Ethical approval for this study was obtained from the Health Research Ethics Committee (Komite Etik Penelitian Kesehatan, KEPK), Universitas Muhammadiyah Malang, Indonesia, under approval number E.5.a/065/KEPKUMM/IV/2025. Before sample collection, permission to conduct the study was obtained from the relevant local authority in Batu City and from the owners or managers of all participating dairy farms. Written or verbal informed consent was obtained from farm owners or responsible personnel before enrollment in the study.

All sampling procedures were performed using non-invasive methods and in accordance with accepted animal welfare principles. Milk samples were collected during routine milking procedures after standard udder hygiene practices, whereas fecal samples were collected as freshly voided material from the ground without restraining or disturbing the animals. Environmental samples, including forage feed, soil, animal drinking water, and hand-wash water, were collected without causing harm, stress, or disruption to the animals or farm activities. No experimental infection, invasive handling, or therapeutic intervention was performed on the animals as part of this study.

All field and laboratory procedures were carried out in accordance with biosafety and biosecurity measures to minimize contamination risks and protect animals, farm workers, and researchers. The confidentiality of farm identity and participant information was maintained throughout the study, and the collected data were used solely for research purposes.

This study used a cross-sectional, surveillance-based approach to examine the presence of MDR and ESBL-producing K. pneumoniae in dairy farm environments. It was carried out between May and August 2025 on select small- to medium-scale dairy farms in Batu City, East Java Province, Indonesia. Batu City is a key dairy-producing region characterized by a high density of smallholder farms, some of which serve as educational agrotourism sites, leading to frequent interactions among livestock, farm workers, visitors, and the surrounding environment.

The study focused on farm-level environmental surveillance rather than investigating clinical cases, with sampling designed to capture multiple on-farm matrixes representing animal-derived products, environmental reservoirs, and human-associated interfaces. This approach was adopted to support a One Health surveillance framework by assessing potential AMR reservoirs and transmission pathways within dairy production systems before environmental dissemination occurs downstream.

During the study period, dairy farms were chosen based on convenience and voluntary participation. Farms in Batu City were identified through local dairy farmer associations and cooperative networks and invited to join the study. A total of 59 dairy farms agreed to participate and were included.

Farms were eligible for inclusion if they: (i) were actively producing raw cow’s milk during the study period; (ii) operated as small- to medium-sized dairy farms; (iii) allowed access for milk and environmental sample collection; and (iv) obtained informed consent from farm owners or responsible workers. Farms were excluded if they were not in active production, declined participation, or did not permit comprehensive sampling across the targeted matrixes.

A single-proportion formula was used to estimate the sample size for cross-sectional prevalence studies. Assuming an expected ESBL prevalence of 11% [29], a 95% confidence level (Z = 1.96), and a precision of 10%, the minimum required sample size was calculated to be 38 farms. The inclusion of 59 dairy farms exceeded this minimum requirement and was considered sufficient to generate baseline surveillance data for K. pneumoniae-producing MDR and ESBL. However, since farm participation was voluntary and non-random, the findings may not be fully representative of all dairy farms in the region, which is acknowledged as a study limitation.

To reduce contamination and ensure reproducibility across farms, sample collection was performed using standardized protocols. Samples were collected once per farm during the study period, with one sample taken for each predefined matrix per farm to broadly represent potential AMR within each dairy farm ecosystem. This method was chosen to support baseline surveillance at the farm-level rather than within-farm quantitative comparisons. To prevent cross-contamination, sterile tools were used for each sample, disposable gloves were changed between different sample types, and all samples were handled aseptically during collection and transport.

Milk samples (30 mL) were collected from individual lactating cows during routine milking after discarding the initial streams of milk and following standard udder hygiene procedures. The samples were stored in sterile containers. Animal drinking water samples (1,000 mL) were collected directly from water troughs used by the cattle, while hand-washing water samples (200 mL) were taken from designated hand-washing stations after milking to represent human-associated environmental exposure on the farm.

Forage feed samples (200 g) were aseptically collected from feed bunks or feed storage areas commonly accessed by cattle. Soil samples (25 g) were taken from areas around cattle housing at an approximate depth of 10 cm and within 1–2 m of the barn to represent the immediate farm environment. Freshly voided fecal samples (10 g) were gathered from the ground to minimize environmental contamination.

Six sample types were collected from each farm, resulting in 354 samples from 59 participating farms. All samples were immediately placed in sterile containers, stored in insulated cool boxes with ice packs, and transported under cold-chain conditions. The samples’ temperatures were monitored using calibrated digital thermometers inside the cool boxes, with readings recorded at the start of transport and upon arrival at the lab to ensure they stayed at approximately 4°C throughout transportation [30]. All samples were delivered to the Veterinary Microbiology and Immunology Laboratory, Faculty of Veterinary Medicine, Universitas Brawijaya, and processed within 6–12 h of collection. Sampling was conducted after obtaining permission from the relevant farm owners and local authorities.

All collected samples were inoculated into Buffered Peptone Water (BPW) (Oxoid, Basingstoke, UK) at a 1:9 ratio and then incubated at 37°C for 24 h. Samples were pre-enriched in BPW with a 1:9 sample-to-broth ratio to improve the recovery of stressed or low-abundance bacteria and to neutralize any inhibitory substances in environmental and milk matrixes. BPW was chosen as a non-selective pre-enrichment medium according to ISO (International Organization for Standardization) and FDA BAM (Food and Drug Administration Bacteriological Analytical Manual) protocols [31].

The BPW isolates were then cultured on MacConkey Agar (MCA) (Oxoid). Colonies displaying mucoid morphology, presumed to be K. pneumoniae, were subsequently subcultured onto Eosin Methylene Blue Agar (EMBA) (Oxoid) and incubated at 37°C for 24 h [32]. Colonies with a pink, mucoid appearance were further identified through Gram staining and biochemical tests, including IMViC (Indole, Methyl Red, Voges–Proskauer, Citrate), Triple Sugar Iron Agar (TSIA), Sulfide Indole Motility (SIM), and urease. All media used for biochemical testing were sourced from Himedia, India [33]. The Gram staining procedure was performed according to Tripathi et al. [33].

K. pneumoniae colonies on MCA appeared pale red to pink, with smooth edges, a convex surface, and a mucoid texture due to the presence of a thick polysaccharide capsule (Figure-1a). Lactose fermentation with subsequent acid production produced pink colonies on this medium, whereas Gram staining confirmed that the isolate was Gram-negative (Figure-1c). Pure isolates presumed to be K. pneumoniae were further confirmed by subculturing onto EMBA followed by incubation at 37°C for 24 h. K. pneumoniae colonies appeared pinkish-black with a mucoid texture on EMBA (Figure-1b). The confirmed isolates were subsequently subjected to biochemical identification using several differential media. On TSIA, K. pneumoniae showed a yellow (acidic) color change with gas production but was negative for hydrogen sulfide (H₂S) production. The urease test showed a positive reaction, with the medium turning pink, indicating urease enzyme activity. In SIM medium, the isolates were negative for H₂S and indole production and were non-motile. The Methyl Red test was negative, whereas the Voges–Proskauer (VP) test was positive, indicating acetoin production. On Simmons citrate agar (SCA), a positive reaction, indicated by a blue coloration, demonstrated the ability to utilize sodium citrate as a carbon source and ammonium phosphate as a nitrogen source. The biochemical test results confirming the presence of K. pneumoniae are presented in Figure 2.

Pure K. pneumoniae isolates were tested for antimicrobial susceptibility using the Kirby–Bauer disk diffusion method on Mueller–Hinton Agar (MHA) (Oxoid). The selected antibiotics represent several antimicrobial classes relevant to dairy cattle treatment, environmental exposure, and public health, supporting a strong MDR classification in line with the One Health surveillance goals. Seven antibiotics commonly used in veterinary and dairy practices in Indonesia were tested, representing different antimicrobial classes, including ampicillin (AMP; 10 μg), cefotaxime (CTX; 30 μg), tetracycline (TE; 30 μg), ciprofloxacin (CIP; 5 μg), streptomycin (S; 10 μg), trimethoprim–sulfamethoxazole (STX; 1.15/23.75 μg), and chloramphenicol (C; 30 μg), all obtained from Oxoid, UK [28, 34].

Antibiotic disks were placed on the agar surface with a center-to-center distance of approximately 25–30 mm, followed by incubation at 35°C ± 2°C for 16–18 h (Figure 3) [35]. Antimicrobial susceptibility testing was performed according to the 2020 Clinical and Laboratory Standards Institute (CLSI) guidelines, including standardized inoculum preparation adjusted to the 0.5 McFarland turbidity standard and controlled incubation conditions.

