279. A Decision Tree Using Clinical Characteristics to Predict a Hospitalized Child’s Risk of a Multidrug-Resistant Gram-Negative Bloodstream Infection
Session: Poster Abstract Session: Pediatric Antimicrobial and Diagnostic Stewardship
Thursday, October 4, 2018
Room: S Poster Hall
Posters
  • Predict MDR poster for IDweek_V2.pdf (667.7 kB)
  • Title: A Decision Tree Using Clinical Characteristics to Predict a Hospitalized Child’s Risk of a Multidrug-Resistant Gram-Negative Bloodstream Infection

    Background: As the threat of multidrug-resistant Gram-negative (MDRGN) bacteria rises, recognizing children at high risk of bloodstream infections with bacteria resistant to commonly prescribed empiric antibiotics is critical.  We developed a decision tree to predict which pediatric bloodstream infections were due to MDRGN bacteria resistant to cefepime or piperacillin-tazobactam, commonly prescribed empiric antibiotics.

    Methods: We conducted a longitudinal retrospective cohort study at the Children’s Hospital of Pittsburgh including all admitted patients with a Gram-negative bloodstream infection from June 2009 to June 2015. Episodes of bloodstream infection were considered unique if at least 30 days had elapsed since the previous bloodstream infection. Logistic regression was performed to identify notable risk factors. A decision tree describing the risk of a MDRGN infection was developed using recursive partitioning based on clinical characteristics available at the time of presentation.

    Results: 689 episodes of Gram-negative bloodstream infections occurred during the study period among 387 patients. 28% of infections were multidrug-resistant (MDR). The decision tree separated patients into higher or lower risk groups based upon history of prior carbapenem treatment for seven or more days, having a prior MDR infection within 6 months, intestinal transplant status, age 3 years or older, and 7 or more prior episodes of bacteremia. The sensitivity to classify high-risk of MDR was 46% and the specificity was 92% based on leave one out cross validation. For patients who had more than one episode, 30% of initially non-MDR infections were subsequently MDR.

    Conclusion: A decision tree using readily available clinical characteristics may be helpful to identify pediatric patients at higher risk of bloodstream infection due to a MDRGN organism resistant to common empirical antibiotic therapy.

    Figure 1. Decision Tree. Risk of Resistance Presented as Percentage.

    Notes:

    Abbreviations: Y- yes; N- no. Grey hexagons represent high-risk of multidrug-resistant (MDR) groups while squares represent low-risk of MDR groups.

     

    Anna Sick-Samuels, MD, MPH, Pediatrics, Johns Hopkins University School of Medicine, Baltimore, MD, Katherine Goodman, JD, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, Glenn Rapsinski, MD, Children's Hospital of Pittsburgh, Pittsburgh, PA, Elizabeth Colantuoni, PhD, Biostatistics, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, Andrew Nowalk, MD, PhD, Pediatrics, Children's Hospital of Pittsburgh of UPMC, Pittsburgh, PA and Pranita Tamma, MD, MHS, Johns Hopkins University School of Medicine, Baltimore, MD

    Disclosures:

    A. Sick-Samuels, None

    K. Goodman, None

    G. Rapsinski, None

    E. Colantuoni, None

    A. Nowalk, None

    P. Tamma, None

    Findings in the abstracts are embargoed until 12:01 a.m. PDT, Wednesday Oct. 3rd with the exception of research findings presented at the IDWeek press conferences.