A multicenter, retrospective, cross-sectional analysis of fresh and processed

Introduction: Objective: To develop a regular expression-based method for identifying cases of Methods: A retrospective cross-sectional analysis was performed using emergency-care electronic health records from 64 United Kingdom-based private veterinary emergency centers between 2018 and 2024. Iteratively refined Python-based regular expressions were developed and validated to identify suspected or confirmed ingestion from free-text clinical narratives. Cases with confounding toxicities or incomplete records were excluded. Demographic, treatment, and outcome data were extracted. Mortality analyses were conducted by species with correction for multiple comparisons. Decision tree models were developed for canines and felines using stratified training and testing datasets with Bayesian hyperparameter optimization. Model performance was assessed using average precision on held-out data. Results: 31,425 probable canine cases and 1,030 probable feline cases were identified. Processed fruit ingestion was more common than fresh fruit ingestion in both species (canines: 61.6% versus 31.2%; felines: 65.0% versus 31.1%), with seasonal clustering in December (canines: 22.0%; felines: 15.5%). Mortality was 1.3%, significantly higher in felines (11.8%) than in canines (0.9%). Fresh fruit ingestion was associated with higher mortality in both species (canine relative risk 3.47; feline relative risk 3.13). Age was associated with mortality in both species, and increased weight was associated with mortality in canines only. Chocolate co-ingestion was not associated with increased mortality. Treatment differed by species, with emesis more common in canines and intravenous fluids more frequent in felines. Decision tree analysis identified decontamination as the strongest predictor of survival. Clinical significance: