Objective: To assess the sensitivity, specificity, positive and negative predictive value (ppv and npv) of CLABSI data reported by TN hospitals into NHSN.
Methods: CLABSI rates for 2008 were calculated by ICU type for each facility. The sample population of facilities consisted of outlier ICUs (i.e., CLABSI rates ≤ 5th and ≥ the 95th percentile vs. NHSN 2006-2007 data) as well as a random selection of non-outlier ICUs. Microbiology laboratories provided line-lists of all positive blood cultures [BC] taken in ICUs in 2008. We aimed to review 16 medical charts with positive BCs at each facility. We over-sampled charts for BCs positive for specific organisms (Candida spp., Staphylococcus aureus, and common skin contaminants [CSC] e.g., coagulase negative staphylococcus). Classification by the TDH was considered the gold standard. We also interviewed infection preventionists in person. We visited ICUs and reviewed denominator data collection, and central line trays/carts.
Results: We reviewed 220 medical charts from 14 facilities (13-17 charts each). Of 59 CLABSIs identified by TDH, 46 were classified as a CLABSI, attributed to that ICU and entered into NHSN. Of 161 positive BCs not meeting the CLABSI definition according to TDH, 4 were in NHSN. The overall sensitivity was 78.0% (range by facility: 40-100%). The overall specificity was 97.5% (range by facility: 80-100%). Overall PPV was 92.0% (range by facility: 67-100%) and NPV was 92.4% (range by facility: 79-100%). “Over-calling” of CLABSIs by IPs was due to classifying as CLABSIs a single isolate of a CSC or 2 CSC whose susceptibilities varied by ≥2 antibiotics. “Under-calling” frequently occurred because IPs attributed positive BCs to another site because the organism was isolated from the other site; however NHSN criteria for infection at the other site were not met. This most commonly occurred for Candida spp which was isolated from the respiratory tract, but criteria for pneumonia were not met. Facilities with low CLABSI rates in general did not have low sensitivities.
Conclusions: The sensitivity, specificity, ppv and npv in our study is not comparable to other validation studies as we used a targeted approach (i.e., outlier facilities and over-sampled for specific organisms). Our targeted approach provided useful data to guide education efforts (especially gaining an understanding of reasons for misclassification). Validation of NHSN data is crucial to ensure confidence in publicly reported data.