Discussion

Our findings suggest that the NHSN surveillance definitions for CLABSI have moderately high sensitivity, specificity, PPV, and NPV compared to the gold standard of physician review, in part because the degree of underreporting (ie, 14% false negatives) was offset by a comparable degree of overreporting (ie, 9% false positives). Underreported cases are not as closely scrutinized in the daily work of infection prevention and represent potentially missed opportunities for the prevention of CLABSIs. Thus, a new quality metric is currently being considered to include all hospital acquired bacteremia, regardless of the presence of a central line. ^{Reference Dantes, Abbo and Anderson4} A major driver of underreporting was the misclassification of BSIs as being secondary to pneumonia, particularly in cases in which pulmonary imaging was the only evidence supporting a diagnosis of pneumonia, accounting for 6 (30%) of 20 of false negatives. However, there are other reasons for chest imaging to be abnormal in hospitalized patients (eg, acute heart failure), and a matching organism in a respiratory specimen could support the attribution of a BSI to pneumonia in the correct clinical context. ^{Reference Zhang, Yang and Makam5} Additionally, 10 (50%) of 20 false negatives had Staphylococcus aureus bacteremia, suggesting that clinicians typically attribute S. aureus (ie, a skin commensal that frequently colonizes catheters) to CLABSI, whereas the NHSN definition may misclassify these BSIs as being secondary to other infections. ^{Reference Hindy, Quintero-Martinez and Lee6}

Even though overreported cases receive similar attention as appropriately reported cases, overreporting is a dissatisfier for frontline healthcare providers whose buy-in is essential for the implementation of key infection prevention practices. The major drivers of overreporting were the misclassification of abdominal infections, pneumonia, and UTIs as NHSN CLABSIs. In clinical practice, BSIs are commonly attributed to translocation in the setting of distension from obstruction (eg, of the colon or common bile duct) or mucosal injury from inflammation (eg, from a gastrointestinal bleed or colitis) in immunocompetent hosts. However, mucosal barrier injury can only be invoked for immunocompromised patients with neutropenia or graft-versus-host disease in the current NHSN definitions. ^{Reference Deitch7} Although stool and tissue cultures are not often obtained in clinical practice, select imaging studies could be used as objective evidence for BSIs from mucosal barrier injury organisms, as proposed in another study. ^{Reference Advani, Murray, Murdzek, Aniskiewicz and Bizzarro3} For pneumonia, BSIs are commonly attributed to a respiratory source using lower thresholds of growth of a matching organism from respiratory cultures (eg, <10³ colony-forming units per milliliter, few, or 1+), particularly in critically ill patients who may be receiving empiric antibiotics before bronchoscopy. ^{Reference Zhang, Yang and Makam5} For UTIs, urinary cultures are frequently used to support a diagnosis of pyelonephritis without imaging. ^{Reference Gupta, Grigoryan and Trautner8,Reference Lacy, Sidhu and Miller9} Moreover, pyelonephritis is not infrequently accompanied by bacteremia, and blood and urine cultures are rarely discordant. ^{Reference Velasco, Martínez and Moreno-Martínez10} Even in the absence of typical symptoms, a positive urine culture might warrant treatment in the presence of neutropenia, chronic obstruction from malignancy, or undifferentiated critical illness.

To our knowledge, this study is one of the few external validation studies of the NHSN surveillance definitions for CLABSI. ^{Reference Advani, Murray, Murdzek, Aniskiewicz and Bizzarro3} The major strength of this study is that both the NHSN and physician review were conducted by multiple individuals, which strengthens the internal validity of the study. However, we were limited by relatively small numbers from a single center and the potentially subjective interpretation of cases by physician review, despite being the gold standard. Larger, multicenter studies could confirm our findings and inform future iterations of the NHSN surveillance definitions. Additionally, we would strongly encourage similar validation procedures of future quality metrics such as hospital-onset bacteremia to support their clinical relevance.