This thesis is an investigation of the clinical features of sick cats and dogs admitted to a teaching hospital intensive care unit (ICU) that predict mortality. The primary goal of this work was to develop a method of applying an accurate and objective measure of illness severity to cats and dogs enrolled in clinical research. Clinical parameters associated with mortality risk were assessed over the 24 hours following ICU admission. The study populations consisted of 600 cats and 810 dogs. Mortality risk was 25.8% for cats and 18.4% for dogs. The incidence of euthanasia as the primary mortality outcome was high in both species (71.6% in cats, 96% in dogs). The study populations were randomly divided into model development and validation cohorts. Logistic regression models were developed on the construction cohorts for each species. The regression models were converted to integer scores, termed the Acute Patient Physiologic and Laboratory Evaluation (APPLE) scores, to facilitate easy manual calculation. The variables selected for the canine models included creatinine, WBC (white blood cell) count, albumin, SpO 2, total bilirubin, mentation score, respiratory rate, age, lactate, glucose, platelet count, and presence of free fluid in a body cavity. The variables selected for the feline models included mentation score, temperature, MAP (mean arterial pressure), lactate, PCV (packed cell volume), urea, chloride, and presence of free fluid in a body cavity. Area under the receiver operator characteristic (AUROC) curves of the models ranged from 0.85 (95% CI=0.80-0.92) to 0.91 (95% CI=0.89-0.96) for the canine models in the validation cohorts, and 0.76 (95% CI=0.72-0.84) to 0.88 (95% CI=0.84-0.96) for the feline models in the validation cohorts. Model calibration was good. Cross validation using only animals that died natural deaths indicated that the impact of euthanasia did not reduce the validity of the models. Two models for both dogs and cats were developed to provide a robust, objective, and user-friendly measure of illness severity and facilitate risk stratification in small animal clinical research. The models were validated on patients independent to those participating in construction.