This paper investigates the contributors to the different outcomes of shelter animals in Los Angeles County Animal Care & Control by using interpretable neural network models. Three interpretable neural networks: Neural Additive Models (NAM), GAMI-Net, and Kolmogorov-Arnold Networks (KAN) are selected and compared against multilayer perceptron on a pipeline of three classification tasks for identifying four animal outcome groups based on breed, sex, length of stay, age, weight, etc. GAMI-Net yielded similar performance as multilayer perceptron in balanced accuracy and Cohen's Kappa, while NAM and KAN performed worse, suggesting a trade-off between performance and interpretability. All interpretable models agree that length of stay is the most important contributor to model performance, while other factors, including age, weight, and readiness for adoption, are also significant.