BACKGROUND: Validation of facility-reported healthcare-associated infection (HAI) data is critical to ensure completeness and accuracy of reporting. It also identifies opportunities for improvement in classification and reporting of HAIs. North Carolina currently has no system for ongoing validation of HAI data. We conducted a validation of 2014 central line-associated bloodstream infection (CLABSI) and laboratory-identified Clostridium difficile(CDI) infection data reported through the National Healthcare Safety Network to assess performance characteristics and identify common reasons for misreporting.
METHODS: We developed a validation protocol based on CDC guidelines. We selected 28 acute care hospitals (ACHs) that reported significantly higher or lower than predicted numbers of CLABSI or CDI events. For CLABSI validation, we randomly selected a subset of positive blood cultures for each ACH and reviewed corresponding medical records onsite to assess for over- or under-reporting. For CDI validation, we compared a list of reported CDI events with a list of all CDI laboratory results submitted by each ACH via secure electronic transfer. We calculated sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) cumulatively and for each facility. We shared these results and identified areas for improvement in aggregate and within each facility.
RESULTS: Of selected ACHs, ten of 14 (71%) participated in CLABSI validation and 8 of 14 (57%) participated in CDI validation, representing approximately 16% of NC reporting hospitals. Facilities were located throughout NC and ranged in size from less than 100 to greater than 400 beds. Overall sensitivity was 79% (95% CI 62-91%) for CLABSI and 60% (95% CI 50-70%) for CDI. Specificity and PPV were 100% for CLABSI. Specificity was 88% (95% CI 79-94) and PPV was 85% (95%CI 73-93%) for CDI. NPV was 97% (95% CI 95-99%) for CLABSI, and 52% (95% CI 42-63%) for CDI. CLABSIs were often not reported due to incorrect classification as bloodstream infections secondary to an alternate primary infection site. The primary reporting error for CDIs was non-report of community-onset events.
CONCLUSIONS: Our validation demonstrated underreporting was more frequent for CDI than CLABSI. While CDI sensitivity was low, CLABSI sensitivity was similar to estimates from other states and to results of an external validation conducted in NC in 2012. Community-onset events do not contribute to the facility’s infection count, but are factored into risk adjustments when comparing a facility to the national baseline. Underreporting these events artificially elevates this metric. We plan to provide education addressing common reporting errors to increase data accuracy.