Should Shelters Bother Assessing Their Dogs?
A recent paper argues shelters waste their time testing dogs for aggression
Posted Aug 19, 2016
Earlier this summer a man in Benicia, CA, was attacked by a dog named Frazier he had adopted just the day before from his local shelter. Leiv Arnesen told TV station KTVU, “It was full on Cujo. One minute he was fine, then I stood up, and he went at me.” Mr Arnesen needed 30 stitches to his arm.
His ordeal immediately led to questions about the shelter: How had they determined that this dog was safe to be placed in a home? The shelter’s veterinarian was quoted as saying, “We couldn’t touch him. We couldn’t get near him to give him a medical exam because the dog would not let us.” But the manager insisted the dog was safe.
This sad incident led to lively discussions in social media and elsewhere about the adequacy of the steps shelters take to identify whether a dog is safe to be adopted. Some have argued that, in an understandable desire to reduce the number of animals they euthanize, certain shelters might be adopting out dogs that are a danger to society. Increased scrutiny focused on the procedures used to determine a dog’s fate.
In a recent issue of the Journal of Veterinary Behavior, Gary Patronek of Tufts University and Janice Bradley of the National Canine Research Council (NCRC), in a paper sponsored by the NCRC, take a radically different point of view. They argue that, rather than trying to refine the tests that are used in shelters to determine whether a dog is suitable for rehoming, shelters should entirely abandon these procedures.
Their logic is tightly argued and mathematically complex, but its essence is not hard to grasp.
To see their point, let’s step away from finding homes for dogs for a moment and think about breast cancer. Every year roughly a quarter million women in the United States are diagnosed with breast cancer. Mammograms have 84% sensitivity in diagnosing this condition. Eighty-four percent sounds pretty good. However, if a woman has 10 annual mammograms, by the end of that decade she has a more than 50% chance of obtaining a false positive result. This much is just arithmetic. If a test for a disease is given to way more people than actually suffer from that condition, then, even if the test is pretty accurate, the small proportion of false positives must add up to a lot of people.
Now let’s go back to thinking about dogs in shelters. Although dogs bite hundreds of thousands of people every year, even if shelters had a test that was 84% accurate in predicting aggression in the home (and presently no test is anywhere close to that level of sensitivity), there would be immense numbers of false positives. Although a false-positive claim of breast cancer must be pretty traumatic, follow-up tests will ultimately lead to an all-clear diagnosis. But for a dog falsely diagnosed as aggressive, the outcome is typically euthanasia – from which there can be no return.
Patronek and Bradley ran the numbers and concluded that, on any reasonable set of assumptions about the frequency of aggression in dogs, and the best-case scenario for the sensitivity of a test that could be used to identify these dogs in the shelter, there would be just as many false positive as ‘true detections’ of dangerous dogs. Hence the title they gave their paper, “No Better than Flipping a Coin.”
Although their logic thus far is impeccable, I think their conclusion, that shelters should give up on any attempt at behaviorally testing the dogs in their care before letting them out into homes, is flawed. Here’s why.
First: Patronek and Bradley assume that shelters already have a watertight method of “… screen[ing] out from adoption obviously dangerous dogs during the intake process.” This would be news to Mr Arnesen in Benicia. The dog that mauled him had apparently been passed as adoptable by the shelter manager, while the shelter vet couldn’t get close to it. That two experienced professionals could reach such divergent conclusions about a dog undermines the idea that serious aggression is so obvious that there is no need for a formal test prior to making a dog available for adoption.
Second: Think again about the breast cancer example. No one test for breast cancer could ever be 100% sensitive and not produce false positives. The solution in human medicine is to have multiple forms of test which converge on a reliable diagnosis. This is surely the approach we should be taking with shelter dogs. By all means (as Patronek and Bradley advocate), if a relinquishing owner is available, let’s ask them about the dog (though in our experience, people giving up pets are reluctant to ‘fess up to a dog’s less stellar qualities, and most dogs in shelters were not surrendered by prior owners); if overt aggressive behaviors are visible in day-to-day interactions with the veterinarian or staff, a good shelter should be noting those in the dog’s record. But it would be crazy not to be working on refining behavioral tests that predict behavior in the home. Just because the tests currently available are not very good, that does not mean, as Patronek and Bradley conclude, that the ambition to develop a better test is misguided.
Third: Patronek and Bradley only focus on aggression. In Anna Karenina, Tolstoy famously wrote, “All happy families are alike; each unhappy family is unhappy in its own way.” We don’t know whether he was including the family pets when he wrote that, but the principle certainly applies across the species boundary. People abandon their pets for many reasons: not just aggression. And conversely, many people keep their dogs, even though there may be issues with problem behaviors. How and why people and dogs form a bond is a complex phenomenon which is presently only very poorly understood. Behavioral testing could be helpful in matching dogs to homes, quite independent of any question of identifying dangerous dogs.
Employers today deploy aptitude tests to find the people best suited for their open positions. There is no principled reason why tests could not be developed that could identify, not just whether a dog is likely to be a danger to people, but what kind of homes a dog might thrive in: homes with other dogs, with children, with cats, with a lot or little opportunity for outside exercise, and so on…
Psychological testing started in the early twentieth century. Early practitioners pushed out tests that were poorly validated, and, after an early boom, by the 1920s there was an inevitable bust. Some argued that psychological testing was impossible and the field fell into disrepute. Gradually over the ensuing decades, however, a formal scientific basis was developed and we now have a science of testing people which is deployed to widespread advantage.
The situation with dog behavioral testing is somewhat like that of human psychological testing a century ago. The fact that the tests available to shelters now are of limited utility is a reason to refine, not abandon, them.
The dog that mauled Mr Arnesen in Benicio was euthanized. Let us strive for a future where, if any dogs are going to be put to sleep in shelters (and hopefully far fewer than at present), dogs like Frazier are identified and euthanized before and not after they have sent a person to the hospital.
One more thing...
After I published this piece, a reader wrote and pointed out that my analysis - following Patronek and Bradley's - only considers the risk to the dog. What about risk to human families of adopting a dangerous dog?
Patronek and Bradley assumed in their article that 16% of dogs placed in homes are a potential danger to their human hosts. Using Patronek and Bradley's assumptions about the impact of behavioral testing on the rate of identification of these dogs, my correspondent calculates that tests presently in use, reduce that risk to just 3%. That's an 80% reduction in the placement of dangerous dogs. Let me say that again with emphasis, present behavioral tests, inadequate as they are, reduce the placement of dangerous dogs into human homes by 80%. That's certainly "better than flipping a coin."