Although a reference quality-control strain (e.g., Escherichia coli ATCC 25922) was not included during routine testing, strict adherence to CLSI-recommended procedures was applied to minimize technical variability. Isolates resistant to ≥3 antimicrobial classes were classified as MDR [36].

The double-disk synergy test (DDST) is a phenotypic assay used to identify K. pneumoniae isolates that produce ESBL [37]. MDR isolates were tested with the Kirby–Bauer disk diffusion method on MHA following CLSI guidelines [28]. This test involved ceftazidime (CAZ; 30 μg), cefotaxime (CTX; 30 μg), and amoxicillin–clavulanate (AMC; 30/10 μg), placed 20 mm apart from the amoxicillin–clavulanate disk [35]. Positive ESBL results were confirmed by the formation of a clear zone with a characteristic keyhole pattern between β-lactam antibiotic disks and AMC [28]. ESBL confirmation was performed using E. coli ATCC 25922 as a negative control and K. pneumoniae ATCC 700603 as a positive control (Figure 4).

Quality assurance procedures were followed during sample collection, transportation, and laboratory analysis to ensure data accuracy and consistency in methods. All sampling equipment and laboratory consumables were sterilized and used only once per sample to prevent cross-contamination. Culture media and reagents were prepared according to the manufacturer’s instructions and ISO and FDA BAM guidelines. Media sterility was confirmed by incubating uninoculated control plates alongside test samples.

Laboratory instruments, including incubators and refrigerators, were routinely calibrated and monitored to ensure proper functioning. Bacterial suspensions used for antimicrobial susceptibility testing were standardized with a 0.5 McFarland turbidity standard. Quality control for antimicrobial susceptibility testing was conducted according to CLSI guidelines. A randomly selected subset of isolates was re-tested to confirm reproducibility. All laboratory procedures were performed by trained personnel following standardized operating procedures. Data entry and laboratory records were cross-checked to reduce transcription errors and ensure consistency between laboratory results and analytical datasets.

Data were analyzed descriptively. Prevalence estimates were calculated as proportions and presented with 95% confidence intervals. Binomial proportion confidence intervals were calculated to quantify the precision of the prevalence estimates. Statistical analyses were performed using Microsoft Excel version 21 (Microsoft Corp., Redmond, WA, USA).

Isolation and phenotypic identification of samples collected from 59 dairy farms in Batu City showed that K. pneumoniae was found in 61.3% of all samples (217/354; 95% CI: 56.1–66.3). The prevalence of K. pneumoniae differed across sample types. The highest prevalence was seen in forage feed (84.7%; 50/59; 95% CI: 73.5–92.4), followed by animal drinking water (74.6%; 44/59; 95% CI: 62.9–83.8), hand-wash water (71.2%; 42/59; 95% CI: 59.1–81.2), and soil (71.2%; 42/59; 95% CI: 59.1–81.2). A moderate prevalence was found in milk samples (50.8%; 30/59; 95% CI: 37.6–64.0), while the lowest prevalence appeared in fecal samples (15.3%; 9/59; 95% CI: 7.3–27.0) (Table 1).

Antimicrobial susceptibility testing of K. pneumoniae isolates showed varying resistance patterns to different antimicrobial agents (Table 2). The highest resistance was seen with ampicillin (89.4%; 194/217), followed by streptomycin (71.4%; 155/217) and cefotaxime (38.3%; 83/217). Lower resistance rates were noted for sulfamethoxazole–trimethoprim (12.4%; 27/217), chloramphenicol (12.4%; 27/217), tetracycline (9.2%; 20/217), and ciprofloxacin (2.8%; 6/217). Resistance profiles varied among different sample types. Isolates from forage feed generally showed higher resistance rates to several antibiotics, especially ampicillin, streptomycin, and cefotaxime, compared to isolates from other sources. In contrast, resistance to ciprofloxacin and tetracycline remained consistently low across all sample types (Table 2).

