Effect of a dairy farmworker stewardship training program on antimicrobial drug usage in dairy cows

Antimicrobial use (AMU) is critical to preserving animal health and welfare. However, the development of antimicrobial resistance (AMR) represents a public health threat. Although most antimicrobials used on the farm require a veterinarian prescription, farmwork-ers make daily on-farm treatment decisions. Therefore, farmworker training is vital to promote responsible AMU. This research project aimed to evaluate the impact of farmworker antimicrobial stewardship (AMS) training on the quantity of AMU on dairy farms in Ohio and California. We hypothesized that farms receiving AMS training would have reduced AMU in dairy cattle compared with farms where training wasn’t administered. We designed a quasi-experimental study with 18 conventional dairy farms enrolled in Ohio and California. Twelve farms received AMS training and 6 farms did not. AMS program included a 12-week training focused on accurate identification of cows requiring antimicrobial treatment. We quantified AMU by collecting used drug containers, manually counted by researchers during farm visits. Treatment incidence (TI) was calculated using animal daily-doses (ADD), and Poisson regression models were used to analyze the TI data. Disease incidence rate (DIR) in lactating cows was 2.2, 1.5, 1.0, 0.4, 0.3, 0.2, and 0.03/1,000 cow-days for mastitis, lameness, metritis, pneumonia, retained placenta, diarrheas, and other diseases (e.g., conjunctivitis, injuries), respectively. The highest TI by antimicrobial class was cephalosporin (5.9), penicillin (5.2), tetracyclines (0.4), lincosamides (0.2), and sulfonamides (0.1 ADD/1,000 cow-days). Among the trained farms using a within-treatment group analysis, no significant differences were observed in TI pre-intervention (10.9) compared with TI post-intervention (10.3) ADD/1,000 cow-days. Treatment incidence from the training group was (10.8) and although numerically lower, it was not significantly different compared with TI in the control groups at (13.9) ADD/1,000 cow-days (rate ratio = 0.77, CIs = 0.25–2.38). Future research on AMS should incorporate farmworker training with social science approaches to overcome barriers and promote on-farm responsible use.


INTRODUCTION
Antimicrobial drug use in food animals has been advocated to preserve animal health and welfare and to improve animal productivity and food safety (Coombe et al., 2019).However, the use of antimicrobials represents a public health threat by increasing antimicrobial resistance (AMR) (Cattaneo et al., 2009).Hence, the use of antimicrobials in food animal production is scrutinized based on the risk of selection of drug-resistant bacteria and the potential for dissemination to human populations through food products of animal origin (Brault et al., 2019).Therefore, improving and promoting the appropriate use of antimicrobial drugs has become a top priority in securing their effectiveness for use in both animal and human medicine.
Given the importance of this shared resource, securing a sustainable and robust US livestock system will require improvement of antimicrobial stewardship (AMS) in food animals.Some important steps toward achieving this goal include providing farmworkers with knowledge and training to increase their ability to accurately and effectively identify and treat animals.Previously, implementation of a training program in a calf production system focused on accurate identification resulted in significant reductions in AMU (Pempek et al., 2022).Additionally, benchmarking can be a useful tool in AMS programs.Through benchmarking, dairy producers evaluate their present methods in comparison to similar dairy farms to find possible areas of improvement (von Keyserlingk et al., 2012).This approach could provide dairy producers with insightful feedback and motivate judicious practices to reduce antimicrobial use (AMU) at their farms (Mills et al., 2018;Schrag et al., 2022).The combination of well-designed stewardship training and feedback using benchmark reports may be an effective stewardship strategy to improve responsible on-farm use of antimicrobial drugs.
Data related to the use of on-farm antimicrobials are necessary to understand current treatment practices and to promote the responsible use of these drugs at the farm level.However, quantification of antimicrobials is labor intensive, time-consuming, and requires a serious level of commitment from the farm personnel (Lardé et al., 2021).Additionally, the lack of standardization and interpretation of the metrics makes the task of tracking on-farm AMU difficult (Lardé et al., 2020).Efforts to evaluate antimicrobial usage at the farm level in the US are limited, particularly on large dairy farms.Some of the published studies include work done in Pennsylvania (Redding et al., 2019;Sawant et al., 2005), and Wisconsin farms (de Campos et al., 2021;Pol & Ruegg, 2007).However, on-farm AMU has been extensively described in other parts of the world, including Canada (Brault et al., 2019;Nobrega et al., 2017;Saini et al., 2012), Switzerland (Menéndez González et al., 2010), Belgium (Stevens et al., 2016), Austria (Firth et al., 2017), and the United Kingdom (Hyde et al., 2017;Rees et al., 2021).The first objective of this study was to describe and quantify AMU across large dairy farms located in Ohio and California.The second objective was to evaluate the impact of farmworker AMS training and benchmarking through the quantification of on-farm AMU.We hypothesized that dairy farms that received the AMS educational training (intervention group) would have reduced on-farm AMU compared with the farms in the control group not receiving the proposed training.

