There's a debate playing out in the pages of Today's Veterinary Business this month, and it's worth paying attention to.

On one side: Dr. Kate Elden, chief medical officer at Dutch, a veterinary telemedicine company. Her April 9 op-ed opens with a striking statistic: 75 million pet owners skipped or declined veterinary care last year due to cost or access. Veterinary costs have outpaced inflation every year since 2019, she argues, and 74% of pet owners can't get a same-day appointment. The technology to fix this exists. What's missing, in her framing, is the political will. California passed a bill in January 2024 that expanded telemedicine and allowed virtual care relationships to be established without a prior in-person exam. The result: medical consultations in the state grew by more than 13,000%. One in three new telemedicine appointments now originate there. To Elden, that's proof of concept. The demand was always there. The vets were ready. The law was the only thing in the way.

Dutch CEO Joe Spector also made an appearance on a podcast with Tucker Carlson with the extremely salacious title, ‘The Private Equity Veterinary Scam Making You Poorer and Killing Your Pets.’

In response, two practice-side voices offered a sharply different reading. Dr. Matthew Salois, an economist and former AVMA chief economist, and Dr. Link Welborn, who owns eight independent small animal practices, don't dispute that access and cost are real problems. They dispute the proposed solution.

Their critique centers on what they call a "pharmacy-first" model. Dutch, they argue, is not simply a telemedicine service — it's a retail business that monetizes the prescription relationship, with medical judgment following the transaction rather than leading it. In-practice pharmacy revenue isn't inefficiency to be disrupted, they write. It's a cross-subsidy that supports the most labor-intensive parts of care: longer appointments, preventive counseling, staff wages. When online platforms extract that revenue, they don't make care cheaper. They weaken the economic floor that makes comprehensive care possible.

The piece invokes the Trojan horse, which might be a bit heavy-handed, but the underlying point is serious. Removing friction is not the same as removing risk. The physical exam requirement for establishing a veterinarian-client-patient relationship isn't protectionism dressed up as medicine. In a species that can't self-report, subtle findings matter, and errors can be irreversible. Telemedicine that extends an established relationship is one thing. Telemedicine that replaces the diagnostic foundation of that relationship is another.

What's useful about this exchange is that it surfaces a tension the profession will have to resolve — probably legislatively per region, whether it wants to or not. Virtual VCPR legislation is currently under consideration in eight states, including Connecticut, Massachusetts, Michigan, and Pennsylvania. Colorado already has it. The California data will be cited often by Dutch and its allies. The questions about what gets lost when convenience becomes the organizing principle of care will keep being asked by everyone else.

There is no way we are taking Dutch’s side after their recent attacks on the profession. But both sides are pointing at something real. Millions of pets without care is a genuine problem. And so is a business model that captures the high-margin prescription transaction while shifting risk downstream to the client, the patient, and the profession. Those two things can both be true at once, and any policy that pretends otherwise deserves scrutiny.

Epilepsy is a common neurological disorder in dogs, and more often than not, practitioners can diagnose it without the benefit of objective confirmation. The standard approach that relies on clinical history, ruling out other causes, and physical and neurological examination findings works well enough in clear-cut generalized seizure cases. But it works less well when the presentation is subtler, or when what looks like a seizure might be something else entirely.

That's the gap Dr. Fiona James is working to close. A veterinary neurologist and associate professor at the Ontario Veterinary College at the University of Guelph, James has been adapting electroencephalography — long the gold standard for human epilepsy diagnosis — for use in dogs. The technical challenge has never been that EEG doesn't work in dogs. It's that the variation in skull shape across breeds makes standardized electrode placement nearly impossible. James and her team have addressed this directly, using 3D dog skull models to map optimal sensor placement for reliable diagnostic readings.

The clinical payoff is meaningful. EEG can detect non-generalized and subclinical seizure presentations that account for roughly one-third of all canine epilepsy cases. These are the cases most likely to be missed on history alone, and most likely to result in a diagnostic shrug and a prescription written on incomplete information. EEG can also differentiate true epilepsy from conditions that can look like it such as paroxysmal dyskinesias, behavioral events, or vestibular episodes. In research settings, wireless video-EEG in unsedated dogs has successfully confirmed or excluded an epilepsy diagnosis in roughly 72% of cases — a meaningful improvement over the current clinical baseline.

What makes James's work particularly practical is the ambulatory piece. Her team has developed a backpack system dogs can wear at home, housing wireless EEG equipment while the patient goes about their normal routine. The device captures data over 48 hours, and even if no seizure occurs during that window, the EEG can still identify inter-ictal epileptiform activity that supports an epilepsy diagnosis. That matters because seizure diaries, the current owner-reported substitute, are notoriously unreliable.

There's a translational dimension here worth noting. Dogs are increasingly recognized as a naturally occurring model for human epilepsy research, sharing not only the clinical presentation but the underlying neuropathology and pharmacokinetics that make them useful for testing therapeutics that may eventually benefit people. The better we get at characterizing canine epilepsy — its subtypes, its inter-ictal signatures, its response to treatment — the more useful that data becomes on both sides of the species boundary.

The technology is not yet in wide clinical use. Fewer than half of veterinary neurologists currently perform EEG, with equipment cost, limited case volume, and lack of standardized protocols cited as the main barriers. What James's work is building toward is a validated, repeatable protocol that could make EEG a practical diagnostic tool in veterinary neurology rather than a research curiosity. The home-worn backpack system is a concrete step in that direction.

For the last couple of years I have been giving an evolving version of a lecture titled “How competent is your co-pilot?” It is an introduction to AI, how diagnostic models are built, and what we know about commercial systems. The answer to the title question is… no one knows. These systems could be highly capable or they might fall on their face more often than not. Certainly the recent paper about radiology systems we covered a couple weeks ago suggests their AI competency is lacking.

David Brundage, a veterinary AI researcher at UW Madison decided to pull data on how transparent theses systems actually are. In other words, he looked at whether companies are giving veterinarians enough information to trust the tools they are selling.

The answer, according to this new audit, is mostly no. The study reviewed 71 commercial veterinary AI products available in North America and scored their public transparency using a 25-point framework built around things like training data, validation, safety, and usability. The average transparency score was 6.4%. Even worse, 63.3% of vendors did not publicly disclose a single meaningful metric. Nothing on sensitivity. Nothing on specificity. Nothing on confidence intervals. In many cases, nothing on who or what the model was even trained on.

That should make people uneasy, especially since many of these products are creeping beyond documentation and into differential diagnosis, triage, and treatment support.

To me, that is the real issue. Veterinary AI companies are often selling confidence before they have earned trust. And the risk does not sit with them. It sits with the veterinarian. If you use the tool, you own the decision. You carry the legal and ethical burden. So if a company wants you to rely on its product in a clinical setting, it should have to show its work.

Veterinary medicine does not need less AI. It needs better AI, and a lot more honesty about what these tools can and cannot do. Right now, too many companies are asking clinicians to take a leap of faith when what they should be offering is data.

-RBA

Here are some of the other stories that caught our eye and we're following this week from around the veterinary world and animal kingdom: