Centers for Disease Control (Georgia headquarters)
Many public organizations amass a significant volume of data on human health in the United States, including the Centers for Disease Control and Prevention (Georgia headquarters, above). For companion animal health, no such public entity exists.
This is the second of two parts. Read part 1.
In human medicine, organizations small and large — from county public health departments to the Centers for Disease Control and Prevention to the World Health Organization — possess great quantities of information on the state of human health. In veterinary medicine, no similar repository exists.
Governments don't collect comprehensive data on small animal health. The effort, if done at all, is largely left to private entities such as pet insurers and consolidated practice groups, which, even if they are large, don't encompass everyone.
The advent of digital recordkeeping has vastly improved the prospects and possibilities for gathering and analyzing veterinary medical data for the common good. So far, though, widespread electronic-record data collection and crunching in the U.S. has been mostly limited to private companies collecting and selling information for commercial use. The larger universe of digital patient records — with the exception of those at teaching hospitals — is largely untapped for real-time health tracking or epidemiological studies.
In an example of what a large practice owner can do, Mars Petcare announced in April the start of a 3-year study of the behavior and health history of participating canine patients at Banfield Pet Hospital, touting it as "one of the largest tech-enabled studies" of its kind. With more than 1,000 practices, Banfield is the largest general-practice chain in the world and one of four U.S. practice brands owned by Mars Inc.
Called the Pet Insight Project, the study uses GPS activity trackers from Whistle, a subsidiary of Mars, to monitor the behavior of enrolled dogs. Combining that information with health data from the dogs' Banfield medical records, the study aims to identify links between behavior and health.
"We hope there will be findings throughout the course of the project that enable new technologies for pet owners and veterinarians that improve the way we care for our pets, as well as opportunities for other researchers to use the information to advance veterinary medicine," Carrie Deverell, senior director for Mars Petcare's Global Affairs, said by email to the VIN News Service.
There has been interest in aggregating and analyzing large quantities of practice records for medical research since veterinarians began the shift from paper to digital recordkeeping in the 1980s. Banfield has used its own extensive database of medical records for research for more than 20 years.
The only disease of companion animals that the federal government tracks is rabies. The United States Department of Agriculture tracks some diseases in farm animals, using compulsory reporting systems.
Even since before the advent of practice software, veterinary schools shared medical records. In its heyday, the Veterinary Medical Databases (VMDB), which originated in 1964, collected medical records from 27 veterinary school teaching hospitals, according to Dr. Loren Schultz at the University of Missouri, which has housed the database since 2014. Today, the number of participating schools is down to five, owing to budgets cuts at universities. The main database includes records of 600,000 patients (including dogs, cats, horses, cows, birds and more) since August 1987. The last time a study using VMDB information was published was 2012.
There are also some public data-driven initiatives centered around specific diseases, such as the Canine Influenza Virus Surveillance Network at Cornell University College of Veterinary Medicine. Using test results from participating diagnostic labs around the country, the network tallies cases of canine influenza.
The Pet Insight Project is a reminder that for the moment, private enterprise has the corner on big-data research in veterinary medicine.
Public database: good idea, many obstacles
UK doubles down on data collection
Experts in the small field of veterinary clinical informatics, the practice of aggregating and analyzing veterinary data to achieve better health outcomes, say medical records collected by practice information management software hold promise for everything from predicting future occurrences of food-borne toxicants to understanding vaccine reactions to getting a clearer picture of antimicrobial resistance (a topic Banfield explored in a recent study).
But their enthusiasm is tempered by the challenges. The creation of a large data-sharing effort for research is hindered by practice information management systems (sometimes referred to by the acronym PIMS) that aren't focused on research; a lack of a shared, standard language that makes comparisons among records doable; and the absence of a carrot or a stick for getting veterinarians and industry invested.
Scores of practice information management systems, pet-owner portals and marketing programs are available to veterinarians. Designed to help practitioners run and promote their practices, send bills and reminders, track inventory, etc., they collect major quantities of potentially rich medical data. But they were not created with research use in mind.
"Most of the problems revolve around our practice management software, if the practice uses practice management software at all," said Dr. Jon Lustgarten, president of the Association for Veterinary Informatics (AVI), which aims to expand the use of clinical information to improve the quality and safety of patient care. Some practices still depend on paper charts and use PIMS as an advanced cash register only, he said.
Even when a practice is paperless, Lustgarten added, the data captured by practice management software isn't structured for what he calls "informatics recall."
That's partly because the software's purpose is quite different — it's designed to run a practice. It's also because there is no standardized coding or terminology widely in use in veterinary medicine.