Isolates resistant to three or more classes of antimicrobial agents were classified as MDR [38]. Among the 217 K. pneumoniae isolates, 35.0% (76/217) were MDR (Table 3). Resistance to three antimicrobial classes was the most common MDR pattern, accounting for 61.8% (47/76) of MDR isolates, followed by resistance to five classes (21.1%; 16/76), four classes (14.5%; 11/76), and six classes (2.6%; 2/76). No isolates showed resistance to any of the seven antimicrobial classes tested. The distribution of MDR isolates across sample types varied based on the number of resistant antimicrobial classes. MDR isolates were most often found in forage feed, hand-wash water, and animal drinking water, while isolates resistant to five or six antimicrobial classes were mainly detected in soil and milk samples (Table 3).

MDR was defined as resistance to three or more classes of antimicrobial agents. Percentages within rows represent the distribution of MDR isolates across sample types for each resistance category. Subtotal percentages represent the overall distribution of MDR isolates. Percentages were calculated using the total number of MDR isolates (n = 76) as the denominator. Values 3–7 indicate the number of different antimicrobial classes to which isolates exhibit resistance.

Among the 76 MDR K. pneumoniae isolates, 21.1% (16/76; 95% CI: 13.1–31.9) were confirmed as ESBL producers using DDST (Table 4). The highest prevalence of ESBL-producing isolates was found in forage feed samples (7.9%; 6/76; 95% CI: 3.7–15.9), followed by animal drinking water (6.6%; 5/76; 95% CI: 2.9–14.5), soil (3.9%; 3/76; 95% CI: 1.3–10.8), hand-wash water (1.3%; 1/76; 95% CI: 0.2–7.1), and milk (1.3%; 1/76; 95% CI: 0.2–7.1). No ESBL-producing isolates were detected in fecal samples (0%; 95% CI: 0.0–4.8).

This study shows a high presence of K. pneumoniae across various matrixes within dairy cattle farms in Batu City, suggesting widespread environmental contamination throughout the farm ecosystem. K. pneumoniae was found in more than half of all samples collected, emphasizing the extensive circulation of this organism across animal-related, environmental, and human-contact interfaces. The dominance of K. pneumoniae in forage feed, drinking water, hand-wash water, and soil highlights the significance of non-animal matrixes as key reservoirs that may support bacterial survival and spread within dairy production systems [39–41].

Compared with previous studies conducted in the Republic of the Congo and China, which reported much lower K. pneumoniae detection rates, the higher occurrence observed in this study suggests differences in context related to farm management practices, environmental hygiene, and patterns of antimicrobial use [6, 42]. While variations in sampling design and matrixes affect direct comparisons across studies, the consistently high detection across different sample types in this study indicates that dairy farms in Batu City represent a complex microbial ecosystem where K. pneumoniae is widespread. This finding is especially relevant in tourism-integrated dairy farms, where frequent interactions between animals, farm environments, and the public may heighten exposure and cross-contamination risks [43].

Antimicrobial susceptibility testing showed high resistance levels to antibiotics commonly used in food-producing animals, especially ampicillin and streptomycin. These results agree with reports from Indonesia and other Southeast Asian nations, where penicillins, sulfonamides, tetracyclines, and quinolones are frequently used in livestock production [24]. Studies from Southeast Asia consistently indicate that extensive and sometimes inappropriate use of antimicrobials in dairy farming leads to higher resistance rates, particularly to β-lactam antibiotics. The similarity between the resistance patterns observed in this study and those reported in wastewater surveillance suggests that dairy farms serve as significant upstream sources of AMR, contributing to environmental contamination downstream [28]. However, unlike wastewater research, the current farm-level, multimatrix approach allows for earlier detection of resistance sources before they are diluted and dispersed.

The proportion of MDR K. pneumoniae identified in this study further highlights the severity of AMR in dairy farm environments. Differences in MDR prevalence reported across various regions may reflect variations in sample matrixes, animal health, herd management, and antimicrobial use [44–46]. In East Java, where dairy cattle populations and antibiotic use are among the highest in Indonesia, selective pressure from frequent antimicrobial application may significantly contribute to MDR emergence [47]. The use of antibiotics for prophylaxis, metaphylaxis, and nonbacterial conditions, including agents classified as veterinary critically important antimicrobial agents, likely worsens the selection and persistence of resistant strains [24, 47].

The detection of ESBL-producing K. pneumoniae among MDR isolates raises an additional public health concern. ESBL producers were mainly found in environmental samples, especially forage feed and water sources, rather than being limited to animal-derived specimens. This distribution highlights the farm environment as a key point for the spread of AMR. Feed contamination, in particular, can occur through contact with soil, water, or poorly sanitized storage facilities, leading to repeated exposure of animals to resistant bacteria [48, 49]. These findings indicate that feed management and storage practices are critical points for controlling AMR spread within dairy systems.

