Statistical Triage Model for Feline Infectious Diseases in a Veterinary Isolation Unit: The Case of Feline Immunodeficiency and Leukemia Viruses.

Abstract

Isolation units in veterinary hospitals are essential for managing infectious diseases, but limited resources in staff, equipment and budget make timely triage difficult. In this scenario, data-driven models, if trained with high-quality data and presenting good accuracy, might be a rapid and low-cost first step before more costly diagnostics are deployed to unclear cases. In this study, we aimed at constructing simple triage models for two common feline infections using routine-collected data from 1211 cats admitted to a Biological Isolation and Containment Unit (BICU) in Portugal, of which 640 were eligible for analysis. Our study focused on infections caused by Feline Immunodeficiency and Leukemia viruses (FIV and FeLV) due to their potential to induce immunosuppression and increased susceptibility to secondary diseases. Our model training was based on data from 134 FIV-seropositive cats, 126 FeLV-seropositive, and 504 confirmed non-cases (i.e., controls) of these infections diagnosed by rapid immunomigration assays and ELISA. Significant triage factors of FIV infections included older age, outdoor access, and concomitant disorders whereas, for FeLV, infections, they were mixed-breed status, concomitant conditions, and low hematocrit. The models showed areas under the receiver operating characteristic curve of 0.70 and 0.69 for FIV and FeLV infections, respectively, with sensitivities and specificities estimated above ≥65%. This performance, being far from sensitivities and specificities above 90%, motivates further research to find additional triage factors to improve model utility.