A novel severity scoring system for dogs with heatstroke

Abstract

Objective: To develop a statistically-derived scoring system that can aid in severity assessment and outcome prediction for dogs with heatstroke. Design: Retrospective study. Setting: Veterinary teaching hospital. Animals: One hundred twenty-six client-owned dogs diagnosed with heatstroke. Interventions: None. Measurements and Main Results: Logistic regression analysis was performed to identify clinicopathologic variables, available in the first 24 hours of hospitalization, which were associated with outcome (P ≤ 0.1). These were subjected to further analyses. In Model A, continuous variables were divided into quartiles, and logistic regression was performed to yield quartile-specific odds ratios (ORs) for the outcome. Model A was developed, assigning weighted values to each quartile, based on its corresponding OR for the outcome. An individual predictive score was calculated for each dog by summating all weighted values. Model B was a multivariable logistic regression model. Receiver operator characteristic (ROC) analyses were performed to assess models' performance and to calculate sensitivity, specificity, and optimal cutoff points. The overall mortality rate was 53%. The total predictive score (Model A) was negatively and significantly (P < 0.001) associated with probability of survival. The areas under the ROC curve for Models A and B were 0.92 and 0.86, respectively. The optimal cutoff score for Model A was 35.0, corresponding to sensitivity of 93% and specificity of 86%, correctly classifying 90% of the cases. Conclusions and Clinical Relevance: The proposed models are applicable, allowing objective assessment of the severity and prognosis of heatstroke in dogs; however, they should be validated further in an independent cohort, and used cautiously for assessment of individual cases.