NHSN Definitions for CLABSI: Preventing the Unpreventable?
There is a nice Concise Communication in the May issue of ICHE out of the University of Michigan (Go Blue). The investigators were interested in the specificity of NHSN CLABSI definitions in their Critical Care Medicine Unit (CCMU) and their SICU during the period after implementation of CLABSI bundles. They excluded from review BSIs that were determined to be central line-related BSIs (CLRBSI) since they required positive cath-tip cultures or differential time to positivity. They then had all CLABSIs, as determined by unit-designated infection preventionists using NHSN definitions, reviewed further by two non-blinded ID physicians.
To cut to the chase, since this post shouldn't be longer than the original concise communication, there were 30 non-secondary BSIs in the CCMU, 10 were CLRBSI and 20 were CLABSI. Of the 20 CLABSI, 9 (45%) were considered to have central line sources, 9 were considered contaminated blood cultures and 2 were considered transient post-op BSIs. Of the 8 non-secondary BSIs in the SICU, there were no CLRBSI and 5 (63%) were intra-abdominal sources and 1 unknown non central-line source leaving 2 confirmed CLABSI. Thus, of the original 38 IP-reported CLABSIs in the two ICUs, ID physicians confirmed 21, leaving them with an positive predictive value of 55%; close to flipping a coin.
Now, I didn't read the discussion section of the paper (who has the time!), however, since they only looked at IP determined CLABSIs they can't report sensitivity or specificity; in fact they didn't calculate the PPV, I did that. Given that so many CLABSIs appeared to be contaminated blood cultures, efforts could now be directed to preventing those, but the clinical benefit to the patient would likely be small apart from a few avoiding unnecessary antibiotic exposure. Efforts spent limiting those non-BSI CLABSIs, I suspect will be undertaken since we all have to get to zero. It's just sad that those efforts won't help our patients.
To cut to the chase, since this post shouldn't be longer than the original concise communication, there were 30 non-secondary BSIs in the CCMU, 10 were CLRBSI and 20 were CLABSI. Of the 20 CLABSI, 9 (45%) were considered to have central line sources, 9 were considered contaminated blood cultures and 2 were considered transient post-op BSIs. Of the 8 non-secondary BSIs in the SICU, there were no CLRBSI and 5 (63%) were intra-abdominal sources and 1 unknown non central-line source leaving 2 confirmed CLABSI. Thus, of the original 38 IP-reported CLABSIs in the two ICUs, ID physicians confirmed 21, leaving them with an positive predictive value of 55%; close to flipping a coin.
Now, I didn't read the discussion section of the paper (who has the time!), however, since they only looked at IP determined CLABSIs they can't report sensitivity or specificity; in fact they didn't calculate the PPV, I did that. Given that so many CLABSIs appeared to be contaminated blood cultures, efforts could now be directed to preventing those, but the clinical benefit to the patient would likely be small apart from a few avoiding unnecessary antibiotic exposure. Efforts spent limiting those non-BSI CLABSIs, I suspect will be undertaken since we all have to get to zero. It's just sad that those efforts won't help our patients.
I just reviewed the CLABSIs for our healthcare system for 2010. On chart review, 3 of the 9 infections involved single positive blood cultures. Two appeared to be due to skin contaminants (VRE), and one a transient GI-related candidemia, Semiquantitative catheter tip cultures were negative.
ReplyDeleteGary Kravitz MD
St. Paul Infectious Disease Assoc.
Wouldn't the CLABSI definition have much greater specificty if >1 positive blood culture was required for "recognized" pathogens?
Agree. The specificity would be improved by requiring >1 positive culture. Alternatively, the definition could be re-written to exclude blood cultures that grow gut flora.
ReplyDeleteGary Kravitz replies:
ReplyDeleteI just reviewed our CLABSI data in Minnesota and found that only 9/15 cases meeting the NHSN surveillance definition my clinical definition (+PPV 60%; very similar to Univ of Michigan). The PPV rose to 100% if positive cultures obtained in the first 48 hr after catheter placement were excluded, > 1 blood culture was required for "recognized pathogens" and neutropenic patients were excluded.
Using these modifications, I think we could then focus on preventing that which is preventable.
Recent studies concur with this Michigan study:
ReplyDeleteThe survey based study from Niedner in Michigan,(AJIC) the Bachman, Melchreit & Rodriguez from the Connecticut DPH, (AJIC) and Lin, Hota & Kahn in Chicago (JAMA) where they used a computer algorithm vs IP determination
My thought when reading these studies is that with any self reported data, there is room for major variance. The economic influences (CMS reimbursement issues), regulatory pressures (CMS, Joint Commission and state survey) and human factors involved including FTE availability, time constraints, pride in facility and pride in job and software standardization have an impact.
Preventing HAI requires vigilance and best practice at the point of care, as well as from careful analysis of data. I'm afraid too many IPs are stuck behind computers.
Deb
I thought these quotes were important.
"Strength of belief in infection prevention myth and general clinical knowledge deficits: a factor most commonly believed to contribute to CA-BSI occurrences was patient risk factors, not central line maintenance or insertion practices", ( Niedner, et al., page 585)
"The median infection preventionist rate 3.3 (interquartile range [IQR], 2.0-4.5)
computer algorithm central line–associated BSI rate 9.0 (IQR, 6.3-11.3) infections per 1000 central line–days, respectively. Overall correlation between computer algorithm and infection preventionist rates was weak (p=0.34)" (Lin, et al. p. 2035)
"While 100% of surveyed infection control professionals (ICPs) indicate using the CDC definition for
CA-BSI, survey vignettes testing how this definition is applied indicate that, strictly speaking, none actually
do. All 10 of the ICPs included or excluded at least 1 line type in a manner inconsistent with the CDC definition, affecting both numerators and denominators
for rate calculations." (Niedner, et al. page 590-591)
(and the only reason I have time to read the entire studies is that I'm doing a paper on the state of the science related to infection and infection prevention in elders, specifically in long term care LOL)