How Accurate Are Bed Bug Detection Dogs? What the Studies Actually Say

Ask how accurate bed bug detection dogs are and you will get one of two answers, depending on who is selling what. There are two peer-reviewed studies on bed bug detection dog accuracy, both published in the Journal of Economic Entomology, and almost every claim about canine accuracy, for or against, traces back to one of them.

The first is from the University of Florida in 2008 (Pfiester, Koehler and Pereira). In controlled tests, trained dogs found live bed bugs 97.5 percent of the time, with no false alerts on other insects such as cockroaches, ants and termites. In hotel room trials with hidden vials they found live bugs 98 percent of the time and did not alert on rooms with nothing in them. This is the number canine companies quote.

The second is from Rutgers University in 2014 (Cooper, Wang and Singh). The researchers took working bed bug detection dog teams from commercial companies into occupied apartment buildings in New Jersey and compared their alerts against the true infestation status, established by two weeks of interceptor traps plus visual inspection. The average detection rate across teams was 44 percent, and the average false positive rate, meaning an alert with no bed bugs present, was 15 percent. This is the number pest control companies quote when they want to talk you out of a canine inspection.

Both studies are real and both were done properly. Anyone who cites only one of them is selling you something.

The Florida study tested a small number of dogs trained by the researchers themselves, under conditions the researchers controlled. It answers the question: can a dog be trained to find bed bugs reliably? The answer is yes, to a very high standard.

The Rutgers study tested eleven commercial teams as they actually worked, in real apartments with clutter, cooking smells, low-level infestations and paying clients. It answers a different question: how good is the average team you can hire? The answer, at least in that market in 2014, was not very.

The detail that matters is buried in the range, not the average. Detection rates across teams ran from 10 percent to 100 percent. False alert rates ran from zero to 57 percent. One team in the study went 100 percent detection with no false alerts on its first day. The same handler dismissed three of four alerts on day two, deciding the dog was playing him for a reward, and by day three the team's detection rate was 17 percent. Same dog. Same building. Different handler decisions.

Every company in the Rutgers study claimed 95 percent accuracy in its own in-house training exercises. Their field results said otherwise. The gap between what a team scores on hides it set up itself and what it scores in a stranger's apartment is the whole story of canine accuracy.

The sceptics are right that hiring a bed bug detection dog team is not automatically accurate. A poorly maintained team can miss most of what is there and alert on things that are not. The Rutgers study documents that clearly, and it should be read by anyone in this business.

They are wrong about what it means. The study did not find that dogs cannot detect bed bugs. It found that the dog and handler are the variable. The best teams in the field performed at the level the Florida lab work predicted. The worst performed at a level where you would be better off with a flashlight. Neither result says anything about a team you have not tested.

There is also a comparison the sceptics leave out. To establish the true infestation status of each apartment, the Rutgers researchers did not rely on visual inspection. They used two weeks of interceptor traps as well, because a visual inspection alone misses low-level infestations. The tool being offered instead of a dog, a technician looking at a mattress, is the one the researchers themselves would not trust to define the answer.

The Rutgers authors ended their paper with recommendations, and they read like a checklist for choosing a bed bug detection dog company. Teams should be tested in real infested buildings, not just on hides they set themselves. Training should include searches where no bed bugs are present, so a dog is not rewarded for alerting in every room. Alerts that cannot be confirmed should not be reported as positive. And inspections should be refused when conditions make a reliable search impossible, such as a unit sprayed the day before.

Notice that every item on that list is about the handler and the programme, not the breed of dog. The Florida researchers said the same thing from the other direction: dogs work best with the same handler every time and need reinforcement training almost daily. A dog that was excellent when certified will drift if nobody keeps it sharp.

One more finding is worth knowing. Rutgers found no relationship between a team's certification status and its detection rate. A certificate proves a team passed a test on one day. It does not prove the team has been maintained since. Ask what the team has done in the last month, not the last year.

Our teams search every area twice, and an alert counts as a finding only when it repeats in the same location on both sweeps. A single-sweep response is recorded as interest, not as a positive. After the dog is out of the room, the handler goes back to each alert location by hand and with a flashlight looking for live bugs, eggs, shed skins or fresh spotting, and the report says separately what the dog indicated and what the handler saw.

We detect and report. We do not sell treatment, so a positive result earns us nothing beyond the inspection fee, and a negative result costs us nothing. That removes the incentive the Rutgers study was worried about when it warned against reporting unconfirmed alerts.

We ask about treatment history before every search, because chemical treatment inside the previous fourteen days reduces detectable odour. Where the conditions would compromise the search we say so before it starts and recommend waiting, and the pre-inspection conditions are recorded in every report so a reader knows what the result is worth.

The rest of the answer is training, and it is the subject of its own guide, because the handler's half of the team is the half most people never ask about.

Not "how accurate are your dogs." Everyone answers 95 percent. Ask instead: when did the team last run a blind search, meaning one where the handler did not know whether anything was hidden or where, and what happened? A team that runs those regularly will tell you without hesitating. A team that does not will change the subject.

So, how accurate are bed bug detection dogs? As accurate as the team working them. The research supports canine detection done properly and it supports scepticism about canine detection done badly. Those are not contradictory positions, and a company that holds both is the one to hire.