MATERIALS AND METHODS
All the materials and study procedures were approved by The Ohio State University Institutional Review Board (#2019B047).The study was conducted from September 2020 to March 2022.
Using 18 large conventional dairy farms in Ohio and California, we designed a quasi-experimental field trial to quantify on-farm AMU in adult dairy cows.Quasiexperimental field studies are designed to assess interventions without using randomization of the participant subjects (Maciejewski, 2020).The original sample size estimation for AMU calculations was conducted to find a significant change between 2-time points for a group of farms.Six farms would provide sufficient power (1-β = 0.83) to demonstrate a significant difference (α = 0.05) between 2 time points for the sampled farms (12 observations) assuming 1) a mean TI of 0.3 and 0.15 doses per cow per day at each time point, 2) a standard deviation of 0.10, and 3) a mean correlation of > 0.5 for the TI between within-herd observations (SAS, v. 9.4, Cary, NC) (Pempek et al., 2022).Farms that were willing to participate in this study were selected if they were large conventional dairy farms (i.e., not producing organically certified milk with > 500 lactating cows), had accessible dairy herd management software, and were within 4 h distance of either Columbus, OH or Davis, CA.Farmer owners were initially contacted through phone calls based on a pre-existing relationship with members of the research team and were asked to nominate other farm owners to be later contacted by the research team.If the farm owner showed interest, the research team met with the farm owner and the farmworkers to explain the purpose and benefits of the study.The farmworkers that voluntarily agreed to participate were enrolled in the study.A total of 26 farms in Ohio or California were originally contacted by the research team.A total of 8 farms declined participation for various reasons.Twelve farms (6 in each state) were allocated to receive the educational intervention (intervention group) based on their willingness to participate in the didactic training offered by the research team.Farmworkers from the intervention group who provided antimicrobial treatments on a regular basis were invited to participate in a 12-week training program focused on the accurate identification of cows requiring antimicrobial therapy (Garzon et al., 2023).During the 12-week training program, farmworkers and a research team member (RPG in Ohio and AG in California) met at the farm every other week for approximately 60 min per session to discuss the AMS educational material (Figure 1).The rest of the farms (control group) did not receive any kind of training throughout the duration of the study (180 d).This group of farms was selected based on their availability and willingness to be assigned to this group.An owner offered to enroll 2 farms; therefore, one farm was placed in the intervention group and the other in the control group.Another farm did not have a complete team of farmworkers at the time of the intervention, so they were willing to receive the training afterward.Finally, the rest of the farms did not mind being placed in this group and receiving stewardship training after the AMU collection period was completed.
Additionally, we manually quantified on-farm AMU using the inventory of empty drug containers (EDC) method before (30 d), during (60 d), and after (90 d) Portillo-Gonzalez et al.: Antimicrobial quantification in adult dairy cattle the implementation of the educational training (Figure 1).The EDC was the only method applied by the research team to collect antimicrobials used from the enrolled dairy farms.We located bins in places convenient to collect empty pharmaceutical containers (i.e., drug rooms, milking parlors, hospital pens, etc.).Farmworkers were instructed to discard all empty antimicrobial drug containers used in adult dairy cattle inside the bins.After the first 12 weeks (when the training program was completed), farm visits were made once a month to continue quantifying on-farm AMU for the next 3 mo (Figure 1).The quantification of on-farm AMU lasted for a period of 6 mo per farm.Each empty antimicrobial drug container was counted and recorded on a drug tally sheet and later entered into a spreadsheet.

