1 Introduction

Enterobacter species have emerged as important opportunistic pathogens in both community and hospital settings, where they are associated with urinary tract infections, wound infections, respiratory infections, and bloodstream infections (Davin-Regli et al., 2019). Their clinical importance has increased steadily over the past decade because of their ability to survive in diverse healthcare environments, colonize vulnerable patients, and rapidly acquire resistance to multiple classes of antimicrobial agents (Annavajhala et al., 2019). Among the clinically relevant species, Enterobacter cloacae and related members of the E. cloacae complex are especially concerning because they possess inducible chromosomal AmpC β-lactamases and can develop broader resistance during antibiotic exposure (Hu et al. 2022, Tamma et al., 2019, Barceló et al., 2024). Recent surveillance data from multiple regions confirm that E. cloacae complex accounts for 40–55% of all Enterobacter genus isolates, making it the clinically dominant member of this genus worldwide (Intra et al., 2023, Ganbold et al., 2023). The Enterobacter cloacae complex is also classified among the ESKAPE group of pathogens, which are frequently associated with multidrug-resistant nosocomial infections and are listed by the World Health Organization as priority pathogens requiring urgent attention (Oliveira et al., 2020, Antimicrobial Resistance Collaborators 2022). The growing burden of antimicrobial resistance in Enterobacter presents a major therapeutic challenge. Resistance to cephalosporins, β-lactam/β-lactamase inhibitor combinations, fluoroquinolones, aminoglycosides, and even carbapenems has been increasingly reported, narrowing treatment options and complicating empirical therapy (Ma et al., 2023, Dong et al., 2022). The emergence and global dissemination of carbapenem-resistant Enterobacter cloacae complex variants have been widely documented and pose particular concern, as carbapenems are considered last-resort agents for treating multidrug-resistant Gram-negative infections (Zheng et al., 2024, Khamis et al., 2022). This problem is particularly serious in tertiary care hospitals, where prolonged admissions, invasive procedures, broad-spectrum antibiotic use, and critically ill patients create strong selection pressure for multidrug-resistant organisms (Langford et al., 2023). The COVID-19 pandemic further exacerbated antimicrobial resistance by increasing broad-spectrum antibiotic usage and disrupting infection prevention practices, resulting in notable increases in hospital-onset resistant infections (Centers for Disease Control and Prevention 2022, Suleiman et al., 2023). Surveillance of local resistance patterns is therefore essential for guiding empirical treatment, supporting stewardship programmes, and identifying emerging resistance phenotypes before they become entrenched (Tamma et al., 2026).

Despite the clinical significance of Enterobacter, long-term institutional datasets remain limited, especially those integrating taxon distribution, specimen epidemiology, multidrug-resistant phenotypes, and temporal resistance trends (Chen et al., 2022). Predictive methods are increasingly studied in AMR surveillance (Kim et al., 2022, Sakagianni et al., 2023, Yuan et al., 2025), but the present dataset contains only nine annual observations; therefore, its 2026 component is presented as a simple exploratory weighted linear-regression extrapolation rather than as machine learning. In the Gulf region, Enterobacter resistance data remain sparse, with many studies focusing on other Enterobacterales or on carbapenemase characterization (Khamis et al., 2022, Dortet et al., 2012, Sonnevend et al., 2015, Thomsen et al., 2024). The present study aimed to characterize the epidemiology and antimicrobial resistance patterns of non-duplicate isolates recovered over nine years at a secondary care centre in northern Oman, evaluate MDRO trends and acquisition-source differences, and examine limited short-term statistical projections for selected resistance indicators.

2 Materials and methods

2.1 Study design and setting

This study was designed as a retrospective, laboratory-based surveillance analysis of non-duplicate isolates recovered over a nine-year period (2017–2025) from a hospital in Sohar, northern Oman. The surveillance dataset comprised 1,805 isolates from samples received from different clinical departments. The study addressed four related objectives: (i) describe annual taxon and specimen distributions; (ii) quantify overall and subgroup-specific antimicrobial resistance profiles; (iii) examine temporal, acquisition-source, and demographic patterns in MDRO phenotypes; and (iv) generate a limited exploratory 2026 weighted linear-regression extrapolation for selected resistance indicators. No validated predictive model was intended.

