available, prompt organism identification and susceptibility testing, and susceptibility-guided de-escalation rather than broad anti-CRAB or anti-CRPA coverage for all patients. When linezolid is being considered for enterococcal infection, unexpected non-susceptibility should trigger laboratory verification. The study’s limitations are candidly acknowledged: it was a single-center, retrospective analysis capturing 99.1 percent inpatient data, it lacked linked clinical details such as comorbidities, prior antibiotic exposure, and outcomes, it could not reliably distinguish true infection from colonization, particularly for ICU non-fermenters, and phenotypic results were not confirmed by reference methods. Because patient-level therapy and outcomes were not linked, the data cannot be used to judge stewardship effectiveness or treatment success. What the study does deliver is a template for how institutions outside major metropolitan centers can generate actionable, age- and setting-stratified resistance intelligence, and a reminder that in the fight against antimicrobial resistance, the most useful map is often the one drawn at the level of a single hospital’s own walls.

Subject of Research: Antimicrobial resistance epidemiology and surveillance of clinical bacterial isolates at a tertiary hospital in southwestern China

Article Title: Epidemiological characteristics and antimicrobial resistance profiles of clinical bacterial isolates in a tertiary hospital in Zigong, southwestern China, 2023–2025

Article References: Yu, J., Chen, Y., You, X., & Zhang, S. (2026). Epidemiological characteristics and antimicrobial resistance profiles of clinical bacterial isolates in a tertiary hospital in Zigong, southwestern China, 2023–2025. Frontiers in Cellular and Infection Microbiology, 16, Article 1947772. https://doi.org/10.3389/fcimb.2026.1947772

Image Credits: AI Generated

DOI: 10.3389/fcimb.2026.1947772

Keywords: antimicrobial resistance, carbapenem-resistant Enterobacterales, carbapenem-resistant Acinetobacter baumannii, carbapenem-resistant Pseudomonas aeruginosa, Haemophilus influenzae, linezolid non-susceptibility, Enterococcus faecalis, antimicrobial surveillance, intensive care units, CLSI breakpoints, tertiary hospital, China