Educational Materials and Stewardship Intervention
We designed a farmworker-oriented AMS educational intervention that contained benchmark reports, summary tables, a hands-on session, and 6 didactic prerecorded modules in English and Spanish (Garzon et al., 2023).The pre-recorded modules include pictures and videos of clinical case vignettes.The first 2 modules explain basic concepts such as antibiotic resistance, selective pressure, extra-label drug use, and how to interpret treatment protocols with a disease-scoring system to make therapeutic decisions.The third module explained the importance of identification of high-risk animals, and how to examine sick cows.The last 3 modules explained in detail how to recognize early disease onset, common clinical signs, differentiation of disease severity levels, and how to make accurate therapeutic decisions for cases of metritis, mastitis, and lameness in adult dairy cattle (Table 1).Along with the training material, we created summary tables to highlight the most relevant aspects of each module through pictures and charts.This information was printed, laminated, and kept in drug rooms or hospital pens as a quick reference guide.Live animal sessions that included the participation of trained farmworkers and the researcher delivering the AMS training were additionally held to exchange experiences, improve engagement, and increase knowledge retention.Finally, benchmark reports were made to measure, evaluate, and compare on-farm antimicrobial usage to gradually improve farm performance (Flor et al., 2022).Benchmark reports showing farm treatment incidence were estimated at the farm level on a 30-d interval at 3 different time points (before, during, and after the educational intervention was completed) (Figure 1).The disease incidence rate (DIR) for the entire study period was calculated from 13 farms that had complete on-farm disease records.DIR was defined as the number of health-recorded events found within the on-farm treatment records for the enrolled dairy farms during the intervention period (6 mo).We extracted all the health event records from the dairy management software during the intervention time.The health and/or treatment events included mastitis, metritis, interdigital necrobacillosis, retained placenta, diarrhea, pneumonia, and others (e.g., injuries, conjunctivitis, clostridial disease).Thus, all the events per farm were divided by the total cow-days at risk during the same period.Cow-days (denominator) at risk were estimated based on the number of adult cows indicated as present in the herd (discounted culled cow days) by the dairy management software within the 180-d collection period.Antimicrobial use data were calculated using the animal daily dose (ADD) metric (Jensen et al., 2004).
The maximum AMU dosage defined on the label was used for each active ingredient for the ADD calculation.For a combination of active ingredients (e.g., Penicillin G procaine and novobiocin), we used the substance in higher concentration to calculate ADD (World Health Organization, 2022).Following the guidelines of the European Medicines Agency for intramammary (IMM) products, one IMM application represents 1 ADD for lactating-cow therapy and 4 IMM applications represent 1 ADD for dry-cow therapy (Medicines Agency, 2015).The ADD numerator was obtained as the product of the number of units administered to the animal (e.g., mg, mL) multiplied by the concentration of the active ingredient.The ADD denominator was obtained by multiplying the on-label recommended dose by 600 kg of standard weight per adult cow.We obtained the somatic cell count (SCC) information per farm from the dairy management software.The SCC records were transformed to somatic cell score (SCS) (Čítek et al., 2022).

Statistical Analysis
AMU data were transferred from the drug tally sheet to an Excel spreadsheet (Microsoft Excel, Microsoft Corp., Redmond, WA), and later imported to SAS (SAS®, version 9.4 SAS Institute Inc., Cary, NC).All the AMU data entered were visually evaluated for potential entry errors.We summarized the data through descriptive statistics to obtain the mean TI by antimicrobial class, route of administration, and changes over time and relative to the intervention.Within the analysis, we considered the farm as the experimental unit, while the unit of observation was calculated as 3 consecutive farm-level measurements of TI.A Poisson regression mixed model was used to analyze the AMU data.The dependent variable was the TI (ADD/1,000 cow-days), and the independent variables included type of farm (intervention or control farms), state (Ohio or California), and time (pre, during, and post-intervention).Additionally, an interaction term (type of farms × time) was kept in the model.Fixed effects were tested using Student's t-test with degrees of freedom estimated using the Kenward-Roger method.The lack of independence of the data was addressed by including a random residual effect for farms (Schabenberger, 2005).The incidence rate ratio and confidence intervals for the TI for the intervention group relative to the control group were calculated.

RESULTS
Eighteen farms were enrolled in the study.Among the 12 farms allocated to the intervention, 31 farmworkers were initially enrolled in the training program.Six farmworkers were excluded from the study after missing one or more training sessions.Therefore, training information was collected from 25 participants who attended all the training sessions voluntarily.At least one farmworker from each enrolled farm remained for the duration of the study.Thirteen farmworkers were enrolled in Ohio and 12 in California for an average of 2 participants per enrolled dairy farm (range: 1-5).Twenty-two males and 3 females represent 88% and 12% of participants, respectively.Seventeen (68%) of the participants chose Spanish knowledge assessments, and 8 (32%) chose English assessments.

