Clinical Predictors of Positive Postoperative Blood Cultures. Issue 2 (February 2018)

• Record Type: Journal Article
• Authors: Copeland-Halperin, Libby R.; Stodghill, Joshua; Emery, Erica; Trickey, Amber W.; Dort, Jonathan
• Is Part Of: Annals of surgery. Volume 267:Issue 2(2018:Feb.)
• Publication Date: 2018-02
• Subjects: bacteremia -- cost-benefit analysis -- postoperative care; Surgery -- Periodicals; 617.005
• DOI: 10.1097/SLA.0000000000002077
• Languages: English
• ISSNs: 0003-4932

Abstract

Abstract : Objective: To define clinical features of surgical patients in whom postoperative blood cultures are likely to identify pathogens. Background: Bacteremia is a worrisome postoperative complication and blood cultures (BCx) are routinely used for evaluation of postoperative bacteremia, but are costly and not always diagnostic. Better methods are needed to select patients in whom BCx identify pathogens. Methods: We reviewed records of patients ≥18 years old with BCx drawn ⩽10 days after surgery in 2013 seeking independent predictors of positive cultures by simple and multiple logistic regression models with statistical significance at α = 0.05. Results: Of 1804 BCx, excluding contaminants yielded 1780 cultures among 746 patients for analysis. The yield was low, with only 4% identifying potential pathogens. Positive BCx were most common after cardiac, ear/nose/throat, obstetric, and urologic procedures [odds ratio (OR) =10.3, P < 0.001 vs low-yield procedures: eg, gynecologic, neurosurgical, plastic surgical, podiatric, transplant]. Cultures more often grew pathogens when drawn in association with higher peak temperature (Tmax, P = 0.001) and longer interval from procedure to Tmax ( P = 0.001). Antibiotic therapy at time of culture reduced yield (2.9% with vs 5.5% without antibiotics, P = 0.007). Multivariable logistic regression analysis found antibiotics at culture, procedure specialty, Tmax, and postoperative timing of Tmax were associated with blood culture results. Conclusions: Ordering blood cultures based on fever or another single predictor inconsistently identifies pathogens. Our dataset, the largest available, identify clinical predictors in the first 10 postoperative days to guide identification of patients with bacteremia.