Choosing which metrics to use when reporting antimicrobial use information to veterinarians in the Canadian swine industry

Abstract

The objective of this study was to evaluate preferences for various metrics and denominators among Canadian swine veterinarians, in order to improve reporting of antimicrobial use (AMU) information to these stakeholders and to facilitate enhanced stewardship decisions. An online survey was made available to swine veterinarians across Canada; 12 responses (estimated response rate 17.6%) were submitted and analyzed. Responses represented veterinarians from every major pig-producing province and from a range of year of graduation from veterinary college. Participants self-evaluated their understanding of dose-based metrics as higher than weight- and frequency-based metrics and interpreted most results of AMU analyses correctly. Participants preferred dose-based metrics over others, and had various objectives for AMU information, including improving AMU on their clients’ farms and enabling comparisons with other farms. The results are useful to those making decisions about which AMU metrics to use in reports targeted to swine veterinarians.

Introduction

Growing public health concern about the development of antimicrobial resistance (AMR) resulting from the use of antimicrobials in food animals, including pigs, has led to an increasing focus on the collection and analysis of antimicrobial use (AMU) information at the farm-level (1–3). Veterinarians can use the results of farm-level AMU analyses for many purposes. These include evaluating their prescribing and use practices in comparisons among herds or with the industry average, and developing and demonstrating good antimicrobial stewardship practices (4). However, the results of AMU analyses are only useful to veterinarians if they are understood and accessible (5).

Antimicrobial use information can be described in quantitative terms using various metrics (6,7). These metrics include frequency- or count-based metrics (e.g., the percentage of rations on a farm that are medicated with antimicrobials, the prevalence or incidence of administration of a given antimicrobial), weight-based metrics (e.g., total kilograms of antimicrobial used), and dose-based metrics (e.g., defined daily doses for animals) (8). Weight-based and dose-based metrics may be further refined by applying a denominator that adjusts the metric by size and/or weight of the population of animals at risk of treatment and, in some cases, by the days at risk (or length of the production cycle) (6). When characterizing AMU, choosing which metrics and denominators to use is an important decision that may be affected by factors such as the intended audience.

Limited research has been done to assess the degree of understanding of AMU metrics among veterinarians, or on their preferences for and comfort level with them. Benedict et al (9) administered a questionnaire to beef industry stakeholders, including veterinarians and producers, about various AMU measures and their perceived accuracy and clarity. Dose-based measures standardized to 1000 animal-days were perceived to be accurate and appropriate for reporting comparisons in use and relationships with AMR, and 64% of participants correctly interpreted this measure (9).

In Canada, there are approximately 7700 pig farms (10), an estimated 68 veterinarians practicing predominately swine medicine and herd health (Anne Deckert, personal communication, 2017), and an unknown number of veterinarians in mixed practice with swine clients. The Canadian Integrated Program for Antimicrobial Resistance Surveillance (CIPARS) has been collecting and analyzing AMU data from sentinel grower-finisher herds across Canada since 2006 (82 herds in 2017) (11,12). This information is available for swine veterinarians and producers to help make antimicrobial stewardship decisions. Knowledge of veterinary understanding of and preferences for AMU metrics is of use to those who communicate AMU information to veterinarians and other stakeholders involved in antimicrobial stewardship. The objective of this study was to evaluate preferences for various AMU metrics and denominators among Canadian swine veterinarians to improve reporting of AMU information to these stakeholders and to facilitate antimicrobial stewardship decisions. We anticipated that simpler metrics (e.g., frequency-based) would be better understood and preferred over more involved metrics (e.g., dose-based).

Materials and methods

The study was designed to evaluate the AMU metric and denominator preferences for both swine veterinarians and producers. Producers were invited, in a manner similar to that described below for the veterinarians, to participate in the study. However, the producer response was too limited to allow meaningful analysis. This paper, therefore, focuses entirely on the veterinary survey and responses.

Questionnaire development

An anonymous, online questionnaire was developed using the Qualtrics Research Core Platform (Qualtrics, Provo, Utah, USA). As an alternative, a mail-based paper version was made available with pre-addressed, stamped envelopes. The questionnaire was available in English and French (the official national languages of Canada). The French version was reviewed by a French-speaking industry expert familiar with Québécois French. The questionnaire is included in supplementary materials available from the authors.

