Shelter Medicine Meets Analytics
What is Shelter Medicine?
Let’s start by clarifying the distinction between Shelter Medicine and the typical veterinary care your dog receives when you bring him into your local vet for his annual check-up. The Cummings School of Veterinary Medicine at Tufts University defines Shelter Medicine as, “a field of veterinary medicine dedicated to the care of homeless animals in shelters or other facilities dedicated to finding them new homes.”
The first formal shelter medicine classes for veterinarians didn't take place until the 1990’s and the University of California, Davis’s Veterinary Medicine program was the first to create a shelter medicine residency program in 2001. Since that time, UC Davis established the Koret Shelter Medicine Program. Within Koret, numerous different programs have been established to facilitate both the practices of shelter medicine and awareness of the benefit of those practices.
Give Us Big Data ‘Meow’
In today’s world of big data and machine learning, the potential impact analytics can, is, and will have on the field of Shelter Medicine is overwhelming. Koret’s Capacity for Care (C4C) program is driven by the idea that shelter capacities are dependent on more factors than the number of kennels located within a facility. Currently, they work with the shelter’s across the nation to aid in their determination of an ideal capacity for their facility that maximized the number of animals processed while minimizing the number of undesirable outcomes (e.g., euthanasia). The process of determining the optimal capacity level for facilities is a huge opportunity for analytics to be leveraged. Additionally, there are a couple of other areas within this domain in which analytical techniques can be used to both optimize and predict different aspects of an animal shelter’s operations.
1. Capacity Optimization
Using an optimization model to recommend capacity levels for animal shelters based on a number of factors has the potential to revolutionize the way shelter administration looks at processing levels. Categorizing cats and dogs by their level of medical need in addition to their average length of stay in a shelter is data that is currently being used to develop such a model that will ideally allow any shelter to better understand what their optimal capacity level is and therefore direct their steps in the right direction to achieve such levels.
2. Clustering for Comparisons
Also currently in development is a clustering system that will allow animal shelters to compare their capacity levels to other shelters that have similar operation practices. This clustering is reliant on shelters being grouped based on a mix of qualitative and quantitative attributes including, but not limited to, the average annual number of animals processed and type of shelter ranging from non-profit rescue to a municipal facility.
By allowing animal shelters to compare themselves with one another, best practices and goals can be shared. Additionally, as animal shelters slowly move into the age of big data and data-driven solutions, shelters behind in the times can compare their objective with similar shelters that ideally already have a functioning data system.
3. Predictive Modeling for Epidemiology Based Issues
One of the main reasons why, in some cases, reducing the number of animals held in a shelter at any given time can actually increase the annual number of animals processed is the spread of disease within shelters. For example, feline herpesvirus type 1 (FHV-1) is transmitted commonly via, “sharing litter boxes, food, and water dishes with an infected cat, as well as by mutual grooming.” While it is a well-known fact that FHV-1 is one of the most commonly transmitted diseases within shelters, requiring additional medical attention for shelter animals, there is little data-based research used to predict the spread.
Using a regression best technique such as logistic regression to predict the probability of an animal being infected, or training a model to predict the rate of spread would provide valuable information for shelters to help them better predict resource needs. While Koret strives to raise awareness for C4C, many government-run or subsidized shelters have little to no control over their intake. Using analytics, not only can seemingly unpredictable intake be predicted, but consequences such as diseases spreading via shared resources can all be predicted.
Before We Can Move Fur-ward
Currently, there is no universally adapted software or system to record capacity levels of animal shelters across the United States. While some programs work to collect data from shelters they often come into contact with, without a more robust database potentially valuable insights are lost.
Attempting to increase the amount of data science-driven aspects of animal shelter management and shelter medicine comes with the risk of dehumanizing the process. Throughout numerous conversations with members of the Koret Shelter Medicine and my own colleagues, we are constantly halted in our brainstorming sessions after momentarily losing track of the end goal in an effort to maximize efficiency. The overarching goal of shelter medicine is to better serve the animal population that interacts with shelter systems. In order to pursue this mission, there are ethical lines that cannot be forfeited.
The impact of the three above mentioned opportunities for shelter medicine to blend with data science/analytics goes beyond accomplishing the immediate impact of aiding animals. The long-run effects of utilizing analytics in the shelter medicine domain include raising awareness of shelter medicine best practices in addition to promoting a more strategic approach to shelter administration that can save potentially millions of animal lives moving forward.