2.2 Data source and isolate selection

Microbiology records were reviewed to extract all available isolates reported within the historical Enterobacter surveillance stream during the study period. Only non-duplicate isolates were included, with one isolate per patient per analytical episode retained to minimize repeated-sampling bias. Available variables included sex, age, age-group category, department/clinical area, specimen type, acquisition source, year of isolation, taxon identification, susceptibility profile, and MDRO phenotype classification. Isolates were recovered from urine, pus/wound, blood, respiratory, soft tissue, catheter, body fluid, and other specimen categories. Age groups were 0–18, 19–40, 41–60, 61–80, and > 80 years. Age was unavailable for two isolates; age-stratified analyses therefore used n = 1,803.

2.3 Species identification and antimicrobial susceptibility testing

Species identification was performed using standard biochemical and/or automated microbiological methods, including VITEK-2, according to routine laboratory practice during each year. Historical records included labels that have since been reclassified: Enterobacter aerogenes is reported in this manuscript as Klebsiella aerogenes and Enterobacter sakazakii as Cronobacter sakazakii, with historical names retained where needed to link revised terminology to the source database. Some isolates remained at genus level when species-level identification was not reliably achieved. Antimicrobial susceptibility testing (AST) results were compiled from the final susceptible/intermediate/resistant interpretations available in the retrospective laboratory dataset. The dataset did not contain sufficiently complete raw MIC/inhibition-zone values or year-specific breakpoint metadata to reconstruct which CLSI/EUCAST version was applied to every historical result; therefore, results could not be retrospectively harmonized to a single breakpoint set (Clinical and Laboratory Standards Institute 2025, European Committee on Antimicrobial Susceptibility Testing 2026). Because testing was clinically directed, not all antibiotics were tested against every isolate and denominators vary by agent; temporal comparisons may therefore reflect both true epidemiological change and changes in testing practice or breakpoint application.

2.4 Definition of MDRO phenotypes and analytical framework

MDRO analysis used the single primary resistance-phenotype label available in the extracted record. For reporting, the six analytical categories (MDR, CRE, ESBL, laboratory-recorded AmpC phenotype, CPO, and Other) were treated as mutually exclusive; no isolate was counted in more than one category. This analytical exclusivity does not imply that resistance mechanisms are biologically mutually exclusive. MDR denoted isolates meeting the operational criterion of resistance to three or more antimicrobial classes without a separately coded specific phenotype, whereas CRE, ESBL, AmpC, and CPO categories reflected the primary phenotype recorded by the laboratory. The available retrospective dataset did not provide a uniform, verifiable year-specific record of the AmpC confirmation procedure (for example, cefoxitin screening, induction testing, disk approximation, or molecular confirmation). No molecular confirmation was performed as part of this retrospective analysis; consequently, AmpC findings are reported as a laboratory-recorded phenotype rather than as definitive mechanistic confirmation. Comparative resistance analyses were performed by hospital department and acquisition source.

2.5 Statistical analysis

Descriptive statistics summarized isolate counts, proportions, and distributions across taxa, specimen types, departments, and years. Resistance proportions were reported with Wilson 95% confidence intervals. Temporal resistance patterns for selected antibiotics were assessed using Spearman rank correlation, with significance defined as p < 0.05. Acquisition-source comparisons used Pearson chi-square tests. Age-stratified MDRO prevalence was analyzed using unadjusted odds ratios (ORs) with 95% confidence intervals, using 0–18 years as the reference category; Fisher’s exact test was used where appropriate. The ordered age-group Cochran-Armitage trend test yielded Z = 7.68, p < 0.001 (Z^2 = 59.02, 1 df), while the overall Pearson chi-square across age groups was chi-square = 59.57, df = 4, p < 0.001. The Cochran-Armitage calculation was verified in Python 3.11 using NumPy/SciPy. Multivariable adjustment was not possible because comorbidity, prior antimicrobial exposure, healthcare-contact intensity, department, and specimen covariates were not consistently available.

2.6 Exploratory 2026 weighted linear-regression projections

Weighted linear regression was applied to annual resistance rates from 2017 to 2025, with greater weight assigned to 2022–2025. With only nine annual observations and no out-of-sample validation, this analysis was treated as a simple exploratory time-series extrapolation, not as machine learning and not as a clinical prediction tool. To reduce instability, reporting was limited to the six key antibiotic series used in the temporal-trend analysis. Sparse or non-calculable series, including tigecycline in 2025 (n = 0 tested), and species/MDRO count projections were not reported.

