194. Increasing the Reliability of Automated Surveillance for Central Line-Associated Bloodstream Infections
Session: Poster Abstract Session: Catheter-associated BSIs
Thursday, October 3, 2013
Room: The Moscone Center: Poster Hall C
Posters
  • IDWeek_NICER_Poster_Final.pdf (199.5 kB)
  • Objective: To increase reliability of an existing algorithm with new rules to better identify bloodstream infections that are secondary to other hospital-acquired infections, as defined by the National Healthcare Safety Network.

    Background: An automated surveillance algorithm has been previously described by our group, and is used to identify central line-associated bloodstream infections (CLABSIs) outside of the intensive care units (ICUs). 

    Methods: Patients with positive blood cultures in 17 ICUs between 1/1/2011 and 6/30/2011 were reviewed.  CLABSI determinations for these patients were based on two sources: manual surveillance by Infection Preventionists (IPs), and automated surveillance from the existing algorithm.  Discrepancies, in which CLABSI events were confirmed by one surveillance method, but not the other, were evaluated by IPs to determine root causes.  Secondary infection sites were identified in the majority of discrepant cases.  Based on these findings, new rules to identify secondary sites were added to the algorithm.  Sensitivity, specificity, predictive values, and kappa were calculated for each of the new models.

    Results: Of 643 positive blood cultures reviewed, 68 (10.6%) were identified as CLABSIs by automated surveillance, while 38 (5.9%) were confirmed by manual surveillance.  New rules were trialed to identify organisms as CLABSIs if they did not meet one, or a combination of, the following: (1) matching organisms (by genus and species) cultured from any other site; (2) any organisms cultured from a sterile site; (3) any organisms cultured from skin/wound (SW); (4) anyorganisms cultured from the respiratory tract (RT).  Additionally, a variety of exclusions related to yeast were trialed with rule 1, as yeast may represent colonization, rather than infection, when cultured from the respiratory tract or skin.  The best-fit model included new rules 1 and 2 (Table 1).

    Table 1: Performance of New Rules for CLABSI Prediction

     

    Predicted CLABSIs

    N (%)

    Sensitivity

    %

    Specificity

    %

    PPV

    %

    NPV

    %

    Kappa

    Existing Algorithm

    68 (10.6)

    78.9

    93.7

    44.1

    98.6

    .530

    Best-Fit Model: Rule 1 (excluding yeast when cultured from RT or SW) + Rule 2

    54 (8.4)

    81.6

    96.2

    57.4

    98.8

    .650

    Conclusion: Improvements to the existing algorithm increased reliability, and improved its usefulness in tracking CLABSIs outside of the ICUs.

    Rachael Snyders, MPH, BSN, RN1, Kathleen Mcmullen, MPH, CIC1, Ashleigh Goris, MPH, BSN, RN2, Joshua A Doherty, BS2 and Keith F. Woeltje, MD, PhD2,3, (1)Patient Safety and Quality, Barnes-Jewish Hospital, St. Louis, MO, (2)BJC HealthCare, St. Louis, MO, (3)Washington University School of Medicine, St. Louis, MO

    Disclosures:

    R. Snyders, None

    K. Mcmullen, None

    A. Goris, None

    J. A. Doherty, None

    K. F. Woeltje, None

    Findings in the abstracts are embargoed until 12:01 a.m. PST, Oct. 2nd with the exception of research findings presented at the IDWeek press conferences.