The questionnaire had 4 sections: Research Summary and Consent, Understanding, Preferences, and Demographics. As part of the Research Summary, consenting participants were asked to identify their occupation and to self-evaluate their understanding of AMU metrics and denominators before being introduced to the various metrics in the Understanding section.

The Understanding section included a general introduction to AMU metrics and a short description, with examples, of frequency-, weight-, and dose-based metrics and denominators. Fictitious results from analyses using each type of metric were provided and participants were asked questions to test their understanding. Each question included an “Unsure” response option. The AMU metrics included the percentage of farms using an antimicrobial, the percentage of rations medicated with antimicrobials, and the percentage of days medicated (frequency-based metrics), the total kilograms of antimicrobials used, and milligrams of antimicrobials used per kg of pig (weight-based metrics), and the defined daily doses for animals per pig or per 1000 pig-days (dose-based metrics) (Table 1). Three questions were included in this section, with 1 question each for frequency-, weight-, and dose-based metrics.

Table 1.

The antimicrobial use metrics used in the study including formulas and abbreviations.

Antimicrobial use metrics	Abbreviation	Formula for calculation
Frequency-based
Percentage of farms	% farms	$\frac{\text{Number\hspace{0.17em}of\hspace{0.17em}farms\hspace{0.17em}using\hspace{0.17em}an\hspace{0.17em}antimicrobial}}{\text{Total\hspace{0.17em}number\hspace{0.17em}of\hspace{0.17em}farms}}$
Percentage of rations medicated	% rations	$\frac{\text{Number\hspace{0.17em}of\hspace{0.17em}rations\hspace{0.17em}medicated\hspace{0.17em}with\hspace{0.17em}antimicrobials}}{\text{Total\hspace{0.17em}number\hspace{0.17em}of\hspace{0.17em}rations}}$
Percentage of days medicated	% days	$\frac{\text{Number\hspace{0.17em}of\hspace{0.17em}days\hspace{0.17em}medicated\hspace{0.17em}with\hspace{0.17em}antimicrobials}}{\text{Total\hspace{0.17em}days\hspace{0.17em}in\hspace{0.17em}the\hspace{0.17em}production\hspace{0.17em}cycle}}$
Weight-based
Kilograms used	kg	$\mathrm{\Sigma }\hspace{0.17em}\text{kg\hspace{0.17em}of\hspace{0.17em}antimicrobial\hspace{0.17em}used}$
Milligrams adjusted by animal biomass	mg/kgpig	$\frac{\mathrm{\Sigma }\hspace{0.17em}\text{mg\hspace{0.17em}of\hspace{0.17em}antimicrobial\hspace{0.17em}used}}{\text{Number\hspace{0.17em}of\hspace{0.17em}pigs}\times \text{standard\hspace{0.17em}weight\hspace{0.17em}at\hspace{0.17em}treatment\hspace{0.17em}}(\text{kg})}$
Dose-based
Number of Canadian defined daily doses for animals	nDDDvetCA	$\frac{\text{mg\hspace{0.17em}of\hspace{0.17em}antimicrobial\hspace{0.17em}used}}{\text{DDDvetCA\hspace{0.17em}}\left(\frac{\text{mg}}{\text{kg}}\right)\times \text{standard\hspace{0.17em}weight\hspace{0.17em}at\hspace{0.17em}treatment\hspace{0.17em}}(\text{kg})}$
Canadian defined daily doses for animals adjusted by animal numbers	DDDvetCA/pig	$\frac{\text{nDDDvetCA}}{\text{Number\hspace{0.17em}of\hspace{0.17em}pigs}}$
Canadian defined daily doses for animals adjusted by animal-time	DDDvetCA/1000 pig-days	$\left(\frac{\text{nDDDvetCA}}{\text{Number\hspace{0.17em}of\hspace{0.17em}pigs}\times \text{Number\hspace{0.17em}days\hspace{0.17em}at\hspace{0.17em}risk}}\right)\times 1000$

The Preferences section included 3 questions about which metrics participants preferred and deemed most useful, factors that influenced their preferences, and their needs for AMU information. Two optional open-text questions allowed participants to provide additional information.

