TITLE:
Data-Driven Insights into Dog Adoption Outcomes and Shelter Length of Stay
AUTHORS:
Sameh Shamroukh
KEYWORDS:
Animal Shelter, Dog Adoption, Length of Stay, Logistic Regression, Random Forest, Cox Proportional Hazards, Animal Welfare
JOURNAL NAME:
Voice of the Publisher,
Vol.12 No.4,
October
28,
2026
ABSTRACT: Animal shelters face ongoing challenges in managing dog populations while maximizing successful outcomes and minimizing length of stay (LOS). This study analyzes factors associated with adoption versus euthanasia among Sonoma County dog shelter episodes restricted to these two recorded outcomes, using the Sonoma County Animal Shelter Intake and Outcome (ASIO) dataset (County of Sonoma Department of Health Services, 2026), comprising approximately 6000 canine records after filtering. Because return-to-owner, transfer, and other outcomes were excluded, the reported classification probabilities apply only to the adoption-versus-euthanasia comparison. They should not be interpreted as adoption probabilities for all shelter intakes. Key demographic and intake-related variables—including age, size, breed, coat color, sex, intake type, intake subtype, and intake condition—were examined using statistical and machine-learning approaches. Logistic regression and random forest models were applied to classify adoption versus euthanasia. In contrast, a Cox proportional hazards model was used to evaluate time to any recorded shelter exit in the restricted sample. Results indicate that intake condition, age, and size are the most influential predictors of adoption versus euthanasia within the restricted two-outcome sample. Dogs entering shelters in good condition and younger or smaller dogs, particularly puppies, were more often classified as adopted rather than euthanized and generally exited the shelter sooner. Conversely, older dogs, larger dogs, and those with treatable or unknown intake conditions were more often classified as euthanized rather than adopted and were associated with slower shelter exit. Model performance was strong across methods, with both classification approaches achieving 84% - 85% accuracy and ROC AUC values above 0.91. These findings highlight the importance of early identification of dogs at elevated risk of euthanasia or prolonged shelter stays within the restricted analytical population. By leveraging data-driven insights, the Sonoma County shelter system can implement targeted interventions—such as enhanced visibility, foster placement, or behavioral support—to improve outcomes, reduce overcrowding, and enhance animal welfare. The findings should not be generalized to return-to-owner, transfer, or other excluded outcome pathways.