307 Automated Surveillance for Catheter-Associated Bloodstream Infections Outside of the Intensive Care Unit

Saturday, April 2, 2011: 3:15 PM
Cortez Ballroom (Hilton Anatole)
Kathleen M. McMullen, MPH , Barnes Jewish Hospital, St. Charles, MO
Ashleigh J. Goris, MPH, CIC , Progress West HealthCare Center, O Fallon, MO
Joshua A. Doherty, BS , BJC Healthcare, St. Louis, MO
Diane Hopkins-Broyles, RN, MSN, CIC , BJC HealthCare, St. Louis, MO
Keith F. Woeltje, MD, PhD , Washington University School of Medicine, St. Louis, MO

Background: Manual surveillance for catheter-associated bloodstream infections (CLABSIs) by infection prevention practitioners is time consuming and often limited to Intensive Care Units (ICUs).  An automated surveillance system using existing databases with patient level variables and microbiology data has been previously described by our group. 

Objective: To update this automated system with the newer Centers for Disease Control and Prevention National Healthcare Safety Network (NHSN) CLABSI definition and to further increase reliability of the algorithm.

Methods: Patients with a positive blood culture in four non-ICU wards at Barnes-Jewish Hospital between July 1, 2005 and Dec 31, 2006 were evaluated.  CLABSI determination for these patients was made via two sources; a manual chart review and automated review from electronically available data.  Agreement between these two sources was used to develop the ‘best-fit' electronic algorithm that used the following rules to identify a CLABSI: (1) culture positive > 48 hours after admission; (2) organism was not a common skin contaminant OR was either confirmed by a duplicate culture within 5 days or was treated with vancomycin; (3) patient had a CVC; (4) organism was not grown from a wound, urine, respiratory, sterile, or non-sterile site anytime during the admission after the positive blood culture.  The algorithm was updated to meet the newer NHSN CLABSI definitions with a change to rule 2: organism not a common skin contaminant OR is confirmed by a duplicate culture within 3 days.  Further improvements to the algorithm were investigated, including rules with presence of fever and varieties of timing for rule 4.  Sensitivity, specificity, predictive values, and Pearson's correlation were calculated for the various new rule sets, using manual chart review as the reference standard.

Results: During the study period, 391 positive blood cultures from 331 patients were evaluated. Eight five (22%) of these were confirmed to be CLABSI by manual chart review.  The best fit model included the original rules 1 and 3, the updated rule 2, and rule 4 changed to ‘organism was not grown from a wound, urine, respiratory, sterile, or non-sterile site during the hospital admission before positive blood culture' (Table 1).  CLABSIs remain slightly over predicted compared to manual chart review.

Conclusions: Further refinement of the electronic algorithm attained compliance with new NHSN definitions and increased accuracy overall, improving its usefulness to track non-ICU CLABSI trends.


Table 1:  Performance of Alternative Methods for CABSI Prediction

Rules

Predicted

CABSIs

Sensitivity

%

Specificity

%

PPV

%

NPV

%

Pearson's

Correlation

1234

80

87.1

98.0

92.5

97.0

.875

1234 (4-before positive)

90

95.2

97.5

90.0

99.2

.908

1234 (2-no fever)

92

87.1

95.1

80.4

96.9

.797

1234 (2-no fever, 4-before positive)

101

94.1

94.3

79.2

98.6

.829