Objective: To estimate the sensitivity and specificity of eye temperature (ET) and rectal temperature (RT) for detecting abnormal temperature in dogs. Methods: This single-shelter observational study enrolled dogs without clinical ocular disease. Tests were dichotomized using prespecified reference intervals. A hierarchical Bayesian latent-class model included 2 subpopulations (puppy and adult), dog-level random intercept, and between-test dependence terms. Posterior draws estimated predictive values across pretest probabilities, and prior sensitivity analysis was conducted to evaluate the robustness of the results. Results: Utilizing 238 visit-level observations from 121 dogs (89 puppy visits; 149 adult visits), both tests demonstrated high specificity but low-to-moderate sensitivity. Posterior median sensitivities were 0.40 for ET and 0.32 for RT, while specificities were 0.94 for ET and 0.99 for RT. The parallel rule (OR any positive) maximized negative predictive value, whereas the series rule (AND both positive) maximized positive predictive value across observed prevalences. Conclusions: Ocular thermography and rectal thermometry demonstrated similar diagnostic performance. For both tests, given that the dog has an abnormal core body temperature, the reading is more likely to be wrong. Given the dog has a normal core body temperature, the reading is likely to be correct. Abnormal temperature readings can inform clinical judgement about the likelihood that the dog's core temperature is abnormal. However, normal readings are not informative about what was already known based on clinical judgment. Clinical Relevance: Abnormal results from either test increase the probability of abnormal body temperature, whereas normal readings should not override clinical suspicion of abnormal temperature status.
Rostami et al. (Mon,) studied this question.