The Demographics section comprised 5 questions: their province(s) of employment; highest education level, how comfortable they felt explaining the various AMU metrics to their swine clients, whether their veterinary practice is limited to pigs, and their decade of graduation (to avoid collecting potentially identifying information).

The questionnaire was piloted by graduate students involved in AMR studies, including 3 veterinarians, resulting in modifications to improve clarity and reduce ambiguity. In addition, 2 veterinary researchers knowledgeable in the field of AMU provided a scientific review of the study design and the questionnaire as part of the ethics review process. The University of Guelph (#18-06-013) and Health Canada’s (#2018-0010) research ethics boards reviewed and approved the study and the questionnaire.

Questionnaire administration

Invitations to complete the online questionnaire were distributed across Canada between November 2018 and June 2019. Paper copies of the survey were made available in 2019 for those who preferred this format. Invitations to participate were e-mailed by the industry and veterinary associations listed in Table 2 to their members, and in some cases, posted on social media. Paper invitations were also distributed at veterinary meetings and conferences, including the Western Canadian Association of Swine Veterinarians’ annual conference, the Canadian Association of Swine Veterinarians’ Annual General Meeting, monthly meetings of Ontario and Quebec’s provincial swine veterinary associations, as well as some smaller, local meetings. At some meetings, a brief verbal invitation to participate was given. Our goal was to target all actively practicing swine veterinarians. Some associations distributed the invitations more than once, acting as reminders to participate.

Table 2.

Swine industry and veterinary associations that assisted with distributing invitations to participate in the survey.

Association	Level
Industry
Canadian Pork Council	National
Ontario Pork	Provincial
Manitoba Pork	Provincial
Veterinary
Canadian Association of Swine Veterinarians	National
Ontario Association of Swine Veterinarians	Provincial
Association des vétérinaires en industrie animale du Québec	Provincial
Western Canadian Association of Swine Veterinarians	Provincial

Analysis

The data were analyzed using R 3.6.0 (13), packages dplyr, tidyr, and reshape2 (14–16). Only submitted responses were included in the analysis. The number and proportion of responses to each question were determined.

Results

Sixteen practicing veterinarians agreed to participate, 12 of whom completed and submitted the survey. Assuming all swine veterinarians in Canada received a survey invitation and the estimate of 68 veterinarians practicing predominately swine medicine and herd health is correct, the response rate was 17.6%. Most participants were from the highest pig producing province in Canada (Quebec), followed by Ontario and Alberta (Table 3). Most participants worked with primarily swine farm clients, completed the questionnaire in English, and graduated from veterinary college prior to 2010 (Table 3). Two veterinarians completed the paper version of the questionnaire, whereas the remainder completed the online version.

Table 3.

Descriptive information about the veterinary participants (N = 12), including the province(s) in which they work with the number of pig farms in the province for comparison, the language in which they completed the questionnaire, the type of veterinary practice, and the decade in which they graduated from veterinary college.

Information about the participants	n (number of pig farms in the provincea)
Primary province of workb
British Columbia	1 (790)
Alberta	3 (1140)
Saskatchewan	1 (625)
Manitoba	1 (580)
Ontario	3 (2630)
Quebec	6 (1915)
Language preference
English	8
French	4
Type of practice
Primarily pigs	10
Pigs and other species	2
Year of graduation
1970 to 1979	3
1980 to 1989	4
1990 to 1999	1
2000 to 2009	3
2010 to 2019	1

^aAs of January 1, 2018 (Statistics Canada. Table 32-10-0202-01 Hogs statistics, number of farms reporting and average number of hogs per farm, semi-annual).

^bSome veterinarians worked in more than 1 province.

Participants were asked to rate their understanding of frequency-, weight-, and dose-based metrics, as well as biomass and animal-time denominators, on a scale of very poor to excellent (Figure 1). Among the AMU metrics, responses ranged from poor to excellent, with dose-based metrics having the highest number of excellent responses (Figure 1). Participants’ responses were similar between the 2 types of denominators (Figure 1). When the AMU metrics and denominators were ordered according to the number of good and excellent responses, dose-based metrics had the highest number of responses (9/12), followed by frequency-based metrics (8/12), animal-days denominators (8/12), biomass denominators (7/12), and weight-based metrics (6/12). Five out of 12 (42%) participants rated their understanding of all metrics and denominators as good or excellent, whereas 2/12 (17%) participants did not have any good or excellent ratings for any metrics or denominators. The remaining 5/12 (42%) participants had mixed responses.

