Pet lifesaving dashboard: Map methodology

Sources for shelter data include the Shelter Pet Data Alliance and other voluntary data submissions, public websites, and government-provided data. Shelters that operate multiple locations may show data in aggregate or broken down by location. Their information will be presented on the map as it is reported to us. While Best Friends attempts to validate data sources, we cannot guarantee the accuracy of these sources.

The list of brick-and-mortar shelters nationwide is maintained year-round by Best Friends to track closures, new organizations, and government contract changes.
 

We consider current data to be 12 consecutive months of data within the past 24 months. While much of the data will align with a calendar year, any 12 consecutive months of data are valid because they sufficiently account for seasonality. This also allows for the display of data more recent than the past full calendar year. Only shelters with current data will display specific data. While older data is not visible, all shelters will display whether they are no-kill.

Projections utilizing old data are based on stratified regression modeling of collected data. Community data is used to adjust year-over-year trends, specifically the CDC’s Social Vulnerability Index, county population, and county rurality for intake and both Social Vulnerability Index and organization type for non-live outcomes.

For any shelter that has not provided 12 consecutive months of data within the past 24 months but has provided at least one older calendar year of data, a machine learning model was used to predict whether that shelter is no-kill. The model was trained on collected 2024 data, and predictors included previous years’ shelter data and community characteristics at the state and county levels available through the CDC, U.S. Census Bureau, and other publicly available sources. Model review and validation was completed by two external, independent, uncompensated university professors who are experts in statistical modeling techniques:

Evaluation of Best Friends Animal Society’s Predictive Models for No-Kill Shelters in the United States by Dr. Adam Feltz, Uyen Hoang, and Jenna Holt, The University of Oklahoma

Best Friends Animal Society: Report on Estimation Validity of Regression Modeling by Courtney Paulson, Ph.D., University of New Hampshire

The model was found to be over 85% accurate and highly reliable at predicting whether a shelter is no-kill. All estimates are replaced with current collected data when it is received.

Data at the dog or cat level is visible when designated within each shelter’s data. This species-level data is blank if a shelter’s data is reported with dogs and cats combined or if the shelter included at least some of their data without specifying species. Some shelters may also only take in either dogs or cats, resulting in the other species data being blank.
 

A gross save rate calculation is used for individual shelters:

Owner-requested euthanasia is included in both intake and non-live outcomes when the pet is taken directly by the shelter and not through a public clinic.
 

The no-kill gap is the number of cats and dogs who would have had to be saved to achieve the 90% save rate threshold. Best Friends applies the following calculation to determine that gap:

Owner-requested euthanasia is included in both intake and non-live outcomes when the pet is taken directly by the shelter and not through a public clinic.

The resulting number of dogs and cats killed is the focus of Best Friends efforts to achieve the no-kill benchmark and does not include the estimated 10% who are euthanized based on presumed humane decisions related to irremediable suffering from medical or behavioral conditions.
 

A “shelter” is defined as a brick-and-mortar facility with open adoption hours for the public.

• For organizations without a government contract, they must take in more than 200 animals annually, be open to the public at least two days a week, and have a non-residential sheltering facility.
   
• For government animal services and organizations with a government contract, they must take in more than 20 animals annually and cannot transfer all their animals to another shelter. In limited situations a foster-based rescue group without a brick-and-mortar facility can be included if they hold a government contract.

Shelters operated by sovereign nations are not included in the map in consideration with the Bureau of Indian Affairs (BIA) declaring that “federally recognized tribes are recognized as possessing certain inherent rights of self-government (i.e., tribal sovereignty).”
 

Shelters that meet the no-kill benchmark of 90% will be shown as no-kill using either current data or estimated data using projections based on older data. Nearly no-kill shelters are those that do not yet meet the no-kill benchmark but have a save rate of at least 85%. Shelters designated as not no-kill have a save rate below 85%.

While the 90% benchmark offers a meaningful measurement by which to gauge the progress of shelters, we recognize that there may be special circumstances in which a community could be successfully implementing no-kill principles and practices but not reach a 90% save rate.

In these rare cases, the shelters that does not meet the statistical benchmark can obtain a no-kill designation in the pet lifesaving dashboard by displaying the following statement on their website and/or giving Best Friends permission to display the information on the shelter’s page of the pet lifesaving dashboard:

"Our shelter has committed to saving all savable pets entering our care. We do not euthanize healthy or treatable pets even at an owner’s request. We only euthanize a pet if:

• A veterinarian or trained medical staff under guidelines set by a veterinarian has deemed the prognosis to be poor or grave, there is no chance of recovering an acceptable quality of life, or
   
• It would be clearly inhumane or unsafe not to do so immediately, or
   
• In cases of irremediable canine aggression, (1) a veterinarian has eliminated medical treatment as a solution; (2) rehabilitation by a specialist in canine behavior has failed; and (3) staff and public safety cannot be reasonably assured, or other management protocols seriously compromise quality of life."

The goal is for every shelter to make a clear commitment to lifesaving and transparency (being honest and open about their data and operations) while striving for no-kill rather than simply working to obtain a no-kill designation.

Any shelter seeking an exception should contact the local regional Best Friends team:

• Pacific region (Washington, Oregon, California, Alaska, Hawaii): pacific@bestfriends.org
   
• Mountain West region (Montana, Idaho, Wyoming, Nevada, Utah, Colorado, Arizona, New Mexico): mountainwest@bestfriends.org
   
• Great Plains region (North Dakota, South Dakota, Minnesota, Nebraska, Iowa, Kansas, Missouri): greatplains@bestfriends.org
   
• South Central region (Oklahoma, Arkansas, Texas, Louisiana): scregion@bestfriends.org
   
• Midwest region (Wisconsin, Illinois, Michigan, Indiana, Ohio, West Virginia): midwest@bestfriends.org
   
• Northeast region (Maine, New Hampshire, Vermont, New York, Massachusetts, Connecticut, Rhode Island, New Jersey, Delaware, Pennsylvania): northeast@bestfriends.org
   
• Mid-Atlantic region (Maryland, D.C., Virginia, North Carolina, Kentucky, Tennessee): midatlantic@bestfriends.org
   
• Southeast region (Mississippi, Alabama, Georgia, South Carolina, Florida): seregion@bestfriends.org