394 Blood Culture Contamination: A Cluster-randomized Cross-over Trial Evaluating the Comparative Effectiveness of Three Skin Antiseptic Interventions

Saturday, March 20, 2010: 11:45 AM
Centennial I-II (Hyatt Regency Atlanta)
Laraine Washer, MD , University of Michigan, Ann Arbor, MI
Carol Chenoweth, MD , University of Michigan, Ann Arbor, MI
Hae-won Kim, MD , University of Michigan, Ann Arbor, MI
Anurag Malani, MD , St Joseph's Mercy Hospital, Ypsilanti, MI
James Riddell, MD , University of Michigan, Ann Arbor, MI
Latoya Kuhn, MPH , Ann Arbor VA Medical Center, Ann Arbor, MI
Mary Rogers, PhD, MS , University of Michigan and Ann Arbor VA Medical Center, Ann Arbor, MI
Bernard Noeyack , University of Michigan, Ann Arbor, MI
Harry Neusius, MS, MBA , University of Michigan, Ann Arbor, MI
Sanjay Saint, MD, MPH , University of Michigan and Ann Arbor VA Medical Center, Ann Arbor, MI
Scott Flanders, MD , University of Michigan, Ann Arbor, MI

Background: Contaminated blood cultures result in unnecessary antibiotic use, increased length of stay, and additional laboratory tests, all of which increase healthcare costs.  While the optimal skin antisepsis agent for reducing blood culture contamination is not known, the cost of various agents differs substantially.

Objective: To determine the relative rates of blood culture contamination for 3 skin antisepsis interventions – 10% povidone iodine aqueous solution (PI), 2% iodine tincture (IT) and 2% chlorhexidine gluconate in 70% isopropyl alcohol (CHG) – when used by dedicated phlebotomy teams to obtain peripheral blood cultures in adult non-ICU medical and surgical patients.

Methods:   We employed a randomized cluster design using hospital floor as the unit of randomization, skin antisepsis agent as the intervention variable, and rate of contaminated blood cultures as the primary outcome.  Each antisepsis intervention was used for 5 months (PI x 5 months, IT x 5 months, CHG x 5months) on 3 separate floors with the sequence differing between floors in a cross-over design.  Dedicated phlebotomy teams collected all peripheral blood cultures using a standard protocol with skin antisepsis varying according to study design.  Random auditing to ensure compliance with the intervention was performed. Blood cultures were processed by the hospital microbiology lab according to standard protocols.  Each positive blood culture result was evaluated by 2 attending infectious diseases (ID) physicians blinded to the intervention and scored as true positive or contaminated blood culture using standardized definitions.  A third attending ID physician evaluated discordant results.   

Results: A total of 12,968 peripheral blood culture sets were evaluated over 15 months, of which 740 (5.7%) were positive.  There were 101 contaminated cultures which represented 13.6% of all positive cultures. Predominant organisms in contaminated cultures were coagulase negative staphylococcus (76.2%), micrococcus (8.9%), and Bacillus species (6.9%).   The intention to treat rates of contaminated blood cultures were similar among the 3 antiseptics (p=0.22), with 0.60% with PI (95% CI, 0.39-0.88), 0.80% with IT (95% CI, 0.55-1.11), and 0.93% with CHG (95% CI, 0.67-1.26).

Conclusions: The rate of contaminated blood cultures obtained by peripheral venipuncture did not vary significantly with antiseptic agent. The choice of antiseptic agent does not appear to impact contamination rate when blood cultures are obtained by a dedicated phlebotomy team, and perhaps should be based instead on economic considerations. To this end, PI appears to be the current agent of choice given that its acquisition cost was the lowest of the 3 agents investigated.