Antimicrobial resistance is often reported as a single, ominous number: the share of hospital bacteria that no longer respond to first-line drugs. But a new five-year surveillance study from a tertiary general hospital in Luoyang, China, argues that this aggregate figure can mask as much as it reveals. By tracking more than ten thousand clinical bacterial isolates between 2021 and 2025, researchers found that nearly half qualified as multidrug-resistant organisms, yet the trajectories of the individual resistant species beneath that headline number diverged dramatically — one climbing steeply, another surging and then bending downward, and others holding frustratingly steady. The work, published in BMC Infectious Diseases, offers a methodological blueprint for how hospitals might read their own resistance data with far greater precision.

The research team, led by Jin Yao of the Department of Infection Prevention and Public Health Management at The Second Affiliated Hospital of Henan University of Science and Technology, conducted a retrospective, laboratory-based surveillance study covering all inpatient bacterial isolates collected over five calendar years. To avoid inflating counts with repeated cultures from the same patient, the investigators applied a 30-day first-isolate rule: only the first detection of a given organism from a patient within any 30-day window was included in the analysis. This deduplication step, a standard but crucial practice in surveillance epidemiology, ensures that proportions reflect genuine shifts in the bacterial population rather than artifacts of repeated sampling of chronically colonized patients.

The scale of the resistance burden was substantial. Among 10,708 non-duplicate clinical isolates, 4,850 — or 45.29 percent — met the aggregate definition of a multidrug-resistant organism, which included isolates satisfying general multidrug-resistance criteria as well as nationally prioritized resistance phenotype definitions. Within that aggregate, 1,265 isolates belonged to the five priority resistance phenotypes that anchor global and national surveillance programs: methicillin-resistant Staphylococcus aureus (MRSA), carbapenem-resistant Klebsiella pneumoniae (CRKP), carbapenem-resistant Pseudomonas aeruginosa (CRPA), carbapenem-resistant Acinetobacter baumannii (CRAB), and carbapenem-resistant Escherichia coli (CREC). Carbapenems are often described as last-line antibiotics, reserved for infections that defy other treatments, which is why resistance to them carries particular weight in hospital settings.

Here the study’s central methodological point comes into focus. The aggregate MDRO proportion was calculated using all included isolates as the denominator, but the species-specific resistance proportions used species-specific denominators — the number of isolates of that particular bacterium. This distinction matters because the two measures answer different questions. The aggregate figure describes what fraction of the total bacterial burden is multidrug-resistant; the species-specific figure describes, for example, what fraction of all S. aureus isolates are methicillin-resistant. A hospital could see its aggregate MDRO proportion fall simply because susceptible species became more common, even while resistance within a dangerous species quietly worsened. Tracking both, the authors argue, provides complementary surveillance information that neither measure alone can supply.

The species-specific results painted a strikingly varied picture. Carbapenem-resistant Acinetobacter baumannii was the most alarming, with 53.67 percent of all A. baumannii isolates resistant to carbapenems — meaning that, at this hospital, a majority of infections caused by this opportunistic pathogen had already exhausted one of the most powerful drug classes available. MRSA followed at 38.35 percent, a reminder that gram-positive resistance remains a persistent fixture decades after methicillin resistance first emerged. Carbapenem-resistant Pseudomonas aeruginosa stood at 21.03 percent, carbapenem-resistant Klebsiella pneumoniae at 13.70 percent, and carbapenem-resistant Escherichia coli at just 0.75 percent — a reassuringly low figure for a phenotype that has caused crises elsewhere in the world.

The temporal analysis was where the study broke the most methodological ground. Rather than fitting a single straight line through five years of quarterly data, the researchers used data-driven single-breakpoint segmented Poisson regression, a statistical technique that searches the time series for a genuine change point and estimates separate trends before and after it. Poisson regression is well suited to count data such as monthly or quarterly isolate numbers, and the segmented approach allows the data themselves — guided by model-fit criteria such as the Akaike information criterion — to reveal whether a single trend line is an honest summary or a distortion. The team also applied Poisson generalized linear models to monthly data to detect both overall monthly trends and calendar-month seasonality.

