What is it about?
Length of stay (LOS) in animal shelters affects animal welfare, adoption chances, shelter crowding, staffing needs, and costs. When dogs stay longer, shelters become overcrowded even if intake numbers are stable or declining. For this reason, shelters need reliable ways to track how LOS changes over time—by month, quarter, or year—and how it compares across shelters. Traditional LOS calculations can be misleading. Animal stays do not align neatly with calendar periods, and many dogs are still in the shelter when data are analyzed. Dogs with long stays are in the shelter during multiple reporting periods but appear in only one period’s calculation – or are excluded entirely if they haven’t left the shelter. In this study, we measure the distribution of dogs’ length of stay using the Kaplan–Meier and Cox regression methods. These statistical tools are designed for time-to-event data and correctly account for dogs whose shelter stays overlap multiple reporting periods, including those still in residence. Using data from the largest animal shelter in Orange County, California, these methods provide a more accurate and informative picture of LOS. Rather than producing a single average, the analysis describes the probability that an animal remains in the shelter on each day of its stay. This allows meaningful comparisons across calendar periods and reveals whether LOS is increasing or decreasing across the entire population of dogs—not just those with short stays. The analysis also makes it possible to identify changes in LOS and associate them with shelter operations and policies. Among other findings, LOS remained elevated after the COVID-19 pandemic in this shelter, a finding related to the continuation of pandemic-era restrictions on visitor access to dog kennels.
Featured Image
Photo by Sasha Sashina on Unsplash
Why is it important?
This research provides a practical framework that shelter administrators, animal welfare organizations, and local governments can use to better understand length of stay and how it responds to operational and community changes. More accurate measurement supports data-driven decisions that improve animal welfare, reduce overcrowding, and allocate shelter resources more effectively. Simple averages, while convenient, can obscure the fact that shelter overcrowding is often driven by a small percentage of dogs with very long stays—the tail of the LOS distribution. The methods used in this study capture the full distribution, not just the average, allowing earlier detection of emerging problems. Because these methods enable fair comparisons across time periods and across shelters, they improve evaluation of staffing levels, policy interventions, and operational changes that affect animal welfare. They can be applied in any large shelter using routinely collected data.
Perspectives
This research originated in my observations at the Orange County animal shelter. In the period following the COVID-19 pandemic, the shelter was more crowded despite having fewer intakes than before the pandemic. This is a classic consequence of increasing length of stay: when each dog remains longer, the shelter fills up even if fewer animals are coming in. At the time, this change was not clearly recognized. This experience highlighted how underutilized LOS analysis is and how growing problems can go undetected if LOS is not analyzed regularly and rigorously. During that same period, many dogs had very long stays and were still in the shelter. If LOS is calculated only by averaging completed stays, those dogs do not count—yet they provide the strongest evidence that stays are getting longer. Waiting until they leave the shelter delays recognition of the problem. This was a problem that the shelter had an opportunity to act on. In 2023, restrictions on visitor access to dog kennels—originally introduced during the pandemic—were still in place. A pilot program allowed me to study how limited visitor access affected adoptions. The results were compelling and supported the conclusion that barriers to adoption were contributing to longer stays. I am pleased to report that two members of the Orange County Board of Supervisors, Janet Nguyen and Vicente Sarmiento, proposed restoring visitor access to pre-pandemic levels, and the Board unanimously approved the change.
Michael Mavrovouniotis
Social Compassion / SCIL
Read the Original
This page is a summary of: Use of Kaplan-Meier and Cox regressions in the distribution of length of stay in animal shelters for pre-specified calendar periods: Definition, computation, and examples of dog length of stay in orange county California, PLOS One, January 2026, PLOS,
DOI: 10.1371/journal.pone.0342102.
You can read the full text:
Resources
SCIL Volunteer Data Scientist Publishes Article on "Length of Stay" in Animal Shelters
Social Compassion in Legislation announcement on the research paper, making the study accessible to a general audience.
The Effect of Kennel Viewing on the Adoption of Slow-Track Dogs
Adoption rates were impacted by the restrictions on kennel access post-COVID-19, especially for slow-track dogs, who account for a minority of intakes but the majority of animals in care at any given time. This reinforces the importance of computing the entire distribution of lengths of stay and not just an average.
Comparison of the Number of Dog Adoptions in a Pilot Program That Restored Limited Visitor Access to Kennels
Viewable dogs are more likely to be adopted. It's important to allow visitors to view dogs in their kennels, in order to speed up adoptions and reduce length of stay.
Contributors
The following have contributed to this page
