Evaluation of a score designed to predict sepsis in foals

Objective: To evaluate the accuracy of a published score designed to predict sepsis in foals in a clinical setting and to evaluate the association of clinical and clinicopathological variables with sepsis. Design: Observational study. Retrospective for data from 1998. Prospective in 1999-2001. Setting: Foal intensive care unit of a university hospital. Animals: Client-owned foals of less than 10 days of age, presenting from 1998 to 2001. Interventions: None. Measurements and main results: Data from the history and physical examination, together with admission hematology, biochemistry and arterial blood gas analysis were used to generate the published sepsis scores. The same data were investigated for their statistical relationship with sepsis. The presence or absence of sepsis was determined from blood culture, culture of sites of suspected local infection, clinical course and/or post-mortem examination. The modified sepsis score was calculated for 168 foals, which were classified as septic (86), non-septic (45) or not possible to classify (37). The modified sepsis score correctly predicted sepsis in 58 out of 86 foals and non-sepsis in 34 out of 45 foals, resulting in a sensitivity of 67%, a specificity of 76%, a positive predictive value of 84% and a negative predictive value of 55%. Abnormal neutrophil cytology, an immunoglobulin concentration of less than 400 mg/dl, and low blood glucose concentration had the strongest association with sepsis. Conclusions: The low negative predictive value of the sepsis score limited its clinical utility The sepsis score should not be used to define sepsis in clinical studies, unless previously validated in the study center.