Herd Characteristics
The 18 enrolled dairy farms were distributed equally across Ohio (9 farms) and California (9 farms).In Ohio, farms were in the northwest (n = 4), northeast (n = 3), and southwest (n = 2) of the state.In California, farms were in northern California (n = 6) and in the Central Valley region (n = 3).Together, the enrolled farms during the entire study period (180 d) housed 42,386 lactating cows, with an average of 2,355 lactating cows per farm (range: 631-6,100) (Table 2).The mean number of lactating cows enrolled in Ohio and California farms was 2,136 (range: 631 -5,579) and 2,573 (range: 791 -6,100), respectively.All the farms enrolled in the study used free-stalls and milked twice per day (Table 2).The farms housed either Holstein-Friesian cows, Holstein and Jersey cross, or Jersey cattle (Table 2).The mean milk production from 15 farms with available milk production records was 11,170 kg/cow per year (range: 8,893 -13,926 kg/cow per year; n = 15); with a mean bulk milk somatic cell score SCS of 4.1 (range: 3.0 -5.0) (Table 2).
The impact of the training on the knowledge of the farmworkers has been previously reported (Garzon et al., 2023).Briefly, there was a significant knowledge increase among trained farmworkers obtained through the AMS 12-week training program.This knowledge improvement represented a 32% increase in knowledge transferred among trained farmworkers after the AMS intervention was completed (Garzon et al., 2023).

Description of Antimicrobial Use in Ohio and California Dairy Farms
The first objective of this manuscript was to describe antimicrobial use on large dairy farms in Ohio and California.The overall mean TI across all farms was 11.8, ranging from 1.7 to 71.6 ADD/1,000 cow-days, with a large AMU variation among all farms (Figure 2).The mean (median) TI by the Ohio farms was 15.3 (6.4) compared with 8.3 (6.9) ADD/1,000 cow-days of the California farms (Figure 3).The Poisson regression mixed model showed that the difference in the mean TI for the entire study period between states was not statistically significant (P = 0.22).However, specifically for parenteral administration, there was a significant difference between the states (Ohio and California) (P = 0.05).The ADD for the California farms was 0.19 times lower than the ADD for the Ohio farms.
The enrolled dairy farms across Ohio and California used 10 different active ingredients.Seven active ingredients (6 β-lactams and one lincosamide) were administered IMM to treat clinical and subclinical cases of mastitis in lactating and dry cows, respectively.The rest were used by parenteral and/or intrauterine administration.All the enrolled dairy farms used IMM antimicrobials, and 17 farms used parenteral antimicrobials.No oral administration of antimicrobials was reported by the enrolled dairy farms.Ohio farms' mean TI for parenteral antimicrobial use was significantly higher (7.6 ADD/1,000 cow-days) compared with the California farms (1.8 ADD/1,000 cow-days) (P = 0.05).For IMM administration both groups of farms showed similar results.For lactating cow therapy, Ohio farms mean TI was (5.6 ADD/1,000 cow-days) and for the California farms was (5.3 ADD/1,000 cow-days).Likewise, for dry-cow therapy, Ohio farms showed a mean TI of (1.9 ADD/1,000 cow-days) and the California group showed (1.3 ADD/1,000 cow-days).California farms did not report intrauterine antimicrobial use (Table 3).Additionally, antimicrobial classes such as penicillins, cephalosporins, tetracyclines, and sulfonamides showed a mean TI of (7.5, 6.8, 0.6, and 0.1 ADD/1,000 cow-days) for the Ohio farms and (3.0, 4.8, 0.3, and 0.01 ADD/1,000 cow-days) for the California farms, respectively (Table 3).
Among the cephalosporin class, ceftiofur was the most common active ingredient used by the enrolled dairy farms.It was used in 16 farms as a parenteral  antimicrobial (1.3 ADD/1,000 cow-days), in 14 farms to treat clinical cases of mastitis in lactating cows (2.9 ADD/1,000 cow-days), and in 7 farms as a dry-off therapy (0.5 ADD/1,000 cow-days).Within the penicillin class, penicillin G procaine was the most frequent active ingredient used within this group of drugs.Penicillin G procaine was used in 10 farms as a parenteral antimicrobial (2.3 ADD/1,000 cow-days), in 1 farm to treat clinical cases of mastitis (0.2 ADD/1,000 cow-days), and in 4 farms to treat subclinical cases of mastitis at dry-off (0.2 ADD/1,000 cow-days).For the tetracycline class, oxytetracycline was used in 10 farms as a parenteral antimicrobial (0.3 ADD/1,000 cow-days), and in 2 farms intrauterine (0.1 ADD/1,000 cow-days).
Regarding the sulfonamides class, sulfadimethoxine was used parenterally in 2 farms (0.1 ADD/1,000 cowdays).Finally, among the lincosamides class, pirlimycin was used in 4 farms to treat clinical cases of mastitis in lactating cows (0.2 ADD/1,000 cow-days) (Table 4).By route of administration, the highest mean TI was obtained by the IMM application 7.1 (5.5 lactating cows and 1.6 dry-cow therapy), followed by parenteral 4.6, and intrauterine 0.1 mean ADD/1,000 cow days.For IMM administration, ceftiofur showed the highest mean TI (2.9 ADD/1,000 cow-days) among the 7 different active ingredients used to treat clinical cases of mastitis in lactating dairy cows.However, for treating subclinical cases of mastitis at the time of dry-off, cloxacillin was the active ingredient that showed the highest mean TI (0.9 ADD/1,000 cow-days) among the 3 different therapeutic substances used during the dry period.For parenteral administration penicillin G procaine showed the highest mean TI (2.3 ADD/1,000 cow-days) among 5 different active ingredients used by the enrolled dairy farms.Finally, only 2 farms within the Ohio group exhibited the use of intrauterine boluses of oxytetracycline (0.1 ADD/1,000 cow-days) to treat cases of post-calving metritis (Table 4).

