Relationship between hygienic management practices of smallholder dairy farms and the distribution of Gram-negative mastitis pathogens along with their antibiogram in Bangladesh Authors: Swarnali Akter, Sadia Tasnuva, Monika Akter Runa, Md Mamunur Rahman, Ataur Rahman, Ashiqul Islam, Md Monirul Islam, Sayra Tasnin Sharmy, Mst Sonia Parvin, Md Taohidul Islam Published: 2026 ABSTRACT Highlights•Higher occurrence of Escherichia coli(20.9%) and K. pneumoniae (27.7%) was observed in dairy farms.•30% of E. coliand 42% of K. pneumoniae isolates were multidrug resistant.•Highest resistance was observed against ampicillin, tetracycline and ceftriaxone, ampicillin and amoxicillin-clavulanic acid.•Higher KAP score related to udder health, hygiene, and mastitis, as well as higher biosecurity scores were not consistently associated with lower occurrence of E. coliand K. pneumoniae.Keywords: Bovine mastitis, Dairy farms, BAUMT, KAP score, Gram-negative mastitis pathogens, Antibiogram Introduction Bovine mastitis is one of the most important infectious diseases affecting dairy cows worldwide including Bangladesh, and poses a major constraint to dairy productivity through reduced milk production and quality, increased treatment and control costs, and premature culling (Cheng & Han, 2020; Sumon et al., 2020; Zigo et al., 2019, 2020). Bovine mastitis is defined as inflammation of the mammary glands parenchyma characterized by physical, chemical, and bacteriological changes in milk, as well as pathological changes of udder tissues (Radostits et al., 2007). It is broadly classified as clinical or subclinical: clinical mastitis presents with visible abnormalities in the udder (e.g., swelling, heat, pain) and/or milk (e.g., clots, discoloration), often accompanied by reduced milk production and in severe cases, systemic signs such as fever and loss of appetite, whereas subclinical mastitis (SCM) shows no visible signs in the udder or milk but is characterized by increased somatic cell count, reduced milk yield, and altered milk composition (Radostits et al., 2007). SCM is far more prevalent and economically damaging due to its silent nature and its role in maintaining infection within herds (Shafeeq et al., 2021; Sumon et al., 2017; Tommasoni et al., 2023). Mastitis is a multi-etiological disease resulting from complex interactions among pathogens, host factors, and the farm environment (Benić et al., 2018; Kibebew, 2017). Over 140 microbial species have been isolated from the bovine mammary gland with pathogens broadly categorized as contagious (e.g., Streptococcus agalactiae, Staphylococcus aureus, and Mycoplasma spp.) and environmental, including Gram-negative bacteria such as E. coli and Klebsiella pneumoniae (Cvetnić et al., 2016, 2022). Gram-negative mastitis pathogens primarily originate from environmental sources, including manure, bedding, soil, water, and feed, and their transmission is strongly associated with poor housing conditions, inadequate udder hygiene, contaminated milking equipment, and improper milking practices (Cobirka et al., 2020). Although these organisms are predominantly environmental and rely mainly on fecal-oral transmission for persistence within dairy systems, they can survive briefly on teat skin, milking liners, and other equipment surfaces. Unhygienic milking practices like failure to wash and sanitize milker’s hands between cows, use of shared and/or inadequately disinfected towels, and insufficient pre- and post-milking teat disinfection can facilitate cow-to-cow transmission of pathogens (Ashraf et al., 2017; Singh & Ramachandran, 2020). Consequently, environmental contamination combined with poor milking hygiene plays a central role in the persistence and dissemination of Gram-negative mastitis pathogens within dairy herds. In Bangladesh, the dairy sector is dominated by smallholder farms, typically rearing two to three cows under low-input management systems (Datta et al., 2019). It is estimated that there are about 1.4 million small-scale dairy farms across the country, contributing significantly to rural livelihoods and household nutrition (Sultana et al., 2024). Despite this importance, mastitis remains widespread: a national review reported prevalence of 11.2% for clinical mastitis and 39.1% for subclinical mastitis (Samad, 2022). Most smallholder dairy farms rely on manual milking, where hygienic practices are often inadequate. Milkers commonly wash their hands only before milking the first cow and do not wash or disinfect their hands between cows; in some cases, hands are merely dipped into the same residual water used to wash previous udders (Abebe et al., 2023). Such practices create ideal conditions for the transmission of Gram-negative pathogens, via contaminated hands or wash water. As Gram-negative pathogens are frequently present on the udder and teat surfaces of infected cows, it can be readily transferred to uninfected quarters or cows during milking. In addition, irregular shed cleaning, limited use of pre- and post-milking teat disinfection, shared towels for udder drying, and poor manure management further increase exposure to environmental Gram-negative pathogens (Abebe et al., 2023; Mbindyo et al., 2020). Recent evidence from studies conducted in Bangladesh indicates that >80% of dairy farms practice poor biosecurity and herd health management, highlighting critical gaps in farmers’ awareness and hygiene practices (Bushra et al., 2024). The growing emergence of antimicrobial resistance (AMR) among Gram-negative mastitis pathogens poses a significant threat to effective disease management, particularly in smallholder dairy systems where antibiotic use is often empirical and poorly regulated (Kovačević et al., 2022; Tomanić et al., 2023). In Bangladesh, where smallholder farms dominate the dairy sector, limited veterinary oversight further increases the risk of inappropriate antimicrobial use and the development of resistant bacterial strains (Al Emon et al., 2024). Despite this, there is a lack of comprehensive data on the farm-level occurrence and AMR profiles of key mastitis-associated pathogens such as E. coli and K. pneumoniae, and their relationship with management practices in these production systems. Such evidence is necessary to guide targeted mastitis prevention strategies, inform evidence-based treatment guidelines, and implement biosecurity interventions in smallholder dairy farms. It was hypothesized that farms with poorer udder hygiene, mastitis control practices, and weaker biosecurity measures would have a higher occurrence of E. coli