Evaluation of a prognostic scoring system for dogs managed with hemodialysis

OBJECTIVE:To investigate prognostic models in a cohort of dogs with acute kidney injury (AKI) and acute on chronic kidney disease (AKI/CKD) managed by hemodialysis. DESIGN:Retrospective study from July 2011 to November 2014. SETTING:University Veterinary Teaching Hospital. ANIMALS:Forty dogs with historical, clinical, imaging, and laboratory findings consistent with AKI or AKI/CKD managed with intermittent hemodialysis were included. INTERVENTIONS:Scoring system models previously established by Segev et al for outcome prediction in dogs with AKI were applied to all dogs. RESULTS:Models A, B, and C correctly classified outcomes in 68%, 83%, and 85% of cases, respectively. In our cohort Model A showed sensitivity of 58% and specificity of 86%, Model B showed sensitivity of 79% and specificity of 87%, Model C showed sensitivity of 86% and specificity of 84%. The presence of anuria (P < 0.0002), respiratory complications (P < 0.0001), disseminated intravascular coagulation (DIC) (P = 0.0004), grade of AKI (P = 0.0023), pancreatitis (P = 0.0001), and systemic inflammatory response syndrome (SIRS) (P = 0.0001) was significantly higher in nonsurvivors compared with survivors. CONCLUSIONS:In our cohort of patients, Segev's model C showed the best sensitivity and specificity for predicting prognosis, while model A had lower sensitivity. In our cohort of dialysis patients, the presence of respiratory complications, DIC, SIRS, and pancreatitis at hospitalization, were correlated with a poor prognosis.