Impact of Antimicrobial Stewardship Training on Quantified Antimicrobial Use
The second objective of the manuscript was to determine the impact of antimicrobial stewardship training on the quantification of antimicrobials used at the farm.
The farms in the control group showed a mean TI (13.9 ADD/1,000 cow-days) compared with the intervention group (10.8 ADD/1,000 cow-days); (rate ratio = 0.77; CI: 0.25-2.38)for the entire study period (180 d) (Figure 4).For parenteral use, the intervention farms showed a mean TI (3.9 ADD/1,000 cow-days) compared with the control farms (6.0 ADD/1,000 cowdays).Similarly, for lactation therapy, the intervention farms showed a mean TI (5.1 ADD/1,000 cow-days) compared with the control farms (6.1 ADD/1,000 cowdays).Finally, for dry cow therapy and intrauterine use, the mean TI for the intervention farms was (1.6 and 0.2 ADD/1,000 cow-days) compared with the control farms (1.8 and 0.01 ADD/1,000 cow-days), respectively.The Poisson regression mixed model showed that the difference in the mean TI between the group (intervention and control) for the whole study period was not statistically significant (P = 0.63).
The Ohio farms estimate (15.3 ADD/1,000 cowdays), although not significantly different were almost twice as large as the California farms (8.3 ADD/1,000 cow-days).However, the Poisson regression mixed model showed that for parenteral antimicrobial administration only, there was a statistically significant difference between the states (P = 0.05).Additionally, no significant differences in parenteral administration were found between the farm group (intervention and control) or before, during, and after the intervention was completed.
The mean TI for the farms in the post-intervention group was 10.3 (range = 1.7 -47.1 ADD/1,000 cowdays) compared with the mean TI pre-intervention 10.1 (range = 1.7 −60.3 ADD/1,000 cow-days) and the mean TI during-intervention 11.8 ADD/1,000 cowdays (range = 1.7 -71.6 ADD/1,000 cow-days) (Figure 5).However, the Poisson regression mixed model showed that the mean TI post-intervention was not significantly different compared with the mean TI preintervention (P = 0.96, CI: −0.36 -0.34) or the mean TI during-intervention (P = 0.45, CI: −0.21 −0.48) (Figure 5).Thus, only 3 of the 12 intervention farms showed a lower mean TI post-intervention compared with the mean TI pre and during-intervention.Additionally, among 4 different antimicrobial classes used by the intervention farms (n = 12), only the penicillin class showed a numerically lower mean TI after the intervention was completed.Antimicrobial classes such as cephalosporins, tetracyclines, and lincosamides did not show a reduction in the mean TI post-intervention.The sulfonamides class was not used by the intervention farms (Table 5).