and K. pneumoniae including antimicrobial-resistant strains. Therefore, the present study was undertaken to determine the occurrence and antimicrobial resistance profiles of these Gram-negative mastitis pathogens in dairy farms and cows in Bangladesh, and to evaluate their relationship with knowledge, attitude, and practices scores related to udder health, hygiene, and mastitis, as well as biosecurity. Materials and methods Selection of upazilas and dairy farms and cows A cross-sectional study was conducted over a one-year period from July 2023 to January 2024. Two upazilas (sub-districts) were randomly selected: Trishal in Mymensingh district and Shahjadpur in Sirajganj district. These two districts were randomly selected among 12 districts, which are known as dairy zones in Bangladesh viz. Sirajganj, Pabna, Rangpur, Bogura, Jashore, Satkhira, Munshiganj, Chattogram, Madaripur-Tekerhat, Mymensingh, Gazipur, Dhaka-Savar (Jagdish, 2021; Parvez, 2018; UNIDO, 2019). A total of 12 randomly selected smallholder dairy farms, comprising six from each upazila, were included for data and sample collection. The list and location of the dairy farms were taken from the respective upazila livestock office and veterinary hospital through consultation with veterinarian surgeon and/or upazila livestock officer. For farms to be included in the study, they had to have at least two lactating crossbred cows, and the farmers had to be willing to take part in the study. From each farm, two lactating cows were randomly selected when more than two were available; if only two lactating cows were present, both were included and sampled. Data collection Data were collected using a pretested structured questionnaire comprising both closed- and open-ended questions. It consisted of three sections: (i) respondent and farm demographic characteristics, (ii) knowledge, attitudes, and practices (KAP) related to udder health, hygiene, and mastitis, and (iii) perceptions and practices related to farm biosecurity. The final questionnaire contained 150 items, of which 73 assessed KAP regarding udder health, hygiene, and mastitis, while 52 addressed biosecurity perceptions and practices. KAP and biosecurity items were recorded as binary responses (yes/no). In addition, cow-level information was collected, including age, breed, parity, stage of lactation, history of mastitis during the previous lactation, and average daily milk yield. The questionnaire was developed in English and administered in Bangla during interviews to ensure respondent comprehension. Data collection was conducted through face-to-face interviews with farm owners or animal attendants, direct observation of farms, examination of cows and milk, and review of farm records when available. Responses were recorded electronically using a tablet device embedded with the KoboCollect application (https://kf.kobotoolbox.org). Clinical examination of udder and teats Clinical examination of the udder and teats was performed by a trained veterinarian (the first author) following a standardized examination protocol. The udder and teats of each selected cow were examined through visual inspection and gentle palpation to detect abnormalities such as swelling, redness, heat, pain, fibrosis, or abnormal milk secretion. Abnormal milk was defined as the presence of clots, flakes, pus, watery secretion or blood-tinged milk. Cows showing visible signs of udder inflammation and/or producing abnormal milk were classified as having clinical mastitis. However, no cases of clinical mastitis were detected during the data and sample collection. Sample collection, transportation and processing A total of 206 samples, comprising individual quarter milk samples (n = 96), teat and udder swabs (n = 96), and floor swabs (n = 14), were collected from 12 randomly selected dairy farms across two upazilas. For each cow, four quarter milk samples were collected. In addition, four composite swab samples per cow were obtained, one from each quarter. For each quarter, a teat-end swab and the corresponding udder skin swab were collected separately and then pooled to form a single composite quarter swab. One floor swab was collected from each farm as part of the standard procedure. However, two farms in Trishal upazila had two separate small sheds and floor swabs were also taken from those sheds, resulting in a total of 14 floor swab samples. All samples were transported within 24 h to the Population Medicine and AMR Laboratory, Department of Medicine, Bangladesh Agricultural University, in an icebox containing ice packs, and were processed immediately for the detection of E. coli and K. pneumoniae. Milk samples from apparently healthy quarters were collected aseptically from individual quarter. Prior to sampling, teat end and orifice were thoroughly cleansed with a cotton pad soaked in 70% ethyl alcohol. The first three to four streams of milk were discarded to minimize the risk of contamination. The collection vial was then held at a 45° angle to prevent debris from entering. The teat was directed toward a sterilized Falcon tube, and approximately 15 mL of milk was collected aseptically from each quarter and stored in a cooler box. Teat swab samples were collected aseptically from each quarter of the udder of lactating cows separately prior to milk sample collection to minimize the risk of cross-contamination. Teat swab samples were collected by using sterile cotton swab stick, with strict hygiene measures maintained throughout the procedure. The swab stick was gently rubbed across the teat orifice and along the side of each teat and corresponding udder surface, carefully avoiding contact with hair or the cow’s flank. To prevent cross-contamination between teats, a new sterile swab was used for each teat, and hands were disinfected or gloves were changed between sampling of different quarters. Immediately after swabbing, the swab sticks were placed into individual sterile Falcon tubes containing 5 mL of nutrient broth (NB) for transport and subsequent analysis. Floor swabs were collected from various areas of floor within the farm sheds of the selected dairy farms using sterile cotton swabs. Three replicate samples from each shed floor covering different places were collected and then pooled. These swabs were immediately placed in Falcon tubes containing 5 mL of nutrient broth and carefully labeled. On-site screening of subclinical mastitis Similar to the California Mastitis Test (CMT), the BAU Mastitis Test (BAUMT), which was developed by the Population Medicine and AMR Laboratory of the Department