Evaluation of early clinical failure criteria for gram-negative bloodstream infections

Objectives: The aim was the development of early clinical failure criteria (ECFC) to predict unfavourable outcomes in patients with Gram-negative bloodstream infections (GN-BSI). Methods: Adults with community-onset GN-BSI who survived hospitalization for >= 72 hr at Prisma Health-Midlands hospitals in Columbia, SC, USA from January 1, 2010 to June 30, 2015 were identified. Multivariable logistic regression was used to examine the association between clinical variables between 72 and 96 hr after GN-BSI and unfavourable outcomes (28-day mortality or hospital length of stay >14 days from GN-BSI onset). Results: Among 766 patients, 225 (29%) had unfavourable outcomes. After adjustments for Charlson Comorbidity Index and appropriateness of empirical antimicrobial therapy in multivariable model, predictors of unfavourable outcomes included systolic blood pressure <100 mmHg or vasopressor use (adjusted odds ratio (aOR) 1.8, 95% confidence interval (CI) 1.2-2.9), heart rate >100 beats/minute (aOR 1.7, 95% CI 1.1-2.5), respiratory rate >= 22 breaths/minute or mechanical ventilation (aOR 2.1, 95% CI 1.4-3.3), altered mental status (aOR 4.5, 95% CI 2.8-7.1), and white blood cell count >12 000/mm(3) (aOR 2.7, 95% CI 1.8-4.1) between 72 and 96 hr after index GN-BSI. Area under receiver operating characteristic curve of ECFC model in predicting unfavourable outcomes was 0.77 (0.84 and 0.71 in predicting 28-day mortality and prolonged hospitalization, respectively). Conclusions: Risk of 28-day mortality or prolonged hospitalization can be estimated between 72 and 96 hr after GN-BSI using ECFC. These criteria may have clinical utility in management of GN-BSI and may improve methodology of future investigations assessing response to antimicrobial therapy based on a standard evidence-based definition of early clinical failure. (c) 2019 European Society of Clinical Microbiology and Infectious Diseases. Published by Elsevier Ltd. All rights reserved.