DISCUSSION
On-farm antimicrobial use quantification is a fundamental first step to evaluate the implications that adopting AMS practices may have on AMU on dairy farms.However, the quantification of AMU is challenging and time-consuming.Thus, few studies have quantified on-farm AMU to assess the efficacy of educational AMS programs.Therefore, this research project provides novel information on AMU on large dairy farms in Ohio and California, documents the impact of farmworker AMS training on those metrics, and provides a future model for testing antimicrobial stewardship interventions.
The dairy farms enrolled in this study between Ohio and California housed a larger number of lactating cows per farm compared with the average per farm found in a similar manuscript that included 40 large dairy farms in Wisconsin (de Campos et al., 2021).Additionally, the average herd somatic cell score for the enrolled dairy farms was slightly higher compared with the average SCS previously reported by the New York Quality Milk Production Services (QMPS) using composite milk samples from commercial dairy herds (Shook et al., 2017).
In our study, we quantified AMU based on the empty drug container inventory method only.This method consists of collecting discarded packaging (infusion tubes, bottles, bags, boxes) of antimicrobials that were used on adult dairy cattle (Redding et al., 2014).The inventory of empty drug containers is considered a reliable method for gathering AMU data at the farm level, especially when compared with farm drug records alone (Carson et al., 2008;Saini et al., 2012;Stevens et al., Portillo-Gonzalez et al.: Antimicrobial quantification in adult dairy cattle Mean TI: Mean treatment incidence (ADD/1,000 cow-days).
2016).Using ADD as the metric, Nobrega et al., 2017 reported that the inventory of drugs collection method provided a better AMU estimation compared with onfarm treatment records on Canadian dairy farms (Nobrega et al., 2017).Similarly, using defined course dose (DCD), the total AMU rate was significantly lower and less accurate when using on-farm treatment records and governmental database compared with empty drug container inventory in dairy farms from Quebec, Canada (Lardé et al., 2021).Therefore, finding reliable ways to measure AMU data at the farm level is a critical step to understanding the relationship between the use of on-farm antimicrobials and the development and transmission of resistant bacteria (Lardé et al., 2020).We selected the dose-based metric called ADD (animal daily doses) to calculate on-farm AMU.ADD has been described as a unit of measurement to report the average dose per day of the antimicrobial used in animals (Grave et al., 1999).This metric has been widely used for the quantification of AMU in veterinary medicine; therefore, results can be easily comparable to what has been previously published in the literature (Humphry et al., 2021;Mills et al., 2018).
Cephalosporins, followed by penicillins, were the most common antimicrobial classes used by the enrolled   dairy farms.Similar findings using ADD as the selected metric were previously reported in other research studies performed in Canada and the US (de Campos et al., 2021;Saini et al., 2012).Within these classes, ceftiofur and penicillin were the most frequently used active ingredients by the enrolled dairy farms.They were applied by 2 different routes of administration, parenteral and IMM, in 17 and 11 farms, respectively.Together, their use represents 94% (11.1/11.8) of the overall TI used in adult dairy cows.Data from previous research have already reported ceftiofur among the most common antimicrobials used to treat health conditions in adult dairy cows (Pol & Ruegg, 2007;Stevens et al., 2016).Common reasons for the frequent use of this drug are due to the extended-spectrum, variable options for frequency of treatment depending on the indication and drug formulation (single use, 24-h intervals or 72 h.intervals), and no milk withholding for the parenteral products when used as per label (Durel et al., 2019).The penicillin class was represented by natural penicillins (penicillin G), aminopenicillins (ampicillin, amoxicillin, and hetacillin), and cloxacillin.Penicillin is a bactericidal, time-dependent β-lactam antibiotic that is frequently used in dairy farms because of its lower price, efficacy against Gram-positive bacteria, and over-the-counter sales status.However, it is important to mention that since June of 2023 throughout the US, all medically important antimicrobials currently available over the counter require veterinary oversight (medical prescriptions) to be used in animals.The data collection in Ohio occurred before the implementation of this intervention.A similar rule was previously implemented in 2018 and in place at the time of data collection for California herds.
The Ohio farms estimate, although not significantly different, were almost twice as large as the California farms.All the antimicrobial classes except for lincosamides, showed a numerically higher mean TI for all routes (parenteral, intrauterine, and IMM for lactating and dry-cow therapies) in the Ohio farms compared with their counterpart in California.Although the observed difference in the mean TI was meaningful, a larger study would be necessary to confirm the result since the TI was not significantly different between states (P = 0.22).It is important to mention the presence of an influential observation in the Ohio data set that might have affected the mean TI in this group of farms.However, the observation might represent a natural variation of antimicrobial use in this farm, since this farm also had the highest disease incidence for diarrhea, retained placenta, and metritis and the second highest for mastitis.When the influential observation was removed from the data set, the overall mean TI  The mean TI from the Ohio farms agrees with findings from Wisconsin (15.8 mean ADD/1,000 cow-days; (de Campos et al., 2021) and Canada (14.3 mean ADD/1,000 cow-days; (Saini et al., 2012).However, using a self-reported survey, the TI on 235 dairy farms in Pennsylvania was lower (4.2 mean ADD/1,000 cowdays) (Redding et al., 2019).Several factors were at play that could have explained the relatively lower mean TI exhibited by the Pennsylvania farms.First, the presence of a large amount of Mennonite and Amish farmers in this state, who have been previously identified to use antimicrobials less often (Schewe & Brock, 2018).Second, the lack of farm treatment records along with insufficient recall of treatment instances from the self-surveyed farmers might have also played a role in the low rate of TI found in this study (Redding et al., 2019).
A numerical change was observed in the mean TI of the trained farms after the intervention was completed.The numerical change in TI shown by the farms in the intervention group may be in part explained by the knowledge transferred obtained through the AMS 12-week training program.During this educational program, farmworkers experienced a significant knowledge increase after the training was completed (Garzon et al., 2023).However, despite the improvement in knowledge transferred among the trained farms, there was no significant reduction in on-farm AMU shown by the intervention farms compared with the control farms (P = 0.99).Since our AMS intervention was focused on the accurate identification of animals that require antimicrobial therapy and intentionally did not suggest changes in on-farm treatment protocols, farmworkers may have continued to adhere to their existent on-farm treatment routines.Additionally, the significant variability in treatment accuracy among the enrolled dairy farms might explain the nonsignificant effects.Some farms may have had less sensitive case definitions, and the intervention may have resulted in improved sensitivity and appropriately higher AMU.Conversely, the farms that used a higher level of antimicrobials and with less specific case definitions could have had reduced AMU as a result of the intervention.Therefore, it is possible that the opposing results on farms with different levels of diagnostic sensitivity resulted in smaller magnitude of change over time.Previous research in a calf production system directly measured antimicrobial treatment accuracy of farmworkers relative to the farm veterinarian (Pempek et al., 2022), however, this measurement was not possible since we did not assess farmworkers' adherence to farm-written treatment protocols.Future studies should use a multipronged approach to measure changes in AMU accuracy on farms.