3 Results

A total of 1,805 non-duplicate isolates were recovered over the nine-year surveillance period (2017–2025), spanning ten hospital departments. Statistical summaries included Wilson 95% confidence intervals, Spearman temporal correlations, chi-square comparisons, unadjusted age-stratified odds ratios, a Cochran-Armitage trend test, and limited weighted linear-regression extrapolation for selected antibiotic resistance rates. Age analyses used n = 1,803 because two isolates lacked age data.

3.1 Baseline demographic and clinical characteristics

The study cohort showed near-equal sex distribution and a wide age range, with the greatest proportion in the 19–40 years group (Table 1). The mean age was 44.3 ± 24.6 years (median 45, IQR 25–64). Overall, 27.6% of isolates harboured multidrug-resistant organism (MDRO) phenotypes.

Table 1 Baseline demographic and clinical characteristics of patients and historically reported Enterobacter-group isolates (2017–2025)
Full size table

The study cohort represents a broad patient population across community and hospital settings. Urine and pus/wound specimens together accounted for 54.0% of isolates. Overall MDRO prevalence was 27.6%, indicating a substantial resistance burden during the nine-year surveillance period.

3.2 Overall distribution of specimen sources

Nineteen distinct specimen types were identified (Table 2). Urine, pus/wound, and blood collectively accounted for 69.4% of all isolates, while respiratory and soft tissue specimens contributed a further 16.0%.

Table 2 Overall distribution of specimen sources in the historical Enterobacter surveillance cohort (n = 1,805)
Full size table

Urine was the predominant specimen source, followed by pus/wound and blood. Blood isolates represented 15.4% of the cohort. The diversity of specimen types - including respiratory, soft tissue, catheter, and body-fluid samples - indicates that the surveillance dataset captured isolates from multiple clinical contexts; it should not be interpreted as a denominator-based incidence estimate of infection.

3.3 Overall species distribution

Enterobacter cloacae accounted for 47.0% of the cohort, followed by 21.3% unspeciated Enterobacter records. Klebsiella aerogenes (historically E. aerogenes) and Cronobacter sakazakii (historically E. sakazakii) accounted for 18.0% and 13.1%, respectively (Table 3). Because 21.3% of records lacked species-level identification, species-specific proportions and temporal shifts should be interpreted cautiously.

Table 3 Overall distribution of taxa in the historical Enterobacter surveillance cohort (n = 1,805)
Full size table

Species identification used standard biochemical and/or automated methods, including VITEK-2. Current nomenclature is used for reclassified taxa, with historical database names shown in parentheses. The 21.3% genus-level group represents isolates without reliable species-level identification; this substantial proportion limits species-specific epidemiological inference and may partly reflect earlier phenotypic identification practices.

The predominance of E. cloacae remains clinically relevant because members of the E. cloacae complex can express inducible chromosomal AmpC beta-lactamases (Davin-Regli et al., 2019, Tamma et al., 2019, Barceló et al., 2024). However, historical species assignments based on biochemical or automated systems should not be treated as equivalent to modern MALDI-TOF MS or genomic speciation, particularly when more than one fifth of records remained at genus level.

3.4 Annual distribution of Enterobacter-group taxa (2017–2025)

The annual composition showed an increasing proportion of E. cloacae after 2022 and declining genus-level and C. sakazakii records (Fig. 1). These changes are descriptive and may reflect both epidemiological shifts and changes in identification or reporting practice.

Fig. 1
Fig. 1
Full size image

Annual distribution of historically reported Enterobacter-group taxa, 2017–2025. Panel A shows absolute annual counts; Panel B shows proportional annual composition. Current names are used for K. aerogenes and C. sakazakii; these correspond to historical database labels E. aerogenes and E. sakazakii. “Other taxa” includes E. cloacae complex, E. amnigenus, and P. gergoviae

3.5 Overall antimicrobial resistance profile

Resistance estimates varied substantially across agents, but direct cross-drug comparison is constrained by selective testing and variable denominators (Table 4). Near-universal resistance to aminopenicillins and high cefuroxime resistance were observed; several low-resistance estimates came from much smaller tested subsets.

