Accuracy of electronic surveillance of catheter-associated urinary tract infection at an academic medical center

Abstract

Objective: To develop and validate a methodology for electronic surveillance of catheter-associated urinary tract infections (CAUTIs).

Design: Diagnostic accuracy study.

Setting: A 425-bed university hospital.

Subjects: A total of 1,695 unique inpatient encounters from November 2009 through November 2010 with a high clinical suspicion of CAUTI.

Methods: An algorithm was developed to identify incident CAUTIs from electronic health records (EHRs) on the basis of the Centers for Disease Control and Prevention (CDC) surveillance definition. CAUTIs identified by electronic surveillance were compared with the reference standard of manual surveillance by infection preventionists. To determine diagnostic accuracy, we created 2 × 2 tables, one unadjusted and one adjusted for misclassification using chart review and case adjudication. Unadjusted and adjusted test statistics (percent agreement, sensitivity, specificity, positive predictive value [PPV], negative predictive value [NPV], and κ) were calculated.

Results: Electronic surveillance identified 64 CAUTIs compared with manual surveillance, which identified 19 CAUTIs for 97% agreement, 79% sensitivity, 97% sensitivity, 23% PPV, 100% NPV, and κ of .33. Compared with the reference standard adjusted for misclassification, which identified 55 CAUTIs, electronic surveillance had 98% agreement, 80% sensitivity, 99% specificity, 69% PPV, 99% NPV, and κ of .71.

Conclusion: The electronic surveillance methodology had a high NPV and a low PPV compared with the reference standard, indicating a role of the electronic algorithm in screening data sets to exclude cases. However, the PPV markedly improved compared with the reference standard adjusted for misclassification, suggesting a future role in surveillance with improvements in EHRs.