2008 HSR&D National Meeting Abstract
3117 — Objective Simplified Surveillance Criteria Improve Comparisons of Central-Line Associated Blood Stream Infection (CLA-BSI) Rates between Institutions
Greene T (VA Salt Lake City Health Care System), Samore MH
(VA Salt Lake City Health Care System), Mayer J
(VA Salt Lake City Health Care System), Sauer B
(VA Salt Lake City Health Care System), Rubin MA
(VA Salt Lake City Health Care System)
Application of subjective clinical surveillance (SCS) criteria to identify CLA-BSI requires subjective judgments by Infection Control Practitioners (ICPs), which may adversely affect reliability. By comparison, objective simplified surveillance (OSS) criteria based solely on microbiologic data provide increased reliability, but may lead to higher misclassification rates. We characterize circumstances under which OSS criteria outperform SCS criteria when the goal is to compare CLA-BSI rates between institutions.
We use calculations based on variances of random variables to approximate the maximum ratio in the misclassification rate (defined as the unweighted average of the false positive and false negative rates) between OSS and SCS criteria such that OSS provides a higher probability of correctly ranking CLA-BSI rates across institutions. This ratio depends heavily on a) the numbers of episodes with positive blood cultures reviewed for a possible CLA-BSI, and b) the ratio between the two criteria in the average of the between-Institution Coefficients of Variation (I-CV) for false positive rates and false negative rates. The I-CV is expected to be larger for SCS than OSS due to subjective judgments of ICPs.
Under one set of realistic assumptions, if the OSS and SCS have I-CVs of 5% and 15%, respectively, and 100 episodes are evaluated, OSS outperforms SCS as long as the OSS misclassification rate is no more than 15% larger than the SCS misclassification rate. If 300 episodes are evaluated under the same conditions, OSS outperforms SCS as long as the OSS misclassification rate is no more than 35% larger than the SCS misclassification rate. If the I-CVs are 5% and 25% for OSS and SCS, and 100 (or 300) episodes are evaluated, OSS outperforms SCS when the OSS misclassification rate is no more than 39% (or 73%) larger than the SCS misclassification rate.
Use of OSS criteria should provide more accurate rankings of CLA-BSI rates among institutions even if OSS has substantially larger misclassification rates than SCS, particularly when large numbers of episodes are evaluated.
Use of simplified objective criteria are likely to improve comparisons of CLA-BSI rates between institutions compared to subjective clinical criteria.