Validation of a Sampling Method to Collect Exposure Data for Central-Line-Associated Bloodstream Infections

Infect Control Hosp Epidemiol. 2016 May;37(5):549-54. doi: 10.1017/ice.2015.344. Epub 2016 Jan 13.

Abstract

Objective: Surveillance of central-line-associated bloodstream infections requires the labor-intensive counting of central-line days (CLDs). This workload could be reduced by sampling. Our objective was to evaluate the accuracy of various sampling strategies in the estimation of CLDs in intensive care units (ICUs) and to establish a set of rules to identify optimal sampling strategies depending on ICU characteristics.

Design: Analyses of existing data collected according to the European protocol for patient-based surveillance of ICU-acquired infections in Belgium between 2004 and 2012.

Setting and participants: CLD data were reported by 56 ICUs in 39 hospitals during 364 trimesters.

Methods: We compared estimated CLD data obtained from weekly and monthly sampling schemes with the observed exhaustive CLD data over the trimester by assessing the CLD percentage error (ie, observed CLDs - estimated CLDs/observed CLDs). We identified predictors of improved accuracy using linear mixed models.

Results: When sampling once per week or 3 times per month, 80% of ICU trimesters had a CLD percentage error within 10%. When sampling twice per week, this was >90% of ICU trimesters. Sampling on Tuesdays provided the best estimations. In the linear mixed model, the observed CLD count was the best predictor for a smaller percentage error. The following sampling strategies provided an estimate within 10% of the actual CLD for 97% of the ICU trimesters with 90% confidence: 3 times per month in an ICU with >650 CLDs per trimester or each Tuesday in an ICU with >480 CLDs per trimester.

Conclusion: Sampling of CLDs provides an acceptable alternative to daily collection of CLD data.

Publication types

  • Multicenter Study
  • Research Support, Non-U.S. Gov't
  • Validation Study

MeSH terms

  • Bacteremia / epidemiology*
  • Belgium / epidemiology
  • Catheterization, Central Venous / adverse effects*
  • Cross Infection / epidemiology*
  • Hospitals / statistics & numerical data*
  • Humans
  • Intensive Care Units / statistics & numerical data*
  • Linear Models
  • Selection Bias
  • Time Factors