of Medicine, Faculty of Veterinary Science, BAU (Tanni et al., 2021), was used to screen milk samples for subclinical mastitis (SCM). The BAUMT was performed immediately after milk sample collection from each quarter, following teat swab sampling. For each test, 1 mL of milk and an equal volume of BAUMT reagent were added to individual cups of a four-cup shallow paddle and mixed gently using a circular motion for a few seconds. Reactions peaked within 10 s and were interpreted based on visible changes in milk viscosity and consistency. Results were scored as negative (0); trace (±), characterized by very slight thread formation or precipitation that disappeared upon stirring; weakly positive (+), showing distinct precipitation and thread formation without gel formation; distinctly positive (++), indicated by increased viscosity with gel formation; and strongly positive (+++), characterized by thick gel formation adhering to the bottom of the paddle cup. A BAUMT score of 1+ in any quarter was considered SCM-positive (Tanni et al., 2021). A cow was considered SCM-positive if at least one of its quarters tested positive and a farm was classified as positive if at least one sampled cow was positive. Isolation and identification of Gram-negative mastitis pathogens Gram-negative mastitis pathogens (E. coli and K. pneumoniae) were isolated and identified using standard bacteriological culture techniques, Gram staining, and biochemical tests, following ISO guidelines (ISO, 2011). The swab samples (teat swab and floor swab) collected in test tube containing NB were directly kept in incubator at 37 °C for 24 h. In case of milk samples, 1 mL sample was transferred to the sterile test tube containing 5 mL of NB and incubated at 37 °C for 24 h. A loopful of pre-enriched culture broth was streaked in duplicate onto Eosin Methylene Blue (EMB) agar plates for the selective isolation of E. coli and K. pneumoniae and incubated at 37 °C for 24 h. Colonies exhibiting a dark blue coloration with a characteristic metallic sheen were considered presumptive E. coli, whereas large, circular, mucoid colonies with a pink to purple dark center were presumptive K. pneumoniae. From each selective agar plate, three presumptive colonies of E. coli and K. pneumoniae were picked and subcultured to obtain pure colonies. Gram staining of pure cultures revealed short, rod-shaped, motile or non-motile organisms arranged singly, in pairs, or in short chains, consistent with Gram-negative bacteria. Biochemical testing further differentiated the isolates. E. coli isolates were indole-positive, producing a red to pink coloration on the medium surface, and Voges–Proskauer (VP) negative, producing a yellow coloration. In contrast, K. pneumoniae isolates were indole-negative, with a yellow reaction, and VP-positive, producing a red surface coloration. Biochemically confirmed isolates were further validated by Polymerase Chain Reaction (PCR) assay (Dashti et al., 2009). Uniplex PCR targeting the malB promoter gene and 16S rRNA gene was used for the confirmation of E. coli and K. pneumoniae, respectively. For E. coli, primers ECO-1 (5′-GACCTCGGTTTAGTTCACAGA-3′) and ECO-2 (5′-CACACGCTGACGCTGACCA-3′) were used to amplify a 585 bp fragment (Wang & Wang, 1996). For K. pneumoniae, primers K16S Forward (5′-ATTTGAAGAGGTTGCAAACGAT-3′) and K16S Reverse (5′-TTCACTCTGAAGTTTTCTTGTGTTC-3′) were used to amplify a 130 bp fragment (Liu et al., 2008). PCR for E. coli was performed in a 25 μL reaction mixture containing 1 μL of template DNA, 1.5 μL (15 pmol) of each primer, 12.5 μL of PCR master mix (New England BioLabs), and 8.5 μL of nuclease-free water. Amplification was carried out in a Veriti 96-well thermal cycler (Thermo Fisher Scientific) under the following conditions: initial denaturation at 95 °C for 5 min; 35 cycles of denaturation at 94 °C for 1 min, annealing at 58 °C for 1 min, and extension at 72 °C for 1 min; followed by a final extension at 72 °C for 7 min. PCR for K. pneumoniae was performed in a 25 μL reaction mixture containing 1 μL of template DNA, 1 μL (10 pmol) of each primer, 12.5 μL of PCR master mix (New England BioLabs), and 9.5 μL of nuclease-free water. The thermal cycling conditions were: initial denaturation at 94 °C for 5 min; 35 cycles of 94 °C for 30 s, 53 °C for 1.5 min, and 72 °C for 1.5 min; followed by a final extension at 72 °C for 10 min. Positive (containing strains with known E. coli and K. pneumoniae) and negative controls (sterile phosphate-buffered saline) were included in each PCR run. Amplified PCR products were analyzed by electrophoresis on 1.5% UltraPure™ agarose gel stained with ethidium bromide (5 µg/mL), alongside a 100 bp DNA ladder (New England BioLabs) as a molecular size marker. Bands were visualized under a UV transilluminator and documented photographically. All PCR-confirmed E. coli and K. pneumoniae isolates were stored at –20 °C with 50% (v/v) glycerol for further study. Antimicrobial susceptibility testing Antimicrobial susceptibility of E. coli and K. pneumoniae isolates was determined against twelve antimicrobial agents representing ten antimicrobial classes using the Kirby–Bauer disk diffusion method, following the guidelines of the Clinical and Laboratory Standards Institute (CLSI, 2018). Antimicrobial susceptibility testing was performed using commercial antimicrobial disks (Biomaxima, Poland), including fluoroquinolones [ciprofloxacin (CIP, 5 µg)]; cephalosporins [ceftriaxone (CRO, 30 µg) and cefepime (FEP, 30 µg)]; penicillins [ampicillin (AM, 10 µg)]; penicillins with β-lactamase inhibitors [amoxicillin–clavulanic acid (AMC, 30 µg)]; antipseudomonal penicillins with β-lactamase inhibitors [piperacillin–tazobactam (TPZ, 110 µg)]; carbapenems [meropenem (MEM, 10 µg)]; aminoglycosides [gentamicin (CN, 10 µg) and amikacin (AK, 30 µg)]; tetracyclines [tetracycline (T, 30 µg)]; folate pathway inhibitors [trimethoprim–sulfamethoxazole (SXT, 25 µg)]; and macrolides [azithromycin (AZM, 15 µg)]. Briefly, isolates were cultured in NB at 37 °C until they reached a turbidity equivalent to the 0.5 McFarland standard. Subsequently, 150 µL of the test suspension was evenly spread onto Mueller–Hinton agar plates, and antimicrobial disks were placed on the surface. The plates were incubated at 37 °C for 24 h. The susceptibility results were interpreted according to CLSI guidelines (2018). Multidrug-resistant (MDR) isolates were classified following the criteria of Magiorakos et al. (2012), with slight modifications. While Magiorakos et al. (2012) defined “nonsusceptible” to include both