Dr. Stephen Pittenger, a veterinarian in Houston and former president of AVI, explains how the issue might play out in the case of an ear infection, for example:
For starters, the presenting complaint and diagnosis of an ear infection could be assigned a code by the practice but typically is not, he said.
Further, no standard code or description exists for results of in-house cytology to identify the infectious agent. Next, the patient could be prescribed a combination of medications, such as a generic drug and a compounded preparation, which most veterinary systems don't track well.
Finally, owners often don't return for follow-up visits, leaving the record silent on whether the patient's condition improved.
"That is just for a simple ear infection," Pittenger said. "You can guess how many holes there are in a complex case."
It is difficult to derive meaningful research findings from incomplete or inconsistent records. You end up trying to compare apples and oranges, or kiwis, grapes and apricots.
"Without standardized nomenclature (a common coding language), it gets really complex to combine and share and manage multiple sources of data," said Dr. Liz Lund, former senior director of research at Banfield, who is now a consultant specializing in digital health records and big-data research for corporations and nonprofits. "Without being able to share and manage and contrast, you lose a lot of potential value."
A standard language for recordkeeping does exist for veterinary medicine, should the profession ever agree to use one.
The hard work of creating and maintaining a nomenclature has been done by veterinarians at the Virginia-Maryland College of Veterinary Medicine, the only university in the country that offers a degree in veterinary medical informatics. Faculty members Drs. Jeffrey Wilcke and Julie Green established the Veterinary Terminology Services Laboratory in 2005 to support the National Animal Health Laboratory Network, which is underwritten by the USDA.
The veterinary-specific vocabulary is an extension of the Systematized Nomenclature of Medical Clinical Terms (SNOMED), one of several nomenclatures used in human medicine. Today, the VetSCT, as it is known, has a place in the U.S. National Library of Medicine. It is the only veterinary terminology among more than 100 terminologies in the library.
But creating a standard terminology is just the beginning. Getting veterinary medicine to adopt the terminology comprehensively enough to make a difference on the research side is another thing altogether.
There has been some interest and progress: The Veterinary Terminology Services Laboratory created SNOMED subsets for the American Animal Hospital Association and the American Association of Equine Practitioners. It also created a Small Animal Specialty subset funded by BluePearl Veterinary Partners.
And some major practice information management systems — including Cornerstone, Impromed, Woofware and Vetter Software — have incorporated those subsets. But for most veterinarians, even those using participating software, adopting the standardized recordkeeping language is optional.
Wilcke said, he's frustrated by the hurdle of adoption. "If my profession can't see where to go with this, I can't make them."
The problem is a lack of incentives, he believes. "No one has clearly demonstrated that veterinarians can benefit financially from the hard work associated with terminology standards," Wilcke said.
Lund said veterinarians also need to see a medical value. "I think that's what the hype around big data is missing in terms of the value proposition," Lund said, "is that anything we turn around and ask the veterinarians to do … the veterinary team really must perceive and understand that it is going to help them help the individual pet."
In human medicine, Medicare/Medicaid provide an incentive for adopting uniform codes for reimbursement. Experts agreed that wider use of pet insurance could do the same on the veterinary side. But Lund worries that the tendency to manipulate codes for insurance reimbursement seen in human medicine could distort veterinary data down the line.
Wilcke suggested that veterinarians have been negatively influenced by doctor friends, who are irritated by the demands of insurance reimbursement on the human side. "Vets have a dark view of what it means to code a medical record," he said.
The best hope for getting veterinarians on the same page is a research breakthrough using big data, Wilcke said, and "for small animal medicine to learn something they could not have learned otherwise."
At the same time, Lustgarten said, veterinary medicine has an advantage over human medicine in extracting useful information from big data. "A huge wave is coming in veterinary medicine of data value, because we don't have those privacy laws or sharing silos that are legally required in human medicine," he said. "The potential for aggregating across a lot of data-intake points is pretty huge."
Meanwhile, in the data-sharing trenches much is still to be done. Dr. Ashley Zehnder, a postdoctoral fellow at Stanford University, studies cancer in companion exotic animals. Her work led her into the world of veterinary medical records, as she attempted to improve data-sharing from veterinary clinical records. Her hope, she said, was "to align veterinary data-sharing with similar efforts in human health care."
Over nine months, she followed a steep learning curve during consultations with experts at AVI, Virginia-Maryland College of Veterinary Medicine, VetCompass in the United Kingdom (see sidebar), and VINx, a business that offers tools for client communications and practice management. Zehnder approached the major commercial veterinary data repositories to cultivate sharing. So far, she said she has an agreement with VetData, a large veterinary data aggregator that is owned by Vets First Choice, to explore how they might provide records for her database of tumors in exotic animals.
