Surveillance length and validity of benchmarks for central line-associated bloodstream infection incidence rates in intensive care units.

Abstract

INTRODUCTION: Several national and regional central line-associated bloodstream infections (CLABSI) surveillance programs do not require continuous hospital participation. We evaluated the effect of different hospital participation requirements on the validity of annual CLABSI incidence rate benchmarks for intensive care units (ICUs). METHODS: We estimated the annual pooled CLABSI incidence rates for both a real regional (<100 ICUs) and a simulated national (600 ICUs) surveillance program, which were used as a reference for the simulations. We simulated scenarios where the annual surveillance participation was randomly or non-randomly reduced. Each scenario's annual pooled CLABSI incidence rate was estimated and compared to the reference rates in terms of validity, bias, and proportion of simulation iterations that presented valid estimates (ideal if ≥ 90%). RESULTS: All random scenarios generated valid CLABSI incidence rates estimates (bias -0.37 to 0.07 CLABSI/1000 CVC-days), while non-random scenarios presented a wide range of valid estimates (0 to 100%) and higher bias (-2.18 to 1.27 CLABSI/1000 CVC-days). In random scenarios, the higher the number of participating ICUs, the shorter the participation required to generate ≥ 90% valid replicates. While participation requirements in a countrywide program ranged from 3 to 13 surveillance blocks (1 block = 28 days), requirements for a regional program ranged from 9 to 13 blocks. CONCLUSIONS: Based on the results of our model of national CLABSI reporting, the shortening of participation requirements may be suitable for nationwide ICU CLABSI surveillance programs if participation months are randomly chosen. However, our regional models showed that regional programs should opt for continuous participation to avoid biased benchmarks.