Figure 1.

Self-reported understanding of different types of antimicrobial use measures and denominators at the start of the questionnaire among 12 swine veterinarians in Canada.

In the Understanding section, the frequency-based questions were answered correctly by all participants. Eleven of the 12 participants (92%) answered the weight-based question correctly. One participant skipped both dose-based questions, with 10 out of 11 participants (91%) giving the correct answer to the question about the number of defined daily doses per pig, and 11 out of 11 participants (100%) giving the correct answer to the question about the number of defined daily doses per 1000 pig-days.

In the Preference section, participants were asked to select which metrics and denominators they preferred or regarded as most useful when receiving information about AMU on their clients’ farms (Figure 2). Participants were able to choose more than one preferred metric. Dose-based metrics were chosen more often than frequency- and weight-based metrics. Biomass and animal-days denominators were chosen an equal number of times.

Figure 2.

Self-reported preferences for different types of metrics and denominators used in the analyses of antimicrobial use data among 12 swine veterinarians in Canada.

Participants were asked to rate the importance of various factors that may influence their preferred choice, from not at all important to extremely important (Figure 3). The ability to make comparisons in AMU, detect changes in use, and to account for antimicrobial dose, received the highest number of very and extremely important ratings. The factor with the widest range of ratings was accounting for the number of animals (Figure 3). Simplicity, ability to detect changes in AMU, and accounting for the dose of the antimicrobial all received ratings between moderately and extremely important (Figure 3). When asked to indicate any additional factors influencing their preferences, respondents commented that reporting by class of antimicrobial was important to them (n = 3), and reporting AMU using measurements that account for dose made more sense to them than those that are weight-based only (n = 2). One respondent suggested that weight-based metrics are often reported without context, leading to a focus on reducing the total amount of antimicrobial use by weight (versus a focus on prudent use or good antimicrobial stewardship practices).

Figure 3.

The importance of various factors influencing preferences for antimicrobial use (AMU) metrics and denominators among 12 swine veterinarians in Canada.

Participants were asked to indicate which listed objectives they would like to achieve when provided with quantitative AMU data. They were also asked to provide any additional objectives that were not listed. All 12 respondents selected the objective of improving AMU on their clients’ farm. Eighty-three percent (10/12) of respondents wanted to be able to compare AMU on their client’s farm with provincial reports on AMU on swine farms and detect changes in AMU on their client’s farm. Sixty-seven percent (8/12) of respondents wanted to be able to compare AMU on their client’s farm with other farms, and 42% (5/12) of respondents were interested in providing data for research. Additional objectives provided by the respondents included comparing AMU among a group of like-minded producers and evaluating AMU by class of antimicrobial and stage of production.

An additional question for veterinarians, added to the demographics section, asked participants to rate their level of comfort in explaining the results of AMU analyses, using the various AMU metrics and denominators, on a scale of very uncomfortable to extremely comfortable (Figure 4). No participants felt very uncomfortable explaining any of the metrics or denominators (Figure 4). Only the 2 denominators received ratings of somewhat uncomfortable. Frequency- and weight-based metrics received the highest number of extremely comfortable ratings, whereas dose-based metrics received more mixed ratings, ranging between neither comfortable nor uncomfortable and extremely comfortable (Figure 4).

Figure 4.

Self-reported comfort level with explaining various antimicrobial use metrics and denominators to swine producer clients, among 12 swine veterinarians in Canada.

Discussion

This survey indicated that veterinarians in the swine industry have a basic to very good understanding of the metrics that are used to report AMU information, and identified their preference for dose-based metrics over others, in contrast with our original expectations. The survey also provided some insight into why veterinarians prefer dose-based metrics.