Table 4 Overall antimicrobial resistance profile of Enterobacter isolates with 95% confidence intervals
Full size table

Resistance was 93.7% for the second-generation cephalosporin cefuroxime. Among third-generation cephalosporins, resistance ranged from 37.7% for cefotaxime to 67.2% for ceftazidime (ceftriaxone 66.8%). Tigecycline showed 98.1% susceptibility (n = 156) and cefepime 95.5% susceptibility (n = 67), but both were selectively tested in small subsets. Colistin should not be described as one of the most active agents: only 62.1% of tested isolates were susceptible and 35.2% were intermediate (n = 182). These subset estimates are not directly comparable with agents tested in substantially larger numbers.

3.6 Temporal trends in resistance to key antibiotics (2017–2025)

Temporal trends were assessed using Spearman rank correlation across all nine annual observations (2017–2025), with significance defined as p < 0.05 (Fig. 2). Four of six key antibiotics showed statistically significant declining correlations; ciprofloxacin showed a non-significant decline and cefotaxime was stable.

Fig. 2
Fig. 2
Full size image

Observed resistance patterns for six key antibiotics, 2017–2025. Panel A compares resistance rates in 2017, 2021, and 2025; Panel B shows Spearman rank-correlation coefficients calculated from all nine annual observations. No 2026 projections are shown in the figure

Table 5 Spearman trend statistics:
Full size table

The significant downward temporal correlations for meropenem, piperacillin-tazobactam, amikacin, and gentamicin are encouraging surveillance observations. However, the study did not include antibiotic-consumption data, stewardship-intervention intensity, formulary changes, or other time-varying confounders; these declines therefore cannot be attributed causally to stewardship or formulary restrictions. The non-significant cefotaxime trend likewise should not be interpreted as evidence of persistent plasmid transmission without molecular data.

3.7 Comparative resistance rates by department and acquisition source

Table 5 presents resistance profiles across clinical departments (Panel A) and acquisition source categories (Panel B), enabling simultaneous ecological and acquisition-based interpretation.

Table 5 Comparative resistance rates (%) for selected antibiotics by department and acquisition source
Full size table

In Panel A, the COVID-19 ward had the highest observed resistance percentages for the listed agents, but this subgroup was small (n = 40) and no case-mix adjustment was performed. In Panel B, statistically significant acquisition-source differences were observed for cefotaxime, meropenem, imipenem, piperacillin-tazobactam, and co-trimoxazole. Hospital-acquired isolates generally had higher resistance than community-acquired isolates for several agents, but not uniformly: ciprofloxacin resistance was highest in the undetermined group (63.9% vs. 57.4% hospital and 53.3% community). Acquisition-source findings should therefore be interpreted antibiotic by antibiotic rather than as a blanket pattern.

3.8 MDRO prevalence by age group

MDRO prevalence increased across ordered age groups (Table 5). The Cochran-Armitage trend test yielded Z = 7.68, p < 0.001, and the Pearson chi-square across age groups was chi-square = 59.57, df = 4, p < 0.001. Age-group analysis included 1,803 isolates because age was missing for two isolates.

Table 6 MDRO prevalence by age group with odds ratios (OR) and 95% confidence intervals
Full size table

The unadjusted odds of an MDRO phenotype were 4.11-fold higher in patients aged > 80 years than in the 0-18-year reference group (OR 4.11, 95% CI 2.53–6.67, p < 0.001). This is an association, not evidence that age itself causes resistance: comorbidity, prior antimicrobial exposure, healthcare contact, department, and specimen type could confound the observed gradient. Accordingly, these data do not directly support age-specific empirical prescribing rules without multivariable validation.

3.9 Annual distribution of MDRO phenotypes (2017–2025)

The analytical MDRO composition changed over time (Fig. 3). The MDR label dominated during 2017–2022 but was not recorded in 2024–2025; CRE remained present across the study period. The laboratory-recorded AmpC phenotype increased from 5 isolates in 2023 to 57 in 2025 and was the largest recorded phenotype in 2025. Because the confirmatory method was not uniformly recoverable, this should not be interpreted as proof of a true mechanistic AmpC surge.

