Using a Canine Companion Training Curriculum to Reduce Return Rates in a No-Kill Animal Shelter Environment

1
Using a Canine Companion Training Curriculum to
Reduce Return Rates in a No-Kill Animal Shelter
Environment

• Lisa R. McCluskey, Alexandra Tellier,
• Elizabeth L. Fay, Jeffrey J. Klunk and Kevin Small

2
Introduction

• The original idea for this work was attending Sue
  Sternbergs 2011 APDT talk regarding dog trainers
  volunteering in shelters
• Wanted to quantify the effect of instituting a
  training program at a shelter (and hopefully
  influence facility policies/culture)
• One important goal of animal shelters is reducing
  the return rates both for tangible (e.g.,
  space) and intangible (e.g., heart warming
  stories, community philanthropy) reasons
• Preventing returns is particularly pertinent in
  no-kill shelter environments as space tends to be
  highly constrained

3
Specific Aims

• The aim of this study is to measure the
  association between dog training and return rates
  in a no-kill shelter
• Odds Ratio What are the odds of a successfully
  adopted dog having received training relative to
  the odds of a returned dog having received
  training?
• Risk Ratio What is the probability that a dog
  receiving no training will be returned relative
  to a dog that has received training?
• Survival Function How long would you expect a
  dog from a specific population to stay adopted?
• Proportional Hazards Ratio How long would you
  expect a dog with a specific property value
  (e.g., breed) to remain adopted relative to other
  dogs with a different value for that property?

4
Study Parameters

• Population Dogs admitted to WARL before 7/31/12
  and adopted between the dates of 10/26/11 and
  7/31/12
• Exposure Dog training did the dog attend at
  least one training class before or after it was
  adopted?
• Control Dogs that received no training from the
  program
• Outcome Did the dog get returned or not?
• Censored Data Any dog with an unknown intake
  date, adoption date, gender, intake type, intake
  age, or breed was excluded from the study
  (missing at random).

5
Worcester Animal Rescue League (WARL)

• 100 year-old
• No-kill
• Non-profit
• 88 of intake are strays

• Wards at WARL
• Adoption Pavilion
• Boarding for the Public
• City Pound
• Boarding also used for emergency housing
• PetPoint Animal Management System (software)
• Animal Rescue League of Boston Behavioral
  Assessment

6
Canine Companion Training Curriculum (CCTC)

• Credentialed dog training and behavior specialist
• Two free, one hour sessions per week
• Group training, on-site, drop-in
• Canine Training Curriculum
• Basic Manners
• Intermediate Tricks
• Rally Working Trials (RWT)
• AKC CGC Skills

7
Training Class Practice Session
8
Population Breed Characteristics

• WARL classes tend to attract larger dogs
  particularly bully and guarding breeds

Breed Family Examples
Bully/Guarding Amer. Bulldog/Pit Bull Terrier
Ratters Dachshund, Rat Terrier
Terriers Cairn, Schnauzer, Scottie
Sm. Hunt/Retrieve Cocker Spaniel, Mini Poodle
Sm. Companion Bichon, Boston Terrier, Pug
Herding Sheltie, Corgi, Border Collie
Ancient Spitz Samoyed, Husky, Shiba Inu
Scenthounds Beagle, Bassett, Coonhound
Lg. Hunt/Retrieve Labrador, Golden, Pointer
Sighthounds Greyhound, Whippet, Afghan
Large Working Rottweiler, Doberman, GSD
9
Age/Gender Characteristics

• Very little difference in age of dogs (at intake)
  between dogs that attended, didnt attend
  classes shelter tends to take in adult dogs
• Classes tend to skew toward male dogs shelter is
  near 50/50 overall

10
(Nested) Case-Control Study

• Cases are dogs that have been returned
• Controls are dogs that have not been returned
• Returned dogs are 3.17 times less likely to have
  attended training
• Dogs that have attended training are 2.63 times
  more likely to stay in their adoptive home

Returned Not Returned Odds Ratio Risk Ratio
Training 2 56 3.17 2.63
No Training 19 168 3.17 2.63
11
Matching by Breed

• Cases are dogs that have been returned
• Controls are dogs of same breed as cases, but
  have not been returned
• Returned dogs are 2.81 times less likely to have
  attended training than a dog of the same breed
• Dogs that have attended training are 2.40 times
  more likely to stay in their adoptive home than a
  untrained dog of the same breed

Returned Not Returned Odds Ratio Risk Ratio
Training 2 48 2.81 2.40
No Training 19 162 2.81 2.40
12
Case-Cohort Study

• Cases are dogs that have been returned
• Controls are dogs that have not been returned who
  were adopted within one week of a corresponding
  case
• Returned dogs are 3.36 times less likely to have
  attended training than untrained dogs adopted at
  the same time
• Dogs that have attended training are 2.69 times
  more likely to stay in their adoptive home than
  an untrained dog adopted at the same time

Returned Not Returned Odds Ratio Risk Ratio
Training 2 40 3.36 2.69
No Training 19 116 3.36 2.69
13
Survival Analysis (Kaplan/Meier)

• What is the probability that a dog remains
  adopted for a specified number of days?

Weibull estimate for right-censored data
14
Cox Proportional Hazards Regression

• Explores the relationship between time in
  adoptive home and several explanatory variables
  simultaneously
• A hazard function is the probability that an
  individual will experience an specified event
  (e.g. being returned) within a small interval of
  time
• Note that we are not estimating the baseline
  hazard function, but the increase in hazard by
  possessing specific properties
• For our study, baseline hazard is for an
  untrained, female, adult (at admission),
  bully/guarding dog

15
Cox Model Results

• Baseline is untrained, adult, bully/guarding,
  female dog
• Dogs who receive training are 3.11 times less
  likely to be returned for a given time period

Covariate Coefficient exp(-Coef) p-Value
Received Training -1.135 3.110 lt 0.001
Age Juvenile -0.125 1.134 gt 0.1
Age Unweaned 1.558 0.211 lt 0.001
Age Young Adult -0.243 1.275 gt 0.1
Gender Male 0.134 0.875 gt 0.1
16
Cox Model Results

• Baseline is untrained, adult, bully/guarding,
  female dog

Covariate Breed Coefficient exp(-Coef) p-Value
Ratter -0.531 1.701 lt 0.05
Terrier -0.376 1.457 gt 0.1
Sm. Hunt/Retrieve -0.130 1.139 gt 0.1
Sm. Companion -0.394 1.484 gt 0.1
Herding 0.703 0.495 gt 0.1
Ancient Spitz 0.026 0.974 gt 0.1
Scenthounds -0.853 2.348 lt 0.01
Lg. Hunt/Retrieve -1.447 4.250 lt 0.001
Lg. Working 0.098 0.907 gt 0.1
17
Other Potential Covariates

• Dog Characteristics
• Jurisdiction of Origin
• Altered before Intake
• Reason for Surrender
• Adopter Demographics
• Gender of Primary Caretaker
• Family Structure (e.g., children, elderly)
• Neighborhood of Residence
• Need more dog data to explore variables further

18
Quantifying the Effect
19
Conclusions

• We provide preliminary, but strong evidence that
  a canine companion training curriculum (CCTC) can
  significantly reduce the adoption return rate in
  a shelter
• The recurring theme seems to be trained dogs are
  three times less likely to be returned than
  untrained dogs
• There are also important intangible (for the time
  being) benefits including increased community
  involvement and volunteer enthusiasm

20
Future Work

• With more data and longer follow-up, a cohort
  study framework might be more conclusive (in
  progress)
• Also, confounders could be better managed if we
  had more data
• A standard one-number definition of outcome
  success should be defined (e.g., 5-year survival
  for cancer)
• The same methodology could be applied to
  euthanasia rates and adoption rates with
  sufficient data
• We would like to perform a multi-center study
  (please contact us if you have PetPoint data you
  would like to share!)

21
Final Words

• Volunteer in your local shelter!
• The dogs need you
• and they will make you an even better
  administrator ?

• Discuss with local dog training professionals how
  an on-site shelter program will likely reduce
  return rates and keep dogs in their adoptive
  homes longer.

22
Thank You!www.moongazer.com/training