Bayesian evaluation of sensitivity and specificity of blood culture media and hypoglycemia in sepsis‐suspected calves

Abstract

Abstract Background Sepsis is a life‐threatening condition for which critically important antimicrobials are often indicated. The value of blood culture for sepsis is indisputable, but appropriate guidelines on sampling and interpretation are currently lacking in cattle. Objective Compare the diagnostic accuracy of 2 blood culture media (pediatric plus [PP] and plus aerobic [PA]) and hypoglycemia for bacteremia detection. Estimate the contamination risk of blood cultures in critically ill calves. Animals One hundred twenty‐six critically ill calves, 0 to 114 days. Methods Retrospective cross‐sectional study in which the performance of PP, PA and hypoglycemia to diagnose sepsis was assessed using a Bayesian latent class model. A Cox proportional hazards model was used to compare time to positivity (TTP). Potential contamination was descriptively analyzed. Isolates were considered relevant when they were; member of the Enterobacterales, isolated from both blood cultures vials, or well‐known, significant bovine pathogens. Results The sensitivities for PP, PA, and hypoglycemia were higher when excluding assumed contaminants; 68.7% (95% credibility interval = 30.5%‐93.7%), 87.5% (47.0%‐99.5%), and 61.3% (49.7%‐72.4%), respectively. Specificity was estimated at 95.1% (82.2%‐99.7%), 94.2% (80.7%‐99.7%), and 72.4% (64.6%‐79.6%), respectively. Out of 121 interpretable samples, 14.9% grew a presumed contaminant in PA, PP, or both. There was no significant difference in the TTP between PA and PP. Conclusions and Clinical Importance PA and PP appear to outperform hypoglycemia as diagnostic tests for sepsis. PA seems most sensitive, but a larger sample size is required to verify this. Accuracy increased greatly after excluding assumed contaminants. The type of culture did not influence TTP or the contamination rate.

Keywords

• contamination
• hemoculture
• rational antimicrobial use
• sepsis
• volume