Researchers look to veterinarians, digital records in ambitious data-collection effort
UMN College of Veterinary Medicine photo by Martin Moen
Experts in human and veterinary medicine comprise the University of Minnesota team researching antibiotic use in household pets. They are (from left): Emma Leof Bollig, an epidemiologist; Dr. Amanda Beaudoin, adjunct assistant professor, director of One Health Antibiotic Stewardship, Minnesota Department of Health; and Dr. Jennifer Granick, associate professor, small animal internal medicine.
The evolution of pathogens to resist drugs designed to kill them is widely understood to be one of the biggest public health challenges of our time. One key to curbing antimicrobial resistance is more judicious use of antibiotics in medicine.
But a glaring deficiency hampers the effort for companion animal veterinarians: No one knows how much of the drugs are used in cats and dogs nationally.
The U.S. Food and Drug Administration began to address this lack last year by awarding $200,000 grants to each of two projects that will collect antibiotic-use data from companion animal practices countrywide. Researchers at the University of Minnesota College of Veterinary Medicine are recruiting practices to participate in a survey through which they will report information for all cases seen during one day in August. Another team, at North Carolina State University College of Veterinary Medicine, is designing a method to analyze relevant data in electronic medical records. Each grant is renewable for four additional years.
Together, the projects could create the first comprehensive picture of antibiotic use in American pets.
Human medicine has been concerned about antibiotic resistance going back to the 1960s. (Strictly defined, antibiotics are a subset of antimicrobial drugs but the terms often are used interchangeably.) Engagement on the veterinary front is more recent.
The FDA Center for Veterinary Medicine began in the 1990s to restrict the use in food animals of antimicrobial drugs deemed important in human medicine, leading to a ban in 2017 on the use of the drugs in livestock feed for the sole purpose of spurring growth. In 2020, the agency shifted toward companion animals.
The expanded view is overdue, according to Dr. Jennifer Granick, an associate professor at the University of Minnesota College of Veterinary Medicine. Granick points to a study of the human skin microbiome to illustrate why antibiotic resistance in pets matters to humans. The skin microbiome is the community of microorganisms, such as bacteria and fungi, that live on skin. The study showed that the skin microbiomes of two adults in a household were similar, but substantially more so if there was a child or dog in the house. Granick said they serve as a kind of "mixing vessel."
"You can't look at data like that and think that antibiotic resistance in humans doesn't affect animals and antibiotic resistance in animals doesn't affect humans," Granick said. "We know these relationships exist, but we haven't taken a good hard look at how we are prescribing in companion animal medicine."
The study she is heading aims to do just that. The Minnesota team will collect data using a point prevalence survey, a method in which standardized data is collected across many locations at a given moment in time. In public health, point prevalence surveys have been used to estimate antibiotic use nationally in acute care and long-term care facilities.
In the veterinary project, participating clinics will report one day's worth of patient-visit information, including reason for visit, any diagnostic tests conducted and any antibiotics prescribed, including the indication for the antibiotics. By collecting information on all patients, researchers will be able to estimate the prevalence of antibiotic use. The clinics will enter the data online into a secure web-based application. Each clinic will provide data for one day of their choice between Aug. 16 and 29. The survey will be repeated in 2023.
The researchers plan to recruit a minimum of 100 clinics, 25 each in four regions, but they are hoping 10 times that number participate. "The more people, the better for the profession," Granick said.
The survey results, she said, will provide a needed baseline.
"If we ramp up efforts as a profession, which is happening now, around antibiotic prescribing and antimicrobial stewardship, we can repeat this process a few years down the road and see what progress we've made," she said. "And if we've made progress, that's great; we know what efforts are working. If we haven't made progress, we can see where we need to redouble efforts and provide more support for appropriate prescribing."
Before it landed the FDA grant, the Minnesota team began the work in 2018 with a pilot project collecting surveys once a month at the university's Veterinary Medical Center. That spawned surveys in 19 companion animal general practices in Minnesota and North Dakota, supported by internal funds. Last year, with FDA funding, the team conducted a survey of 14 veterinary teaching hospitals around the country. The teaching hospital surveys will repeat in 2022 and 2024.
The early results taught the researchers to hone their focus. While they will collect information on all antibiotic use, they will also gather more specific data for a subset of diseases. "We know that there are certain areas where there's a lot of prescribing and where there are guidelines available for practitioners," Granick said. "We felt it was important to gather information around those particular disease processes." These include urinary and respiratory infections in cats and dogs and skin infections in dogs.
