Single-center study of interrater agreement in the identification of central line–associated bloodstream infection

Abstract

Interrater reliability of central line-associated bloodstream infection (CLABSI) determination has not been well studied. The present study evaluated interrater reliability between infection preventionists (IPs) for CLABSI- and other bloodstream infection (BSI)-related factors and examined whether any nurse characteristics are associated with interrater reliability. A total of 165 blood cultures were reviewed by 2 IPs assigned at random. Reliability outcomes were CLABSI, infection type (hospital- or community-acquired), presence of a central line, primary versus secondary BSI, secondary source of BSI, and IP-determined source of BSI (primary, secondary, or indeterminate). Kappa coefficients were calculated. Logistic regression was used to evaluate associations between IP characteristics and agreement on diagnosis of CLABSI. CLABSI agreement was moderate in IP pairs (κ = 0.562 ± 0.080) and not associated with IP characteristics. After controlling for IP characteristics associated with secondary outcomes, agreement regarding secondary source was more likely in pairs with a larger absolute difference in years employed (P = .013), and agreement regarding infection source was more likely in pairs with larger differences in years employed and duration of certification (P = .025). The rate of IP agreement regarding CLABSI was moderate and not associated with IP characteristics, reflecting adequate training. Education and reassessment of definitions may promote higher rates of agreement between IPs.