resistant and intermediate isolates, in this study only resistant isolates were considered. Accordingly, isolates resistant to at least one agent in three or more antimicrobial classes were defined as MDR. Scoring of knowledge, attitude and practice (KAP) and farms biosecurity For the assessment of KAP related to udder health, hygiene and mastitis as well as biosecurity perception and practices related to farms biosecurity, a structured scoring system was applied. Each relevant variable under knowledge, attitude, and practices was assigned a score of 1 if the farmer demonstrated adequate knowledge, a positive attitude, or appropriate practice, and 0 if otherwise. Similarly, for farm biosecurity and management practice scoring, each variable was scored as 1 if the practice was present and 0 if absent. To obtain udder health, hygiene and mastitis as well as biosecurity practices score for a dairy farm, the points for each variable were combined. The total score was 125. Statistical analysis Descriptive analysis was conducted to determine the occurrence of bacterial isolates and their antimicrobial resistance percentages. Differences in bacterial occurrence and resistance rates among different sample types and areas were assessed using the chi-square test (Z-test for proportions) and Fisher’s exact test, as appropriate. The association between KAP score related to udder health, hygiene, mastitis and biosecurity with the occurrence of E. coli and K. pneumoniae were assessed by using binomial logistic regression. Five outcome variables were considered: prevalence of E. coli, prevalence of Klebsiella spp., concurrent occurrence of both pathogens, presence of MDR E. coli, presence of MDR K. pneumoniae. For each outcome, the response was defined as the number of positive samples (numerator) among the total number of samples tested per farm (denominator). Explanatory variables included farm-level scores derived from structured questionnaires: i) knowledge (K), attitude (A), and practices (P) related to udder health, hygiene, and mastitis, ii) biosecurity-related perception score, iii) biosecurity practice score. Composite variables were also generated: KAP total score (combined knowledge, attitude, and practices related to udder health and mastitis control), biosecurity total score (combined perception and practices), grand total score (combined KAP and biosecurity scores). All scores were treated as continuous variables, where higher values indicated better knowledge, attitudes, or practices. Associations between explanatory variables (scores) and bacterial occurrence were assessed using binomial logistic regression within the framework of generalized linear models in IBM SPSS. For each outcome, the number of positive samples per farm was modeled as a binomial response with a logit link function: where pi represents the probability that a sample from the ith farm is positive, and Xi represents the explanatory variable (score). Considering the limited number of farms (n = 12), separate univariable models were fitted for each explanatory variable and outcome to avoid model overfitting. Results were expressed as odds ratios (OR) with corresponding 95% confidence intervals (CI). Statistical significance was determined at p < 0.05. Results Occurrence and distribution of E. coli and K. pneumoniae Out of 206 samples analyzed (96 milk, 96 teat swabs, and 14 floor swabs), 20.9% (43/206) were positive for E. coli and 27.7% (57/206) for K. pneumoniae based on cultural and biochemical characteristics. All cultural and biochemically positive isolates were PCR positive (Fig. 1a-b). Among the three sample types, the occurrence of bacteria was significantly higher in floor swabs (64.3% for E. coli and 85.7% for K. pneumoniae), compared to milk samples (12.5% for E. coli and 10.4% for K. pneumoniae) and teat swab (22.9% for E. coli and 36.5% for K. pneumoniae) (Fig. 2). a-b. PCR amplified product of (a) 585 bp from malB promoter gene of E. coli (b) 130 bp from 16s rRNA gene of K. pneumoniae in 1.5% agarose gel and ethidium bromide staining. Legends: M = DNA marker (1 kb plus), Lane 1 = Positive control of E. coli and K. pneumoniae, Lane 2–11 = PCR products of tested E. coli and K. pneumoniae isolates. Prevalence of E. coli and K. pneumoniae isolated from three different types of samples. Area-wise analysis figured that the occurrence of E. coli and K. pneumoniae in milk samples was almost similar (14.6% and 8.3% in Shahjadpur, and 10.4% and 12.5% in Trishal, respectively). No significant difference was observed for the occurrence of E. coli and K. pneumoniae among areas isolated from different samples except E. coli isolated from floor swab samples. For floor swab samples, the occurrence of E. coli was significantly higher in Trishal (75%) than in Shahjadpur (50%) upazila. On the contrary, a higher occurrence of K. pneumoniae was observed in Shahjadpur (100%) compared to Trishal upazila (75%) but it was non-significant (Fig. 3). For teat swab samples, E. coli occurrence was higher in Trishal (29.2%), whereas K. pneumoniae was more prevalent in Shahjadpur upazila (41.7%). Area-wise distribution of E. coli and K. pneumoniae isolated from three different types of samples. Quarter-level distribution of E. coli and K. pneumoniae according to BAUMT status At the quarter level (n = 96), among BAUMT-positive quarters (n = 45, 46.9%), E. coli and K. pneumoniae were each detected in 13.3% (n = 6) quarters, while 22.2% (n = 10) of quarters were positive for at least one of the investigated pathogens. In BAUMT-negative quarters (n = 51), E. coli was identified in 11.8% (n = 6) and K. pneumoniae in 7.8% (n = 4) quarters, with 17.6% (n = 9) of quarters tested positive for either of the pathogens. Overall, E. coli and K. pneumoniae were detected in 12.5% (n = 12) and 10.4% (n = 10) of the 96 quarter samples, respectively, and 19.8% (n = 19) quarters were positive for either of the pathogens (Table 1). Table 1Quarter-level occurrence of E. coli and K. pneumoniae according to BAUMT result status.Table 1 dummy alt textNo. (%) of quarter milk samples (n = 96)E. coli positive (n, %)K. pneumoniae positive (n, %)Pathogen-positive quarters (n, %)BAUMT +ve = 45 (46.9)6 (13.3)a6 (13.3)a10 (22.2)BAUMT –ve = 51 (53.1)6 (11.8)a4 (7.8)a9 (17.6)Total12 (12.5)10 (10.4)19 (19.8)Values with same letters do not differ significantly between pathogen (p ≥ 0.05). [Chi-square test (Z-test for proportions) and Fisher’s exact test (as appropriate) were done for identification of significant variation]; BAUMT- Bangladesh Agricultural University Mastitis