Fig. 3
Fig. 3
Full size image

Annual distribution of mutually exclusive analytical MDRO phenotype categories, 2017–2025. Stacked bars show annual counts and the line shows annual MDRO prevalence. MDR = multidrug-resistant; CRE = carbapenem-resistant Enterobacterales; AmpC = laboratory-recorded AmpC phenotype; ESBL = extended-spectrum beta-lactamase phenotype; CPO = carbapenemase-producing organism. Analytical exclusivity reflects the single primary phenotype label in the extracted record and does not imply biological exclusivity

Changes in recorded phenotype categories may reflect true epidemiology, testing/reporting practices, or both. The disappearance of the MDR label after 2022 cannot be attributed to stewardship, and the increase in the AmpC phenotype may partly reflect changes in detection or coding. AmpC-associated resistance can compromise third-generation cephalosporins and some beta-lactam/beta-lactamase inhibitor regimens, but susceptibility depends on the agent, enzyme expression, and co-resistance (Tamma et al., 2026, Tamma et al., 2022, Hardy et al., 2023). The 48.7% MDRO prevalence in 2025 merits continued surveillance and review of laboratory classification methods.

3.10 Exploratory 2026 weighted linear-regression projections

Weighted linear regression using the nine annual observations was retained only as an exploratory extrapolation for the six key antibiotic resistance series (Table 6). No out-of-sample validation was possible. Species-count and MDRO-count projections were removed because sparse terminal values produced unstable extrapolations, and tigecycline was not projected because no isolates were tested in 2025.

Table 7 Exploratory weighted linear-regression projections for selected antibiotic resistance rates in 2026
Full size table

The remaining 2026 values are hypothesis-generating surveillance signals only. They should not be interpreted as validated forecasts, evidence of future clinical resistance, or a basis for empirical therapy, formulary change, or resource allocation without new observed data.

4 Discussion

Across 2017–2025, the study captured 1,805 isolates in a historical Enterobacter surveillance stream at a secondary care centre in northern Oman. E. cloacae was the most frequent identified taxon (47.0%), consistent with its clinical importance and AmpC biology (Davin-Regli et al., 2019, Tamma et al., 2019, Barceló et al., 2024, Wang et al., 2026). However, 21.3% of records remained at genus level, and two historically used names now belong to other genera: K. aerogenes and C. sakazakii. These features limit species-specific inference and complicate direct comparison with modern surveillance based on MALDI-TOF or genomic identification. Urine was the most common specimen (31.6%), followed by wound and blood, broadly consistent with prior Enterobacter surveillance (Annavajhala et al., 2019, Elbehiry et al., 2024).

Overall imipenem and meropenem resistance were approximately 24%, while meropenem resistance declined from 32.1% in 2017 to 8.8% in 2025 and similar downward correlations were seen for piperacillin-tazobactam, amikacin, and gentamicin. These temporal patterns are descriptive: the study did not measure antimicrobial consumption, stewardship-intervention intensity, formulary changes, infection-control implementation, or case-mix shifts, so no causal stewardship effect can be inferred. In the acquisition-source subcohort, hospital-acquired isolates generally showed higher resistance for several drugs, but ciprofloxacin was highest in the undetermined group. A 2026 study from a newly established regional hospital in Turkey similarly found greater beta-lactam and fluoroquinolone resistance and more carbapenemase production among hospital-acquired Gram-negative isolates, with carbapenem resistance concentrated particularly in K. pneumoniae and E. cloacae complex (Tozluyurt, 2026). The agreement in direction across two different geographies supports the importance of early hospital surveillance, while differences in organism mix, testing methods, hospital maturity, and patient populations preclude direct prevalence comparison. Regional evidence from Saudi Arabia also documents a substantial burden of carbapenem-resistant Enterobacterales (Alshehri et al., 2024), while coordinated infection-control and antimicrobial-stewardship programmes have been associated with control of carbapenem-resistant Enterobacterales in other centres (Cornistein et al., 2023). These external observations provide context for the present temporal trends but do not establish their cause.

Age-stratified MDRO prevalence increased from 16.7% in patients aged 0–18 years to 45.2% in those > 80 years (OR 4.11, 95% CI 2.53–6.67). Because this analysis was unadjusted, age may be acting as a marker for comorbidity, prior antimicrobial exposure, healthcare contact, department, or specimen type (Papazachariou et al., 2024, Fernández-García et al., 2021, Al-Otaibi et al., 2023). Similarly, the rise in the laboratory-recorded AmpC phenotype from 5 isolates in 2023 to 57 in 2025 is an important surveillance signal but not definitive evidence of a mechanistic AmpC outbreak because a uniform confirmatory method and molecular data were unavailable. Modern treatment guidance for AmpC-producing Enterobacterales emphasizes organism- and susceptibility-specific therapy rather than assuming class-wide resistance (Tamma et al., 2026, Tamma et al., 2022, Hardy et al., 2023). Long-term hospital cohort data further show that MDRO epidemiology and associated risk factors can vary across calendar periods and COVID-19 policy phases (Zhou et al., 2026), and large hospital datasets have documented pandemic-era changes in antibiotic-resistant infection burden (Kadri et al., 2025). These observations reinforce the need to interpret the present unadjusted age gradient and temporal trends in light of potential time-varying confounding.

Therapeutically, this study did not test newer beta-lactam/beta-lactamase inhibitor agents or define carbapenemase genotypes, so treatment implications must be drawn from external Enterobacterales evidence rather than from the local dataset. In 603 clinical CRE isolates, Tamma et al., reported high in vitro susceptibility to ceftazidime-avibactam (95%), meropenem-vaborbactam (92%), imipenem-relebactam (84%), and cefiderocol (92%), with activity varying by carbapenemase mechanism (Tamma et al., 2023). For 593 metallo-beta-lactamase-producing Enterobacterales collected in France in 2024, aztreonam-avibactam and cefepime-zidebactam showed 98.0% and 96.6% susceptibility, whereas cefepime-taniborbactam and cefiderocol were less active overall, particularly among NDM producers (Emeraud et al., 2026). In a separate in vitro study of 57 blaNDM-producing E. coli isolates, aztreonam-avibactam outperformed aztreonam, cefiderocol, and the investigational cefiderocol-avibactam combination (Tozluyurt, 2024). Current IDSA guidance identifies aztreonam-avibactam and cefiderocol as preferred options for invasive NDM-producing Enterobacterales, but these recommendations depend on mechanism and susceptibility testing and cannot be directly extrapolated to this cohort without molecular characterization (Tamma et al., 2026).

The exploratory 2026 weighted linear-regression output was therefore deliberately de-emphasized. With only nine annual observations and no out-of-sample validation, it is a statistical extrapolation rather than a machine-learning model. Sparse series and zero-denominator outputs were suppressed, and the remaining projections should be treated as hypothesis-generating signals for surveillance rather than actionable forecasts (Kim et al., 2022, Sakagianni et al., 2023, Yuan et al., 2025).

5 Study limitations

Several limitations are important. First, this was a retrospective single-centre laboratory surveillance study and cannot establish causality. Second, 21.3% of isolates were identified only to genus level, and historical taxonomic labels reduce species-level precision. Third, AST was clinically directed, denominators varied by agent, and the dataset lacked sufficiently complete raw MIC/zone values and year-specific breakpoint metadata for retrospective harmonization; breakpoint revisions and changing testing practices may therefore influence temporal trends. Fourth, the available retrospective data did not contain a uniform, verifiable confirmatory method or molecular characterization for AmpC, ESBL, CRE, or CPO across all years. Fifth, acquisition-source analyses used an available-case subcohort and cefotaxime had a smaller denominator (n = 294) than other agents. Sixth, age associations were unadjusted for comorbidity, prior antibiotics, healthcare exposure, department, and specimen type. Seventh, antibiotic-consumption and stewardship-intervention data were unavailable, precluding attribution of resistance declines to stewardship. Finally, the 2026 weighted linear-regression analysis used only nine annual observations without external or out-of-sample validation and should not be used for clinical decision-making.

6 Conclusion

This nine-year surveillance study documents a substantial but changing antimicrobial-resistance burden in the historical Enterobacter surveillance cohort at a secondary care centre in Oman. Declining resistance to meropenem, piperacillin-tazobactam, amikacin, and gentamicin was observed, but these trends cannot be attributed to stewardship without antibiotic-consumption or intervention data. The increase in the laboratory-recorded AmpC phenotype, persistent cephalosporin resistance, and higher unadjusted MDRO prevalence in older patients justify continued laboratory and epidemiological surveillance. Future work should use standardized species identification, standardized AST denominators and retrospective breakpoint harmonization where raw data permit, molecular resistance characterization, antibiotic-use data, and multivariable analyses. The exploratory 2026 regression outputs are hypothesis-generating only and should not be used as standalone clinical forecasts.