Number Needed to Ban: a new tool for calculating the benefits of banning particular dog breeds

A study published this month in the Journal of the American Veterinary Association (JAVMA) takes on the issue of whether breed-specific legislation (BSL) is effective. BSL is a tool used by some communities to attempt to reduce injuries from dog bites. The idea is that particular breeds of dogs are responsible for more than their share of injuries, so banning or otherwise controlling those breeds will result in a reduction in injuries. The group of breeds collectively known as “pit bulls” receive the most attention today, though other breeds (Rottweilers, Dobermans, German Shepherds) have received attention in the past.

But does BSL actually work? Experts say no; how the dog is trained and managed is a better predictor of aggression than its breed. Nevertheless, new BSL continues to be enacted. So why do legislators reach for this tool?


The authors of “Use of a number-needed-to-ban calculation to illustrate limitations of breed-specific legislation in increasing the risk of dog bite-related injury” believe that BSL’s appeal comes from:

• Misperception of risk. Poor reporting of the number of dog bites that occur and of their severity makes it very difficult for the public to get a handle on how often they occur.
• Stereotyping and misinformation. The media may portray particular breeds as especially aggressive, in the face of scientific studies which suggest that they are not.
• Erroneous beliefs about efficacy of BSL. There is currently no evidence for the effectiveness of BSL, but  there is evidence to suggest that it is ineffective.

The authors hope to provide a tool for use in understanding the effectiveness of BSL, and they hijack some terminology from the medical community to do so. “Number needed to treat” (NNT) is a concept used to understand the effectiveness of a particular medication or therapy. For example, you have a patient showing signs of a stroke. Should you give him tPA (tissue plasminogen activator)? One measure you might use in making this decision is NNT. How many similar patients would you treat with tPA, on average, before you saw one patient improve? A smaller NNT implies a more effective therapy. In human medicine, we expect the NNT of an effective therapy to be in the tens or at most hundreds.

The authors suggested evaluating BSL’s effectiveness using a “number needed to ban” (NNB) concept. If BSL is implemented in a particular community, how many dogs will need to be banned (removed from the community) before one dog bite (or dog bite related injury, or dog bite related fatality) is prevented?

The authors point out that because our knowledge of the true prevalence of dog bites is so poor (many are never reported), this calculation is hard to do. I think the important thing to understand is that what they are offering is a tool that can be applied to different statistics. After all, dog bite prevalence will vary among different communities. This tool can be used to understand the possible benefit of BSL in different communities. It’s an algorithm to apply to a variety of data inputs!

However, the paper would have been really unsatisfying without some numbers, so they applied their algorithm to some statistics (much appreciated, because I hate arithmetic).

• Based on the reported number of dog bite related emergency department visits, 5,128 dogs would have to be banned to prevent a single emergency department visit in one year.
• In Kansas City, 4,255 dogs would have to be banned to prevent a single emergency department visit in one year.
• 30,663 dogs would need to be banned to prevent a single reconstructive surgery in one year.
• 109,495 dogs would need to be banned to prevent a single hospitalization in one year.
• 59,523 dogs would need to be banned to prevent a single insurance claim in one year.

The authors note that these calculations were based on legislation banning a particular breed or breeds entirely. For legislation which simply requires that dogs of a particular breed(s) be muzzled while in public, these numbers would be even higher, because such legislation would not prevent bites on private property (which is where many of them occur).

It is the authors’ hope that “easily understood communication tools, such as NNB, can help put the lack of efficacy of BSL into perspective and narrow the perception gap.” This is a great tool and I hope we see it used more. I am concerned that proponents of BSL will argue that any tool is only as good as the data put in to it, and that the lack of reliable reporting of dog bites will mean that this tool isn’t itself reliable. However, as long as we are focusing on enacting BSL instead of focusing on understanding the true problem, our data will continue to be flawed. This article represents a step forward in understanding data about the causes of dog bites. Our next step is improving the accuracy of that data.

Patronek GJ, Slater M, & Marder A (2010). Use of a number-needed-to-ban calculation to illustrate limitations of breed-specific legislation in decreasing the risk of dog bite-related injury. Journal of the American Veterinary Medical Association, 237 (7), 788-92 PMID: 20919843