The environmental persistence of ESBL-producing K. pneumoniae may further contribute to its widespread spread. K. pneumoniae is known for its ability to survive in various environmental conditions, including soil and water, and to form biofilms that boost resistance to environmental stressors and disinfectants [50, 51]. This persistence allows resistant strains to stay viable in farm environments even without direct antimicrobial pressure, increasing the chance of long-term survival and circulation. The presence of ESBL-producing isolates in multiple environmental compartments suggests that dairy farms could act as stable reservoirs supporting the persistence of antimicrobial resistance [52].

Beyond persistence, the co-occurrence of MDR and ESBL-producing K. pneumoniae in environmental matrixes raises concerns about horizontal gene transfer (HGT). ESBL genes are often found on mobile genetic elements, such as plasmids, which can be exchanged through conjugation, transformation, or transduction among bacterial populations [19, 53]. Environments with high microbial density, like contaminated feed, water, and soil, offer favorable conditions for genetic exchange [54]. Therefore, the detection of ESBL producers across environmental, animal-associated, and human-contact matrixes indicates an increased potential for interspecies and intercompartmental spread of resistance determinants within dairy farm ecosystems.

The concurrent detection of MDR- and ESBL-producing K. pneumoniae across animal-, environmental-, and human-related interfaces underscores the interconnected transmission pathways within dairy farm ecosystems. These results emphasize the need to adopt a One Health approach, recognizing that AMR control extends beyond animals to include feed safety, water quality, environmental sanitation, and responsible antimicrobial use [43, 55]. The risk of zoonotic and environmental transmission in tourism-linked dairy farms, where livestock production overlaps with public access, further heightens the public health significance of these findings.

Overall, this study offers several new contributions to research on AMR in dairy production systems. It provides a farm-level, multimatrix assessment of K. pneumoniae that identifies upstream resistance reservoirs often missed in clinical or wastewater studies. Finding forage feed as a key reservoir of ESBL-producing K. pneumoniae highlights an overlooked pathway for spreading resistance. Additionally, by examining animal-derived, environmental, and human-related matrixes simultaneously, the study reveals interconnected transmission interfaces aligned with a One Health approach. Although ESBL detection was limited to phenotypic methods, the results give valuable baseline data for Indonesian dairy farms and underline the need for future molecular and longitudinal research to better understand resistance patterns and guide targeted intervention efforts [52, 56, 57].

This study has several limitations to consider when interpreting the results. First, the cross-sectional design offers a snapshot of AMR at a single point in time and cannot assess changes over time or seasonal variations in the occurrence and resistance patterns of K. pneumoniae. Second, ESBL detection depended solely on phenotypic confirmation with DDST, without molecular analysis of ESBL-related genes. The lack of molecular typing restricts the ability to identify specific resistance genes, clonal relationships, and transmission pathways.

Third, the study was conducted in a single geographic area and involved one-time sampling per farm, which may limit the applicability of the findings to other dairy production systems with different management practices or environmental conditions. Additionally, although environmental and animal-associated matrixes were thoroughly assessed, human carriage was not evaluated, preventing direct conclusions about zoonotic transmission to farm workers or visitors. Furthermore, the study did not quantitatively analyze farm-level risk factors, such as hygiene practices, water source quality, or feed storage conditions, which may affect the distribution and persistence of AMR bacteria.

This study relied on phenotypic AMR testing following the CLSI guidelines; however, the lack of a reference quality-control strain in the AMR test (e.g., E. coli ATCC 25922) is a methodological limitation that should be addressed in future studies to further enhance laboratory quality assurance.

Despite these limitations, the current study provides valuable baseline data on the prevalence of MDR and ESBL-producing K. pneumoniae across different matrixes within dairy farm environments. Future research should focus on longitudinal and seasonal sampling to identify temporal trends, incorporate molecular techniques to analyze resistance genes and mobile genetic elements, and include human sampling to better understand zoonotic and environmental transmission pathways. Combining farm-level risk factor assessments with whole-genome sequencing would further enhance One Health–based surveillance and support targeted intervention strategies to reduce AMR in dairy production systems.