Limitations
The farms enrolled in this study were not randomly selected.Instead, the selection was based on convenience sampling following specific inclusion criteria designed by the research team.Since we heavily rely on producers' participation, the interest expressed by them to be part of this study was a key element for inclusion.For this reason, we might have incurred selection bias based on the lack of randomization of the target population in both states.To quantify on-farm AMU we used 600 kg of standard weight per adult cow.This might have caused an underestimation of TI for heavier animals and the opposite for lighter ones.Nevertheless, the use of the standard weight is a common practice when there are no records available of the actual animal weight at the farm level (Mills et al., 2018).The EDC inventory as a method of collection on-farm AMU has its limitations since it was unlikely to have collected every one of the empty antimicrobial packages and some partially used antimicrobial bottles were most probably kept on the drug room shelves.Therefore, the lack of this data might have created an underestimation of the AMU TI at the farm level.Finally, no standardized protocols or guidelines were used for disease identification at the farm level.This issue must have led to inconsistency in the data obtained from the on-farm treatment records.Therefore, careful considerations should be applied when interpreting the results obtained from the disease incidence rate.

CONCLUSIONS
This study estimated the TI for 18 conventional dairy farms located in 2 different states Ohio and California.

Portillo-Gonzalez et al.: Antimicrobial quantification in adult dairy cattle
There was substantial variation in the number of antimicrobials used by the enrolled dairy farms.Mastitis was the main health condition for on-farm AMU, and cephalosporins followed by penicillin were the most frequently used class of antimicrobials by the enrolled dairy farms.Even though not significantly different, the Ohio farms estimates were almost twice as large as the estimates obtained from the farms in California.Additionally, no significant change in on-farm AMU was shown by the intervention farms compared with the control farms.Training for farmworkers to improve disease detection accuracy is a necessary component of AMS, but sustained progress will require a holistic approach.Additionally, the inclusion of psycho-social models that study the interaction between farmworkers' beliefs, societal-level factors, and environmental elements could help to shape human behavior and improve the responsible use of antimicrobials at the farm level.