Test; n = Number of quarter samples. Quarter-level occurrence of E. coli and K. pneumoniae according to BAUMT result status. Values with same letters do not differ significantly between pathogen (p ≥ 0.05). [Chi-square test (Z-test for proportions) and Fisher’s exact test (as appropriate) were done for identification of significant variation]; BAUMT- Bangladesh Agricultural University Mastitis Test; n = Number of quarter samples. At the individual cow level (n = 24), 47.1% (n = 8) of cows were positive for E. coli, 35.3% (n = 6) for K. pneumoniae, and 52.9% (n = 9) for at least one of the pathogens. At the farm level (n = 12), E. coli was detected in 58.3% (n = 7) of farms, K. pneumoniae in 50% (n = 6), and at least one of the pathogens in 58.3% (n = 7) of farms (Table 2). Table 2Cow-level and farm-level occurrence of E. coli and K. pneumoniae according to BAUMT result status.Table 2 dummy alt textNo. (%) of milk samplesE. coli positive (n, %)K. pneumoniae positive (n, %)Pathogen-positive (n, %)Cow-level (n = 24)BAUMT +ve = 17 (70.8)8 (47.1)a6 (35.3)a9 (52.9)BAUMT –ve = 7 (29.2)4 (57.1)a3 (42.9)a5 (71.4)Total12 (50.0)9 (37.5)14 (58.3)Farm-level (n = 12)BAUMT +ve = 12 (100)7 (58.3)a6 (50)a7 (58.3)Total7 (58.3)6 (50)7 (58.3)Values with same letters do not differ significantly between pathogen (p ≥ 0.05). [Chi-square test (Z-test for proportions) and Fisher’s exact test (as appropriate) were done for identification of significant variation]; BAUMT- Bangladesh Agricultural University Mastitis Test; n = Number of quarter samples. Cow-level and farm-level occurrence of E. coli and K. pneumoniae according to BAUMT result status. Values with same letters do not differ significantly between pathogen (p ≥ 0.05). [Chi-square test (Z-test for proportions) and Fisher’s exact test (as appropriate) were done for identification of significant variation]; BAUMT- Bangladesh Agricultural University Mastitis Test; n = Number of quarter samples. Antimicrobial resistance profile of E. coli and K. pneumoniae Antimicrobial susceptibility testing revealed resistance patterns of E. coli and K. pneumoniae isolates to individual antimicrobial agents (Fig. 4–5). Among E. coli, the highest resistance was observed against ampicillin (41.9%, n = 18), followed by tetracycline and ceftriaxone (23.3%, n = 10 each), and gentamicin (20.9%, n = 9). The lowest resistance was recorded against piperacillin-tazobactam (4.7%, n = 2) and ciprofloxacin (7.0%, n = 3). None of the E. coli isolates were resistant to meropenem (Fig. 4). Similarly, K. pneumoniae isolates showed highest resistance to ampicillin (64.9%, n = 37) and amoxicilin-clavulanic acid (49.1%, n = 28), with reduced resistance to ciprofloxacin (1.8%, n = 1), pipercillin-tazobactam, amikacin, and meropenem (3.5% each, n = 2) (Fig. 5). Phenotypic resistance pattern of E. coli to number of antimicrobials and antimicrobial classes isolated from different types of samples. Phenotypic resistance pattern of K. pneumoniae to number of antimicrobials and antimicrobial classes isolated from different types of samples. The UpSet plot further illustrates the combinations of resistance to individual antimicrobials. Among E. coli, common resistance profiles included AM-CRO-T-CN-FEP-SXT-AMC (n = 1), AM-CRO-T-FEP-AK-SXT-CIP (n = 1), AM-CRO-T-CN-FEP-TPZ (n = 1), and AM-CRO-T-FEP-SXT-CIP (n = 1) (Fig. 4). For K. pneumoniae, predominant resistance profiles were AM-AMC-SXT-CRO (n = 5), AM-AMC-SXT (n = 5), AM-AMC (n = 5), followed by less frequent patterns such as AM-SXT-CRO-MEM and AM-AZM (n = 2 each) (Fig. 5). Multidrug resistance (MDR) was observed in 30.2% (n = 13) of E. coli and 42.1% (n = 24) of K. pneumoniae isolates. Among the sample types, the prevalence of MDR E. coli was significantly higher in floor swabs (44.4%, n = 4) and teat swab (31.8%, n = 7) compared with milk (16.7%, n = 2) samples. For K. pneumoniae, 50% of the isolates recovered from both milk (n = 5) and floor swab (n = 6) samples were identified as MDR (Fig. 6). The common MDR patterns in E. coli were P-AG-Ceph-Tet-FPI-BLI (n = 1) and P-AG-Ceph-Tet-FPI-FQ (n = 1). In K. pneumoniae, the predominant patterns were P-BLI-FPI-Ceph (n = 5) and P-BLI-FPI (n = 5) followed by P-FPI-Ceph-Car, P-BLI-AG and P-BLI-Ceph (n = 2 each) (Fig. 4–5). Multidrug resistance pattern of E. coli and K. pneumoniae observed among different types of samples. Sample-type-wise analysis of AMR patterns revealed that, ampicillin resistance was consistently the highest in both E. coli and K. pneumoniae, with particularly higher in K. pneumoniae from milk (90%) and floor swabs (83.3%) samples. In milk samples, E. coli showed moderate resistance to gentamicin and amikacin (25%) but low resistance to tetracycline and amoxicillin-clavulanic acid (8.3%), and remained fully susceptible to ciprofloxacin, ceftriaxone, cefepime, piperacillin-tazobactam, meropenem. K. pneumoniae exhibited very high resistance to ampicillin but low resistance (10%) to most other antimicrobials, with no resistance to ciprofloxacin, amikacin, or piperacillin-tazobactam (Table 3). Table 3Individual antimicrobial resistance pattern of E. coli and K. pneumoniae isolated from three different types of samples.Table 3 dummy alt textAntimicrobial classAntimicrobial agentsNo. (%) of isolatesE. coli (n = 43)Klebsiella pneumoniae (n = 57)MilkTeat swabFloor swabMilkTeat swabFloor swabFluoroquinolonesCiprofloxacin02 (9.1)1 (11.1)01 (2.9)0CephalosporinsCeftriaxone07 (31.8)3 (33.3)3 (30.0)11 (31.4)1 (8.3)Cefepime06 (27.3)2 (22.2)2 (20.0)2 (5.7)0PenicillinsAmpicillin4 (33.3)9 (40.9)5 (55.6)9 (90.0)18 (51.4)10 (83.3)Penicillins + β-lactamase inhibitorsAmoxicilin-Clavulanic acid1 (8.3)2 (9.1)1 (11.1)3 (30.0)17 (48.6)8 (66.7)Antipseudomonal penicillins + β-lactamase inhibitorsPiperacillin-Tazobactam01 (4.5)1 (11.1)01 (2.9)1 (8.3)CarbapenemsMeropenem0001 (10.0)1 (2.9)0AminoglycosidesGentamicin3 (25.0)4 (18.2)2 (22.2)1 (10.0)4 (11.4)2 (16.7)Amikacin3 (25.0)4 (18.2)1 (11.1)01 (2.9)1 (8.3)TetracyclinesTetracycline1 (8.3)6 (27.3)3 (33.3)1 (10.0)2 (5.7)1 (8.3)Folate pathway inhibitorsTrimethoprim-Sulphamethoxazole2 (16.7)4 (18.2)1 (11.1)3 (30.0)8 (22.9)5 (41.7)MacrolidesAzithromycin2 (16.7)2 (9.1)1 (11.1)1 (10.0)4 (11.4)1 (8.3)n = No. of isolates. Individual antimicrobial resistance pattern of E. coli and K. pneumoniae isolated from three different types of samples. n = No. of isolates. In teat swab samples, E. coli isolates exhibited the highest resistance to ampicillin (40.9%) and ceftriaxone (31.8%), while resistance to piperacillin-tazobactam was lowest (4.5%). K. pneumoniae showed high resistance