Another area for focus, she said, is the prescribing of metronidazole for acute diarrhea. Granick cited research that suggests metronidazole does not hasten the resolution of acute diarrhea and may negatively impact the biome. Understanding how it's prescribed can lay the basis for future guidelines on its use.
Even before guidelines are created or modified, Granick said, just collecting the data and making it available to the veterinary community can be powerful. "There is good evidence in human medicine that even without heroic antimicrobial stewardship interventions, just providing data to practitioners can change their behavior," she explained.
Employing artificial intelligence
Photo courtesy of Dr. Cristina Lanzas
Dr. Cristina Lanzas leads a research project at the University of North Carolina College of Veterinary Medicine that will tap electronic medical records to understand antibiotic use in compantion animals nationwide. It's a key first step to curbing misuse, she said: "If you don't measure something, it is very difficult to have any change."
Dr. Cristina Lanzas agrees that data on its own is transformative. Lanzas leads the North Carolina State team that aims to create a nationwide digital surveillance system that will automatically collect from electronic medical records any information related to antimicrobial drug prescribing. The records are housed in practice information management software, or PIMS.
Extracting information from veterinary medical records has interested researchers for many years, but the lack of uniformity in recordkeeping has been an obstacle to studies like this one.
That's changing, according to Lanzas. "We have all this potential for data, but definitely we have had to wait [for] advances on artificial intelligence to help us harvest it in an efficient way," she explained. "Some of the things we are proposing maybe five years ago would not have been possible."
An associate professor in infectious diseases, Lanzas has crunched antimicrobial-drug data contained in human medical records, a uniform system driven by Medicare/Medicaid that has proven beneficial to researchers. Veterinary records lack that consistency.
Something as seemingly simple as identifying which antibiotic was prescribed can be tricky. The prescribing veterinarian may use an abbreviation, the brand name or a generic name, or introduce typos into the record.
"How [do] you reconcile that all these different names mean the same thing? That's when new techniques of natural language processing come in," Lanzas said.
Natural language processing enables artificial intelligence to make sense of idiosyncratic human language. This is especially important when it comes to veterinarians' notes in the medical record, known to data analysts as "free text," which may contain important answers to the question of why an antibiotic was or was not prescribed.
"Understanding why people are prescribing antibiotics based on the electronic records is very, very complex," Lanzas said.
Another challenge for researchers is the difficulty of gaining access to the wide variety of PIMS in use today. NC State has overcome this obstacle by partnering with a commercial company that already interacts with and extracts data from a broad array of practice management software within its existing data agreements. The data is aggregated and cannot be traced back to individual practices, Lanzas said.
Like Minnesota, the NC State team has piloted its project, in its case using records from the teaching hospital on campus and some private practices in the state. The researchers are currently studying what they get when they extract data from the PIMS relating to urinary tract infections. They ask: Are we seeing what we think we should be seeing? Essentially, they are double-checking the AI to make sure it is extracting and interpreting the information correctly.
Once the NC State researchers confirm the system is working as intended, they will expand data collection to general practice, specialty and emergency hospitals around the country for ongoing surveillance. Lanzas said they have not identified a target number of practices yet.
The data will be made available to other researchers through a web-based portal. Ideally, Lanzas said, there will be opportunities for practices to benchmark their antimicrobial-use against national trends.
"People want to do better when it comes to antibiotic stewardship," she said, "but they just don't know where they are."
The team at the University of Minnesota is also developing an automated system to collect antimicrobial-use data from PIMS, as part of its FDA-funded research. They are in the early stages of creating what they're calling the Companion Animal Veterinary Surveillance Network, with help from colleagues at the University of Liverpool in England.
Researchers in the United Kingdom have pioneered the use of practice data for research. Both the Small Animal Veterinary Surveillance Network at the University of Liverpool and the Veterinary Companion Animal Surveillance System at the Royal Veterinary College are already collecting and sharing data from PIMS on the use of antimicrobials.
Meanwhile, the NC State and Minnesota teams meet regularly and collaborate to move their projects forward.
"Being able to use electronic records for research, that is one of those last frontiers," Lanzas said. "If we want to do research on this scale, we really have to figure it out."
This story has been updated to include a contact email for the University of Minnesota College of Veterinary Medicine survey.