The questions that tested participants’ understanding of AMU metrics were largely answered correctly, indicating that participants were able to interpret the results of AMU analyses using each type of metric. One participant skipped the dose-based question, which may indicate uncertainty in how to answer the question, although an “unsure” option was available. Although the participants rated their understanding of denominators lower than the numerators, they were still able to correctly interpret results when denominators were used. This suggests that although practicing swine veterinarians may not have the same comfort level with AMU metrics as those who frequently work with them, they do have sufficient knowledge about the various components of the metric (e.g., frequencies, doses, animal weights) to interpret and make use of the information. In a future study, it would be useful to investigate what aspects of the denominators were less well understood, such as denominator calculation, denominator purpose, or other elements.

Compared to Benedict et al (9), a higher percentage of participants (91% and 100% versus 64%) were able to correctly interpret AMU information using dose-based metrics (9). However, differences in methodology may at least partially explain the different results. In our study, we tested participants’ understanding by providing them with a chart displaying AMU information using dose-based metrics and asked them to identify the incorrect interpretation. By providing a visual representation of the data, the participants’ ability to interpret charts was also tested. In Benedict et al (9), participants were provided with AMU information using dose-based metrics in text form and were asked to interpret the information in their own words. Asking participants to interpret AMU information in their own words was likely more challenging than asking them to choose the incorrect interpretation from a list of options. In addition, the Benedict et al study (9) was targeted to the beef industry, including non-veterinarians, and was distributed across 10 countries (including Canada). It is also possible that veterinarians, including the swine veterinarians responding to our survey, have, from ongoing exposure, become more comfortable with AMU metrics over time.

We also measured participants’ understanding of each metric by assessing how comfortable they felt explaining AMU information using metrics to their clients. In this case, the results differed from their self-reported understanding, as frequency-and weight-based metrics received higher comfort ratings than dose-based metrics. Participants may have more experience and practice explaining frequency- and weight-based metrics since these metrics have been in use in veterinary medicine for longer than dose-based metrics. These findings may also be related to the degree of complexity of each metric, with the more straightforward frequency and weight-based metrics being easier to explain than dose-based metrics. Regardless, the participants expressed a basic level of comfort with all metrics.

The participants’ preference for dose-based metrics was contrary to our expectation that more straightforward metrics, such as those that are frequency- or weight-based, would be preferred and better understood. When examining the ratings for the factors that influenced the participants’ preferences, it was evident that the ability to make comparisons in use, whether between farms or with regional/national averages, was highly rated, as was the ability to detect changes in use. These factors may explain why dose-based metrics were preferred, as they adjust for differences in antimicrobial dose, facilitating comparisons among various types of antimicrobials are used or when dosing differs between farms. Dose-based indicators also include adjustments for animal numbers and weight, and when the animal-time denominator is used, for days at risk. These adjustments enhance comparisons and reduce the number of factors that can affect the reporting of changes in AMU over time.

Although the importance of factors such as the ability to make comparisons in use and to detect changes in antimicrobial use were highly rated, the importance of factors that involve denominator preference received lower ratings (e.g., accounts for days at risk, accounts for number of animals). These findings were unexpected, as making appropriate comparisons relies on the use of denominators to adjust quantitative estimates of AMU by variables that contextualize differences among herds (i.e., animal numbers, animal weights, antimicrobial dosages, time at risk). These common variables can also vary over time in the same herd and controlling for them can help identify true changes in AMU for other reasons. These findings, in addition to the lower ratings for understanding, could indicate a need for education on the function of denominators in AMU indicators, or it could simply indicate a relative difference in the importance of the factors.

The ability to make comparisons with AMU data and to detect changes in AMU also rated highly as overall objectives of AMU analyses. It was encouraging to note that most participants wanted to use AMU information to improve AMU on their clients’ farms, which confirms the importance of collecting farm-level AMU information for AMU stewardship purposes. When preparing AMU reports for veterinary stakeholders, it would also be useful to keep in mind their additional objective of describing AMU by class of antimicrobial. It was encouraging that almost half of participants were interested in providing data for research purposes, as these data are needed to aid in the development of AMU policies and stewardship guidelines, and in the study of associations between AMU and AMR.

As the response rate was low, caution is needed in generalizing the results more broadly beyond our group of participating swine veterinarians. Although only 12 veterinarians participated, we received responses from every major pig-producing province in Canada. Quebec is the highest pig-producing province in Canada, whereas Ontario has the greatest number of pig farms (10,17). There are no available data on the number of actively practicing swine veterinarians in each province in Canada and whether the number of veterinarians practicing predominately swine herd health is proportional to the number of farms or the number of pigs in the province is unknown. It is difficult to determine, then, if the number of participants from each province represented the distribution of swine veterinarians across Canada. It is possible that Quebec was over-represented in this study, and Ontario and the western provinces under-represented.