to ampicillin (51.4%) and amoxicillin-clavulanic acid (48.6%), whereas very low resistance (2.9%) was observed against amikacin, meropenem, piperacillin-tazobactam, and ciprofloxacin (Table 3). In floor swab samples, resistance was generally higher, particularly for E. coli (ampicillin: 55.6%) and K. pneumoniae (ampicillin: 83.3%; amoxicillin-clavulanic acid: 66.7%). Importantly, no resistance to meropenem was observed in both organisms (Table 3). Relationship of KAP score about udder health, hygiene, mastitis and biosecurity with the occurrence of E. coli and K. pneumoniae Binomial logistic regression analysis demonstrated that practice score related to udder health, hygiene, and mastitis (OR = 1.095, p = 0.015), biosecurity perception score (OR = 1.575, p = 0.027), KAP total score (OR = 1.097, p = 0.011), and grand score (OR = 1.058, p = 0.022) showed significant positive associations with E. coli positivity. Knowledge score showed a borderline association (OR = 1.783, p = 0.068) (Table 4). Table 4Association between knowledge, attitude, and practice scores related to udder health, hygiene, mastitis, and biosecurity associated with the occurrence of E. coli and K. pneumoniae in dairy cows and farms.Table 4 dummy alt textScoresE. coli OR (95% CI)p-valueK. pneumoniae OR (95% CI)p-valueCo-occurrence OR (95% CI)p-valueMDR E. coli OR (95% CI)p-valueMDR K. pneumoniae OR (95% CI)p-valueKnowledge (K)1.78 (0.96–3.32)0.0681.33 (0.76–2.33)0.3100.97 (0.41–2.29)0.9531.48 (0.45–4.88)0.5210.92 (0.39–2.13)0.838Attitude (A)0.83 (0.54–1.27)0.3810.75 (0.51–1.09)0.1350.85 (0.47–1.52)0.5770.91 (0.42–2.00)0.8200.59 (0.32–1.13)0.113Practices (P)1.09 (1.02–1.18)0.0151.03 (0.96–1.09)0.4351.12 (1.01–1.24)0.0390.98 (0.85–1.14)0.8231.14 (0.99–1.29)0.058KAP total score1.09 (1.02–1.18)0.0111.02 (0.96–1.09)0.5061.11 (1.00–1.22)0.0460.99 (0.85–1.14)0.8571.11 (0.98–1.27)0.112Biosecurity perception1.58 (1.05–2.35)0.0271.56 (1.07–2.29)0.0211.18 (0.66–2.12)0.5771.11 (0.54–2.26)0.7830.94 (0.52–1.72)0.847Biosecurity practices1.04 (0.94–1.16)0.4370.99 (0.91–1.09)0.9391.09 (0.94–1.26)0.2431.07 (0.88–1.30)0.5081.30 (1.09–1.56)0.004Biosecurity total1.08 (0.97–1.20)0.1821.02 (0.93–1.13)0.6331.11 (0.95–1.30)0.1811.09 (0.88–1.36)0.4281.38 (1.11–1.71)0.004Grand total1.06 (1.01–1.11)0.0221.02 (0.97–1.06)0.5091.07 (0.99–1.15)0.0541.01 (0.92–1.12)0.8041.12 (1.02–1.23)0.014 Association between knowledge, attitude, and practice scores related to udder health, hygiene, mastitis, and biosecurity associated with the occurrence of E. coli and K. pneumoniae in dairy cows and farms. For K. pneumoniae, biosecurity perception score was the only variable significantly associated with positivity (OR = 1.563, p = 0.021). Separate scores of knowledge, attitude, and practice related to udder health, hygiene, and mastitis, and all composite indices (KAP total, biosecurity total, and grand score) were not significantly associated with K. pneumoniae positivity (p > 0.05) (Table 4). In the model assessing concurrent occurrence of E. coli and K. pneumoniae, practice score related to udder health, hygiene, and mastitis (OR = 1.115, p = 0.039) and KAP total score (OR = 1.106, p = 0.046) were significantly associated with increased odds of co-occurrence. The grand score showed a borderline association (OR = 1.069, p = 0.054). Other variables were not statistically significant (Table 4). Binomial logistic regression analysis also revealed that biosecurity practices score (OR = 1.30, p = 0.004), biosecurity total score (OR = 1.38, p = 0.004), and grand total score (OR = 1.12, p = 0.014) were significantly associated with MDR K. pneumoniae positivity. On the other hand, practices score related to udder health, hygiene, and mastitis showed a borderline association (OR = 1.14, p = 0.058). No significant association were found incase of MDR E. coli positivity and other variables (knowledge, attitude, KAP total and biosecurity perception score) associated with MDR K. pneumoniae positivity (p > 0.05) (Table 4). Discussion This study provides updated information on the prevalence and AMR profiles of Gram-negative mastitis pathogens (E. coli and K. pneumoniae) in dairy farms, and their relationship with hygienic practices of the farms, which will be utilized to establish an integrated AMR surveillance system and to make it easier to assess AMR prevention and control measures in the dairy farms of Bangladesh. The occurrence of E. coli and K. pneumoniae was 12.5% and 10.4%, respectively in milk samples. The occurrence of E. coli in milk samples in the present study is in agreement with the previous reports of Hasan et al. (2022) who found 7.6% prevalence of E. coli in milk samples from dairy cows in Bangladesh. The prevalence of K. pneumoniae in the current study was lower than the study of Salauddin et al. (2019) who reported a higher (62.5%) prevalence of Klebsiella spp. in clinical mastitis milk samples from Rangpur district in Bangladesh. However, variable prevalence of E. coli (30% to 71%) and K. pneumoniae (26% to 30%) in different types of samples from dairy farms and cows was reported by earlier researchers (Bag et al., 2021; Nahar et al., 2021, 2023; Rana et al., 2022; Saddam et al., 2023), likely reflecting differences in geographical location, environmental conditions, breed and feeding regimen, farm management practices, wide distribution of pathogens in soil, manure, contaminated udder and bedding materials and immune status of the animals. E. coli and K. pneumoniae are widely distributed in the farm environment, including bedding, soil, dung, and other organic materials, and are well adapted to persist under such conditions. This facilitates contamination of the udder and may lead to the establishment of subclinical infections of varying duration (Cheng & Han, 2020; Cvetnić et al., 2022; Sahu et al., 2024). The area-wise analysis indicated that the occurrence of E. coli and K. pneumoniae in milk samples was relatively comparable between the studied locations, suggesting a broadly similar level of contamination across dairy systems in both Shahjadpur and Trishal. This similarity may reflect common management practices, milking hygiene conditions, and environmental exposures shared by smallholder farms in these regions. The lack of significant variation in milk samples implies that the direct contamination of milk by these Gram-negative pathogens is not strongly influenced by geographic location alone, but rather by underlying farm-level practices. Quarter-level analysis demonstrated that E. coli and K. pneumoniae were detected in both BAUMT-positive