Figure 1 .
Figure 1.Timeline for antimicrobial use quantification and antimicrobial stewardship training sessions.The educational component was delivered within the first 4 sessions.Meanwhile, the quantification of antimicrobial use and benchmark reports were performed for all the farms (intervention and controls) before, during, and after the educational intervention was completed.
of antibiotic used × drug concentration n dose mg/kg × standard animal weight( ) ( )Treatment incidence (TI) was estimated through the ADD used per adult cow divided by the total number of cow days at risk (discounted culled cow days) within 30-d intervals and expressed by 1,000 cow-days.Cow-Portillo-Gonzalez et al.: Antimicrobial quantification in adult dairy cattle Table 1.Learning objectives that described the content discussed with farmworkers per training session during the educational intervention.The educational sessions were held at the farm premises once every other week for approximately 60 min Training modules Learning objectives Antibiotic resistance Recognize the existence of good and bad bacteria within the animal Identify common diseases on dairy farms caused by bacteria and viruses Define antibiotics, and recognize that antibiotics are only effective against bacteria Give examples of antibiotics, vaccines, anti-inflammatories, antiparasitics, and hormones Describe the concept of "antibiotic resistance" and "selective pressure" Give examples of how antibiotic resistance spreads from animals to humans Treatment protocols Describe the importance of having treatment protocols Describe the components of treatment protocols Interpret the different components of pharmaceutical drug labels Define "Extra-Label Drug Use (ELDU)" Interpret a protocol that has disease scoring system to make a therapeutic decision Clinical mastitis Define clinical mastitis Identify cows with signs of clinical mastitis Recognize the importance of collecting milk samples appropriately for bacteriological culture Distinguish between different severities of clinical cases of mastitis Determined when antimicrobials are needed Recognize the importance of fluids and anti-inflammatory drugs Recognize non-antimicrobial management options Give examples of clinical cases of mastitis using different levels of disease severity Metritis post-calving Describe what normally occurs after calving Define a clinical case of post-calving metritis Identify an abnormal vaginal discharge to differentiate a cow with metritis Recognize the clinical signs associated with metritis Determined when antimicrobials are needed Give examples of clinical cases of post-calving metritis using different levels of disease severity Lameness Identify lame cows Distinguish different levels of disease severity Distinguish non-bacterial causes from bacterial causes of lameness Distinguish different treatment strategies for non-bacterial and bacterial causes of lameness Give examples of clinical cases of lameness using different levels of disease severitydays (denominator) at risk were estimated based on the number of adult cows indicated as present in the herd (discounted culled cow days) by the dairy management software within the 30-d collection period.
Portillo-Gonzalez et al.: Antimicrobial quantification in adult dairy cattle Portillo-Gonzalez et al.: Antimicrobial quantification in adult dairy cattle Figure 2. Mean Treatment incidence (animal daily dose ADD/1,000 cow-days) by farm.The mean treatment incidence included intervention and control farms in both states (OH and CA) for the whole study period (180 d).The overall mean treatment incidence for all the farms (n = 18) was included as a reference.The black bar represents the standard deviation of the mean.

Figure 3 .
Figure 3. Mean treatment incidence (animal daily dose ADD/1,000 cow-days) by state (Ohio n = 9 and California n = 9).The mean treatment incidence included control and intervention farms in both states for the whole study period (180 d).X = mean treatment incidence (Ohio = 15.3 and California = 8.3) Solid horizontal line = median (Ohio = 6.4 and California = 6.9)Top of box = third quartile (Ohio = 23.7 and California = 12.2) Bottom of the box = first quartile (Ohio = 4.5 and California = 5.4) Solid dot = Outlier (within the Ohio farms only) Portillo-Gonzalez et al.: Antimicrobial quantification in adult dairy cattle

Table 2 .
Farm characteristics and production parameters by the enrolled dairy farms (n = 18).Farms were enrolled from September 2020 to March 2022, and 12 farms received antimicrobial stewardship training materials ("Training")

Table 4 .
Treatment incidence (animal daily doses ADD/1,000 cow-days) by antimicrobial classes and route of administration from 18 conventional dairy farms in Ohio (n = 9) and California (n = 9)

Table 5 .
(Abdelfattah et al., 2022)imicrobial class before, during, and after the implementation of the antimicrobial stewardship (AMS) the Ohio farms was 9.6 ADD/1,000 cow-days, which was similar compared with the California farms' 8.3 ADD/1,000 cow-days.Additionally, differences in state regulations may have influenced our TI findings; in California Senate Bill 27 (SB 27) passed in January 2018 (Food and Agricultural Code, 2015) made California the first state to require veterinary drug prescriptions of "medically important antimicrobials (MIA)" used in livestock operations.SB 27 banned all MIA ("critically important, highly important, and important antimicrobials") to be sold over the counter (OTC) for livestock.Furthermore, SB 27 requires creating and implementing AMS guidelines, overseeing husbandry practices, and tracking antimicrobial sales to provide feedback to farm owners and veterinarians(Abdelfattah et al., 2022).There was no similar bill or law in the state of Ohio by the time the antimicrobial use information was collected (from September 2020 to March 2022).