The results validated the effort. Clear statistical support for an internal breakpoint emerged for only one phenotype: CRKP, after the first quarter of 2023. Before that turning point, carbapenem resistance in K. pneumoniae was escalating at a breathtaking pace — an increase of 21.21 percent per quarter (95 percent confidence interval, 8.22 to 35.76 percent; P less than 0.001). After the breakpoint, the trend reversed, declining by 4.48 percent per quarter, though this downward movement fell just short of conventional statistical significance (95 percent confidence interval, −8.85 to 0.10 percent; P = 0.055). Had the analysts fit a single linear trend across the full period, this dramatic rise-and-partial-retreat pattern would have been smeared into a misleadingly modest average slope, potentially obscuring both the crisis that built through 2023 and the subsequent intervention-era improvement.

CRPA told a different story altogether. Rather than a breakpoint, carbapenem-resistant P. aeruginosa showed a steady, unbroken upward climb of 2.04 percent per month (95 percent confidence interval, 1.05 to 3.04 percent; P less than 0.001) — a relentless trajectory that compounded over the study period into a substantial overall increase. Compounding that concern, CRPA also displayed significant calendar-month variation (P = 0.003), meaning resistance detections rose and fell with the seasons in a statistically detectable rhythm. Seasonality in resistant infections can reflect seasonal shifts in antibiotic prescribing, hospital admission patterns, environmental reservoirs, or the ecology of the pathogen itself, and identifying such rhythms can help infection-control teams anticipate high-risk periods. By contrast, MRSA, CRKP, and CRAB showed no significant monthly trends, underscoring how differently these organisms behave even within the same hospital walls.

The aggregate picture, meanwhile, offered a cautiously encouraging frame. Aggregate MDRO proportions held at approximately 48 to 50 percent during 2021 through 2023 and were lower in 2024 and 2025. Yet the species-level findings make clear that this aggregate improvement was not uniform: the post-2023 decline in CRKP contributed to it, while CRPA’s steady monthly climb continued unabated. A hospital administrator reading only the aggregate number might have concluded that the resistance problem was broadly improving; the segmented, species-specific analysis reveals instead a mixed landscape in which one crisis receded while another quietly deepened. This is precisely the interpretive gap the authors set out to close.

The implications extend well beyond a single institution in Henan Province. Surveillance frameworks such as the World Health Organization’s Global Antimicrobial Resistance and Use Surveillance System and China’s own national antimicrobial resistance surveillance system depend on hospitals reporting resistance data that policymakers can interpret correctly. This study demonstrates that the choice of denominator, the temporal scale of analysis, and the statistical model all shape what the data appear to say. Integrating species-specific denominators with analyses at multiple temporal scales — annual, quarterly, and monthly, with formal breakpoint detection — the authors conclude, may improve surveillance interpretation and better inform both infection prevention programs and antimicrobial stewardship efforts. As carbapenem-resistant organisms continue to rank among the most feared threats in modern medicine, the lesson from Luoyang is that the shape of a resistance epidemic matters as much as its size, and that only fine-grained, statistically honest analysis can reveal that shape in time to act.

Subject of Research: Five-year laboratory surveillance of multidrug-resistant organism proportions and species-specific antimicrobial resistance trends in a tertiary hospital

Article Title: Aggregate MDRO proportion and species-specific resistance trends in a tertiary general hospital: a five-year laboratory-based surveillance study of priority resistance phenotypes

Article References: Yao, J., Wang, P., Wang, X., Guo, Z., & Huang, X. (2026). Aggregate MDRO proportion and species-specific resistance trends in a tertiary general hospital: a five-year laboratory-based surveillance study of priority resistance phenotypes. BMC Infectious Diseases. https://doi.org/10.1186/s12879-026-14602-x

Image Credits: AI Generated

DOI: 10.1186/s12879-026-14602-x

Keywords: antimicrobial resistance, multidrug-resistant organisms, carbapenem-resistant Klebsiella pneumoniae, carbapenem-resistant Pseudomonas aeruginosa, carbapenem-resistant Acinetobacter baumannii, MRSA, hospital surveillance, segmented Poisson regression, temporal trends, infection prevention, antimicrobial stewardship, clinical microbiology