and BAUMT-negative quarters, with slightly greater occurrence in BAUMT-positive samples. Both pathogens were identified at comparable frequencies among mastitis-test–positive quarters, and 19.8% of all examined quarters harbored at least one of the investigated bacteria. The presence of these environmental coliforms in BAUMT-positive quarters supports their established role in bovine mastitis pathogenesis, particularly under suboptimal hygiene and housing conditions (Cheng & Han, 2020; Hogan & Smith, 2012). Importantly, their detection in BAUMT-negative quarters suggests possible subclinical colonization, early infection stages, or intermittent bacterial shedding, highlighting the dynamic nature of intramammary infections and the limitations of relying solely on screening test status for pathogen detection. At the individual cow level, 47.1% and 35.3% of cows were positive for E. coli and K. pneumoniae, respectively, with 52.9% harboring at least one pathogen, indicating substantial exposure to environmental coliforms. This relatively high cow-level occurrence underscores the widespread exposure of dairy cattle to environmental coliforms, which are recognized as major opportunistic mastitis pathogens (Hogan & Smith, 2012). At the farm level, the detection of E. coli (58.3%) and K. pneumoniae (50%) further suggests that these pathogens are widely distributed and likely endemic under prevailing management conditions (Alawneh et al., 2020; Morales-Ubaldo et al., 2023). However, it should be noted that this study did not assess the concomitant presence of other mastitis-associated pathogens in relation to SCC. Therefore, the observed associations between BAUMT results and bacterial detection should be interpreted cautiously, as SCC levels may also be influenced by mixed infections involving other undetected pathogens. Future studies incorporating comprehensive pathogen profiling would be valuable to better elucidate the relationship between co-infections and SCC dynamics. The AMR profiles observed in this study highlight a growing threat to effective mastitis management. The challenge was not negligible as some of the antibiotics tested in this study are classified by the World Health Organization (WHO, 2024) as medically important in human medicine. Meropenem is authorized for human use only, whereas ceftriaxone and ciprofloxacin are categorized as Highest Priority Critically Important Antimicrobials (HPCIAs). Furthermore, amoxicillin-clavulanic acid is classified as highly important antibiotic. E. coli isolates exhibited the highest resistance against ampicillin (41.9%), whereas K. pneumoniae showed even higher resistance to both ampicillin (64.9%) and amoxicillin-clavulanic acid (49.1%). These findings are in line with previous studies documenting high resistance of Gram-negative mastitis pathogens to β-lactam antibiotics, likely due to the widespread and often indiscriminate use of these drugs in veterinary practice (Bag et al., 2021; Nahar et al., 2021). Importantly, both pathogens displayed low resistance to critically important antimicrobials such as meropenem, amikacin, and ciprofloxacin, reflecting limited use of these agents in veterinary practice in Bangladesh. However, their use in food animals must be avoided to preserve clinical utility in human medicine. Multidrug resistance was observed in 30.2% of E. coli and 42.1% of K. pneumoniae isolates, with the highest levels detected in floor swabs. The MDR patterns observed in E. coli and K. pneumoniae likely reflect the routine and often empirical use of broad-spectrum antibiotics for mastitis treatment in dairy farms of the study region. Penicillins, tetracyclines, aminoglycosides, and third-generation cephalosporins are among the most commonly used antimicrobials in dairy practice of Bangladesh due to their wide availability, low cost, and frequent use without prior culture and susceptibility testing (Hosain et al., 2022). Consistent with this practice, the highest resistance in E. coli was detected against ampicillin, tetracycline, ceftriaxone, and gentamicin, with the most prevalent MDR combinations involving penicillin-aminoglycosides-cephalosporins-tetracycline-folate pathway inhibitors-penicillins-β-lactamase inhibitors. Similarly, K. pneumoniae exhibited dominant MDR profiles characterized by co-resistance to penicillin-penicillins-β-lactamase inhibitors-folate pathway inhibitors-cephalosporins, followed by penicillin-penicillins-β-lactamase inhibitors-folate pathway inhibitors. The environmental concentration of MDR strains poses a serious risk for persistence, horizontal gene transfer, and reinfection in dairy herds (Shoaib et al., 2023). Previous studies have also reported higher MDR prevalence in environmental isolates compared to those from milk, reinforcing the importance of environmental sanitation in AMR control strategies (Bag et al., 2021; Benić et al., 2018; Cvetnić et al., 2016; Kovačević et al., 2022; Salauddin et al., 2019). Treating patients infected with MDR pathogens particularly those with severe or complicated infections remains a major challenge in developing countries like Bangladesh where the capacity to develop or access new antimicrobial agents is limited. In this context, routine and systematic documentation of antimicrobial susceptibility testing is essential to guide rational therapy and promote the judicious use of existing antimicrobial agents, representing a critical strategy to mitigate the growing threat of multidrug resistance. The present study evaluated the association between knowledge, attitude, practices (KAP) related to udder health, hygiene, and practices; and biosecurity-related perceptions and practices with the occurrence of mastitis-associated pathogens. Contrary to expectations, several variables including practice score related to udder health, hygiene, and mastitis, biosecurity perception, and composite indices (KAP total, biosecurity total, and grand score) were positively associated with E. coli positivity and concurrent occurrence with K. pneumoniae. This unexpected direction of association may be explained by reverse causality, a known limitation of cross-sectional studies. Farms experiencing higher levels of mastitis or bacterial contamination may subsequently improve their knowledge, awareness, and reported practices, resulting in higher KAP scores alongside higher pathogen prevalence. Environmental mastitis pathogens such as E. coli are widely recognized to persist in farm environments and remain difficult to control despite improved management practices (Bradley, 2002; Hogan & Smith, 2012). Similarly, K. pneumoniae is strongly associated with environmental sources such as bedding, manure, and contaminated housing systems. Their occurrence is often driven more by environmental and management conditions than by farmer knowledge alone (Munoz et al., 2007; Zadoks et al., 2011). The finding that only biosecurity perception was significantly associated with K. pneumoniae positivity may reflect increased awareness among farmers in response to existing contamination rather than effective control. Another plausible explanation is the knowledge–practice gap, where increased knowledge does not necessarily translate into effective implementation of control measures. Previous studies have shown that structural factors such as housing design, hygiene infrastructure, and resource availability play a critical role in mastitis epidemiology and may outweigh the influence of individual-level practices (Barkema et al., 1999; Dufour et al., 2012). Therefore, higher KAP scores may reflect awareness without corresponding improvements in actual farm conditions. Additionally, farmers with higher knowledge may report practices more accurately or in a socially desirable manner, which may not correspond to consistent on-farm implementation. This limitation is common in questionnaire-based KAP studies. The relatively small number of farms (n = 12) is another important limitation. Small sample sizes can lead to unstable regression estimates, reduced statistical power, and unexpected directions of association. The significant positive associations of biosecurity practices score, biosecurity total score, and grand total score with MDR K. pneumoniae positivity appear paradoxical but may reflect the co-existence of intensive biosecurity measures with higher antimicrobial usage on well-managed farms, imposing selective pressure that favors MDR strain emergence. This phenomenon has been documented in intensive livestock systems, where higher biosecurity scores often correlate with greater therapeutic and prophylactic antibiotic use, inadvertently promoting resistance selection (Van Boeckel et al., 2015). K. pneumoniae possesses remarkable genomic plasticity, readily acquiring resistance determinants through mobile genetic elements even under stringent biosecurity conditions (Wyres & Holt, 2018). The borderline association for udder health and hygiene practices score further suggests that management intensity rather than awareness alone drives MDR K. pneumoniae dynamics. These findings collectively highlight that biosecurity interventions must be coupled with robust antimicrobial stewardship programs to effectively reduce MDR pathogen burden on dairy farms. Overall, the findings suggest that improved knowledge and perceived practices alone may not be sufficient to reduce the occurrence of environmental mastitis pathogens without effective implementation and structural support. Future research using larger sample sizes, longitudinal study designs, and objective measurement of farm practices is recommended to better understand these relationships. Conclusion The study revealed a high prevalence of E. coli and K. pneumoniae in dairy farms, with considerable multidrug resistance, particularly against commonly used antibiotics such as ampicillin, tetracycline, ceftriaxone, and amoxicillin-clavulanic acid. The findings further demonstrated that higher KAP score related to udder health, hygiene, and mastitis, as well as higher biosecurity scores were not consistently associated with lower occurrence of E. coli and K. pneumoniae. Several positive associations were observed with E. coli positivity, concurrent occurrence of both pathogens, and MDR K. pneumoniae, suggesting that knowledge and self-reported practices alone may be insufficient to reduce environmental mastitis pathogens at farm level without parallel antimicrobial stewardship and integrated herd health management. For treatment of mastitis caused by Gram-negative bacteria, antimicrobial therapy should be guided by antimicrobial susceptibility testing; based on the present findings, piperacillin-tazobactam followed by ciprofloxacin may be considered where justified and in accordance with national treatment guidelines. Future studies with large sample sizes and longitudinal designs are recommended to better clarify causal relationships and the gap between reported practices and actual farm conditions. Data availability The datasets generated and analyzed during the current study are available from the corresponding author on reasonable request. Ethical statement Before data collection, the dairy farmers involved in the study were informed about the objectives and procedures to ensure their voluntary participation. Verbal consent was obtained from farm owners or animal attendants prior to conducting the interviews. Milk samples were collected in accordance with the Animal Welfare Guidelines, ensuring no harm to the animals. The study protocol was reviewed and approved by the Animal Welfare and Experimentation Ethics Committee of Bangladesh Agricultural University, Mymensingh (AWEEC/BAU/2023(5), dated 04/03/2023). Funding The authors acknowledge the partial financial support of the Livestock and Dairy Development Project (LDDP) [RP-C-01–11] of the Department of Livestock Services, Bangladesh. CRediT authorship contribution statement Swarnali Akter: Writing – original draft, Methodology, Investigation, Formal analysis. Sadia Tasnuva: Writing – review & editing, Resources, Methodology. Monika Akter Runa: Writing – review & editing, Resources, Methodology. Md. Mamunur Rahman: Writing – review & editing, Resources, Methodology. Ataur Rahman: Writing – review & editing, Resources, Methodology. Ashiqul Islam: Writing – review & editing, Resources, Methodology. Md. Monirul Islam: Writing – review & editing, Resources, Methodology. Sayra Tasnin Sharmy: Writing – review & editing, Resources, Methodology. Mst. Sonia Parvin: Writing – review & editing, Investigation, Formal analysis. Md. Taohidul Islam: Writing – review & editing, Visualization, Validation, Supervision, Software, Project administration, Funding acquisition, Formal analysis, Data curation, Conceptualization. Declaration of competing interest The authors declare no competing interests. Acknowledgment The authors acknowledge the cooperation of the dairy farmers of the respective areas during this study. References Associated Data Data Availability Statement The datasets generated and analyzed during the current study are available from the corresponding author on reasonable request.