Construction of a Diagnostic Prediction Model for Feline Nasal and Nasopharyngeal Diseases in Japan Using Noninvasive Examinations.

Abstract

Background: Although feline nasal and nasopharyngeal diseases (NNDs) often require advanced tests under general anaesthesia for definitive diagnosis, not all patients can undergo them.

Objectives: This study aimed to construct diagnostic prediction models for feline NNDs in Japan using noninvasive examinations, signalment and history.

Methods: Seventy-nine cats diagnosed with NNDs, including representative diseases in Japan-nasal and nasopharyngeal tumours (NNT), rhinitis (RS) and nasopharyngeal stenosis (NPS)-were retrospectively investigated to construct prediction models (model group, GM). Thirty-nine cats diagnosed were prospectively investigated to validate their efficacy (validation group, GV). Three predictive models were developed: Models 1 and 2 were manually constructed, with Model 1 designed to predict NNT, RS and NPS individually and Model 2 distinguishing between these diseases. Model 3 was constructed using least absolute shrinkage and selection operator logistic regression. Sensitivity, indicating the ability to identify cases of each disease, and specificity, reflecting the ability to exclude other diseases, were used to assess performance.

Results: In Model 1 of the GV, the sensitivity and specificity for NNT, RS and NPS were 1.00 and 0.73, 0.62 and 0.96 and 0.78 and 0.97, respectively. In Model 2 of the GV, the values were 0.94 and 0.86 for NNT, 0.77 and 0.92 for RS and 0.75 and 0.94 for NPS. In Model 3 of the GV, they were 0.94 and 0.05 for NNT, 0.25 and 1.00 for RS and 0.13 and 0.84 for NPS.

Conclusions: The diagnostic prediction models, particularly Models 1 and 2, could help estimate whether advanced tests are necessary.