Predictors for Identifying Burn Sepsis and Performance vs Existing Criteria

• Published in: Journal of burn care & research 2018-10, Vol.39 (6), p.982-988
• Main Authors: Hill, David M, Percy, Matthew D, Velamuri, Sai R, Lanfranco, Julio, Romero Legro, Ivan, Sinclair, Scott E, Hickerson, William L
• Format: Article
• Language: English
• Subjects: Adolescent; Adult; Burn Units; Burns - complications; Diagnosis, Differential; Female; Humans; Male; Middle Aged; Predictive Value of Tests; Retrospective Studies; Sensitivity and Specificity; Sepsis - diagnosis; Sepsis - microbiology
• ISSN: 1559-047X; 1559-0488
• DOI: 10.1093/jbcr/iry022

Sepsis is the largest cause of mortality in thermally injured patients. Traditional systemic inflammatory response syndrome (SIRS) criteria do not aid diagnosis of sepsis in burn centers. Studies have attempted identification of the best indicators of sepsis in the thermal injured patient, but predictive variables are inconsistent across the various studies. Currently, consensus guidelines lack evidential support as to which patients will benefit most from prompt antimicrobial therapy. The purpose of our study was to evaluate novel diagnostic parameters for thermal injured patients with known sepsis and compare these parameters with existing diagnostic criteria. This study was a retrospective, electronic medical record review. Baseline demographics were analyzed utilizing chi-square, Mann-Whitney U test, or t test. Each patient served as their own control. Generalized linear mixed modeling was utilized for univariable and multivariable analysis. Several models with ≤6 variables each were built with the top performing variables. Performance was analyzed using area under receiver operating curves, sensitivity, specificity, positive predictive value, and negative predictive value. Three hundred and ninety-nine patients were screened. Twenty-nine patients remained after exclusions, leaving 198 blood culture results (62 positive) for analysis. Forty variables were statistically significant during univariable analysis. From multivariable analysis, the best performing model was: Temperature > 39°C or < 36°C, heart rate > 130 beats/min, 10% decrease in mean arterial pressure, and gastric residual volumes twice the feeding rate. Meeting at least one variable of the presented model best identified incidence of sepsis with positive bloodstream infections and outperformed current models in our patients.