Overall, her experience has not been encouraging, she said. "I think it is becoming more technologically feasible to share data and groups are working on it, but it will take a long time," she said. "It's like corralling a million cats, each with their own incentives."
The Banfield approach
Corralling cats appears to be something that a corporation such as Banfield can manage with the help of proprietary software, a standard language and a belief in the value of the research.
Banfield claims to have the largest electronic veterinary medical records database in the world, with more than 25 million unique records (comprising 700 million clinical notes and more than 2 billion medical observations/treatments). According to a spokesperson for Banfield, its data has been used in more than 50 peer-reviewed publications. The company has collaborated with the federal government and issues free public reports, including an annual State of Pet Health Report and Veterinary Emerging Topics Report.
"Banfield got the importance of veterinary data from a very, very early start," Lustgarten said.
Back when it was still a single stand-alone hospital in the 1980s, Banfield began developing a proprietary practice management and electronic veterinary health record called PetWare, which collects and stores medical records in a uniform way with retrieval for research in mind. Patented in 2000, PetWare does not use the standard terminologies developed by the Veterinary Medical Informatics Lab, which means that it likely would be hard to integrate their data with that of outside clinics.
The company also committed early to informatics. In 2005, Banfield began devoting resources to mining, analyzing and deriving insights from its data; the practice continues to this day.
Over the years, Banfield has shared its data with researchers at universities, especially Purdue University, and government agencies, including the Department of Homeland Security. Together, Purdue and Banfield received a $1.3 million CDC grant to pilot a real-time small animal surveillance network in 2003.
After the world's largest pet food recall caused by melamine contamination in 2007, Banfield partnered with researchers at the University of California, Davis, to see what could be learned from the event and how the veterinary industry might respond to a future such event. In an article published in Peer J in 2016, the authors, mostly employees of Mars, cited the collaboration as an example of "proof-in-concept of using veterinary hospital databases to detect aberrant events in space and time."
Early this year, Banfield signed a memorandum of understanding with the FDA to share "Banfield data relating to disease prevalence in dogs and/or cats" as part of an effort to define "minor-use" drugs in order expedite their development. Minor-use drugs are used for diseases that occur infrequently or in limited geographic areas and in only a small number of animals annually.
At times, Banfield's ideas for sharing data for research have run into resistance. In 2009, for example, VIN News reported that Banfield was in talks to share practice data with the UC Davis School of Veterinary Medicine as part of a proposed Center for Pet Nutrition and Food Safety. Veterinarians objected to the partnership, which included UC Davis paying half of the salary of two veterinarians working at a Banfield clinic. Opponents included students, who cited concerns about Banfield's medical protocols, which they worried were "more geared to generating profits than to practicing sound medicine."
The resistance was effective. As Dr. Bennie L. Osburn, who was dean at the time, told VIN News in a recent email: "Because there was local resistance from local small animal veterinarians, we stepped back to reconsider other options."
Banfield also has faced questions about the quality of its clinical data. In a letter to the Journal of the American Veterinary Medical Association, veterinary medicine professors at the University of Pennsylvania and UC Davis, as well as Drs. Paul Pion and Mark Rishniw, cardiology specialists with VIN, questioned the validity of a 2009 study that suggested an important association between periodontal disease and cardiovascular events in dogs. The observational study was based on Banfield practice records.
The letter challenged the accuracy of the underlying diagnoses used in the study, which, they said, did not correlate "with the established knowledge of cardiac disease in dogs."
In a response, the study authors defended the diagnoses of the primary care veterinarians that were used in the study. However, they conceded that the study did not provide sufficient evidence to establish a causal relationship. "Additional population-based studies are needed to demonstrate consistency of results and these will probably require large clinical data bases similar to that created by Banfield," they wrote.
Looking back at the study and its aftermath, Lund, who worked for Banfield at the time but was not an author on the paper, said the back and forth about the study "speaks to the complexity and nuance of using big data." There are myriad issues when working with systems that are generally created for clinical rather than research use, she said, including, for example, the fact that the Petware clinical information system can account for diagnostic status (confirmed, suspected and tentative diagnoses) but the research may not have accounted for diagnostic status in the analysis of the records.
"Using large databases — big data — is not just pushing a button and waiting for the answers to spit out," Lund said. "Big data analytics use epidemiologic methods, among other approaches. People think these methods and the interpretation of results are so simple but it's messy, hard and complex."
Part 1: With merger, Schein, Vets First Choice to amass clinic data