Our sample represented veterinarians with a range of years of practice experience, graduating from the 1970s to the current decade. Since the issue of AMR is a growing concern, we could also speculate that the amount of instruction on AMR and antimicrobial stewardship varied with year of graduation from veterinary college. However, with the increased profile of the AMU/AMR issue, it is also possible that the sampled veterinarians have acquired updated information through continuing education or personal study.

Factors that may have affected our response rate include interest in the topic and the presence of survey fatigue among the survey population, recruitment methods, aspects of questionnaire design such as mode (by mail, online, or mixed), length, and the use of incentives (18–21). The subject of AMU in animal production is a topic of current interest worldwide due to growing concerns about AMR, particularly to veterinarians, who are antimicrobial stewards (4). Due to its more technical nature, there may be less interest in how AMU is described and reported, leading to difficulties engaging veterinarians on this subject. However, having some understanding of how to interpret quantitative AMU information can help veterinarians make full use of AMU reports. Survey (or respondent) fatigue likely also contributed to the low response rate, since the number of veterinarians in Canada who practice swine herd health is relatively small, and they receive many requests to participate in research. We hoped to address survey fatigue by providing a choice of survey modes (online or paper) and by attempting to keep the time to complete the questionnaire to less than 20 min. Although unconditional incentives have been shown to improve survey response rates (18,22), difficulties administering such an incentive using an online questionnaire format precluded their use in this study.

Indirect methods of participant recruitment were used, due to difficulties obtaining the names and contact information of veterinarians practicing swine herd health in Canada as a result of privacy legislation and individual organization’s privacy codes (23,24). We attempted to reach every veterinarian practicing swine herd health in Canada by using multiple indirect methods of recruitment and asking associations to send or post reminders to participate. However, it is possible that this method of recruitment was less effective than direct mailings and reminders in soliciting participation.

Acknowledging the low response rate in this survey and the complex reasons why it may have occurred, a related question is whether non-response bias could be present in our data (21,25). In our survey, the primary factor that may contribute to non-response bias is the topic itself. Due to the specific and technical nature of the topic, it is possible that the responses we received were from veterinarians with pre-existing knowledge and interest in the subject. As a result, our results may show a higher level of understanding and comfort than would be expected among swine veterinarians in general. Preferences for AMU metrics may also be different between those who have experience and interest in the topic and those with less interest. It is possible, then, that our results are more reflective of a sub-group of swine veterinarians with an interest in and experience with measuring AMU. It is also reasonable to assume that these swine veterinarians with an interest in AMU are more likely than others to engage in detailed AMU and antimicrobial stewardship discussion with their clients and, therefore, make use of the study findings.

In conclusion, our study provided some insight into swine veterinarians’ understanding of and preferences for AMU metrics, with the caveat that our small sample size and response rate limited our ability to generalize the findings broadly. It would be useful to repeat the study, with other groups of veterinarians, to determine if our observations hold true for swine veterinarians in other areas, or for veterinarians with other species’ focus. We determined that participating swine veterinarians preferred and understood dose-based metrics, although they may benefit from resources to enable them to explain them to clients in plain language. Moreover, study participants indicated that they want/need farm level AMU information for stewardship purposes, to make AMU comparisons, and to monitor for changes in use. We believe the results of this study are useful to those making decisions about which metrics to use in AMU reports targeted to veterinarians. By keeping in mind veterinary preferences and objectives for AMU information when metric choices are made and by providing resources to help veterinarians explain these metrics, veterinary uptake and use of the information may be improved, which may ultimately influence the stewardship decisions made by both veterinarians and producers. It may also be helpful to emphasize the importance of and the function of denominators in AMU reports. For future studies of this type, expanding the survey to cover additional food animal sectors may help to improve the response rate and to determine if perceptions in the swine industry are reflected in other food animal industries. In addition, using other methods of engaging veterinarians and the farming community such as workshops at conferences, interactive presentations at producer meetings, focus groups, or interviews, may be helpful.